diff --git a/.github/workflows/cloud-push-deploy.yml b/.github/workflows/cloud-push-deploy.yml new file mode 100644 index 00000000000..9290b4ab2ce --- /dev/null +++ b/.github/workflows/cloud-push-deploy.yml @@ -0,0 +1,340 @@ +name: Deploy Push Gateway Production + +on: + workflow_dispatch: + inputs: + confirmation: + description: Enter DEPLOY_PUSH_GATEWAY to shift production traffic + required: true + type: string + +permissions: + contents: read + id-token: write + +# The gateway applies its own schema at startup against the shared Cloud SQL instance, so a +# deploy is a connection-budget rollout and belongs in the same serialized group as the relay. +concurrency: + group: production-cloud-sql-rollout + cancel-in-progress: false + +defaults: + run: + working-directory: cloud + +jobs: + deploy: + if: >- + ${{ vars.ORCA_CLOUD_OPERATIONS_ENABLED == 'true' && + github.ref == 'refs/heads/main' }} + runs-on: blacksmith-2vcpu-ubuntu-2204 + environment: production + env: + GCP_PROJECT_ID: onorca-cloud + GCP_REGION: ${{ vars.PRODUCTION_GCP_REGION }} + SERVICE_NAME: orca-cloud-push + REPOSITORY_ID: orca-cloud + IMAGE_NAME: push + PUSH_ORIGIN: https://push.onorca.dev + PUSH_RUNTIME_SERVICE_ACCOUNT: orca-cloud-push@onorca-cloud.iam.gserviceaccount.com + # Scaling the serving revision must already hold, matching push_min_instances and + # push_max_instances. Terraform owns both, and the candidate inherits them from the + # service, so this deploy never passes a scaling flag: doing so would write a + # Terraform-owned field that `lifecycle.ignore_changes` does not cover, and a later + # `push_max_instances` raise would then be reverted by every deploy. These two values + # are the expected shape, asserted before the candidate is created and again on the + # candidate itself, so a deploy that would change the gateway's Cloud SQL draw fails. + PUSH_MIN_INSTANCES: 1 + PUSH_MAX_INSTANCES: 2 + CONFIRMATION: ${{ inputs.confirmation }} + steps: + - uses: actions/checkout@v4 + + - name: Require the explicit deploy confirmation + shell: bash + run: | + set -euo pipefail + test "${CONFIRMATION}" = DEPLOY_PUSH_GATEWAY + + - uses: google-github-actions/auth@v2 + with: + workload_identity_provider: ${{ vars.PRODUCTION_GCP_RELAY_DEPLOY_WORKLOAD_IDENTITY_PROVIDER }} + service_account: ${{ vars.PRODUCTION_GCP_RELAY_DEPLOY_SERVICE_ACCOUNT }} + + - uses: google-github-actions/setup-gcloud@v2 + + - uses: docker/setup-buildx-action@v3 + + - name: Configure Docker auth + run: gcloud auth configure-docker "${GCP_REGION}-docker.pkg.dev" --quiet + + # Why: the build runs before the lease. Artifact Registry is not the Cloud SQL instance, + # and a multi-minute image build inside the lease blocks every relay deploy and rehome for + # its duration. The lease below covers exactly the connection-budget window: deploy, probe, + # shift. + - name: Build and publish the immutable gateway image + shell: bash + run: | + set -euo pipefail + image_tag="${GCP_REGION}-docker.pkg.dev/${GCP_PROJECT_ID}/${REPOSITORY_ID}/${IMAGE_NAME}:sha-${GITHUB_SHA}" + docker build -f apps/push/Dockerfile -t "${image_tag}" . + docker push "${image_tag}" + digest="$(gcloud artifacts docker images describe "${image_tag}" \ + --format='value(image_summary.digest)')" + [[ "${digest}" =~ ^sha256:[a-f0-9]{64}$ ]] + echo "IMAGE=${GCP_REGION}-docker.pkg.dev/${GCP_PROJECT_ID}/${REPOSITORY_ID}/${IMAGE_NAME}@${digest}" \ + >> "${GITHUB_ENV}" + echo "IMAGE_DIGEST=${digest}" >> "${GITHUB_ENV}" + + # Held across the deploy, not just a separate schema step: the gateway opens its pool and + # applies its schema while the new revision starts, so the revision is the schema step. + - uses: ./.github/actions/cloud-sql-rollout-lease + with: + bucket: onorca-cloud-terraform-state + object: terraform/state/cloud-sql-rollout/production.lock + + # Why: the candidate inherits the serving revision's scaling. A serving revision that has + # drifted below the floor would hand the candidate a cold start on every notification, and + # one that has drifted above the ceiling would hand it a larger Cloud SQL draw than the + # rollout lease was taken for. Refuse to inherit either rather than latch it. + - name: Record the serving revision and require its Terraform-owned scaling + shell: bash + run: | + set -euo pipefail + serving="$(gcloud run services describe "${SERVICE_NAME}" \ + --project "${GCP_PROJECT_ID}" --region "${GCP_REGION}" --format=json \ + | jq -r '[.status.traffic[] | select((.percent // 0) > 0)] + | if length == 1 and .[0].percent == 100 then .[0].revisionName else empty end')" + test -n "${serving}" + floor="$(gcloud run revisions describe "${serving}" \ + --project "${GCP_PROJECT_ID}" --region "${GCP_REGION}" \ + --format="value(metadata.annotations['autoscaling.knative.dev/minScale'])")" + if [[ "${floor:-0}" -lt "${PUSH_MIN_INSTANCES}" ]]; then + echo "serving revision ${serving} holds ${floor:-0} minimum instances," \ + "below ${PUSH_MIN_INSTANCES}; deploying would inherit and latch it." >&2 + echo "Restore the floor first: gcloud run services update ${SERVICE_NAME}" \ + "--region ${GCP_REGION} --min-instances=${PUSH_MIN_INSTANCES}" >&2 + exit 1 + fi + ceiling="$(gcloud run revisions describe "${serving}" \ + --project "${GCP_PROJECT_ID}" --region "${GCP_REGION}" \ + --format="value(metadata.annotations['autoscaling.knative.dev/maxScale'])")" + test "${ceiling}" = "${PUSH_MAX_INSTANCES}" + echo "serving revision ${serving} holds ${floor} minimum and ${ceiling} maximum instances" + echo "ROLLBACK_REVISION=${serving}" >> "${GITHUB_ENV}" + + # No traffic and a per-revision tag: the candidate boots, applies schema, and is probed on + # its own URL while every phone and desktop still reaches the previous revision. + - name: Deploy the candidate revision with no traffic + shell: bash + run: | + set -euo pipefail + tag="c${GITHUB_RUN_ID}-${GITHUB_RUN_ATTEMPT}" + echo "CANDIDATE_TAG=${tag}" >> "${GITHUB_ENV}" + echo "CANDIDATE_REVISION=${SERVICE_NAME}-${tag}" >> "${GITHUB_ENV}" + gcloud run deploy "${SERVICE_NAME}" \ + --project "${GCP_PROJECT_ID}" \ + --region "${GCP_REGION}" \ + --image "${IMAGE}" \ + --tag "${tag}" \ + --revision-suffix "${tag}" \ + --no-traffic \ + --quiet + candidate="$(gcloud run services describe "${SERVICE_NAME}" \ + --project "${GCP_PROJECT_ID}" --region "${GCP_REGION}" --format=json \ + | jq -er --arg tag "${tag}" \ + '[.status.traffic[] | select(.tag == $tag)] + | if length == 1 then .[0] else error("tagged candidate is not unique") end')" + test "$(jq -r '.revisionName' <<< "${candidate}")" = "${SERVICE_NAME}-${tag}" + echo "CANDIDATE_URL=$(jq -r '.url' <<< "${candidate}")" >> "${GITHUB_ENV}" + + # A tagged revision is directly addressable and sits outside the service-wide cap, so the + # candidate and the serving revision each draw up to the ceiling during the probe window. + # The lease is taken for exactly that doubling; a candidate that inherited a wider ceiling + # would exceed it, so the inherited scaling is asserted here too. + - name: Require the candidate to serve the exact image and inherited scaling + shell: bash + run: | + set -euo pipefail + served="$(gcloud run revisions describe "${CANDIDATE_REVISION}" \ + --project "${GCP_PROJECT_ID}" --region "${GCP_REGION}" \ + --format='value(spec.containers[0].image)')" + test "${served}" = "${IMAGE}" + test "${CANDIDATE_REVISION}" != "${ROLLBACK_REVISION}" + candidate_ceiling="$(gcloud run revisions describe "${CANDIDATE_REVISION}" \ + --project "${GCP_PROJECT_ID}" --region "${GCP_REGION}" \ + --format="value(metadata.annotations['autoscaling.knative.dev/maxScale'])")" + test "${candidate_ceiling}" = "${PUSH_MAX_INSTANCES}" + + - name: Probe the candidate readiness endpoint + shell: bash + run: | + set -euo pipefail + [[ "${CANDIDATE_URL}" =~ ^https://[^/]+$ ]] + for attempt in $(seq 1 30); do + code="$(curl -sS -o "${RUNNER_TEMP}/push-ready.json" -w '%{http_code}' \ + --max-time 10 "${CANDIDATE_URL}/ready" || true)" + if test "${code}" = 200; then + jq -e . < "${RUNNER_TEMP}/push-ready.json" > /dev/null + echo "candidate ${CANDIDATE_REVISION} is ready after ${attempt} attempt(s)" + exit 0 + fi + echo "attempt ${attempt}: /ready returned ${code}" + sleep 5 + done + echo "candidate ${CANDIDATE_REVISION} never reported ready" >&2 + exit 1 + + # Why: a gateway that boots and answers /ready can still be unable to send. This proves the + # runtime account's FCM grant end to end without delivering anything: validate_only stops + # Google before any push, and the deliberately invalid token means a healthy credential + # answers INVALID_ARGUMENT. PERMISSION_DENIED is the failure this step exists to catch. + # + # Only the four verdicts below are conclusive. A 429, a 5xx, or a transport failure says + # nothing about the credential, so it is retried rather than treated as either answer; a + # denied credential still fails on the first attempt, without burning the retries. + - name: Prove the runtime identity can reach FCM + shell: bash + run: | + set -euo pipefail + token="$(gcloud auth print-access-token \ + --impersonate-service-account "${PUSH_RUNTIME_SERVICE_ACCOUNT}")" + test -n "${token}" + echo "::add-mask::${token}" + body='{"validate_only":true,"message":{"token":"orca-push-deploy-probe-invalid-token","notification":{"title":"Orca","body":"deploy probe"}}}' + for attempt in $(seq 1 5); do + code="$(curl -sS -o "${RUNNER_TEMP}/push-fcm.json" -w '%{http_code}' --max-time 20 \ + -X POST "https://fcm.googleapis.com/v1/projects/${GCP_PROJECT_ID}/messages:send" \ + -H "Authorization: Bearer ${token}" \ + -H 'Content-Type: application/json' \ + --data "${body}" || true)" + status="$(jq -r '.error.status // empty' < "${RUNNER_TEMP}/push-fcm.json" || true)" + echo "attempt ${attempt}: FCM validate-only send returned HTTP ${code} status ${status:-OK}" + if test "${status}" = PERMISSION_DENIED || test "${status}" = INVALID_ARGUMENT || + test "${code}" = 401 || test "${code}" = 403; then + break + fi + sleep 5 + done + if test "${status}" = PERMISSION_DENIED || test "${code}" = 401 || test "${code}" = 403; then + echo "the push runtime identity cannot send through FCM" >&2 + exit 1 + fi + test "${status}" = INVALID_ARGUMENT + + - name: Shift all traffic to the verified candidate + shell: bash + run: | + set -euo pipefail + echo "TRAFFIC_SHIFT_ATTEMPTED=true" >> "${GITHUB_ENV}" + gcloud run services update-traffic "${SERVICE_NAME}" \ + --project "${GCP_PROJECT_ID}" \ + --region "${GCP_REGION}" \ + --to-revisions "${CANDIDATE_REVISION}=100" \ + --quiet + serving="$(gcloud run services describe "${SERVICE_NAME}" \ + --project "${GCP_PROJECT_ID}" --region "${GCP_REGION}" --format=json \ + | jq -r '[.status.traffic[] | select((.percent // 0) > 0)] + | if length == 1 and .[0].percent == 100 then .[0].revisionName else empty end')" + test "${serving}" = "${CANDIDATE_REVISION}" + echo "TRAFFIC_SHIFTED=true" >> "${GITHUB_ENV}" + + # Why: the summary is written before the origin check, not after it. Once traffic has + # moved, the rollback target is the single thing an operator needs, and a summary that only + # appeared on success would be missing in exactly the run that needs it. + - name: Publish the rollout summary + if: ${{ always() && env.CANDIDATE_REVISION != '' && env.ROLLBACK_REVISION != '' }} + shell: bash + run: | + set -euo pipefail + { + echo '### Push gateway rollout' + echo + echo "Revision: \`${CANDIDATE_REVISION}\`" + echo + echo "Image: \`${IMAGE_DIGEST}\`" + echo + echo "Rollback: \`gcloud run services update-traffic ${SERVICE_NAME}" \ + "--region ${GCP_REGION} --to-revisions ${ROLLBACK_REVISION}=100\`" + } >> "${GITHUB_STEP_SUMMARY}" + + - name: Verify the public origin after the shift + shell: bash + run: | + set -euo pipefail + for attempt in $(seq 1 30); do + code="$(curl -sS -o /dev/null -w '%{http_code}' --max-time 10 \ + "${PUSH_ORIGIN}/ready" || true)" + if test "${code}" = 200; then + echo "${PUSH_ORIGIN} is ready after ${attempt} attempt(s)" + exit 0 + fi + echo "attempt ${attempt}: ${PUSH_ORIGIN}/ready returned ${code}" + sleep 5 + done + echo "${PUSH_ORIGIN} never reported ready after the shift" >&2 + exit 1 + + # Why: everything after the shift runs with production on the candidate. A failure there + # is not a failure to deploy, it is a live gateway that has to go back, so the traffic move + # is undone here rather than left to whoever reads the run. + - name: Roll traffic back to the previous revision + if: ${{ (failure() || cancelled()) && env.TRAFFIC_SHIFT_ATTEMPTED == 'true' }} + shell: bash + run: | + set -euo pipefail + test -n "${ROLLBACK_REVISION:-}" + gcloud run services update-traffic "${SERVICE_NAME}" \ + --project "${GCP_PROJECT_ID}" \ + --region "${GCP_REGION}" \ + --to-revisions "${ROLLBACK_REVISION}=100" \ + --quiet + serving="$(gcloud run services describe "${SERVICE_NAME}" \ + --project "${GCP_PROJECT_ID}" --region "${GCP_REGION}" --format=json \ + | jq -r '[.status.traffic[] | select((.percent // 0) > 0)] + | if length == 1 and .[0].percent == 100 then .[0].revisionName else empty end')" + test "${serving}" = "${ROLLBACK_REVISION}" + echo "TRAFFIC_ROLLED_BACK=true" >> "${GITHUB_ENV}" + { + echo + echo '### Push gateway rolled back' + echo + echo "Traffic returned to \`${ROLLBACK_REVISION}\`; the candidate" \ + "\`${CANDIDATE_REVISION}\` no longer serves." + } >> "${GITHUB_STEP_SUMMARY}" + + # Why: a candidate that never took traffic is a revision holding a warm floor and a Cloud + # SQL pool for nothing. Its tag comes off first, because Cloud Run refuses to delete a + # revision a traffic target still names, and clearing CANDIDATE_TAG makes the always() tag + # step below a no-op rather than a second failure. + - name: Delete the rejected candidate revision + if: ${{ (failure() || cancelled()) && (env.TRAFFIC_SHIFT_ATTEMPTED != 'true' || env.TRAFFIC_ROLLED_BACK == 'true') }} + shell: bash + run: | + set -euo pipefail + test -n "${CANDIDATE_REVISION:-}" || exit 0 + if test -n "${CANDIDATE_TAG:-}"; then + gcloud run services update-traffic "${SERVICE_NAME}" \ + --project "${GCP_PROJECT_ID}" \ + --region "${GCP_REGION}" \ + --remove-tags "${CANDIDATE_TAG}" \ + --quiet + echo "CANDIDATE_TAG=" >> "${GITHUB_ENV}" + fi + gcloud run revisions delete "${CANDIDATE_REVISION}" \ + --project "${GCP_PROJECT_ID}" \ + --region "${GCP_REGION}" \ + --quiet + echo "deleted the candidate revision ${CANDIDATE_REVISION}" + + - name: Drop the candidate traffic tag + if: always() + shell: bash + run: | + set -euo pipefail + test -n "${CANDIDATE_TAG:-}" || exit 0 + gcloud run services update-traffic "${SERVICE_NAME}" \ + --project "${GCP_PROJECT_ID}" \ + --region "${GCP_REGION}" \ + --remove-tags "${CANDIDATE_TAG}" \ + --quiet diff --git a/.github/workflows/cloud-verify.yml b/.github/workflows/cloud-verify.yml index e2ba9407ac4..5e24cae76cc 100644 --- a/.github/workflows/cloud-verify.yml +++ b/.github/workflows/cloud-verify.yml @@ -90,6 +90,7 @@ jobs: --health-timeout 5s --health-retries 10 env: + ORCA_PUSH_TEST_DATABASE_URL: postgres://relay_test:relay_test@127.0.0.1:5432/orca_relay_test ORCA_RELAY_TEST_POSTGRES_URL: postgres://relay_test:relay_test@127.0.0.1:5432/orca_relay_test steps: - uses: actions/checkout@v4 diff --git a/.github/workflows/mobile-android-release.yml b/.github/workflows/mobile-android-release.yml index 35e900dc31c..17100c788b8 100644 --- a/.github/workflows/mobile-android-release.yml +++ b/.github/workflows/mobile-android-release.yml @@ -104,11 +104,47 @@ jobs: --clobber \ android/app/build/outputs/apk/release/*.apk else + # Why: release tags live on side branches, so GitHub's automatic + # previous-tag detection reaches back several releases; that body + # already exceeds the 125000-character API limit and grows each + # release. Pin the comparison base and cap the size. + notes_file="$RUNNER_TEMP/android-release-notes.md" + previous_tag="$( + gh release list --repo "$GITHUB_REPOSITORY" --limit 200 --json tagName --jq '.[].tagName' \ + | grep '^mobile-android-v' | grep -Fxv "$tag" | sort -V | tail -1 || true + )" + + if [ -n "$previous_tag" ]; then + # Why: gh writes the JSON error body to stdout on an HTTP error, so a + # non-empty file is not proof of success — gate on exit status. + if ! gh api "repos/$GITHUB_REPOSITORY/releases/generate-notes" -X POST \ + -f tag_name="$tag" \ + -f target_commitish="$GITHUB_SHA" \ + -f previous_tag_name="$previous_tag" \ + --jq .body > "$notes_file"; then + : > "$notes_file" + fi + fi + if [ ! -s "$notes_file" ]; then + printf 'Orca Mobile Android %s\n' "$tag" > "$notes_file" + fi + # Why: reuse the desktop release path's character-safe truncation so a + # multi-byte character cannot be split at the cap. + NOTES_FILE="$notes_file" \ + NOTES_MODULE="$GITHUB_WORKSPACE/config/scripts/create-draft-release.mjs" \ + node --input-type=module -e ' + const { readFileSync, writeFileSync } = await import("node:fs") + const { pathToFileURL } = await import("node:url") + const { truncateReleaseBody } = await import(pathToFileURL(process.env.NOTES_MODULE).href) + const file = process.env.NOTES_FILE + writeFileSync(file, truncateReleaseBody(readFileSync(file, "utf8"))) + ' + gh release create "$tag" \ --repo "$GITHUB_REPOSITORY" \ --title "Orca Mobile Android $tag" \ --prerelease \ --latest=false \ - --generate-notes \ + --notes-file "$notes_file" \ android/app/build/outputs/apk/release/*.apk fi diff --git a/.github/workflows/mobile-ios-release.yml b/.github/workflows/mobile-ios-release.yml index 934b3f694a3..27372260c01 100644 --- a/.github/workflows/mobile-ios-release.yml +++ b/.github/workflows/mobile-ios-release.yml @@ -94,6 +94,13 @@ jobs: run: node -e 'const fs = require("node:fs"); const { expo } = require("./app.json"); fs.appendFileSync(process.env.GITHUB_OUTPUT, `version=${expo.version}\nbuild_number=${expo.ios.buildNumber}\n`)' - name: Expo prebuild + # Why the env var: app.config.js derives the expo-notifications plugin's + # `mode` from it, which is what writes `aps-environment: production` into the + # entitlements. push-token.ts reports a production APNs environment for every + # non-__DEV__ build, so a development entitlement here would leave TestFlight + # and App Store builds registered against a sandbox they never receive from. + env: + ORCA_IOS_APS_ENVIRONMENT: production run: npx expo prebuild --platform ios --no-install - name: Install CocoaPods diff --git a/.github/workflows/packaged-browser-e2e.yml b/.github/workflows/packaged-browser-e2e.yml new file mode 100644 index 00000000000..2a23ac58988 --- /dev/null +++ b/.github/workflows/packaged-browser-e2e.yml @@ -0,0 +1,74 @@ +name: Packaged browser compatibility +on: + workflow_dispatch: + inputs: + ref: + description: Commit SHA or ref to validate (defaults to the selected revision) + type: string + required: false + schedule: + - cron: '20 8 * * 1' + workflow_call: + inputs: + ref: + type: string + required: false +permissions: + contents: read +jobs: + compatibility: + runs-on: ubuntu-latest + timeout-minutes: 25 + steps: + - uses: actions/checkout@v6 + with: + ref: ${{ inputs.ref || github.sha }} + persist-credentials: false + - name: Install headless tools + run: sudo apt-get update && sudo apt-get install -y build-essential openssh-client python3 ripgrep xvfb zsh openbox x11-utils + - uses: ./.github/actions/install-node-dependencies + with: + native-runtime: electron + - name: Download pinned old release + env: + GH_TOKEN: ${{ github.token }} + run: | + gh release download v1.4.188 --repo stablyai/orca --pattern orca-ide_1.4.188_amd64.deb --dir "$RUNNER_TEMP/old-orca" + python3 - <<'PYVERIFY' + import base64,hashlib,os,pathlib,subprocess + root=pathlib.Path(os.environ['RUNNER_TEMP'])/'old-orca' + package=root/'orca-ide_1.4.188_amd64.deb' + expected='uGONFUDfinYggxcT9ac72wnnlofLQaqasDDeP0HWOSqarBwTi1Ax3khmzKUY3vUnvuYOpSCEmsH4InzLZ2vg6g==' + assert base64.b64encode(hashlib.sha512(package.read_bytes()).digest()).decode()==expected + extracted=root/'extracted' + subprocess.run(['dpkg-deb','-x',str(package),str(extracted)],check=True) + executable=extracted/'opt'/'Orca'/'orca-ide' + assert executable.is_file() and os.access(executable,os.X_OK) + with open(os.environ['GITHUB_ENV'],'a') as env: env.write('ORCA_CROSS_VERSION_PACKAGED_EXECUTABLE='+str(executable)+'\n') + print('Verified old package:',executable) + PYVERIFY + - name: Build current Electron app + env: + VITE_EXPOSE_STORE: 'true' + run: | + pnpm run build:relay + pnpm exec electron-vite build --mode e2e + pnpm run build:web-from-renderer + - name: Run both mixed-version directions + env: + PLAYWRIGHT_JSON_OUTPUT_FILE: test-results/packaged-browser-results.json + run: >- + xvfb-run --auto-servernum bash .github/scripts/e2e-with-window-manager.sh + env SKIP_BUILD=1 ORCA_E2E_FORWARD_APP_LOGS=1 + pnpm exec playwright test --config tests/playwright.config.ts + tests/e2e/packaged-mixed-version-browser-placement.spec.ts + --project=electron-headless --workers=1 --retries=0 --repeat-each=3 --reporter=list,json + - name: Require all six compatibility executions + if: always() + run: node config/scripts/verify-packaged-browser-participation.mjs test-results/packaged-browser-results.json + - uses: actions/upload-artifact@v7 + if: always() + with: + name: packaged-mixed-version-audit + path: test-results/ + retention-days: 3 diff --git a/.github/workflows/pr.yml b/.github/workflows/pr.yml index 1fd141ee4a0..9279f35b39f 100644 --- a/.github/workflows/pr.yml +++ b/.github/workflows/pr.yml @@ -858,6 +858,7 @@ jobs: src/main/windows/windows-pty-job.win32.test.ts src/main/windows/windows-host-job.win32.test.ts src/main/windows/windows-process-tree-command-line-patch.test.ts + src/main/windows/windows-process-table-native-addon.win32.test.ts src/main/windows-live-tree-kill.win32.test.ts src/main/wsl/wsl-runner.test.ts src/main/wsl/wsl-guest-environment.test.ts diff --git a/.github/workflows/terminal-ime-e2e.yml b/.github/workflows/terminal-ime-e2e.yml index 1ab905d8783..bd6be26bd27 100644 --- a/.github/workflows/terminal-ime-e2e.yml +++ b/.github/workflows/terminal-ime-e2e.yml @@ -70,3 +70,40 @@ jobs: path: test-results/ retention-days: 7 if-no-files-found: ignore + + linux-wayland: + name: Linux Wayland Hangul terminating digit + runs-on: ubuntu-22.04 + timeout-minutes: 25 + steps: + - uses: actions/checkout@v6 + with: + persist-credentials: false + - name: Install native build, nested compositor and IME tools + run: >- + sudo apt-get update && sudo apt-get install -y + build-essential python3 fonts-noto-cjk dbus-x11 dconf-gsettings-backend + ibus ibus-hangul gnome-shell gnome-settings-daemon libglib2.0-bin + xdotool xvfb x11-utils imagemagick + - uses: ./.github/actions/install-node-dependencies + with: + native-runtime: electron + - name: Build Electron app for E2E + env: + VITE_EXPOSE_STORE: 'true' + run: | + pnpm run build:relay + pnpm exec electron-vite build --mode e2e + pnpm run build:web-from-renderer + - name: Run native Wayland Hangul terminating digit + env: + SKIP_BUILD: '1' + run: node config/scripts/run-terminal-ibus-hangul-e2e.mjs --nested-wayland + - name: Upload Wayland terminal IME evidence + if: always() + uses: actions/upload-artifact@v7 + with: + name: terminal-wayland-ime-evidence + path: test-results/ + retention-days: 7 + if-no-files-found: error diff --git a/.gitignore b/.gitignore index 6722fc5ae54..37519cf04f5 100644 --- a/.gitignore +++ b/.gitignore @@ -107,6 +107,7 @@ docs/** !docs/reference/headless-linux-server.md !docs/reference/ime-regression-checklist.md !docs/reference/linux-glibc-compatibility.md +!docs/reference/mobile-push-contract.md !docs/reference/macos-press-and-hold.md !docs/reference/orcad-operations.md !docs/reference/relay-grace-time-reconfiguration.md diff --git a/README.md b/README.md index 2ae59035da8..7e3540c80f1 100644 --- a/README.md +++ b/README.md @@ -36,7 +36,7 @@ Monitor and steer your agents from your phone — get notified when an agent finishes and send follow-ups from anywhere. -[iOS App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) · [TestFlight](https://testflight.apple.com/join/YjeGMQBA) · [Android APK 0.0.47](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) · [Docs →](https://www.onorca.dev/docs/mobile) +[iOS App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) · [TestFlight](https://testflight.apple.com/join/YjeGMQBA) · [Android APK 0.0.48](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) · [Docs →](https://www.onorca.dev/docs/mobile) @@ -230,7 +230,7 @@ yay -S stably-orca-bin Pair with your desktop app to monitor and steer your agents from your phone. - **iOS:** [Download on the App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) or [join TestFlight](https://testflight.apple.com/join/YjeGMQBA) -- **Android:** [Download APK 0.0.47](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) · [Install guide](https://www.onorca.dev/docs/android-apk) +- **Android:** [Download APK 0.0.48](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) · [Install guide](https://www.onorca.dev/docs/android-apk) --- diff --git a/cloud/README.md b/cloud/README.md index 8ffcd9fa6b3..a2171700bb1 100644 --- a/cloud/README.md +++ b/cloud/README.md @@ -24,6 +24,32 @@ the repository's root [MIT license](../LICENSE). - `apps/relay-ops`: the relay operations console and the incident monitor behind `pnpm ops:relay`, `pnpm incident:relay`, and `pnpm incident:relay-preflight`. +- `apps/push` and `packages/push-contract`: the mobile push gateway that holds + the APNs key and sends to phones through APNs and FCM, and its wire contract. + It is deployed and operated from here but is not part of the relay data path; + see [docs/push-gateway.md](docs/push-gateway.md). + +## Mobile push gateway + +`apps/push` is a separate Cloud Run service from the relay. Phones never hold an +Orca credential for it: the desktop host authenticates with the same X25519 +key it uses for the relay, answering an encrypted challenge to mint a 24 hour +session, then registers each paired phone's native push token and asks the +gateway to push. The gateway coalesces a burst per registration into one +notification, enforces per-host and per-registration quotas, and retires a +registration as soon as Apple or Google reports the token unregistered. + +Storage follows the relay pattern: PostgreSQL in production, SQLite for tests +and local development. Configure it with `ORCA_PUSH_PUBLIC_URL`, +`ORCA_PUSH_DATABASE_URL`, the three APNs variables (`ORCA_PUSH_APNS_KEY`, +`ORCA_PUSH_APNS_KEY_ID`, `ORCA_PUSH_APPLE_TEAM_ID`, all three or none), and +optionally `ORCA_PUSH_APNS_TOPIC`, `ORCA_PUSH_FCM_PROJECT_ID`, and +`ORCA_PUSH_COALESCE_MS`. The FCM credential comes from the runtime service +account, so no key material is configured for Android. The full contract lives +in `docs/reference/mobile-push-contract.md` at the repository root. + +Logging is aggregate counters only. Tokens, notification titles, notification +bodies, and full host fingerprints never reach a log line. ## Infrastructure and operations @@ -38,16 +64,18 @@ the repository's root [MIT license](../LICENSE). - `dev/contracts` and `dev/fixtures`: the checked-in data those contract tests read, including the Terraform root partition. - `docs/`: the relay runbooks, capacity-testing guide, incident-monitor - reference, and the workflow variable reference in `docs/relay-workflows.md`. + reference, the workflow variable reference in `docs/relay-workflows.md`, and + the push gateway runbook in `docs/push-gateway.md`. ## Workflows -The 24 `.github/workflows/cloud-*.yml` workflows are the relay's deploy and -operate surface: publish and deploy the director, roll GCE cell capacity, -operate Asia admission and regional rehoming, prove staging capacity, monitor -production, and power staging up and down. `.github/actions/cloud-sql-rollout-lease` -is the compare-and-swap lease that serializes every rollout against the shared -Cloud SQL instance. +The 25 `.github/workflows/cloud-*.yml` workflows are the deploy and operate +surface: publish and deploy the director, roll GCE cell capacity, operate Asia +admission and regional rehoming, prove staging capacity, monitor production, +power staging up and down, and deploy the mobile push gateway. +`.github/actions/cloud-sql-rollout-lease` is the compare-and-swap lease that +serializes every rollout against the shared Cloud SQL instance, the push +gateway deploy included. Every one of them is inert. Each top-level job is gated on `vars.ORCA_CLOUD_OPERATIONS_ENABLED == 'true'`, a repository variable that is diff --git a/cloud/apps/push/Dockerfile b/cloud/apps/push/Dockerfile new file mode 100644 index 00000000000..efdc85fc404 --- /dev/null +++ b/cloud/apps/push/Dockerfile @@ -0,0 +1,29 @@ +FROM node:24-alpine AS build +WORKDIR /app +RUN corepack enable +COPY package.json pnpm-lock.yaml pnpm-workspace.yaml tsconfig.base.json ./ +COPY packages/push-contract/package.json packages/push-contract/package.json +COPY packages/postgres-schema/package.json packages/postgres-schema/package.json +COPY apps/push/package.json apps/push/package.json +RUN pnpm install --frozen-lockfile +COPY packages/push-contract packages/push-contract +COPY apps/push apps/push +COPY packages/postgres-schema packages/postgres-schema +RUN pnpm --filter @orca-cloud/postgres-schema build && pnpm --filter @orca-cloud/push-contract build && pnpm --filter @orca-cloud/push build + +FROM node:24-alpine AS runtime +ENV NODE_ENV=production +ENV PORT=8080 +WORKDIR /app +RUN corepack enable +COPY package.json pnpm-lock.yaml pnpm-workspace.yaml ./ +COPY packages/push-contract/package.json packages/push-contract/package.json +COPY packages/postgres-schema/package.json packages/postgres-schema/package.json +COPY apps/push/package.json apps/push/package.json +COPY --from=build /app/packages/push-contract/dist packages/push-contract/dist +COPY --from=build /app/packages/postgres-schema/dist packages/postgres-schema/dist +COPY --from=build /app/apps/push/dist apps/push/dist +RUN pnpm install --prod --frozen-lockfile --filter @orca-cloud/push... +USER node +EXPOSE 8080 +CMD ["node", "apps/push/dist/index.js"] diff --git a/cloud/apps/push/package.json b/cloud/apps/push/package.json new file mode 100644 index 00000000000..d84d0af8b25 --- /dev/null +++ b/cloud/apps/push/package.json @@ -0,0 +1,34 @@ +{ + "name": "@orca-cloud/push", + "private": true, + "version": "0.0.0", + "type": "module", + "main": "dist/index.js", + "scripts": { + "build": "pnpm clean && tsc -p tsconfig.build.json", + "clean": "node -e \"require('fs').rmSync('dist', { recursive: true, force: true })\"", + "dev": "tsx watch src/index.ts", + "lint": "tsc -p tsconfig.json --noEmit", + "pretest": "pnpm --filter @orca-cloud/postgres-schema build && pnpm --filter @orca-cloud/push-contract build", + "start": "node dist/index.js", + "test": "vitest run", + "typecheck": "tsc -p tsconfig.json --noEmit" + }, + "dependencies": { + "@hono/node-server": "^1.19.14", + "@orca-cloud/postgres-schema": "workspace:*", + "@orca-cloud/push-contract": "workspace:*", + "google-auth-library": "^10.5.0", + "hono": "^4.12.27", + "pg": "^8.22.0", + "tweetnacl": "^1.0.3", + "zod": "^3.25.76" + }, + "devDependencies": { + "@types/node": "^24.10.0", + "@types/pg": "^8.20.0", + "tsx": "^4.21.0", + "typescript": "^5.9.3", + "vitest": "^4.0.8" + } +} diff --git a/cloud/apps/push/src/apns-authentication-token.ts b/cloud/apps/push/src/apns-authentication-token.ts new file mode 100644 index 00000000000..34def16e86e --- /dev/null +++ b/cloud/apps/push/src/apns-authentication-token.ts @@ -0,0 +1,42 @@ +import { createPrivateKey, type KeyObject, sign } from 'node:crypto' +import type { ApnsCredentials } from './config.js' + +// Apple rejects a provider token older than an hour and throttles reissue +// under about 20 minutes, so 50 minutes is the safe rotation point. +export const APNS_TOKEN_ROTATION_MS = 50 * 60 * 1000 + +function base64UrlJson(value: Record): string { + return Buffer.from(JSON.stringify(value), 'utf8').toString('base64url') +} + +export class ApnsAuthenticationToken { + private readonly privateKey: KeyObject + private cached: { token: string; issuedAtMs: number } | null = null + + constructor( + private readonly credentials: ApnsCredentials, + private readonly now: () => number = Date.now, + private readonly rotationMs: number = APNS_TOKEN_ROTATION_MS + ) { + this.privateKey = createPrivateKey(credentials.keyPem) + } + + value(): string { + const nowMs = this.now() + if (this.cached && nowMs - this.cached.issuedAtMs < this.rotationMs) return this.cached.token + const header = base64UrlJson({ alg: 'ES256', kid: this.credentials.keyId }) + const payload = base64UrlJson({ + iss: this.credentials.teamId, + iat: Math.floor(nowMs / 1000) + }) + const signingInput = `${header}.${payload}` + // ES256 requires the raw r||s pair; Node emits DER unless asked otherwise. + const signature = sign('sha256', Buffer.from(signingInput, 'utf8'), { + key: this.privateKey, + dsaEncoding: 'ieee-p1363' + }).toString('base64url') + const token = `${signingInput}.${signature}` + this.cached = { token, issuedAtMs: nowMs } + return token + } +} diff --git a/cloud/apps/push/src/apns-client.test.ts b/cloud/apps/push/src/apns-client.test.ts new file mode 100644 index 00000000000..c0f312e46e6 --- /dev/null +++ b/cloud/apps/push/src/apns-client.test.ts @@ -0,0 +1,174 @@ +import { generateKeyPairSync } from 'node:crypto' +import { describe, expect, it } from 'vitest' +import { ApnsAuthenticationToken, APNS_TOKEN_ROTATION_MS } from './apns-authentication-token.js' +import { ApnsClient } from './apns-client.js' +import type { ApnsRequest, ApnsResponse } from './apns-http2-transport.js' +import type { ApnsCredentials } from './config.js' +import { buildPushDelivery } from './push-delivery-message.js' + +const HOST = 'abcdefghijklmnop' + +function credentials(): ApnsCredentials { + const { privateKey } = generateKeyPairSync('ec', { + namedCurve: 'P-256', + privateKeyEncoding: { type: 'pkcs8', format: 'pem' }, + publicKeyEncoding: { type: 'spki', format: 'pem' } + }) + return { keyPem: privateKey, keyId: 'ABCDE12345', teamId: 'TEAM123456' } +} + +function delivery(coalescedCount = 1) { + return buildPushDelivery({ + registrationId: 'reg-1', + hostFingerprint: HOST, + notification: { + notificationId: 'note-1', + notificationSeq: 7, + notificationEpoch: 'epoch-1', + source: 'agent-task-complete', + agentState: 'needs-input', + title: 'Agent needs input', + body: 'Waiting on your answer', + worktreeId: 'wt-1' + }, + title: 'Agent needs input', + body: 'Waiting on your answer', + coalescedCount + }) +} + +function fakeTransport(response: ApnsResponse) { + const requests: ApnsRequest[] = [] + return { + requests, + transport: async (request: ApnsRequest): Promise => { + requests.push(request) + return response + } + } +} + +describe('apns authentication token', () => { + it('signs an ES256 provider token and caches it until the rotation point', () => { + let clock = 1_700_000_000_000 + const authentication = new ApnsAuthenticationToken(credentials(), () => clock) + const first = authentication.value() + const [header, payload, signature] = first.split('.') + expect(JSON.parse(Buffer.from(header!, 'base64url').toString('utf8'))).toEqual({ + alg: 'ES256', + kid: 'ABCDE12345' + }) + expect(JSON.parse(Buffer.from(payload!, 'base64url').toString('utf8'))).toEqual({ + iss: 'TEAM123456', + iat: Math.floor(clock / 1000) + }) + expect(Buffer.from(signature!, 'base64url').byteLength).toBe(64) + + clock += APNS_TOKEN_ROTATION_MS - 1 + expect(authentication.value()).toBe(first) + clock += 1 + expect(authentication.value()).not.toBe(first) + }) +}) + +describe('apns client', () => { + it('sends the specified headers, path, and alert body', async () => { + const clock = 1_700_000_000_000 + const fake = fakeTransport({ status: 200, body: '' }) + const client = new ApnsClient({ + topic: 'com.stably.orca.mobile', + credentials: credentials(), + transport: fake.transport, + now: () => clock + }) + await expect( + client.send(delivery(), { token: 'a'.repeat(64), apnsEnvironment: 'production' }) + ).resolves.toEqual({ status: 'sent' }) + const request = fake.requests[0]! + expect(request.host).toBe('api.push.apple.com') + expect(request.path).toBe(`/3/device/${'a'.repeat(64)}`) + expect(request.headers).toMatchObject({ + 'apns-topic': 'com.stably.orca.mobile', + 'apns-push-type': 'alert', + 'apns-priority': '10', + 'apns-expiration': String(Math.floor(clock / 1000) + 4 * 60 * 60), + 'apns-collapse-id': 'note-1' + }) + expect(request.headers.authorization).toMatch(/^bearer /) + expect(JSON.parse(request.body)).toEqual({ + aps: { + alert: { title: 'Agent needs input', body: 'Waiting on your answer' }, + sound: 'default', + 'thread-id': HOST + }, + orca: { + hostFingerprint: HOST, + worktreeId: 'wt-1', + notificationId: 'note-1', + notificationSeq: 7, + notificationEpoch: 'epoch-1', + source: 'agent-task-complete', + agentState: 'needs-input', + coalescedCount: 1 + } + }) + }) + + it('targets the sandbox host and the host collapse id for a summary', async () => { + const fake = fakeTransport({ status: 200, body: '' }) + const client = new ApnsClient({ + topic: 'com.stably.orca.mobile', + credentials: credentials(), + transport: fake.transport + }) + await client.send(delivery(3), { token: 'b'.repeat(64), apnsEnvironment: 'sandbox' }) + expect(fake.requests[0]?.host).toBe('api.sandbox.push.apple.com') + expect(fake.requests[0]?.headers['apns-collapse-id']).toBe(`host:${HOST}`) + }) + + it.each([ + [410, 'Unregistered'], + [400, 'BadDeviceToken'], + [400, 'Unregistered'], + [400, 'DeviceTokenNotForTopic'] + ])('classifies %i %s as a dead token', async (status, reason) => { + const fake = fakeTransport({ status, body: JSON.stringify({ reason }) }) + const client = new ApnsClient({ + topic: 'com.stably.orca.mobile', + credentials: credentials(), + transport: fake.transport + }) + await expect( + client.send(delivery(), { token: 'a'.repeat(64), apnsEnvironment: 'production' }) + ).resolves.toEqual({ status: 'dead', reason }) + }) + + it.each([ + [400, 'PayloadTooLarge'], + [429, 'TooManyRequests'], + [500, 'InternalServerError'] + ])('treats %i %s with the appropriate retry policy', async (status, reason) => { + const fake = fakeTransport({ status, body: JSON.stringify({ reason }) }) + const client = new ApnsClient({ + topic: 'com.stably.orca.mobile', + credentials: credentials(), + transport: fake.transport + }) + await expect( + client.send(delivery(), { token: 'a'.repeat(64), apnsEnvironment: 'production' }) + ).resolves.toEqual({ status: 'error', reason, retryable: status === 429 || status >= 500 }) + }) + + it('reports a transport failure as an error rather than throwing', async () => { + const client = new ApnsClient({ + topic: 'com.stably.orca.mobile', + credentials: credentials(), + transport: async () => { + throw new Error('socket hang up') + } + }) + await expect( + client.send(delivery(), { token: 'a'.repeat(64), apnsEnvironment: 'production' }) + ).resolves.toEqual({ status: 'error', reason: 'Error', retryable: true }) + }) +}) diff --git a/cloud/apps/push/src/apns-client.ts b/cloud/apps/push/src/apns-client.ts new file mode 100644 index 00000000000..767b96e83df --- /dev/null +++ b/cloud/apps/push/src/apns-client.ts @@ -0,0 +1,91 @@ +import { PUSH_LIMITS, type ApnsEnvironment } from '@orca-cloud/push-contract' +import { ApnsAuthenticationToken } from './apns-authentication-token.js' +import type { ApnsTransport } from './apns-http2-transport.js' +import type { ApnsCredentials } from './config.js' +import type { PushDelivery } from './push-delivery-message.js' +import type { PushProviderOutcome } from './push-provider-outcome.js' + +const APNS_HOSTS: Record = { + production: 'api.push.apple.com', + sandbox: 'api.sandbox.push.apple.com' +} + +const DEAD_TOKEN_REASONS = new Set(['BadDeviceToken', 'Unregistered', 'DeviceTokenNotForTopic']) + +export type ApnsClientOptions = { + topic: string + credentials: ApnsCredentials + transport: ApnsTransport + now?: () => number +} + +function readReason(body: string): string { + try { + const parsed = JSON.parse(body) as { reason?: unknown } + return typeof parsed.reason === 'string' ? parsed.reason : 'unknown' + } catch { + return 'unparseable' + } +} + +export function apnsBody(delivery: PushDelivery): string { + return JSON.stringify({ + aps: { + alert: { title: delivery.title, body: delivery.body }, + ...(delivery.sound === false ? {} : { sound: 'default' }), + 'thread-id': delivery.hostFingerprint + }, + orca: delivery.orca + }) +} + +export class ApnsClient { + private readonly authentication: ApnsAuthenticationToken + private readonly now: () => number + + constructor(private readonly options: ApnsClientOptions) { + this.now = options.now ?? Date.now + this.authentication = new ApnsAuthenticationToken(options.credentials, this.now) + } + + async send( + delivery: PushDelivery, + device: { token: string; apnsEnvironment: ApnsEnvironment } + ): Promise { + const expiration = Math.floor(this.now() / 1000) + PUSH_LIMITS.notificationTtlSeconds + let response + try { + response = await this.options.transport({ + host: APNS_HOSTS[device.apnsEnvironment], + path: `/3/device/${device.token}`, + headers: { + authorization: `bearer ${this.authentication.value()}`, + 'apns-topic': this.options.topic, + 'apns-push-type': 'alert', + 'apns-priority': '10', + 'apns-expiration': String(expiration), + 'apns-collapse-id': delivery.collapseId + }, + body: apnsBody(delivery) + }) + } catch (error) { + return { + status: 'error', + reason: error instanceof Error ? error.name : 'transport_failed', + retryable: true + } + } + if (response.status === 200) return { status: 'sent' } + const reason = readReason(response.body) + if (response.status === 410) return { status: 'dead', reason } + if (response.status === 400 && DEAD_TOKEN_REASONS.has(reason)) { + return { status: 'dead', reason } + } + return { + status: 'error', + reason, + retryable: response.status === 429 || response.status >= 500, + ...(response.retryAfterMs === undefined ? {} : { retryAfterMs: response.retryAfterMs }) + } + } +} diff --git a/cloud/apps/push/src/apns-http2-transport.ts b/cloud/apps/push/src/apns-http2-transport.ts new file mode 100644 index 00000000000..167b4d14e38 --- /dev/null +++ b/cloud/apps/push/src/apns-http2-transport.ts @@ -0,0 +1,50 @@ +import { connect, constants, type ClientHttp2Session } from 'node:http2' +import { readApnsStreamResponse, type ApnsResponse } from './apns-stream-response.js' + +export type ApnsRequest = { + host: string + path: string + headers: Record + body: string +} + +export type { ApnsResponse } +export type ApnsTransport = (request: ApnsRequest) => Promise + +// APNs requires HTTP/2 and rewards a long-lived session per host, so sessions +// are cached and only dropped when the socket itself goes away. +export function createApnsHttp2Transport(): ApnsTransport & { close(): void } { + const sessions = new Map() + + const sessionFor = (host: string): ClientHttp2Session => { + const existing = sessions.get(host) + if (existing && !existing.closed && !existing.destroyed) return existing + const session = connect(`https://${host}`) + const forget = (): void => { + if (sessions.get(host) === session) sessions.delete(host) + } + session.on('error', forget) + session.on('close', forget) + sessions.set(host, session) + return session + } + + const transport = async (request: ApnsRequest): Promise => { + const stream = sessionFor(request.host).request({ + ...request.headers, + [constants.HTTP2_HEADER_METHOD]: 'POST', + [constants.HTTP2_HEADER_PATH]: request.path, + [constants.HTTP2_HEADER_AUTHORITY]: request.host, + 'content-type': 'application/json', + 'content-length': String(Buffer.byteLength(request.body)) + }) + return await readApnsStreamResponse(stream, request.body) + } + + return Object.assign(transport, { + close(): void { + for (const session of sessions.values()) session.close() + sessions.clear() + } + }) +} diff --git a/cloud/apps/push/src/apns-session-replacement.test.ts b/cloud/apps/push/src/apns-session-replacement.test.ts new file mode 100644 index 00000000000..2678732ca94 --- /dev/null +++ b/cloud/apps/push/src/apns-session-replacement.test.ts @@ -0,0 +1,45 @@ +import { EventEmitter } from 'node:events' +import { expect, it, vi } from 'vitest' +const mocks = vi.hoisted(() => ({ + connect: vi.fn(), + read: vi.fn(async () => ({ status: 200, body: '' })) +})) +vi.mock('node:http2', async (original) => ({ + ...(await original()), + connect: mocks.connect +})) +vi.mock('./apns-stream-response.js', () => ({ readApnsStreamResponse: mocks.read })) +import { createApnsHttp2Transport } from './apns-http2-transport.js' + +it('keeps the replacement cached when the draining session closes later', async () => { + const sessions: Array< + EventEmitter & { + closed: boolean + destroyed: boolean + request: ReturnType + close: ReturnType + } + > = [] + mocks.connect.mockImplementation(() => { + const session = Object.assign(new EventEmitter(), { + closed: false, + destroyed: false, + request: vi.fn(() => ({})), + close: vi.fn() + }) + sessions.push(session) + return session + }) + const transport = createApnsHttp2Transport() + const request = { host: 'api.push.apple.com', path: '/synthetic', headers: {}, body: '{}' } + await transport(request) + sessions[0]!.closed = true + await transport(request) + sessions[0]!.emit('close') + sessions[0]!.emit('error', new Error('old-session')) + await transport(request) + expect(sessions).toHaveLength(2) + expect(sessions[1]!.request).toHaveBeenCalledTimes(2) + transport.close() + expect(sessions[1]!.close).toHaveBeenCalledOnce() +}) diff --git a/cloud/apps/push/src/apns-stream-response.test.ts b/cloud/apps/push/src/apns-stream-response.test.ts new file mode 100644 index 00000000000..c87b9031ca1 --- /dev/null +++ b/cloud/apps/push/src/apns-stream-response.test.ts @@ -0,0 +1,82 @@ +import { EventEmitter } from 'node:events' +import { describe, expect, it } from 'vitest' +import { readApnsStreamResponse, type ApnsResponseStream } from './apns-stream-response.js' + +type FakeStream = ApnsResponseStream & { + sentBody: string | null + destroyedWith: Error | null + fireTimeout(): void +} + +function fakeApnsStream(): FakeStream { + const emitter = new EventEmitter() as FakeStream + emitter.sentBody = null + emitter.destroyedWith = null + let onTimeout: (() => void) | null = null + emitter.setTimeout = (_ms, callback) => { + onTimeout = callback + } + emitter.destroy = (error?: Error) => { + emitter.destroyedWith = error ?? null + if (error) emitter.emit('error', error) + } + emitter.end = (body: string) => { + emitter.sentBody = body + } + emitter.fireTimeout = () => onTimeout?.() + return emitter +} + +describe('apns stream response', () => { + it('resolves with the status and the concatenated body', async () => { + const stream = fakeApnsStream() + const pending = readApnsStreamResponse(stream, '{"aps":{}}') + expect(stream.sentBody).toBe('{"aps":{}}') + stream.emit('response', { ':status': '200' }) + stream.emit('data', Buffer.from('{"re')) + stream.emit('data', Buffer.from('ason":"ok"}')) + stream.emit('end') + await expect(pending).resolves.toEqual({ status: 200, body: '{"reason":"ok"}' }) + }) + + it('rejects when the peer resets the stream without an end or an error', async () => { + const stream = fakeApnsStream() + const pending = readApnsStreamResponse(stream, 'body') + stream.emit('response', { ':status': '200' }) + // NGHTTP2_NO_ERROR: node emits only 'close', so nothing else would settle. + stream.emit('close') + await expect(pending).rejects.toThrow('apns_stream_closed') + }) + + it('keeps the resolved response when close follows a completed end', async () => { + const stream = fakeApnsStream() + const pending = readApnsStreamResponse(stream, 'body') + stream.emit('response', { ':status': '410' }) + stream.emit('end') + stream.emit('close') + await expect(pending).resolves.toEqual({ status: 410, body: '' }) + }) + + it('keeps the original error when close follows a stream error', async () => { + const stream = fakeApnsStream() + const pending = readApnsStreamResponse(stream, 'body') + stream.emit('error', new Error('socket_hang_up')) + stream.emit('close') + await expect(pending).rejects.toThrow('socket_hang_up') + }) + + it('destroys the stream on timeout and surfaces the timeout error', async () => { + const stream = fakeApnsStream() + const pending = readApnsStreamResponse(stream, 'body', 10) + stream.fireTimeout() + await expect(pending).rejects.toThrow('apns_timeout') + expect(stream.destroyedWith?.message).toBe('apns_timeout') + }) + + it('reports a missing status header as zero rather than NaN', async () => { + const stream = fakeApnsStream() + const pending = readApnsStreamResponse(stream, 'body') + stream.emit('end') + await expect(pending).resolves.toEqual({ status: 0, body: '' }) + }) +}) diff --git a/cloud/apps/push/src/apns-stream-response.ts b/cloud/apps/push/src/apns-stream-response.ts new file mode 100644 index 00000000000..da001a5df31 --- /dev/null +++ b/cloud/apps/push/src/apns-stream-response.ts @@ -0,0 +1,53 @@ +import type { EventEmitter } from 'node:events' +import { providerRetryAfter } from './provider-retry-delay.js' +import { constants } from 'node:http2' + +export type ApnsResponse = { status: number; body: string; retryAfterMs?: number } + +// The subset of ClientHttp2Stream this module drives, so a fake emitter can +// stand in for a real APNs stream in tests. +export type ApnsResponseStream = EventEmitter & { + setTimeout(ms: number, callback: () => void): void + destroy(error?: Error): void + end(body: string): void +} + +export const APNS_REQUEST_TIMEOUT_MS = 10_000 + +export function readApnsStreamResponse( + stream: ApnsResponseStream, + body: string, + timeoutMs = APNS_REQUEST_TIMEOUT_MS +): Promise { + return new Promise((resolve, reject) => { + let settled = false + const settle = (run: () => void): void => { + if (settled) return + settled = true + run() + } + let status = 0 + let retryAfterMs: number | undefined + const chunks: Buffer[] = [] + stream.setTimeout(timeoutMs, () => stream.destroy(new Error('apns_timeout'))) + stream.on('response', (headers: Record) => { + status = Number(headers[constants.HTTP2_HEADER_STATUS] ?? 0) + retryAfterMs = providerRetryAfter(String(headers['retry-after'] ?? '')) + }) + stream.on('data', (chunk: Buffer) => chunks.push(chunk)) + stream.on('error', (error: Error) => settle(() => reject(error))) + stream.on('end', () => + settle(() => + resolve({ + status, + body: Buffer.concat(chunks).toString('utf8'), + ...(retryAfterMs === undefined ? {} : { retryAfterMs }) + }) + ) + ) + // A peer reset with NGHTTP2_NO_ERROR emits neither 'end' nor 'error', which + // would leave the coalescer's delivery pending for the life of the process. + stream.on('close', () => settle(() => reject(new Error('apns_stream_closed')))) + stream.end(body) + }) +} diff --git a/cloud/apps/push/src/canonical-base64.ts b/cloud/apps/push/src/canonical-base64.ts new file mode 100644 index 00000000000..e13ea982cb6 --- /dev/null +++ b/cloud/apps/push/src/canonical-base64.ts @@ -0,0 +1,9 @@ +// Rejects the many base64 spellings of the same bytes: a non-canonical +// encoding would change the transcript the host signs without changing the key. +export function decodeCanonicalBase64(value: string, expectedBytes: number): Buffer | null { + if (!/^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/.test(value)) return null + const decoded = Buffer.from(value, 'base64') + return decoded.byteLength === expectedBytes && decoded.toString('base64') === value + ? decoded + : null +} diff --git a/cloud/apps/push/src/client-ip-rate-limit.test.ts b/cloud/apps/push/src/client-ip-rate-limit.test.ts new file mode 100644 index 00000000000..2fc3734adc2 --- /dev/null +++ b/cloud/apps/push/src/client-ip-rate-limit.test.ts @@ -0,0 +1,145 @@ +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import { Hono } from 'hono' +import { describe, expect, it } from 'vitest' +import { ClientIpRateLimiter, clientIpRateLimit } from './client-ip-rate-limit.js' + +const CAPACITY = PUSH_LIMITS.unauthenticatedRequestsPerMinutePerIp + +function limiterApp(limiter: ClientIpRateLimiter, trustedProxyHops = 0): Hono { + const app = new Hono() + app.post('/probe', clientIpRateLimit(limiter, { trustedProxyHops }), (context) => + context.json({ ok: true }) + ) + return app +} + +describe('client ip rate limiter', () => { + it('admits exactly the per-minute allowance and refuses the next request', () => { + const limiter = new ClientIpRateLimiter({ now: () => 1_000 }) + for (let index = 0; index < CAPACITY; index++) { + expect(limiter.allow('203.0.113.7')).toBe(true) + } + expect(limiter.allow('203.0.113.7')).toBe(false) + }) + + it('keeps one client ip from spending another one budget', () => { + const limiter = new ClientIpRateLimiter({ now: () => 1_000 }) + for (let index = 0; index < CAPACITY; index++) limiter.allow('203.0.113.7') + expect(limiter.allow('203.0.113.7')).toBe(false) + expect(limiter.allow('198.51.100.9')).toBe(true) + }) + + it('refills over the window rather than resetting on a boundary', () => { + let clock = 1_000 + const limiter = new ClientIpRateLimiter({ now: () => clock }) + for (let index = 0; index < CAPACITY; index++) limiter.allow('203.0.113.7') + expect(limiter.allow('203.0.113.7')).toBe(false) + + // Half a window buys back half the allowance, no more. + clock += 30_000 + for (let index = 0; index < CAPACITY / 2; index++) { + expect(limiter.allow('203.0.113.7')).toBe(true) + } + expect(limiter.allow('203.0.113.7')).toBe(false) + }) + + it('bounds what it remembers when a flood of distinct ips arrives', () => { + let clock = 1_000 + const limiter = new ClientIpRateLimiter({ now: () => clock, maxTrackedIps: 8 }) + for (let index = 0; index < 200; index++) { + clock += 1 + limiter.allow(`198.51.100.${index}`) + } + expect(limiter.trackedIpCount()).toBeLessThanOrEqual(8) + }) + + it('answers 429 with a rate_limited body once the bucket is empty', async () => { + const app = limiterApp(new ClientIpRateLimiter({ now: () => 1_000 })) + const headers = { 'x-forwarded-for': '10.0.0.1, 10.0.0.2, 203.0.113.7' } + for (let index = 0; index < CAPACITY; index++) { + expect((await app.request('/probe', { method: 'POST', headers })).status).toBe(200) + } + const limited = await app.request('/probe', { method: 'POST', headers }) + expect(limited.status).toBe(429) + expect(await limited.json()).toEqual({ error: 'rate_limited' }) + }) + + it('buckets on the last forwarded hop, the only one the platform appended', async () => { + const app = limiterApp(new ClientIpRateLimiter({ now: () => 1_000 })) + for (let index = 0; index < CAPACITY; index++) { + await app.request('/probe', { + method: 'POST', + headers: { 'x-forwarded-for': `10.0.0.${index}, 203.0.113.7` } + }) + } + const sameClient = await app.request('/probe', { + method: 'POST', + headers: { 'x-forwarded-for': '10.9.9.9, 203.0.113.7' } + }) + expect(sameClient.status).toBe(429) + const otherClient = await app.request('/probe', { + method: 'POST', + headers: { 'x-forwarded-for': '10.0.0.1, 198.51.100.9' } + }) + expect(otherClient.status).toBe(200) + }) + + it('gives a spoofed left-most hop no escape from the caller own bucket', async () => { + const app = limiterApp(new ClientIpRateLimiter({ now: () => 1_000 })) + // A caller that rewrites its own x-forwarded-for on every request still ends + // up behind the one value Cloud Run appended. + for (let index = 0; index < CAPACITY; index++) { + const allowed = await app.request('/probe', { + method: 'POST', + headers: { 'x-forwarded-for': `198.51.100.${index}, 203.0.113.7` } + }) + expect(allowed.status).toBe(200) + } + const spoofed = await app.request('/probe', { + method: 'POST', + headers: { 'x-forwarded-for': '198.51.100.250, 10.1.1.1, 203.0.113.7' } + }) + expect(spoofed.status).toBe(429) + }) + + it('skips the configured trusted proxies when counting from the right', async () => { + const app = limiterApp(new ClientIpRateLimiter({ now: () => 1_000, capacity: 1 }), 1) + // , , : one trusted hop after the client. + const headers = { 'x-forwarded-for': '203.0.113.7, 10.0.0.1' } + expect((await app.request('/probe', { method: 'POST', headers })).status).toBe(200) + expect((await app.request('/probe', { method: 'POST', headers })).status).toBe(429) + expect( + (await app.request('/probe', { + method: 'POST', + headers: { 'x-forwarded-for': '198.51.100.9, 10.0.0.1' } + })).status + ).toBe(200) + }) + + it('trusts nothing when the header is shorter than the configured depth', async () => { + const app = limiterApp(new ClientIpRateLimiter({ now: () => 1_000, capacity: 1 }), 1) + // Only one hop, so the client value the depth points at does not exist. + const headers = { 'x-forwarded-for': '203.0.113.7' } + expect((await app.request('/probe', { method: 'POST', headers })).status).toBe(200) + expect( + (await app.request('/probe', { + method: 'POST', + headers: { 'x-forwarded-for': '198.51.100.9' } + })).status + ).toBe(429) + }) + + it('falls back to x-real-ip and then to a single shared bucket', async () => { + const app = limiterApp(new ClientIpRateLimiter({ now: () => 1_000, capacity: 1 })) + expect( + (await app.request('/probe', { method: 'POST', headers: { 'x-real-ip': '203.0.113.7' } })) + .status + ).toBe(200) + expect( + (await app.request('/probe', { method: 'POST', headers: { 'x-real-ip': '203.0.113.7' } })) + .status + ).toBe(429) + expect((await app.request('/probe', { method: 'POST' })).status).toBe(200) + expect((await app.request('/probe', { method: 'POST' })).status).toBe(429) + }) +}) diff --git a/cloud/apps/push/src/client-ip-rate-limit.ts b/cloud/apps/push/src/client-ip-rate-limit.ts new file mode 100644 index 00000000000..efc26a7ea78 --- /dev/null +++ b/cloud/apps/push/src/client-ip-rate-limit.ts @@ -0,0 +1,110 @@ +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import type { Context, MiddlewareHandler } from 'hono' + +const REFILL_WINDOW_MS = 60_000 +const MAX_TRACKED_IPS = 10_000 +const UNKNOWN_CLIENT_IP = 'unknown' + +export type ClientIpRateLimiterOptions = { + capacity?: number + windowMs?: number + maxTrackedIps?: number + now?: () => number +} + +type Bucket = { tokens: number; updatedAt: number } + +// Read x-forwarded-for from the right. Cloud Run appends the connecting peer, +// so the last value is the only one it wrote; everything to its left is +// whatever the caller sent and can be a fresh forgery on every request. +// trustedProxyHops is how many appenders sit between Cloud Run and the client +// (0 today, 1 once a load balancer fronts it). A header too short for that +// depth is not trusted at all and falls through to the shared bucket, which +// throttles rather than opens. +export function readClientIp(context: Context, trustedProxyHops = 0): string { + const hops = + context.req + .header('x-forwarded-for') + ?.split(',') + .map((hop) => hop.trim()) + .filter((hop) => hop.length > 0) ?? [] + const client = hops[hops.length - 1 - trustedProxyHops] + if (client) return client + return context.req.header('x-real-ip')?.trim() || UNKNOWN_CLIENT_IP +} + +// In-memory and per-instance on purpose. A shared counter would put a database +// round trip in front of the only routes an attacker can reach unauthenticated, +// and Cloud Run's instance fan-out only loosens the cap by the instance count. +export class ClientIpRateLimiter { + private readonly buckets = new Map() + private readonly capacity: number + private readonly windowMs: number + private readonly maxTrackedIps: number + private readonly now: () => number + + constructor(options: ClientIpRateLimiterOptions = {}) { + this.capacity = options.capacity ?? PUSH_LIMITS.unauthenticatedRequestsPerMinutePerIp + this.windowMs = options.windowMs ?? REFILL_WINDOW_MS + this.maxTrackedIps = options.maxTrackedIps ?? MAX_TRACKED_IPS + this.now = options.now ?? Date.now + } + + allow(clientIp: string): boolean { + const now = this.now() + const tokens = this.tokensAt(this.buckets.get(clientIp), now) + if (tokens < 1) { + this.buckets.set(clientIp, { tokens, updatedAt: now }) + return false + } + this.buckets.set(clientIp, { tokens: tokens - 1, updatedAt: now }) + this.evict(now) + return true + } + + trackedIpCount(): number { + return this.buckets.size + } + + private tokensAt(bucket: Bucket | undefined, now: number): number { + if (!bucket) return this.capacity + const refilled = ((now - bucket.updatedAt) * this.capacity) / this.windowMs + return Math.min(this.capacity, bucket.tokens + Math.max(0, refilled)) + } + + private evict(now: number): void { + if (this.buckets.size <= this.maxTrackedIps) return + // A bucket that has refilled to capacity is indistinguishable from an + // absent one, so dropping it changes no decision. + for (const [clientIp, bucket] of this.buckets) { + if (this.tokensAt(bucket, now) >= this.capacity) this.buckets.delete(clientIp) + } + if (this.buckets.size <= this.maxTrackedIps) return + // A flood of distinct live IPs can still overflow. The least recently seen + // are the least likely to be mid-burst. + const excess = [...this.buckets.entries()] + .sort((left, right) => left[1].updatedAt - right[1].updatedAt) + .slice(0, this.buckets.size - this.maxTrackedIps) + for (const [clientIp] of excess) this.buckets.delete(clientIp) + } +} + +export type ClientIpRateLimitOptions = { + trustedProxyHops?: number + onLimited?: () => void +} + +export function clientIpRateLimit( + limiter: ClientIpRateLimiter, + options: ClientIpRateLimitOptions = {} +): MiddlewareHandler { + const trustedProxyHops = options.trustedProxyHops ?? 0 + return async (context, next) => { + if (!limiter.allow(readClientIp(context, trustedProxyHops))) { + options.onLimited?.() + return context.json({ error: 'rate_limited' }, 429) + } + await next() + return + } +} diff --git a/cloud/apps/push/src/coalescer.test.ts b/cloud/apps/push/src/coalescer.test.ts new file mode 100644 index 00000000000..5fcf8f3342c --- /dev/null +++ b/cloud/apps/push/src/coalescer.test.ts @@ -0,0 +1,173 @@ +import type { PushNotification } from '@orca-cloud/push-contract' +import { describe, expect, it } from 'vitest' +import { PushCoalescer, summaryBody, type CoalescerTimer } from './coalescer.js' +import type { PushDelivery } from './push-delivery-message.js' + +const HOST = 'abcdefghijklmnop' + +function notification(overrides: Partial = {}): PushNotification { + return { + notificationId: 'note-1', + notificationSeq: 1, + notificationEpoch: 'epoch-1', + source: 'agent-task-complete', + agentState: 'needs-input', + title: 'Agent needs input', + body: 'Waiting on your answer', + worktreeId: 'wt-1', + ...overrides + } +} + +// A manual timer queue so a 3s window is exercised without waiting 3s. +function createTimerHarness() { + const pending = new Map void>() + let nextId = 0 + return { + delays: [] as number[], + setTimer(callback: () => void, delayMs: number): CoalescerTimer { + const handle = nextId++ + pending.set(handle, callback) + this.delays.push(delayMs) + return { handle } + }, + clearTimer(timer: CoalescerTimer): void { + pending.delete(timer.handle as number) + }, + fireAll(): void { + for (const callback of [...pending.values()]) callback() + } + } +} + +function createCoalescer(windowMs = 3_000) { + const timers = createTimerHarness() + const delivered: PushDelivery[] = [] + const coalescer = new PushCoalescer({ + windowMs, + deliver: async (delivery) => { + delivered.push(delivery) + }, + setTimer: (callback, delayMs) => timers.setTimer(callback, delayMs), + clearTimer: (timer) => timers.clearTimer(timer) + }) + return { coalescer, delivered, timers } +} + +describe('push coalescer', () => { + it('sends a single event unchanged with the notification collapse id', async () => { + const { coalescer, delivered, timers } = createCoalescer() + coalescer.enqueue({ registrationId: 'reg-1', hostFingerprint: HOST, notification: notification() }) + expect(timers.delays).toEqual([3_000]) + expect(delivered).toHaveLength(0) + await coalescer.flush('reg-1') + expect(delivered).toHaveLength(1) + expect(delivered[0]).toMatchObject({ + registrationId: 'reg-1', + title: 'Agent needs input', + body: 'Waiting on your answer', + collapseId: 'note-1' + }) + expect(delivered[0]?.orca).toMatchObject({ + hostFingerprint: HOST, + notificationId: 'note-1', + notificationSeq: 1, + worktreeId: 'wt-1', + coalescedCount: 1 + }) + }) + + it('falls back to the host collapse id when the event carries no notification id', async () => { + const { coalescer, delivered } = createCoalescer() + const { notificationId: _absent, ...bell } = notification({ source: 'terminal-bell' }) + coalescer.enqueue({ + registrationId: 'reg-1', + hostFingerprint: HOST, + notification: { ...bell, agentState: null } + }) + await coalescer.flush('reg-1') + expect(delivered[0]?.collapseId).toBe(`host:${HOST}`) + expect(delivered[0]?.orca.notificationId).toBeUndefined() + }) + + it('summarises a burst and collapses it under the host id', async () => { + const { coalescer, delivered } = createCoalescer() + for (const seq of [1, 2, 3]) { + coalescer.enqueue({ + registrationId: 'reg-1', + hostFingerprint: HOST, + notification: notification({ notificationId: `note-${seq}`, notificationSeq: seq }) + }) + } + expect(coalescer.pendingCount('reg-1')).toBe(3) + await coalescer.flush('reg-1') + expect(delivered).toHaveLength(1) + expect(delivered[0]).toMatchObject({ + title: 'Orca', + body: '3 agents need attention', + collapseId: `host:${HOST}` + }) + // The data carries the latest event, so a tap still opens the newest work. + expect(delivered[0]?.orca).toMatchObject({ + notificationId: 'note-3', + notificationSeq: 3, + coalescedCount: 3 + }) + }) + + it('says updates when no event in the burst needs input', async () => { + const { coalescer, delivered } = createCoalescer() + for (const seq of [1, 2]) { + coalescer.enqueue({ + registrationId: 'reg-1', + hostFingerprint: HOST, + notification: notification({ notificationSeq: seq, agentState: 'finished' }) + }) + } + await coalescer.flush('reg-1') + expect(delivered[0]?.body).toBe('2 updates') + expect(summaryBody([notification({ agentState: null }), notification({ agentState: null })])) + .toBe('2 updates') + }) + + it('keeps one window per registration', async () => { + const { coalescer, delivered, timers } = createCoalescer() + coalescer.enqueue({ registrationId: 'reg-1', hostFingerprint: HOST, notification: notification() }) + coalescer.enqueue({ registrationId: 'reg-2', hostFingerprint: HOST, notification: notification() }) + coalescer.enqueue({ registrationId: 'reg-1', hostFingerprint: HOST, notification: notification() }) + expect(timers.delays).toHaveLength(2) + await coalescer.flushAll() + expect(delivered.map((delivery) => delivery.registrationId).sort()).toEqual(['reg-1', 'reg-2']) + expect(delivered.find((d) => d.registrationId === 'reg-1')?.orca.coalescedCount).toBe(2) + expect(delivered.find((d) => d.registrationId === 'reg-2')?.orca.coalescedCount).toBe(1) + }) + + it('flushes when the window timer fires and starts a fresh window after', async () => { + const { coalescer, delivered, timers } = createCoalescer() + coalescer.enqueue({ registrationId: 'reg-1', hostFingerprint: HOST, notification: notification() }) + timers.fireAll() + await Promise.resolve() + expect(delivered).toHaveLength(1) + coalescer.enqueue({ registrationId: 'reg-1', hostFingerprint: HOST, notification: notification() }) + expect(coalescer.pendingCount('reg-1')).toBe(1) + await coalescer.flushAll() + expect(delivered).toHaveLength(2) + }) + + it('reports a delivery failure instead of throwing into the caller', async () => { + const failures: unknown[] = [] + const coalescer = new PushCoalescer({ + windowMs: 0, + deliver: async () => { + throw new Error('provider down') + }, + setTimer: () => ({ handle: null }), + clearTimer: () => undefined, + onDeliveryFailed: (error) => failures.push(error) + }) + coalescer.enqueue({ registrationId: 'reg-1', hostFingerprint: HOST, notification: notification() }) + await expect(coalescer.flush('reg-1')).resolves.toBeUndefined() + expect(failures).toHaveLength(1) + coalescer.stop() + }) +}) diff --git a/cloud/apps/push/src/coalescer.ts b/cloud/apps/push/src/coalescer.ts new file mode 100644 index 00000000000..f55b6757418 --- /dev/null +++ b/cloud/apps/push/src/coalescer.ts @@ -0,0 +1,117 @@ +import { PUSH_LIMITS, type PushNotification } from '@orca-cloud/push-contract' +import { buildPushDelivery, type PushDelivery } from './push-delivery-message.js' + +export type CoalescerTimer = { readonly handle: unknown } + +export type PushCoalescerOptions = { + windowMs?: number + deliver: (delivery: PushDelivery) => Promise + setTimer?: (callback: () => void, delayMs: number) => CoalescerTimer + clearTimer?: (timer: CoalescerTimer) => void + onDeliveryFailed?: (error: unknown) => void +} + +type PendingWindow = { + hostFingerprint: string + notifications: PushNotification[] + timer: CoalescerTimer +} + +function defaultSetTimer(callback: () => void, delayMs: number): CoalescerTimer { + const handle = setTimeout(callback, delayMs) + handle.unref?.() + return { handle } +} + +function defaultClearTimer(timer: CoalescerTimer): void { + clearTimeout(timer.handle as NodeJS.Timeout) +} + +export function summaryBody(notifications: readonly PushNotification[]): string { + const count = notifications.length + return notifications.some((notification) => notification.agentState === 'needs-input') + ? `${count} agents need attention` + : `${count} updates` +} + +// Holds sends per registration for one window so a burst of desktop events +// reaches the phone as a single banner instead of a stack of near-duplicates. +export class PushCoalescer { + private readonly deliveries = new Set>() + private stopped = false + private readonly windows = new Map() + private readonly windowMs: number + private readonly setTimer: (callback: () => void, delayMs: number) => CoalescerTimer + private readonly clearTimer: (timer: CoalescerTimer) => void + + constructor(private readonly options: PushCoalescerOptions) { + this.windowMs = options.windowMs ?? PUSH_LIMITS.coalesceWindowMs + this.setTimer = options.setTimer ?? defaultSetTimer + this.clearTimer = options.clearTimer ?? defaultClearTimer + } + + enqueue(input: { + registrationId: string + hostFingerprint: string + notification: PushNotification + }): void { + if (this.stopped) throw new Error('push_coalescer_stopped') + const existing = this.windows.get(input.registrationId) + if (existing) { + existing.notifications.push(input.notification) + return + } + this.windows.set(input.registrationId, { + hostFingerprint: input.hostFingerprint, + notifications: [input.notification], + timer: this.setTimer(() => { + void this.flush(input.registrationId) + }, this.windowMs) + }) + } + + pendingCount(registrationId: string): number { + return this.windows.get(registrationId)?.notifications.length ?? 0 + } + + async flush(registrationId: string): Promise { + const window = this.windows.get(registrationId) + if (!window) return + this.windows.delete(registrationId) + this.clearTimer(window.timer) + const latest = window.notifications.at(-1)! + const coalescedCount = window.notifications.length + const delivery = buildPushDelivery({ + registrationId, + hostFingerprint: window.hostFingerprint, + notification: latest, + title: coalescedCount > 1 ? 'Orca' : latest.title, + body: coalescedCount > 1 ? summaryBody(window.notifications) : latest.body, + coalescedCount + }) + const pending = Promise.resolve() + .then(() => this.options.deliver(delivery)) + .catch((error) => { + this.options.onDeliveryFailed?.(error) + }) + this.deliveries.add(pending) + try { + await pending + } finally { + this.deliveries.delete(pending) + } + } + + async flushAll(): Promise { + do { + await Promise.all([...this.windows.keys()].map((id) => this.flush(id))) + await Promise.all([...this.deliveries]) + } while (this.windows.size || this.deliveries.size) + } + + stop(): void { + this.stopped = true + for (const window of this.windows.values()) this.clearTimer(window.timer) + this.windows.clear() + } +} diff --git a/cloud/apps/push/src/config.test.ts b/cloud/apps/push/src/config.test.ts new file mode 100644 index 00000000000..857022a63a3 --- /dev/null +++ b/cloud/apps/push/src/config.test.ts @@ -0,0 +1,90 @@ +import { generateKeyPairSync } from 'node:crypto' +import { PUSH_DEFAULTS, PUSH_LIMITS } from '@orca-cloud/push-contract' +import { describe, expect, it } from 'vitest' +import { loadPushConfig, PUSH_DATABASE_POOL_MAX } from './config.js' + +function apnsKeyPem(): string { + return generateKeyPairSync('ec', { + namedCurve: 'P-256', + privateKeyEncoding: { type: 'pkcs8', format: 'pem' }, + publicKeyEncoding: { type: 'spki', format: 'pem' } + }).privateKey +} + +const MINIMAL = { ORCA_PUSH_PUBLIC_URL: 'https://push.onorca.dev' } + +describe('push gateway config', () => { + it('applies the documented defaults', () => { + expect(loadPushConfig(MINIMAL)).toEqual({ + port: 8080, + publicUrl: 'https://push.onorca.dev', + databaseUrl: undefined, + dataDir: './data/push', + databasePoolMax: PUSH_DATABASE_POOL_MAX, + apns: undefined, + apnsTopic: PUSH_DEFAULTS.apnsTopic, + fcmProjectId: PUSH_DEFAULTS.fcmProjectId, + coalesceMs: PUSH_LIMITS.coalesceWindowMs, + trustedProxyHops: 0 + }) + }) + + it('reads a full APNs credential and the overridable knobs', () => { + const keyPem = apnsKeyPem() + const config = loadPushConfig({ + ...MINIMAL, + PORT: '9090', + ORCA_PUSH_DATABASE_URL: 'postgres://localhost/orca_push', + ORCA_PUSH_DATA_DIR: '/var/lib/push', + ORCA_PUSH_APNS_KEY: keyPem, + ORCA_PUSH_APNS_KEY_ID: 'ABCDE12345', + ORCA_PUSH_APPLE_TEAM_ID: 'TEAM123456', + ORCA_PUSH_APNS_TOPIC: 'com.stably.orca.mobile.dev', + ORCA_PUSH_FCM_PROJECT_ID: 'onorca-staging', + ORCA_PUSH_COALESCE_MS: '1500', + ORCA_PUSH_TRUSTED_PROXY_HOPS: '1' + }) + expect(config).toMatchObject({ + port: 9090, + databaseUrl: 'postgres://localhost/orca_push', + dataDir: '/var/lib/push', + apns: { keyPem, keyId: 'ABCDE12345', teamId: 'TEAM123456' }, + apnsTopic: 'com.stably.orca.mobile.dev', + trustedProxyHops: 1, + fcmProjectId: 'onorca-staging', + coalesceMs: 1500 + }) + }) + + it('refuses a partial APNs credential', () => { + expect(() => + loadPushConfig({ ...MINIMAL, ORCA_PUSH_APNS_KEY: apnsKeyPem() }) + ).toThrow('configured together') + expect(() => + loadPushConfig({ + ...MINIMAL, + ORCA_PUSH_APNS_KEY: 'not-a-pem', + ORCA_PUSH_APNS_KEY_ID: 'ABCDE12345', + ORCA_PUSH_APPLE_TEAM_ID: 'TEAM123456' + }) + ).toThrow('PEM text') + }) + + it('requires a canonical HTTPS origin outside loopback', () => { + expect(() => loadPushConfig({ ORCA_PUSH_PUBLIC_URL: 'https://push.onorca.dev/v1' })).toThrow( + 'must be an origin' + ) + expect(() => loadPushConfig({ ORCA_PUSH_PUBLIC_URL: 'http://push.onorca.dev' })).toThrow( + 'must use HTTPS' + ) + expect(loadPushConfig({ ORCA_PUSH_PUBLIC_URL: 'http://localhost:8080' }).publicUrl).toBe( + 'http://localhost:8080' + ) + }) + + it('treats an empty optional variable as unset', () => { + expect( + loadPushConfig({ ...MINIMAL, ORCA_PUSH_DATABASE_URL: '', ORCA_PUSH_APNS_KEY_ID: '' }) + ).toMatchObject({ databaseUrl: undefined, apns: undefined }) + }) +}) diff --git a/cloud/apps/push/src/config.ts b/cloud/apps/push/src/config.ts new file mode 100644 index 00000000000..08ec608e528 --- /dev/null +++ b/cloud/apps/push/src/config.ts @@ -0,0 +1,105 @@ +import { PUSH_DEFAULTS, PUSH_LIMITS } from '@orca-cloud/push-contract' +import { z } from 'zod' + +export const PUSH_DATABASE_POOL_MAX = 10 + +const OptionalTextSchema = z.preprocess( + (value) => (value === '' ? undefined : value), + z.string().min(1).optional() +) + +const EnvSchema = z.object({ + PORT: z.coerce.number().int().positive().default(8080), + ORCA_PUSH_PUBLIC_URL: z.string().url(), + ORCA_PUSH_DATABASE_URL: OptionalTextSchema, + ORCA_PUSH_DATA_DIR: z.string().min(1).default('./data/push'), + ORCA_PUSH_DATABASE_POOL_MAX: z.coerce.number().int().positive().max(100).optional(), + ORCA_PUSH_APNS_KEY: OptionalTextSchema, + ORCA_PUSH_APNS_KEY_ID: z.preprocess( + (value) => (value === '' ? undefined : value), + z.string().regex(/^[A-Z0-9]{10}$/).optional() + ), + ORCA_PUSH_APPLE_TEAM_ID: z.preprocess( + (value) => (value === '' ? undefined : value), + z.string().regex(/^[A-Z0-9]{10}$/).optional() + ), + ORCA_PUSH_APNS_TOPIC: z.string().min(1).max(255).default(PUSH_DEFAULTS.apnsTopic), + ORCA_PUSH_FCM_PROJECT_ID: z + .string() + .regex(/^[a-z0-9-]{4,64}$/) + .default(PUSH_DEFAULTS.fcmProjectId), + ORCA_PUSH_COALESCE_MS: z.coerce + .number() + .int() + .nonnegative() + .max(60_000) + .default(PUSH_LIMITS.coalesceWindowMs), + // How many proxies append to x-forwarded-for after the client. 0 is Cloud Run + // alone; raise it to 1 when a load balancer fronts the service. + ORCA_PUSH_TRUSTED_PROXY_HOPS: z.coerce.number().int().nonnegative().max(8).default(0) +}) + +export type ApnsCredentials = { keyPem: string; keyId: string; teamId: string } + +export type PushConfig = { + port: number + publicUrl: string + databaseUrl?: string + dataDir: string + databasePoolMax: number + apns?: ApnsCredentials + apnsTopic: string + fcmProjectId: string + coalesceMs: number + trustedProxyHops: number +} + +function canonicalOrigin(value: string, name: string): string { + const url = new URL(value) + if (url.origin !== value || url.pathname !== '/') throw new Error(`${name} must be an origin`) + const loopback = ['127.0.0.1', 'localhost', '::1', '[::1]'].includes(url.hostname) + if (url.protocol !== 'https:' && !(loopback && url.protocol === 'http:')) { + throw new Error(`${name} must use HTTPS outside loopback development`) + } + return value +} + +// The APNs key, key id, and team id are one credential; a partial set would +// pass startup and then fail every iOS send at runtime. +function readApnsCredentials( + parsed: z.infer +): ApnsCredentials | undefined { + const parts = [ + parsed.ORCA_PUSH_APNS_KEY, + parsed.ORCA_PUSH_APNS_KEY_ID, + parsed.ORCA_PUSH_APPLE_TEAM_ID + ] + const present = parts.filter((value) => value !== undefined).length + if (present === 0) return undefined + if (present !== parts.length) { + throw new Error('APNs key, key id, and team id must be configured together') + } + const keyPem = parsed.ORCA_PUSH_APNS_KEY! + if (!keyPem.includes('-----BEGIN')) throw new Error('ORCA_PUSH_APNS_KEY must be PEM text') + return { + keyPem, + keyId: parsed.ORCA_PUSH_APNS_KEY_ID!, + teamId: parsed.ORCA_PUSH_APPLE_TEAM_ID! + } +} + +export function loadPushConfig(env: NodeJS.ProcessEnv = process.env): PushConfig { + const parsed = EnvSchema.parse(env) + return { + port: parsed.PORT, + publicUrl: canonicalOrigin(parsed.ORCA_PUSH_PUBLIC_URL, 'ORCA_PUSH_PUBLIC_URL'), + databaseUrl: parsed.ORCA_PUSH_DATABASE_URL, + dataDir: parsed.ORCA_PUSH_DATA_DIR, + databasePoolMax: parsed.ORCA_PUSH_DATABASE_POOL_MAX ?? PUSH_DATABASE_POOL_MAX, + apns: readApnsCredentials(parsed), + apnsTopic: parsed.ORCA_PUSH_APNS_TOPIC, + fcmProjectId: parsed.ORCA_PUSH_FCM_PROJECT_ID, + coalesceMs: parsed.ORCA_PUSH_COALESCE_MS, + trustedProxyHops: parsed.ORCA_PUSH_TRUSTED_PROXY_HOPS + } +} diff --git a/cloud/apps/push/src/desktop-host-proof-interop.test.ts b/cloud/apps/push/src/desktop-host-proof-interop.test.ts new file mode 100644 index 00000000000..654423b0de8 --- /dev/null +++ b/cloud/apps/push/src/desktop-host-proof-interop.test.ts @@ -0,0 +1,47 @@ +import { describe, expect, it } from 'vitest' +import { createHmac } from 'node:crypto' +import vector from '../../../packages/push-contract/src/push-host-proof-vector.json' with { type: 'json' } +import { answerPushHostChallenge, createPushHostKeypair } from './host-challenge-answering.test-fixture.js' +import { PushHostChallengeStore } from './host-challenge-store.js' +import { deriveHostFingerprint } from './host-fingerprint.js' +import { openInMemoryPushDatabase } from './push-database.js' + +// Why: the desktop answers challenges in a workspace this one cannot import. +// Both sides replay the same checked-in vector, so a transcript drift on +// either side fails in that side's own suite. +describe('desktop host proof interop', () => { + it('the checked-in vector answers to the same proof the fixture host computes', () => { + const secretKey = new Uint8Array(Buffer.from(vector.hostSecretKeyB64, 'base64')) + const keypair = { publicKey: new Uint8Array(Buffer.from(vector.hostPublicKeyB64, 'base64')), secretKey } + expect(deriveHostFingerprint(keypair.publicKey)).toBe(vector.hostFingerprint) + const proof = answerPushHostChallenge(vector.challenge, { + gatewayOrigin: vector.gatewayOrigin, + keypair, + now: () => vector.issuedAt + 1_000 + }) + const expected = createHmac('sha256', Buffer.from(vector.challengeSecretB64, 'base64')) + .update(Buffer.from('orca-push-host-proof/v1\0ack\0')) + .update(Buffer.from(vector.transcriptB64, 'base64')) + .digest('base64') + expect(proof).toBe(expected) + }) + + it('a live challenge from the store round-trips through the fixture host once', async () => { + const database = await openInMemoryPushDatabase() + const store = new PushHostChallengeStore(database, vector.gatewayOrigin) + const keypair = createPushHostKeypair(11) + const challenge = await store.issue(Buffer.from(keypair.publicKey).toString('base64')) + expect(challenge).not.toBeNull() + const proof = answerPushHostChallenge(challenge!, { gatewayOrigin: vector.gatewayOrigin, keypair }) + expect(proof).not.toBeNull() + expect(await store.verify(challenge!.challengeId, proof!)).toEqual({ + ok: true, + hostFingerprint: deriveHostFingerprint(keypair.publicKey) + }) + expect(await store.verify(challenge!.challengeId, proof!)).toEqual({ + ok: false, + reason: 'already_consumed' + }) + await database.close() + }) +}) diff --git a/cloud/apps/push/src/device-registry-store.test.ts b/cloud/apps/push/src/device-registry-store.test.ts new file mode 100644 index 00000000000..f191112f06a --- /dev/null +++ b/cloud/apps/push/src/device-registry-store.test.ts @@ -0,0 +1,205 @@ +import { PUSH_LIMITS, type PushNotificationFilter } from '@orca-cloud/push-contract' +import { afterEach, beforeEach, describe, expect, it } from 'vitest' +import { PushDeviceRegistryStore, type PushDeviceUpsert } from './device-registry-store.js' +import { openInMemoryPushDatabase, type PushDatabase } from './push-database.js' + +const OWNER = 'abcdefghijklmnop' +const OTHER = 'ponmlkjihgfedcba' +const FILTER: PushNotificationFilter = { + sources: ['agent-task-complete'], + agentStates: ['needs-input'] +} + +describe('push device registry store', () => { + let database: PushDatabase + let clock = 1_700_000_000_000 + let devices: PushDeviceRegistryStore + + beforeEach(async () => { + database = await openInMemoryPushDatabase() + clock = 1_700_000_000_000 + devices = new PushDeviceRegistryStore(database, () => clock) + }) + + afterEach(async () => { + await database.close() + }) + + async function upsertOk(input: PushDeviceUpsert): Promise { + const result = await devices.upsert(input) + if (!result.ok) throw new Error(`unexpected upsert refusal: ${result.reason}`) + return result.registrationId + } + + function androidDevice(deviceId: string): PushDeviceUpsert { + return { + hostFingerprint: OWNER, + deviceId, + platform: 'android', + token: `token-${deviceId}`, + filter: FILTER + } + } + + it('keeps one registration per host and device while replacing the token', async () => { + const first = await upsertOk({ + hostFingerprint: OWNER, + deviceId: 'device-1', + platform: 'ios', + token: 'a'.repeat(64), + apnsEnvironment: 'sandbox', + filter: FILTER + }) + clock += 1_000 + const second = await upsertOk({ + hostFingerprint: OWNER, + deviceId: 'device-1', + platform: 'ios', + token: 'b'.repeat(64), + apnsEnvironment: 'production', + filter: FILTER + }) + expect(second).toBe(first) + const registration = await devices.findById(first) + expect(registration).toMatchObject({ + token: 'b'.repeat(64), + apnsEnvironment: 'production', + dead: false + }) + expect(await devices.list(OWNER)).toHaveLength(1) + }) + + it('revives a registration that a re-registered token replaces', async () => { + const registrationId = await upsertOk({ + hostFingerprint: OWNER, + deviceId: 'device-1', + platform: 'android', + token: 'token-one', + filter: FILTER + }) + await devices.markDead(registrationId) + expect((await devices.findById(registrationId))?.dead).toBe(true) + await upsertOk({ + hostFingerprint: OWNER, + deviceId: 'device-1', + platform: 'android', + token: 'token-two', + filter: FILTER + }) + expect(await devices.findById(registrationId)).toMatchObject({ + token: 'token-two', + dead: false + }) + }) + + it('lets only the owning host delete a registration', async () => { + const registrationId = await upsertOk({ + hostFingerprint: OWNER, + deviceId: 'device-1', + platform: 'android', + token: 'token-one', + filter: FILTER + }) + expect(await devices.deleteOwned(OTHER, registrationId)).toBe(false) + expect(await devices.findById(registrationId)).not.toBeNull() + expect(await devices.deleteOwned(OWNER, registrationId)).toBe(true) + expect(await devices.findById(registrationId)).toBeNull() + }) + + it('scopes lookups and listings to the owning host', async () => { + const owned = await upsertOk({ + hostFingerprint: OWNER, + deviceId: 'device-1', + platform: 'android', + token: 'token-one', + filter: FILTER + }) + const foreign = await upsertOk({ + hostFingerprint: OTHER, + deviceId: 'device-2', + platform: 'android', + token: 'token-two', + filter: FILTER + }) + const found = await devices.findOwned(OWNER, [owned, foreign]) + expect([...found.keys()]).toEqual([owned]) + expect(await devices.list(OTHER)).toEqual([ + { registrationId: foreign, deviceId: 'device-2', platform: 'android', dead: false } + ]) + expect(await devices.findOwned(OWNER, [])).toEqual(new Map()) + }) + + it('refuses a new device once the host reaches its registration cap', async () => { + for (let index = 0; index < PUSH_LIMITS.maxDevicesPerHost; index++) { + await upsertOk(androidDevice(`device-${index}`)) + } + expect(await devices.upsert(androidDevice('one-too-many'))).toEqual({ + ok: false, + reason: 'too_many_devices' + }) + expect(await devices.list(OWNER)).toHaveLength(PUSH_LIMITS.maxDevicesPerHost) + }) + + it('still lets a capped host re-register a device it already owns', async () => { + for (let index = 0; index < PUSH_LIMITS.maxDevicesPerHost; index++) { + await upsertOk(androidDevice(`device-${index}`)) + } + const rotated = await devices.upsert({ ...androidDevice('device-0'), token: 'rotated-token' }) + expect(rotated.ok).toBe(true) + expect(await devices.list(OWNER)).toHaveLength(PUSH_LIMITS.maxDevicesPerHost) + }) + + it('frees a slot when a registration is deleted', async () => { + const first = await upsertOk(androidDevice('device-0')) + for (let index = 1; index < PUSH_LIMITS.maxDevicesPerHost; index++) { + await upsertOk(androidDevice(`device-${index}`)) + } + expect((await devices.upsert(androidDevice('extra'))).ok).toBe(false) + expect(await devices.deleteOwned(OWNER, first)).toBe(true) + expect((await devices.upsert(androidDevice('extra'))).ok).toBe(true) + }) + + it('counts the cap per host, not across the whole table', async () => { + for (let index = 0; index < PUSH_LIMITS.maxDevicesPerHost; index++) { + await upsertOk(androidDevice(`device-${index}`)) + } + expect((await devices.upsert(androidDevice('extra'))).ok).toBe(false) + expect( + (await devices.upsert({ ...androidDevice('device-0'), hostFingerprint: OTHER })).ok + ).toBe(true) + }) + + it('never returns more devices than the list response schema accepts', async () => { + // Straight past the per-host cap, so only the query LIMIT can bound this. + const rows = PUSH_LIMITS.maxDevicesPerListResponse + 5 + for (let index = 0; index < rows; index++) { + await database.query( + `INSERT INTO push_devices (registration_id, host_fingerprint, device_id, platform, token, + filter_json, created_at, updated_at) + VALUES (?, ?, ?, ?, ?, ?, ?, ?)`, + [`reg-${index}`, OWNER, `device-${index}`, 'android', 'token', '{}', clock + index, clock] + ) + } + expect(await devices.list(OWNER)).toHaveLength(PUSH_LIMITS.maxDevicesPerListResponse) + }) + + it('separates the same device id registered against two hosts', async () => { + const first = await upsertOk({ + hostFingerprint: OWNER, + deviceId: 'shared-device', + platform: 'ios', + token: 'a'.repeat(64), + apnsEnvironment: 'sandbox', + filter: FILTER + }) + const second = await upsertOk({ + hostFingerprint: OTHER, + deviceId: 'shared-device', + platform: 'ios', + token: 'c'.repeat(64), + apnsEnvironment: 'sandbox', + filter: FILTER + }) + expect(first).not.toBe(second) + }) +}) diff --git a/cloud/apps/push/src/device-registry-store.ts b/cloud/apps/push/src/device-registry-store.ts new file mode 100644 index 00000000000..9aac22dd25c --- /dev/null +++ b/cloud/apps/push/src/device-registry-store.ts @@ -0,0 +1,185 @@ +import { randomUUID } from 'node:crypto' +import { + PUSH_LIMITS, + type ApnsEnvironment, + type PushDeviceSummary, + type PushNotificationFilter, + type PushPlatform +} from '@orca-cloud/push-contract' +import type { PushDatabase, SqlRow } from './push-database.js' + +const DEVICE_CAP_LOCK_PREFIX = 'orca-push-device-cap:' + +export type PushDeviceRegistration = { + registrationId: string + hostFingerprint: string + deviceId: string + platform: PushPlatform + token: string + apnsEnvironment?: ApnsEnvironment + dead: boolean +} + +export type PushDeviceUpsertResult = + | { ok: true; registrationId: string } + | { ok: false; reason: 'too_many_devices' } + +export type PushDeviceUpsert = { + hostFingerprint: string + deviceId: string + platform: PushPlatform + token: string + apnsEnvironment?: ApnsEnvironment + filter: PushNotificationFilter +} + +function toRegistration(row: SqlRow): PushDeviceRegistration { + const apnsEnvironment = row.apns_environment + return { + registrationId: String(row.registration_id), + hostFingerprint: String(row.host_fingerprint), + deviceId: String(row.device_id), + platform: String(row.platform) as PushPlatform, + token: String(row.token), + ...(apnsEnvironment === null || apnsEnvironment === undefined + ? {} + : { apnsEnvironment: String(apnsEnvironment) as ApnsEnvironment }), + dead: row.dead_at !== null && row.dead_at !== undefined + } +} + +export class PushDeviceRegistryStore { + constructor( + private readonly database: PushDatabase, + private readonly now: () => number = Date.now + ) {} + + // The registration id is stable for a (host, device) pair so a re-registered + // phone keeps the id the desktop already persisted; only the token rotates. + async upsert(input: PushDeviceUpsert): Promise { + const now = this.now() + const filterJson = JSON.stringify(input.filter) + return await this.database.transaction(async (transaction) => { + // deviceId is caller-chosen, so counting and inserting must not interleave + // or a burst of new ids would walk straight past the cap. + await transaction.lockQuotaScope(`${DEVICE_CAP_LOCK_PREFIX}${input.hostFingerprint}`) + const [existing] = await transaction.query( + 'SELECT registration_id FROM push_devices WHERE host_fingerprint = ? AND device_id = ?', + [input.hostFingerprint, input.deviceId] + ) + if (existing) { + const registrationId = String(existing.registration_id) + await transaction.query( + `UPDATE push_devices + SET platform = ?, token = ?, apns_environment = ?, filter_json = ?, + dead_at = NULL, updated_at = ? + WHERE registration_id = ?`, + [ + input.platform, + input.token, + input.apnsEnvironment ?? null, + filterJson, + now, + registrationId + ] + ) + return { ok: true, registrationId } + } + const [countRow] = await transaction.query( + 'SELECT COUNT(*) AS devices FROM push_devices WHERE host_fingerprint = ?', + [input.hostFingerprint] + ) + if (Number(countRow?.devices ?? 0) >= PUSH_LIMITS.maxDevicesPerHost) { + return { ok: false, reason: 'too_many_devices' } + } + const registrationId = randomUUID() + await transaction.query( + `INSERT INTO push_devices + (registration_id, host_fingerprint, device_id, platform, token, apns_environment, + filter_json, dead_at, created_at, updated_at) + VALUES (?, ?, ?, ?, ?, ?, ?, NULL, ?, ?)`, + [ + registrationId, + input.hostFingerprint, + input.deviceId, + input.platform, + input.token, + input.apnsEnvironment ?? null, + filterJson, + now, + now + ] + ) + return { ok: true, registrationId } + }) + } + + async deleteOwned(hostFingerprint: string, registrationId: string): Promise { + const [result] = await this.database.query( + 'DELETE FROM push_devices WHERE registration_id = ? AND host_fingerprint = ?', + [registrationId, hostFingerprint] + ) + return Number(result?.changes ?? 0) > 0 + } + + async list(hostFingerprint: string): Promise { + const rows = await this.database.query( + // Bounded to what PushDeviceListResponseSchema will accept, so an + // oversized table degrades to a truncated list instead of a 500. + `SELECT registration_id, device_id, platform, dead_at + FROM push_devices WHERE host_fingerprint = ? ORDER BY created_at ASC LIMIT ?`, + [hostFingerprint, PUSH_LIMITS.maxDevicesPerListResponse] + ) + return rows.map((row) => ({ + registrationId: String(row.registration_id), + deviceId: String(row.device_id), + platform: String(row.platform) as PushPlatform, + dead: row.dead_at !== null && row.dead_at !== undefined + })) + } + + async findOwned( + hostFingerprint: string, + registrationIds: readonly string[] + ): Promise> { + if (registrationIds.length === 0) return new Map() + const placeholders = registrationIds.map(() => '?').join(', ') + const rows = await this.database.query( + `SELECT registration_id, host_fingerprint, device_id, platform, token, apns_environment, + dead_at + FROM push_devices + WHERE host_fingerprint = ? AND registration_id IN (${placeholders})`, + [hostFingerprint, ...registrationIds] + ) + return new Map( + rows.map((row) => { + const registration = toRegistration(row) + return [registration.registrationId, registration] + }) + ) + } + + async findById(registrationId: string): Promise { + const [row] = await this.database.query( + `SELECT registration_id, host_fingerprint, device_id, platform, token, apns_environment, + dead_at + FROM push_devices WHERE registration_id = ?`, + [registrationId] + ) + return row ? toRegistration(row) : null + } + + async markDead(registrationId: string, observed?: PushDeviceRegistration): Promise { + await this.database.query( + `UPDATE push_devices SET dead_at = ?, updated_at = ? WHERE registration_id = ?${ + observed ? " AND token = ? AND platform = ? AND COALESCE(apns_environment, '') = ?" : '' + }`, + [ + this.now(), + this.now(), + registrationId, + ...(observed ? [observed.token, observed.platform, observed.apnsEnvironment ?? ''] : []) + ] + ) + } +} diff --git a/cloud/apps/push/src/fcm-access-token.ts b/cloud/apps/push/src/fcm-access-token.ts new file mode 100644 index 00000000000..542e0e8d0ed --- /dev/null +++ b/cloud/apps/push/src/fcm-access-token.ts @@ -0,0 +1,15 @@ +import { GoogleAuth } from 'google-auth-library' +import { FCM_SCOPE } from './fcm-client.js' + +// Resolves the runtime service account credential from the GCE metadata server +// in Cloud Run and from GOOGLE_APPLICATION_CREDENTIALS locally; the library +// caches and refreshes the token itself. +export function createFcmAccessTokenProvider(): () => Promise { + const auth = new GoogleAuth({ scopes: [FCM_SCOPE] }) + return async () => { + const client = await auth.getClient() + const token = await client.getAccessToken() + if (!token.token) throw new Error('fcm_access_token_unavailable') + return token.token + } +} diff --git a/cloud/apps/push/src/fcm-client.test.ts b/cloud/apps/push/src/fcm-client.test.ts new file mode 100644 index 00000000000..3069c62032b --- /dev/null +++ b/cloud/apps/push/src/fcm-client.test.ts @@ -0,0 +1,182 @@ +import { createHash } from 'node:crypto' +import { describe, expect, it } from 'vitest' +import { fcmCollapseKey, FcmClient, type FcmRequest, type FcmResponse } from './fcm-client.js' +import { buildPushDelivery } from './push-delivery-message.js' + +const HOST = 'abcdefghijklmnop' +const TOKEN = 'cQ1abcDEF_gh:APA91bZZ-zz0123456789abcdefghijklmnopqrstuvwxyz' + +function delivery(coalescedCount = 1, agentState: 'needs-input' | null = 'needs-input') { + return buildPushDelivery({ + registrationId: 'reg-1', + hostFingerprint: HOST, + notification: { + notificationId: 'note-1', + notificationSeq: 7, + notificationEpoch: 'epoch-1', + source: 'agent-task-complete', + agentState, + title: 'Agent needs input', + body: 'Waiting on your answer', + worktreeId: 'wt-1' + }, + title: coalescedCount > 1 ? 'Orca' : 'Agent needs input', + body: coalescedCount > 1 ? '3 agents need attention' : 'Waiting on your answer', + coalescedCount + }) +} + +function fakeTransport(response: FcmResponse) { + const requests: FcmRequest[] = [] + return { + requests, + transport: async (request: FcmRequest): Promise => { + requests.push(request) + return response + } + } +} + +function client(response: FcmResponse) { + const fake = fakeTransport(response) + return { + fake, + client: new FcmClient({ + projectId: 'onorca-cloud', + accessToken: async () => 'access-token', + transport: fake.transport + }) + } +} + +describe('fcm client', () => { + it('posts the v1 send payload for the configured project', async () => { + const { fake, client: fcm } = client({ status: 200, body: '{"name":"projects/x/messages/1"}' }) + await expect(fcm.send(delivery(), { token: TOKEN })).resolves.toEqual({ status: 'sent' }) + const request = fake.requests[0]! + expect(request.url).toBe('https://fcm.googleapis.com/v1/projects/onorca-cloud/messages:send') + expect(request.accessToken).toBe('access-token') + expect(JSON.parse(request.body)).toEqual({ + message: { + token: TOKEN, + notification: { title: 'Agent needs input', body: 'Waiting on your answer' }, + android: { + priority: 'HIGH', + ttl: '14400s', + collapse_key: createHash('sha256').update('note-1').digest('hex').slice(0, 32), + notification: { channel_id: 'orca-desktop', tag: 'note-1' } + }, + data: { + hostFingerprint: HOST, + worktreeId: 'wt-1', + notificationId: 'note-1', + notificationSeq: '7', + notificationEpoch: 'epoch-1', + source: 'agent-task-complete', + agentState: 'needs-input', + coalescedCount: '1' + } + } + }) + }) + + it('carries every data value as a string and omits a null agent state', async () => { + const { fake, client: fcm } = client({ status: 200, body: '{}' }) + await fcm.send(delivery(3, null), { token: TOKEN }) + const message = JSON.parse(fake.requests[0]!.body) as { + message: { + android: { collapse_key: string; notification: { tag: string } } + data: Record + } + } + expect(Object.values(message.message.data).every((value) => typeof value === 'string')).toBe( + true + ) + expect(message.message.data.agentState).toBeUndefined() + expect(message.message.data.coalescedCount).toBe('3') + expect(message.message.android.notification.tag).toBe(`host:${HOST}`) + expect(message.message.android.collapse_key).toBe(fcmCollapseKey(`host:${HOST}`)) + expect(message.message.android.collapse_key).toHaveLength(32) + }) + + it('passes validate_only through for the deploy probe', async () => { + const { fake, client: fcm } = client({ status: 200, body: '{}' }) + await fcm.send(delivery(), { token: TOKEN }, { validateOnly: true }) + expect(JSON.parse(fake.requests[0]!.body)).toMatchObject({ validate_only: true }) + }) + + it('marks an unregistered token dead from the status or the error detail', async () => { + const byStatus = client({ + status: 404, + body: JSON.stringify({ error: { status: 'UNREGISTERED', message: 'not registered' } }) + }) + await expect(byStatus.client.send(delivery(), { token: TOKEN })).resolves.toEqual({ + status: 'dead', + reason: 'UNREGISTERED' + }) + const byDetail = client({ + status: 404, + body: JSON.stringify({ + error: { + status: 'NOT_FOUND', + message: 'Requested entity was not found.', + details: [{ errorCode: 'UNREGISTERED' }] + } + }) + }) + await expect(byDetail.client.send(delivery(), { token: TOKEN })).resolves.toEqual({ + status: 'dead', + reason: 'UNREGISTERED' + }) + }) + + it('marks an invalid-argument that names the token dead, and others an error', async () => { + const named = client({ + status: 400, + body: JSON.stringify({ + error: { status: 'INVALID_ARGUMENT', message: 'The registration token is not valid.' } + }) + }) + await expect(named.client.send(delivery(), { token: TOKEN })).resolves.toEqual({ + status: 'dead', + reason: 'INVALID_ARGUMENT' + }) + const unnamed = client({ + status: 400, + body: JSON.stringify({ + error: { status: 'INVALID_ARGUMENT', message: 'Invalid value at message.android.ttl' } + }) + }) + await expect(unnamed.client.send(delivery(), { token: TOKEN })).resolves.toEqual({ + status: 'error', + reason: 'INVALID_ARGUMENT', + retryable: false, + retryAfterMs: 10000 + }) + }) + + it('treats a server fault and a transport failure as errors', async () => { + const faulted = client({ + status: 503, + body: JSON.stringify({ error: { status: 'UNAVAILABLE', message: 'backend busy' } }) + }) + await expect(faulted.client.send(delivery(), { token: TOKEN })).resolves.toEqual({ + status: 'error', + reason: 'UNAVAILABLE', + retryable: true, + retryAfterMs: 10000 + }) + const broken = new FcmClient({ + projectId: 'onorca-cloud', + accessToken: async () => 'access-token', + transport: async () => { + throw new Error('ECONNRESET') + } + }) + await expect(broken.send(delivery(), { token: TOKEN })).resolves.toEqual({ + status: 'error', + reason: 'Error', + retryable: true + }) + }) +}) diff --git a/cloud/apps/push/src/fcm-client.ts b/cloud/apps/push/src/fcm-client.ts new file mode 100644 index 00000000000..61c7a997345 --- /dev/null +++ b/cloud/apps/push/src/fcm-client.ts @@ -0,0 +1,138 @@ +import { providerRetryAfter } from './provider-retry-delay.js' +import { createHash } from 'node:crypto' +import { PUSH_DEFAULTS, PUSH_LIMITS } from '@orca-cloud/push-contract' +import { orcaDataStrings, type PushDelivery } from './push-delivery-message.js' +import type { PushProviderOutcome } from './push-provider-outcome.js' + +export const FCM_SCOPE = 'https://www.googleapis.com/auth/firebase.messaging' + +export type FcmRequest = { url: string; accessToken: string; body: string } +export type FcmResponse = { status: number; body: string; retryAfterMs?: number } +export type FcmTransport = (request: FcmRequest) => Promise + +export type FcmClientOptions = { + projectId: string + accessToken: () => Promise + transport: FcmTransport + channelId?: string +} + +type FcmErrorBody = { + error?: { status?: unknown; message?: unknown; details?: { errorCode?: unknown }[] } +} + +// FCM collapse_key is a short opaque string, so the collapse id is hashed +// rather than truncated: truncation would merge unrelated notifications. +export function fcmCollapseKey(collapseId: string): string { + return createHash('sha256').update(collapseId).digest('hex').slice(0, 32) +} + +export function fcmMessageBody(input: { + delivery: PushDelivery + token: string + channelId: string + validateOnly?: boolean +}): string { + const { delivery } = input + return JSON.stringify({ + ...(input.validateOnly ? { validate_only: true } : {}), + message: { + token: input.token, + notification: { title: delivery.title, body: delivery.body }, + android: { + priority: 'HIGH', + ttl: `${PUSH_LIMITS.notificationTtlSeconds}s`, + collapse_key: fcmCollapseKey(delivery.collapseId), + notification: { + channel_id: delivery.sound === false ? `${input.channelId}-silent` : input.channelId, + tag: delivery.collapseId + } + }, + data: orcaDataStrings(delivery.orca) + } + }) +} + +function readFcmError(body: string): { status: string; message: string; errorCodes: string[] } { + try { + const parsed = JSON.parse(body) as FcmErrorBody + return { + status: typeof parsed.error?.status === 'string' ? parsed.error.status : 'unknown', + message: typeof parsed.error?.message === 'string' ? parsed.error.message : '', + errorCodes: (parsed.error?.details ?? []) + .map((detail) => detail.errorCode) + .filter((code): code is string => typeof code === 'string') + } + } catch { + return { status: 'unparseable', message: '', errorCodes: [] } + } +} + +export class FcmClient { + private readonly channelId: string + + constructor(private readonly options: FcmClientOptions) { + this.channelId = options.channelId ?? PUSH_DEFAULTS.androidChannelId + } + + async send( + delivery: PushDelivery, + device: { token: string }, + options: { validateOnly?: boolean } = {} + ): Promise { + let response: FcmResponse + try { + response = await this.options.transport({ + url: `https://fcm.googleapis.com/v1/projects/${this.options.projectId}/messages:send`, + accessToken: await this.options.accessToken(), + body: fcmMessageBody({ + delivery, + token: device.token, + channelId: this.channelId, + ...(options.validateOnly === undefined ? {} : { validateOnly: options.validateOnly }) + }) + }) + } catch (error) { + return { + status: 'error', + reason: error instanceof Error ? error.name : 'transport_failed', + retryable: true + } + } + if (response.status >= 200 && response.status < 300) return { status: 'sent' } + const failure = readFcmError(response.body) + if (failure.status === 'UNREGISTERED' || failure.errorCodes.includes('UNREGISTERED')) { + return { status: 'dead', reason: 'UNREGISTERED' } + } + // A revoked token also surfaces as INVALID_ARGUMENT naming the token field. + if (failure.status === 'INVALID_ARGUMENT' && /\btoken\b/i.test(failure.message)) { + return { status: 'dead', reason: 'INVALID_ARGUMENT' } + } + return { + status: 'error', + reason: failure.status, + retryable: response.status === 429 || response.status >= 500, + retryAfterMs: Math.max(response.status === 429 ? 60_000 : 10_000, response.retryAfterMs ?? 0) + } + } +} + +export function createFcmFetchTransport(fetchImpl: typeof fetch = fetch): FcmTransport { + return async (request) => { + const response = await fetchImpl(request.url, { + method: 'POST', + headers: { + authorization: `Bearer ${request.accessToken}`, + 'content-type': 'application/json' + }, + body: request.body, + redirect: 'error', + signal: AbortSignal.timeout(10_000) + }) + return { + status: response.status, + body: await response.text(), + retryAfterMs: providerRetryAfter(response.headers.get('retry-after') ?? undefined) + } + } +} diff --git a/cloud/apps/push/src/host-challenge-answering.test-fixture.ts b/cloud/apps/push/src/host-challenge-answering.test-fixture.ts new file mode 100644 index 00000000000..4dec1e48c5b --- /dev/null +++ b/cloud/apps/push/src/host-challenge-answering.test-fixture.ts @@ -0,0 +1,163 @@ +import { createHmac, timingSafeEqual } from 'node:crypto' +import { + PUSH_HOST_CHALLENGE_PLAINTEXT_DOMAIN, + PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN, + PUSH_HOST_PROOF_TRANSCRIPT_FIELD_COUNT, + PUSH_LIMITS +} from '@orca-cloud/push-contract' +import nacl from 'tweetnacl' +import { decodeCanonicalBase64 } from './canonical-base64.js' +import { deriveHostFingerprint } from './host-fingerprint.js' + +// The desktop side of the push challenge, written the way the shipped host +// will answer it, so the gateway is exercised against a real box-opening peer. +const textEncoder = new TextEncoder() +const textDecoder = new TextDecoder() + +export type PushHostKeypair = { publicKey: Uint8Array; secretKey: Uint8Array } + +export type PushChallengeWire = { + challengeId: string + gatewayEphemeralPublicKeyB64: string + nonceB64: string + ciphertextB64: string + expiresAt: number +} + +export function createPushHostKeypair(seed?: number): PushHostKeypair { + const pair = + seed === undefined + ? nacl.box.keyPair() + : nacl.box.keyPair.fromSecretKey(new Uint8Array(32).fill(seed)) + return { publicKey: pair.publicKey, secretKey: pair.secretKey } +} + +export function hostPublicKeyB64(keypair: PushHostKeypair): string { + return Buffer.from(keypair.publicKey).toString('base64') +} + +function equal(left: Uint8Array | undefined, right: Uint8Array): boolean { + return Boolean(left && left.byteLength === right.byteLength && timingSafeEqual(left, right)) +} + +function uint64(value: number): Uint8Array { + const bytes = new Uint8Array(8) + new DataView(bytes.buffer).setBigUint64(0, BigInt(value), false) + return bytes +} + +function parseTranscript(transcript: Uint8Array): Map | null { + const fields = new Map() + const view = new DataView(transcript.buffer, transcript.byteOffset, transcript.byteLength) + let offset = 0 + try { + while (offset < transcript.byteLength) { + const nameLength = view.getUint32(offset, false) + offset += 4 + const name = textDecoder.decode(transcript.slice(offset, offset + nameLength)) + offset += nameLength + const valueLength = view.getUint32(offset, false) + offset += 4 + if (fields.has(name) || offset + valueLength > transcript.byteLength) return null + fields.set(name, transcript.slice(offset, offset + valueLength)) + offset += valueLength + } + } catch { + return null + } + return offset === transcript.byteLength ? fields : null +} + +function readUint64(value: Uint8Array | undefined): number | null { + if (!value || value.byteLength !== 8) return null + const parsed = new DataView(value.buffer, value.byteOffset, value.byteLength).getBigUint64(0, false) + return parsed <= BigInt(Number.MAX_SAFE_INTEGER) ? Number(parsed) : null +} + +export type PushHostProofContext = { + gatewayOrigin: string + keypair: PushHostKeypair + now?: () => number + onInvalid?: (reason: string) => void +} + +function validateTranscript( + transcript: Uint8Array, + challenge: PushChallengeWire, + context: PushHostProofContext, + gatewayKey: Uint8Array, + nonce: Uint8Array +): boolean { + const fields = parseTranscript(transcript) + if (!fields || fields.size !== PUSH_HOST_PROOF_TRANSCRIPT_FIELD_COUNT) { + context.onInvalid?.('transcript-structure') + return false + } + const now = (context.now ?? Date.now)() + const issuedAt = readUint64(fields.get('issuedAt')) + const expiresAt = readUint64(fields.get('expiresAt')) + const fingerprint = deriveHostFingerprint(context.keypair.publicKey) + const checks: [string, boolean][] = [ + ['issuedAt-readable', issuedAt !== null], + [ + 'issuedAt-not-future', + issuedAt === null || issuedAt - PUSH_LIMITS.clockSkewToleranceMs <= now + ], + ['not-expired', now - PUSH_LIMITS.clockSkewToleranceMs <= challenge.expiresAt], + ['issuedAt-before-expiry', issuedAt === null || issuedAt <= challenge.expiresAt], + [ + 'window', + issuedAt === null || challenge.expiresAt - issuedAt <= PUSH_LIMITS.challengeTtlMs + ], + ['expiry-consistent', expiresAt === challenge.expiresAt], + ['protocol', equal(fields.get('protocol'), textEncoder.encode(PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN))], + ['version', equal(fields.get('version'), new Uint8Array([1]))], + ['gatewayOrigin', equal(fields.get('gatewayOrigin'), textEncoder.encode(context.gatewayOrigin))], + ['gatewayEphemeralPublicKey', equal(fields.get('gatewayEphemeralPublicKey'), gatewayKey)], + ['challengeNonce', equal(fields.get('challengeNonce'), nonce)], + ['challengeId', equal(fields.get('challengeId'), textEncoder.encode(challenge.challengeId))], + ['hostFingerprint', equal(fields.get('hostFingerprint'), textEncoder.encode(fingerprint))], + ['hostPublicKey', equal(fields.get('hostPublicKey'), context.keypair.publicKey)], + ['issuedAt-value', issuedAt === null || uint64(issuedAt).byteLength === 8] + ] + const failed = checks.filter(([, ok]) => !ok).map(([name]) => name) + if (failed.length === 0) return true + context.onInvalid?.(`transcript:${failed.join('+')}`) + return false +} + +export function answerPushHostChallenge( + challenge: PushChallengeWire, + context: PushHostProofContext +): string | null { + const gatewayKey = decodeCanonicalBase64(challenge.gatewayEphemeralPublicKeyB64, 32) + const nonce = decodeCanonicalBase64(challenge.nonceB64, 24) + const ciphertext = Buffer.from(challenge.ciphertextB64, 'base64') + if (!gatewayKey || !nonce || ciphertext.toString('base64') !== challenge.ciphertextB64) return null + const plaintext = nacl.box.open(ciphertext, nonce, gatewayKey, context.keypair.secretKey) + if (!plaintext) { + context.onInvalid?.('challenge-box-open') + return null + } + const domain = textEncoder.encode(`${PUSH_HOST_CHALLENGE_PLAINTEXT_DOMAIN}\0`) + if ( + !equal(plaintext.slice(0, domain.byteLength), domain) || + plaintext.byteLength < domain.byteLength + 36 + ) { + return null + } + const transcriptLength = new DataView( + plaintext.buffer, + plaintext.byteOffset + domain.byteLength, + 4 + ).getUint32(0, false) + const transcriptStart = domain.byteLength + 4 + const secretStart = transcriptStart + transcriptLength + if (secretStart + 32 !== plaintext.byteLength) return null + const transcript = plaintext.slice(transcriptStart, secretStart) + if (!validateTranscript(transcript, challenge, context, gatewayKey, nonce)) return null + return createHmac('sha256', plaintext.slice(secretStart)) + .update(textEncoder.encode(`${PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN}\0ack\0`)) + .update(transcript) + .digest('base64') +} diff --git a/cloud/apps/push/src/host-challenge-store.test.ts b/cloud/apps/push/src/host-challenge-store.test.ts new file mode 100644 index 00000000000..e3dbcf8389f --- /dev/null +++ b/cloud/apps/push/src/host-challenge-store.test.ts @@ -0,0 +1,245 @@ +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import { afterEach, beforeEach, describe, expect, it } from 'vitest' +import { + answerPushHostChallenge, + createPushHostKeypair, + hostPublicKeyB64 +} from './host-challenge-answering.test-fixture.js' +import { PushHostChallengeStore } from './host-challenge-store.js' +import { deriveHostFingerprint } from './host-fingerprint.js' +import { openInMemoryPushDatabase, type PushDatabase } from './push-database.js' + +const GATEWAY_ORIGIN = 'https://push.onorca.dev' + +describe('push host challenge store', () => { + let database: PushDatabase + let clock = 1_700_000_000_000 + let store: PushHostChallengeStore + + beforeEach(async () => { + database = await openInMemoryPushDatabase() + clock = 1_700_000_000_000 + store = new PushHostChallengeStore(database, GATEWAY_ORIGIN, () => clock) + }) + + afterEach(async () => { + await database.close() + }) + + it('completes a challenge, proof, and consume round trip', async () => { + const host = createPushHostKeypair(1) + const challenge = await store.issue(hostPublicKeyB64(host)) + expect(challenge).not.toBeNull() + expect(challenge!.expiresAt).toBe(clock + PUSH_LIMITS.challengeTtlMs) + expect(challenge!.hostFingerprint).toBe(deriveHostFingerprint(host.publicKey)) + + const proof = answerPushHostChallenge(challenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: host, + now: () => clock + }) + expect(proof).not.toBeNull() + await expect(store.verify(challenge!.challengeId, proof!)).resolves.toEqual({ + ok: true, + hostFingerprint: deriveHostFingerprint(host.publicKey) + }) + const [hostRow] = await database.query('SELECT host_fingerprint, last_seen_at FROM push_hosts') + expect(hostRow?.host_fingerprint).toBe(deriveHostFingerprint(host.publicKey)) + }) + + it('never stores material that reproduces the proof', async () => { + const host = createPushHostKeypair(2) + const challenge = await store.issue(hostPublicKeyB64(host)) + const proof = answerPushHostChallenge(challenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: host, + now: () => clock + }) + const [row] = await database.query('SELECT secret_hash FROM push_challenges') + expect(String(row?.secret_hash)).not.toBe(proof) + expect(Buffer.from(String(row?.secret_hash), 'base64url').byteLength).toBe(32) + }) + + it('rejects a replayed challenge', async () => { + const host = createPushHostKeypair(3) + const challenge = await store.issue(hostPublicKeyB64(host)) + const proof = answerPushHostChallenge(challenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: host, + now: () => clock + })! + await expect(store.verify(challenge!.challengeId, proof)).resolves.toMatchObject({ ok: true }) + await expect(store.verify(challenge!.challengeId, proof)).resolves.toEqual({ + ok: false, + reason: 'already_consumed' + }) + }) + + it('rejects a challenge the moment its own ttl elapses', async () => { + const host = createPushHostKeypair(4) + const challenge = await store.issue(hostPublicKeyB64(host)) + const proof = answerPushHostChallenge(challenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: host, + now: () => clock + })! + clock += PUSH_LIMITS.challengeTtlMs + 1 + await expect(store.verify(challenge!.challengeId, proof)).resolves.toEqual({ + ok: false, + reason: 'expired' + }) + }) + + it('spends no skew tolerance on its own expiry, so the ttl is the whole window', async () => { + const host = createPushHostKeypair(5) + const challenge = await store.issue(hostPublicKeyB64(host)) + const proof = answerPushHostChallenge(challenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: host, + now: () => clock + })! + // A proof that the host would still consider in-window is refused here: the + // gateway issued expires_at against this clock and needs no allowance. + clock += PUSH_LIMITS.challengeTtlMs + PUSH_LIMITS.clockSkewToleranceMs - 1 + await expect(store.verify(challenge!.challengeId, proof)).resolves.toEqual({ + ok: false, + reason: 'expired' + }) + }) + + it('accepts a proof that lands just inside the ttl', async () => { + const host = createPushHostKeypair(26) + const challenge = await store.issue(hostPublicKeyB64(host)) + const proof = answerPushHostChallenge(challenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: host, + now: () => clock + })! + clock += PUSH_LIMITS.challengeTtlMs + await expect(store.verify(challenge!.challengeId, proof)).resolves.toMatchObject({ ok: true }) + }) + + it('keeps an expired row long enough to answer expired rather than unknown', async () => { + const host = createPushHostKeypair(27) + const challenge = await store.issue(hostPublicKeyB64(host)) + const proof = answerPushHostChallenge(challenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: host, + now: () => clock + })! + clock += PUSH_LIMITS.challengeTtlMs + 1 + expect(await store.pruneExpired()).toBe(0) + await expect(store.verify(challenge!.challengeId, proof)).resolves.toEqual({ + ok: false, + reason: 'expired' + }) + }) + + it('refuses a wrong host: the box will not open and a foreign proof will not match', async () => { + const owner = createPushHostKeypair(6) + const intruder = createPushHostKeypair(7) + const ownerChallenge = await store.issue(hostPublicKeyB64(owner)) + expect( + answerPushHostChallenge(ownerChallenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: intruder, + now: () => clock + }) + ).toBeNull() + + const intruderChallenge = await store.issue(hostPublicKeyB64(intruder)) + const intruderProof = answerPushHostChallenge(intruderChallenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: intruder, + now: () => clock + })! + await expect(store.verify(ownerChallenge!.challengeId, intruderProof)).resolves.toEqual({ + ok: false, + reason: 'proof_mismatch' + }) + }) + + it('rejects a proof bound to a different gateway origin', async () => { + const host = createPushHostKeypair(8) + const challenge = await store.issue(hostPublicKeyB64(host)) + const reasons: string[] = [] + expect( + answerPushHostChallenge(challenge!, { + gatewayOrigin: 'https://push.example.test', + keypair: host, + now: () => clock, + onInvalid: (reason) => reasons.push(reason) + }) + ).toBeNull() + expect(reasons.join()).toContain('gatewayOrigin') + }) + + it('rejects an unknown challenge id and a malformed public key', async () => { + await expect(store.verify('missing', Buffer.alloc(32, 9).toString('base64'))).resolves.toEqual({ + ok: false, + reason: 'unknown_challenge' + }) + await expect(store.issue('not-base64!!')).resolves.toBeNull() + await expect(store.issue(Buffer.alloc(31, 1).toString('base64'))).resolves.toBeNull() + }) + + it('creates no host row until a proof succeeds', async () => { + const host = createPushHostKeypair(30) + const challenge = await store.issue(hostPublicKeyB64(host)) + const [beforeProof] = await database.query('SELECT COUNT(*) AS hosts FROM push_hosts') + expect(Number(beforeProof?.hosts)).toBe(0) + + const proof = answerPushHostChallenge(challenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: host, + now: () => clock + })! + await expect(store.verify(challenge!.challengeId, proof)).resolves.toMatchObject({ ok: true }) + const [row] = await database.query('SELECT host_public_key, last_seen_at FROM push_hosts') + expect(row?.host_public_key).toBe(hostPublicKeyB64(host)) + expect(Number(row?.last_seen_at)).toBe(clock) + }) + + it('leaves no host row behind when a challenge is never answered', async () => { + for (let index = 0; index < 5; index++) { + await store.issue(hostPublicKeyB64(createPushHostKeypair(40 + index))) + } + const [row] = await database.query('SELECT COUNT(*) AS hosts FROM push_hosts') + expect(Number(row?.hosts)).toBe(0) + }) + + it('prunes a host past retention only when it has no registration left', async () => { + const stale = createPushHostKeypair(50) + const kept = createPushHostKeypair(51) + for (const host of [stale, kept]) { + const challenge = await store.issue(hostPublicKeyB64(host)) + const proof = answerPushHostChallenge(challenge!, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair: host, + now: () => clock + })! + await store.verify(challenge!.challengeId, proof) + } + await database.query( + `INSERT INTO push_devices (registration_id, host_fingerprint, device_id, platform, token, + filter_json, created_at, updated_at) + VALUES (?, ?, ?, ?, ?, ?, ?, ?)`, + ['reg-1', deriveHostFingerprint(kept.publicKey), 'device-1', 'android', 'token', '{}', clock, clock] + ) + + clock += PUSH_LIMITS.hostRetentionMs + expect(await store.pruneStaleHosts()).toBe(0) + clock += 1 + expect(await store.pruneStaleHosts()).toBe(1) + const [row] = await database.query('SELECT host_fingerprint FROM push_hosts') + expect(row?.host_fingerprint).toBe(deriveHostFingerprint(kept.publicKey)) + }) + + it('prunes challenges that fell out of the skew window', async () => { + const host = createPushHostKeypair(9) + await store.issue(hostPublicKeyB64(host)) + expect(await store.pruneExpired()).toBe(0) + clock += PUSH_LIMITS.challengeTtlMs + PUSH_LIMITS.clockSkewToleranceMs + 1 + expect(await store.pruneExpired()).toBe(1) + }) +}) diff --git a/cloud/apps/push/src/host-challenge-store.ts b/cloud/apps/push/src/host-challenge-store.ts new file mode 100644 index 00000000000..032e5509dbc --- /dev/null +++ b/cloud/apps/push/src/host-challenge-store.ts @@ -0,0 +1,175 @@ +import { createHash, createHmac, randomBytes, randomUUID, timingSafeEqual } from 'node:crypto' +import { + buildPushHostChallengePlaintext, + buildPushHostProofMacInput, + buildPushHostProofTranscript, + PUSH_LIMITS +} from '@orca-cloud/push-contract' +import nacl from 'tweetnacl' +import { decodeCanonicalBase64 } from './canonical-base64.js' +import { deriveHostFingerprint } from './host-fingerprint.js' +import type { PushDatabase } from './push-database.js' + +export type IssuedPushChallenge = { + challengeId: string + gatewayEphemeralPublicKeyB64: string + nonceB64: string + ciphertextB64: string + expiresAt: number + hostFingerprint: string +} + +export type PushProofVerification = + | { ok: true; hostFingerprint: string } + | { ok: false; reason: 'unknown_challenge' | 'already_consumed' | 'expired' | 'proof_mismatch' } + +function sha256(value: Uint8Array): string { + return createHash('sha256').update(value).digest('base64url') +} + +function equalDigest(left: string, right: string): boolean { + const leftBytes = Buffer.from(left) + const rightBytes = Buffer.from(right) + return leftBytes.length === rightBytes.length && timingSafeEqual(leftBytes, rightBytes) +} + +export class PushHostChallengeStore { + constructor( + private readonly database: PushDatabase, + private readonly gatewayOrigin: string, + private readonly now: () => number = Date.now + ) {} + + async issue(hostPublicKeyB64: string): Promise { + const hostPublicKey = decodeCanonicalBase64(hostPublicKeyB64, 32) + if (!hostPublicKey) return null + const hostFingerprint = deriveHostFingerprint(hostPublicKey) + const ephemeral = nacl.box.keyPair() + const challengeNonce = randomBytes(nacl.box.nonceLength) + const challengeSecret = randomBytes(32) + const challengeId = randomUUID() + const issuedAt = this.now() + const expiresAt = issuedAt + PUSH_LIMITS.challengeTtlMs + const transcript = buildPushHostProofTranscript({ + gatewayOrigin: this.gatewayOrigin, + gatewayEphemeralPublicKey: ephemeral.publicKey, + challengeNonce, + challengeId, + issuedAt, + expiresAt, + hostFingerprint, + hostPublicKey + }) + const ciphertext = nacl.box( + buildPushHostChallengePlaintext(transcript, challengeSecret), + challengeNonce, + hostPublicKey, + ephemeral.secretKey + ) + const expectedProof = createHmac('sha256', challengeSecret) + .update(buildPushHostProofMacInput(transcript)) + .digest() + // No push_hosts row yet: issuing is unauthenticated, so anyone could + // otherwise fill the table. The key rides the challenge until verify() proves it. + await this.database.query( + `INSERT INTO push_challenges + (challenge_id, host_fingerprint, host_public_key, secret_hash, transcript, expires_at, + consumed_at) + VALUES (?, ?, ?, ?, ?, ?, NULL)`, + [ + challengeId, + hostFingerprint, + hostPublicKeyB64, + // The stored digest is of the ack the secret produces, never of the + // secret itself: a database reader must not be able to forge a proof. + sha256(expectedProof), + Buffer.from(transcript).toString('base64'), + expiresAt + ] + ) + return { + challengeId, + gatewayEphemeralPublicKeyB64: Buffer.from(ephemeral.publicKey).toString('base64'), + nonceB64: Buffer.from(challengeNonce).toString('base64'), + ciphertextB64: Buffer.from(ciphertext).toString('base64'), + expiresAt, + hostFingerprint + } + } + + async verify(challengeId: string, proofB64: string): Promise { + const proof = decodeCanonicalBase64(proofB64, 32) + return await this.database.transaction(async (transaction) => { + const [row] = await transaction.query( + `SELECT host_fingerprint, host_public_key, secret_hash, expires_at, consumed_at + FROM push_challenges WHERE challenge_id = ?`, + [challengeId] + ) + if (!row) return { ok: false, reason: 'unknown_challenge' } + if (row.consumed_at !== null && row.consumed_at !== undefined) { + return { ok: false, reason: 'already_consumed' } + } + const now = this.now() + // No skew allowance here: the gateway set expires_at from this same clock. + // The tolerance belongs to the host, which validates a foreign timestamp. + if (now > Number(row.expires_at)) return { ok: false, reason: 'expired' } + if (!proof || !equalDigest(sha256(proof), String(row.secret_hash))) { + return { ok: false, reason: 'proof_mismatch' } + } + // Consume under the same predicate the read used, so two concurrent + // proofs for one challenge cannot both mint a session. + const [consumed] = await transaction.query( + 'UPDATE push_challenges SET consumed_at = ? WHERE challenge_id = ? AND consumed_at IS NULL', + [now, challengeId] + ) + if (Number(consumed?.changes ?? 0) !== 1) return { ok: false, reason: 'already_consumed' } + await this.rememberHost( + transaction, + String(row.host_fingerprint), + String(row.host_public_key), + now + ) + return { ok: true, hostFingerprint: String(row.host_fingerprint) } + }) + } + + // Rows outlive the expiry check by the skew tolerance so a late proof reads + // as 'expired' rather than as an unknown challenge. + async pruneExpired(): Promise { + const cutoff = this.now() - PUSH_LIMITS.clockSkewToleranceMs + const [result] = await this.database.query('DELETE FROM push_challenges WHERE expires_at < ?', [ + cutoff + ]) + return Number(result?.changes ?? 0) + } + + // A host that stopped proving and has no registration left is dead weight; + // its public key is recoverable from the desktop on the next challenge. + async pruneStaleHosts(): Promise { + const [result] = await this.database.query( + `DELETE FROM push_hosts + WHERE last_seen_at < ? + AND host_fingerprint NOT IN (SELECT host_fingerprint FROM push_devices)`, + [this.now() - PUSH_LIMITS.hostRetentionMs] + ) + return Number(result?.changes ?? 0) + } + + private async rememberHost( + transaction: PushDatabase, + hostFingerprint: string, + hostPublicKeyB64: string, + now: number + ): Promise { + const [updated] = await transaction.query( + 'UPDATE push_hosts SET last_seen_at = ?, host_public_key = ? WHERE host_fingerprint = ?', + [now, hostPublicKeyB64, hostFingerprint] + ) + if (Number(updated?.changes ?? 0) > 0) return + await transaction.query( + `INSERT INTO push_hosts (host_fingerprint, host_public_key, created_at, last_seen_at) + VALUES (?, ?, ?, ?)`, + [hostFingerprint, hostPublicKeyB64, now, now] + ) + } +} diff --git a/cloud/apps/push/src/host-fingerprint.ts b/cloud/apps/push/src/host-fingerprint.ts new file mode 100644 index 00000000000..955b1ac8ecb --- /dev/null +++ b/cloud/apps/push/src/host-fingerprint.ts @@ -0,0 +1,16 @@ +import { createHash } from 'node:crypto' +import { PUSH_HOST_FINGERPRINT_LENGTH } from '@orca-cloud/push-contract' + +// Identical derivation to deriveRelayHostId on the desktop, so a host and a +// phone reach the same fingerprint from the same X25519 public key. +export function deriveHostFingerprint(hostPublicKey: Uint8Array): string { + return createHash('sha256') + .update(hostPublicKey) + .digest('base64url') + .slice(0, PUSH_HOST_FINGERPRINT_LENGTH) +} + +// Logs may carry at most this much of a fingerprint. +export function fingerprintLogPrefix(hostFingerprint: string): string { + return hostFingerprint.slice(0, 4) +} diff --git a/cloud/apps/push/src/host-session-store.test.ts b/cloud/apps/push/src/host-session-store.test.ts new file mode 100644 index 00000000000..129dba2134c --- /dev/null +++ b/cloud/apps/push/src/host-session-store.test.ts @@ -0,0 +1,70 @@ +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import { afterEach, beforeEach, describe, expect, it } from 'vitest' +import { PushHostSessionStore } from './host-session-store.js' +import { openInMemoryPushDatabase, type PushDatabase } from './push-database.js' + +const HOST = 'abcdefghijklmnop' + +describe('push host session store', () => { + let database: PushDatabase + let clock = 1_700_000_000_000 + let sessions: PushHostSessionStore + + beforeEach(async () => { + database = await openInMemoryPushDatabase() + clock = 1_700_000_000_000 + sessions = new PushHostSessionStore(database, () => clock) + }) + + afterEach(async () => { + await database.close() + }) + + it('mints a 24 hour session and stores only its hash', async () => { + const session = await sessions.create(HOST) + expect(session.expiresAt).toBe(clock + PUSH_LIMITS.sessionTtlMs) + expect(Buffer.from(session.sessionToken, 'base64url').byteLength).toBe(32) + const [row] = await database.query('SELECT token_hash FROM push_sessions') + expect(String(row?.token_hash)).not.toBe(session.sessionToken) + await expect(sessions.resolve(session.sessionToken)).resolves.toMatchObject({ + ok: true, + hostFingerprint: HOST + }) + }) + + it('reports expiry separately from an unknown token', async () => { + const session = await sessions.create(HOST) + clock += PUSH_LIMITS.sessionTtlMs + 1 + await expect(sessions.resolve(session.sessionToken)).resolves.toEqual({ + ok: false, + reason: 'session_expired' + }) + await expect(sessions.resolve('not-a-session')).resolves.toEqual({ + ok: false, + reason: 'unknown_session' + }) + }) + + it('accepts a session on its final millisecond', async () => { + const session = await sessions.create(HOST) + clock += PUSH_LIMITS.sessionTtlMs + await expect(sessions.resolve(session.sessionToken)).resolves.toMatchObject({ ok: true }) + }) + + it('keeps one live session per host and prunes it once expired', async () => { + const first = await sessions.create(HOST) + const second = await sessions.create(HOST) + // The earlier session is gone the moment its host proves again, so a flood + // of proofs leaves one row per host rather than one per proof. + await expect(sessions.resolve(first.sessionToken)).resolves.toEqual({ + ok: false, + reason: 'unknown_session' + }) + await expect(sessions.resolve(second.sessionToken)).resolves.toMatchObject({ ok: true }) + const other = await sessions.create('ponmlkjihgfedcba') + await expect(sessions.resolve(second.sessionToken)).resolves.toMatchObject({ ok: true }) + clock += PUSH_LIMITS.sessionTtlMs + 1 + expect(await sessions.pruneExpired()).toBe(2) + await expect(sessions.resolve(other.sessionToken)).resolves.toMatchObject({ ok: false }) + }) +}) diff --git a/cloud/apps/push/src/host-session-store.ts b/cloud/apps/push/src/host-session-store.ts new file mode 100644 index 00000000000..899bacabcc8 --- /dev/null +++ b/cloud/apps/push/src/host-session-store.ts @@ -0,0 +1,65 @@ +import { createHash, randomBytes } from 'node:crypto' +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import type { PushDatabase } from './push-database.js' + +export type IssuedPushSession = { + sessionToken: string + expiresAt: number + hostFingerprint: string +} + +export type PushSessionLookup = + | { ok: true; hostFingerprint: string; expiresAt: number } + | { ok: false; reason: 'unknown_session' | 'session_expired' } + +function hashSessionToken(sessionToken: string): string { + return createHash('sha256').update(sessionToken).digest('base64url') +} + +export class PushHostSessionStore { + constructor( + private readonly database: PushDatabase, + private readonly now: () => number = Date.now + ) {} + + async create(hostFingerprint: string): Promise { + const sessionToken = randomBytes(32).toString('base64url') + const createdAt = this.now() + const expiresAt = createdAt + PUSH_LIMITS.sessionTtlMs + await this.database.transaction(async (transaction) => { + // Why: a desktop holds one session at a time and only re-proves once it is + // gone, so an earlier row is dead weight. It also bounds the table to one + // row per host however many proofs a self-minted identity answers. + await transaction.lockQuotaScope(`orca-push-session:${hostFingerprint}`) + await transaction.query('DELETE FROM push_sessions WHERE host_fingerprint = ?', [ + hostFingerprint + ]) + await transaction.query( + `INSERT INTO push_sessions (token_hash, host_fingerprint, expires_at, created_at) + VALUES (?, ?, ?, ?)`, + [hashSessionToken(sessionToken), hostFingerprint, expiresAt, createdAt] + ) + }) + return { sessionToken, expiresAt, hostFingerprint } + } + + async resolve(sessionToken: string): Promise { + const [row] = await this.database.query( + 'SELECT host_fingerprint, expires_at FROM push_sessions WHERE token_hash = ?', + [hashSessionToken(sessionToken)] + ) + if (!row) return { ok: false, reason: 'unknown_session' } + const expiresAt = Number(row.expires_at) + // No skew grace here: a 24h session that just expired should be re-minted + // through the challenge, which is cheap and already handled by the host. + if (this.now() > expiresAt) return { ok: false, reason: 'session_expired' } + return { ok: true, hostFingerprint: String(row.host_fingerprint), expiresAt } + } + + async pruneExpired(): Promise { + const [result] = await this.database.query('DELETE FROM push_sessions WHERE expires_at < ?', [ + this.now() + ]) + return Number(result?.changes ?? 0) + } +} diff --git a/cloud/apps/push/src/index.ts b/cloud/apps/push/src/index.ts new file mode 100644 index 00000000000..c3415dc307a --- /dev/null +++ b/cloud/apps/push/src/index.ts @@ -0,0 +1,81 @@ +import { loadPushConfig } from './config.js' +import { openPushDatabase } from './push-database.js' +import { createPushServer } from './push-server.js' + +const CHALLENGE_PRUNE_INTERVAL_MS = 60_000 +const SESSION_PRUNE_INTERVAL_MS = 10 * 60_000 +const SEND_LOG_PRUNE_INTERVAL_MS = 30 * 60_000 +const STALE_HOST_PRUNE_INTERVAL_MS = 30 * 60_000 + +const config = loadPushConfig() +const database = await openPushDatabase({ + ...(config.databaseUrl === undefined ? {} : { databaseUrl: config.databaseUrl }), + dataDir: config.dataDir, + poolMax: config.databasePoolMax, + applicationName: 'orca-push' +}) +const { + server, + challenges, + sessions, + quota, + coalescer, + observability, + closeTransports, + requestDrain +} = createPushServer(config, database) + +function prune(label: string, run: () => Promise, intervalMs: number): NodeJS.Timeout { + const timer = setInterval(() => { + void run().catch((error: unknown) => { + console.warn( + JSON.stringify({ + event: 'orca_push_prune_failed', + target: label, + error: error instanceof Error ? error.name : 'unknown' + }) + ) + }) + }, intervalMs) + timer.unref() + return timer +} + +const timers = [ + prune('challenges', () => challenges.pruneExpired(), CHALLENGE_PRUNE_INTERVAL_MS), + prune('sessions', () => sessions.pruneExpired(), SESSION_PRUNE_INTERVAL_MS), + prune('send_log', () => quota.prune(), SEND_LOG_PRUNE_INTERVAL_MS), + prune('stale_hosts', () => challenges.pruneStaleHosts(), STALE_HOST_PRUNE_INTERVAL_MS) +] +observability.start() + +server.listen(config.port, () => { + console.log(`[orca-push] listening on ${config.publicUrl} (port ${config.port})`) +}) + +let stopping = false +const shutdown = (): void => { + if (stopping) return + stopping = true + for (const timer of timers) clearInterval(timer) + // Cloud Run sends SIGKILL after ten seconds; leave time for explicit cleanup. + const deadline = setTimeout(() => process.exit(1), 9_000) + deadline.unref() + const requests = requestDrain.begin() + const connections = new Promise((resolve) => server.close(() => resolve())) + void Promise.all([requests, connections]) + .then(async () => { + await coalescer.flushAll() + coalescer.stop() + closeTransports() + await database.close() + observability.stop() + clearTimeout(deadline) + }) + .catch(() => { + console.warn(JSON.stringify({ event: 'orca_push_shutdown_failed' })) + process.exitCode = 1 + }) +} +process.once('SIGTERM', shutdown) +process.once('SIGINT', shutdown) diff --git a/cloud/apps/push/src/provider-retry-delay.ts b/cloud/apps/push/src/provider-retry-delay.ts new file mode 100644 index 00000000000..4c77b3c6dc7 --- /dev/null +++ b/cloud/apps/push/src/provider-retry-delay.ts @@ -0,0 +1,9 @@ +export function providerRetryAfter( + value: string | undefined, + now = Date.now() +): number | undefined { + if (!value) return undefined + const seconds = Number(value) + const delay = Number.isFinite(seconds) ? seconds * 1000 : Date.parse(value) - now + return Number.isFinite(delay) ? Math.max(0, delay) : undefined +} diff --git a/cloud/apps/push/src/push-database-postgres-startup.test.ts b/cloud/apps/push/src/push-database-postgres-startup.test.ts new file mode 100644 index 00000000000..181016d062a --- /dev/null +++ b/cloud/apps/push/src/push-database-postgres-startup.test.ts @@ -0,0 +1,89 @@ +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' + +const fakes = vi.hoisted(() => ({ + configs: [] as Array>, + lifecycle: [] as string[], + query: vi.fn(async (_sql: string) => ({ rows: [], rowCount: 0 })), + release: vi.fn() +})) + +vi.mock('pg', () => ({ + default: { + Pool: class { + on = vi.fn() + connect = vi.fn(async () => ({ query: fakes.query, release: fakes.release })) + private readonly label: string + + constructor(config: Record) { + fakes.configs.push(config) + this.label = `max=${String(config.max)} statement_timeout=${String(config.statement_timeout)}` + fakes.lifecycle.push(`open ${this.label}`) + } + + async end(): Promise { + fakes.lifecycle.push(`end ${this.label}`) + } + } + } +})) + +import { openPushDatabase } from './push-database.js' +import { pushSchemaStatements } from './push-schema.js' + +describe('PostgreSQL push gateway startup', () => { + beforeEach(() => { + fakes.configs.length = 0 + fakes.lifecycle.length = 0 + fakes.query.mockClear() + }) + + afterEach(() => { + vi.restoreAllMocks() + }) + + // Why: a CREATE INDEX on a grown table can outlive the 5s request deadline, + // and a schema that inherits it fails every startup at the same statement. + it('applies the schema on an untimed pool that is gone before the serving pool opens', async () => { + const database = await openPushDatabase({ + databaseUrl: 'postgresql://push@localhost:55440/orca_push', + dataDir: '/unused', + poolMax: 2, + applicationName: 'orca-push' + }) + expect(fakes.lifecycle).toEqual([ + 'open max=1 statement_timeout=0', + 'end max=1 statement_timeout=0', + 'open max=2 statement_timeout=5000' + ]) + expect(fakes.configs[0]).toMatchObject({ + application_name: 'orca-push/schema', + lock_timeout: 1_000, + idle_in_transaction_session_timeout: 5_000 + }) + expect( + fakes.query.mock.calls.map(([sql]) => sql).slice(0, pushSchemaStatements().length) + ).toEqual(pushSchemaStatements()) + await database.close() + }) + + it('retries a transaction the pool statement_timeout aborted', async () => { + const database = await openPushDatabase({ + databaseUrl: 'postgresql://push@localhost:55440/orca_push', + dataDir: '/unused' + }) + const warn = vi.spyOn(console, 'warn').mockImplementation(() => undefined) + let attempts = 0 + const result = await database.transaction(async () => { + attempts += 1 + if (attempts === 1) throw Object.assign(new Error('canceling statement'), { code: '57014' }) + return 'done' + }) + expect(result).toBe('done') + expect(attempts).toBe(2) + expect(warn.mock.calls.map(([line]) => String(line))).toEqual([ + expect.stringContaining('"code":"57014"') + ]) + warn.mockRestore() + await database.close() + }) +}) diff --git a/cloud/apps/push/src/push-database.ts b/cloud/apps/push/src/push-database.ts new file mode 100644 index 00000000000..6f8ba88ed1d --- /dev/null +++ b/cloud/apps/push/src/push-database.ts @@ -0,0 +1,275 @@ +import { mkdirSync } from 'node:fs' +import { join } from 'node:path' +import { DatabaseSync } from 'node:sqlite' +import pg from 'pg' +import { applyPostgresSchema } from '@orca-cloud/postgres-schema' +import { ensurePushSessionIndex } from './push-session-schema.js' +import { pushSchemaStatements } from './push-schema.js' + +const POSTGRES_LOCK_TIMEOUT_MS = 1_000 +const POSTGRES_CONNECTION_TIMEOUT_MS = 2_000 +const POSTGRES_STATEMENT_TIMEOUT_MS = 5_000 +const POSTGRES_IDLE_TRANSACTION_TIMEOUT_MS = 5_000 +const POSTGRES_TRANSACTION_ATTEMPTS = 3 +const POSTGRES_RETRY_MAX_DELAY_MS = 25 + +export type SqlRow = Record + +export interface PushDatabase { + readonly dialect: 'sqlite' | 'postgres' + query(sql: string, params?: unknown[]): Promise + transaction(operation: (transaction: PushDatabase) => Promise): Promise + // Serializes every transaction that reads then writes the same identity's + // quota rows. Must be called inside a transaction; it releases at commit. + lockQuotaScope(key: string): Promise + close(): Promise +} + +function postgresSql(sql: string): string { + let index = 0 + return sql.replace(/\?/g, () => `$${++index}`) +} + +function returnsRows(sql: string): boolean { + return /^\s*(select|with)/i.test(sql) || /returning/i.test(sql) +} + +class SqliteTransaction implements PushDatabase { + readonly dialect = 'sqlite' as const + + constructor(protected readonly database: DatabaseSync) {} + + async query(sql: string, params: unknown[] = []): Promise { + const statement = this.database.prepare(sql) + const bound = params.map((value) => (value === undefined ? null : value)) as never[] + if (returnsRows(sql)) return statement.all(...bound) as SqlRow[] + const result = statement.run(...bound) + return [{ changes: Number(result.changes) }] + } + + async transaction(operation: (transaction: PushDatabase) => Promise): Promise { + return await operation(this) + } + + // BEGIN IMMEDIATE already holds the single writer lock for the whole + // transaction, so there is nothing narrower left to take. + async lockQuotaScope(): Promise {} + + async close(): Promise {} +} + +class SqliteDatabase extends SqliteTransaction { + // node:sqlite is synchronous and has no nested transactions, so overlapping + // callers are serialized behind one tail promise instead of racing BEGIN. + private tail: Promise = Promise.resolve() + + override async query(sql: string, params: unknown[] = []): Promise { + await this.tail + return await super.query(sql, params) + } + + override async transaction(operation: (transaction: PushDatabase) => Promise): Promise { + const previous = this.tail + let release!: () => void + this.tail = new Promise((resolve) => (release = resolve)) + await previous + this.database.exec('BEGIN IMMEDIATE') + const transaction = new SqliteTransaction(this.database) + try { + const result = await operation(transaction) + this.database.exec('COMMIT') + return result + } catch (error) { + this.database.exec('ROLLBACK') + throw error + } finally { + release() + } + } + + override async close(): Promise { + await this.tail + this.database.close() + } +} + +class PostgresTransaction implements PushDatabase { + readonly dialect = 'postgres' as const + + constructor(private readonly client: pg.PoolClient) {} + + async query(sql: string, params: unknown[] = []): Promise { + const result = await this.client.query(postgresSql(sql), params) + return returnsRows(sql) ? (result.rows as SqlRow[]) : [{ changes: result.rowCount ?? 0 }] + } + + async transaction(operation: (transaction: PushDatabase) => Promise): Promise { + return await operation(this) + } + + // READ COMMITTED lets a concurrent count-then-insert read the same + // under-quota total, so the identity is serialized for the whole transaction. + async lockQuotaScope(key: string): Promise { + await this.query('SELECT pg_advisory_xact_lock(hashtext(?::text))', [key]) + } + + async close(): Promise {} +} + +function retryablePostgresTransactionError(error: unknown): boolean { + const code = String((error as { code?: unknown }).code) + // 57014 is the pool statement_timeout firing. It aborts the transaction the + // same way a lock timeout does, so it takes the bounded retry path too. + return code === '40P01' || code === '40001' || code === '55P03' || code === '57014' +} + +async function waitForPostgresRetry(): Promise { + const delayMs = Math.floor(Math.random() * (POSTGRES_RETRY_MAX_DELAY_MS + 1)) + await new Promise((resolve) => setTimeout(resolve, delayMs)) +} + +class PostgresDatabase implements PushDatabase { + readonly dialect = 'postgres' as const + + constructor(private readonly pool: pg.Pool) {} + + async query(sql: string, params: unknown[] = []): Promise { + const client = await this.pool.connect() + try { + const result = await client.query(postgresSql(sql), params) + return returnsRows(sql) ? (result.rows as SqlRow[]) : [{ changes: result.rowCount ?? 0 }] + } finally { + client.release() + } + } + + async transaction(operation: (transaction: PushDatabase) => Promise): Promise { + for (let attempt = 1; attempt <= POSTGRES_TRANSACTION_ATTEMPTS; attempt++) { + const client = await this.pool.connect() + try { + await client.query('BEGIN') + const result = await operation(new PostgresTransaction(client)) + await client.query('COMMIT') + return result + } catch (error) { + await client.query('ROLLBACK').catch(() => undefined) + if ( + !retryablePostgresTransactionError(error) || + attempt === POSTGRES_TRANSACTION_ATTEMPTS + ) { + throw error + } + console.warn( + JSON.stringify({ + event: 'orca_push_postgres_transaction_retry', + code: String((error as { code?: unknown }).code), + attempt + }) + ) + } finally { + client.release() + } + // A PostgreSQL transaction is unusable after an abort, so retry all work + // on a fresh pooled client with a small full-jitter delay. + await waitForPostgresRetry() + } + throw new Error('postgres_transaction_retry_exhausted') + } + + // An advisory transaction lock taken outside a transaction is released by the + // implicit commit before the caller reads anything, which protects nothing. + async lockQuotaScope(): Promise { + throw new Error('lock_quota_scope_requires_transaction') + } + + async close(): Promise { + await this.pool.end() + } +} + +async function applySchema(database: PushDatabase): Promise { + for (const statement of pushSchemaStatements()) await database.query(statement) + await ensurePushSessionIndex(database) +} + +// Why: DDL is not a request. A CREATE INDEX on a grown table can legitimately +// outlive the request statement_timeout, and inheriting it would fail every +// startup at the same statement instead of finishing once. One connection of +// its own, closed before the serving pool opens, keeps the untimed session off +// the request path entirely. +async function applySchemaOnUntimedPool( + databaseUrl: string, + applicationName: string | undefined +): Promise { + const pool = new pg.Pool({ + connectionString: databaseUrl, + max: 1, + application_name: applicationName ? `${applicationName}/schema` : undefined, + connectionTimeoutMillis: POSTGRES_CONNECTION_TIMEOUT_MS, + statement_timeout: 0, + lock_timeout: POSTGRES_LOCK_TIMEOUT_MS, + idle_in_transaction_session_timeout: POSTGRES_IDLE_TRANSACTION_TIMEOUT_MS + }) + absorbPostgresIdleClientErrors(pool) + const database = new PostgresDatabase(pool) + try { + await applyPostgresSchema(pushSchemaStatements(), (statement) => database.query(statement), { + eventPrefix: 'orca_push_postgres_schema' + }) + await ensurePushSessionIndex(database) + } finally { + await database.close().catch(() => undefined) + } +} + +export function absorbPostgresIdleClientErrors(pool: Pick): void { + pool.on('error', () => { + // node-postgres removes failed idle clients itself; an unhandled 'error' + // would crash the service and turn a SQL blip into a restart loop. + console.warn('[orca-push] idle PostgreSQL client failed') + }) +} + +export async function openPushDatabase(input: { + databaseUrl?: string + dataDir: string + poolMax?: number + applicationName?: string +}): Promise { + let database: PushDatabase + if (input.databaseUrl) { + await applySchemaOnUntimedPool(input.databaseUrl, input.applicationName) + const pool = new pg.Pool({ + connectionString: input.databaseUrl, + max: input.poolMax ?? 10, + application_name: input.applicationName, + connectionTimeoutMillis: POSTGRES_CONNECTION_TIMEOUT_MS, + statement_timeout: POSTGRES_STATEMENT_TIMEOUT_MS, + lock_timeout: POSTGRES_LOCK_TIMEOUT_MS, + idle_in_transaction_session_timeout: POSTGRES_IDLE_TRANSACTION_TIMEOUT_MS + }) + absorbPostgresIdleClientErrors(pool) + database = new PostgresDatabase(pool) + } else { + mkdirSync(input.dataDir, { recursive: true }) + const sqlite = new DatabaseSync(join(input.dataDir, 'orca-push.sqlite')) + sqlite.exec('PRAGMA journal_mode = WAL; PRAGMA foreign_keys = ON;') + database = new SqliteDatabase(sqlite) + } + if (database.dialect === 'postgres') return database + try { + await applySchema(database) + return database + } catch (error) { + await database.close().catch(() => undefined) + throw error + } +} + +export async function openInMemoryPushDatabase(): Promise { + const sqlite = new DatabaseSync(':memory:') + sqlite.exec('PRAGMA foreign_keys = ON;') + const database = new SqliteDatabase(sqlite) + await applySchema(database) + return database +} diff --git a/cloud/apps/push/src/push-delivery-lifecycle.test.ts b/cloud/apps/push/src/push-delivery-lifecycle.test.ts new file mode 100644 index 00000000000..88d95081515 --- /dev/null +++ b/cloud/apps/push/src/push-delivery-lifecycle.test.ts @@ -0,0 +1,155 @@ +import { afterEach, expect, it, vi } from 'vitest' +import { Hono } from 'hono' +import { PushRequestDrain } from './push-request-drain.js' +import { PushCoalescer } from './coalescer.js' +import { PushDispatcher } from './push-dispatcher.js' +import { PushDeviceRegistryStore } from './device-registry-store.js' +import { openInMemoryPushDatabase, type PushDatabase } from './push-database.js' +import { buildPushDelivery } from './push-delivery-message.js' +import { PushNotificationSchema } from '@orca-cloud/push-contract' +import { notification } from './push-server-harness.test-fixture.js' + +const databases: PushDatabase[] = [] +afterEach(async () => { + await Promise.all(databases.splice(0).map((db) => db.close())) + vi.restoreAllMocks() +}) +const note = PushNotificationSchema.parse(notification()) +const tick = () => new Promise((resolve) => setImmediate(resolve)) +function deferred() { + let resolve!: () => void + const promise = new Promise((done) => { + resolve = done + }) + return { promise, resolve } +} +async function registered() { + const db = await openInMemoryPushDatabase() + databases.push(db) + const devices = new PushDeviceRegistryStore(db) + const input = { + hostFingerprint: 'abcdefghijklmnop', + deviceId: 'device', + platform: 'android' as const, + token: 'old-token', + filter: { sources: [], agentStates: [] } + } + const row = await devices.upsert(input) + if (!row.ok) throw new Error('registration failed') + const delivery = buildPushDelivery({ + registrationId: row.registrationId, + hostFingerprint: input.hostFingerprint, + notification: note, + title: note.title, + body: note.body, + coalescedCount: 1 + }) + return { db, devices, input, delivery } +} + +it('does not retire a refreshed token after the old token fails', async () => { + const h = await registered() + const gate = deferred() + const send = vi.fn(async () => { + await gate.promise + return { status: 'dead', reason: 'UNREGISTERED' } + }) + vi.spyOn(console, 'warn').mockImplementation(() => {}) + const dispatcher = new PushDispatcher({ devices: h.devices, fcm: { send } as never }) + const pending = dispatcher.deliver(h.delivery) + await tick() + await h.devices.upsert({ ...h.input, token: 'replacement-token' }) + gate.resolve() + await pending + expect(await h.devices.findById(h.delivery.registrationId)).toMatchObject({ + token: 'replacement-token', + dead: false + }) +}) + +it('drains timer-triggered deliveries that already left the window map', async () => { + const gate = deferred() + const deliver = vi.fn(() => gate.promise) + const coalescer = new PushCoalescer({ + deliver, + setTimer: () => ({ handle: null }), + clearTimer: () => {} + }) + coalescer.enqueue({ + registrationId: 'reg', + hostFingerprint: 'abcdefghijklmnop', + notification: note + }) + const pending = coalescer.flush('reg') + let drained = false + const drain = coalescer.flushAll().then(() => { + drained = true + }) + await tick() + expect(deliver).toHaveBeenCalledOnce() + expect(drained).toBe(false) + gate.resolve() + await Promise.all([pending, drain]) + expect(drained).toBe(true) +}) + +it('rejects new requests during drain and waits for an admitted handler', async () => { + const gate = deferred() + const requests = new PushRequestDrain() + const app = new Hono().use('*', requests.middleware).post('/send', async (c) => { + await gate.promise + return c.json({ queued: true }) + }) + const pending = app.request('/send', { method: 'POST' }) + await tick() + let drained = false + const drain = requests.begin().then(() => { + drained = true + }) + expect((await app.request('/send', { method: 'POST' })).status).toBe(503) + expect(drained).toBe(false) + gate.resolve() + expect((await pending).status).toBe(200) + await drain + expect(drained).toBe(true) +}) + +it('retries transient failures with the provider delay and stops after success', async () => { + const h = await registered() + vi.spyOn(console, 'warn').mockImplementation(() => {}) + const send = vi + .fn() + .mockResolvedValueOnce({ + status: 'error', + reason: 'UNAVAILABLE', + retryable: true, + retryAfterMs: 10000 + }) + .mockResolvedValue({ status: 'sent' }) + const wait = vi.fn(async (_ms: number) => {}) + await new PushDispatcher({ devices: h.devices, fcm: { send } as never, wait }).deliver(h.delivery) + expect(send).toHaveBeenCalledTimes(2) + expect(wait).toHaveBeenCalledExactlyOnceWith(expect.any(Number)) + expect(wait.mock.calls[0]![0]).toBeGreaterThanOrEqual(10000) +}) + +it('bounds retries and rechecks registration after waiting', async () => { + const h = await registered() + vi.spyOn(console, 'warn').mockImplementation(() => {}) + const send = vi.fn().mockResolvedValue({ status: 'error', reason: 'timeout', retryable: true }) + await new PushDispatcher({ + devices: h.devices, + fcm: { send } as never, + wait: async () => {} + }).deliver(h.delivery) + expect(send).toHaveBeenCalledTimes(3) + send.mockClear() + await new PushDispatcher({ + devices: h.devices, + fcm: { send } as never, + wait: async () => { + await h.devices.deleteOwned(h.input.hostFingerprint, h.delivery.registrationId) + } + }).deliver(h.delivery) + expect(send).toHaveBeenCalledOnce() +}) diff --git a/cloud/apps/push/src/push-delivery-message.ts b/cloud/apps/push/src/push-delivery-message.ts new file mode 100644 index 00000000000..04c0e549286 --- /dev/null +++ b/cloud/apps/push/src/push-delivery-message.ts @@ -0,0 +1,87 @@ +import { PUSH_LIMITS, type PushNotification } from '@orca-cloud/push-contract' + +export type PushOrcaData = { + hostFingerprint: string + worktreeId?: string + notificationId?: string + notificationSeq: number + notificationEpoch: string + source: string + agentState: string | null + coalescedCount: number +} + +export type PushDelivery = { + sound?: boolean + registrationId: string + hostFingerprint: string + title: string + body: string + collapseId: string + orca: PushOrcaData +} + +export function hostCollapseId(hostFingerprint: string): string { + return `host:${hostFingerprint}` +} + +// APNs rejects a collapse id over 64 bytes, and notification ids are opaque +// desktop strings that may be longer or carry multi-byte characters. +export function truncateUtf8(value: string, maxBytes: number): string { + const encoded = Buffer.from(value, 'utf8') + if (encoded.byteLength <= maxBytes) return value + let end = maxBytes + // Walk back off a continuation byte so the cut never splits a code point. + while (end > 0 && (encoded[end]! & 0b1100_0000) === 0b1000_0000) end -= 1 + return encoded.subarray(0, end).toString('utf8') +} + +export function collapseIdFor( + notification: PushNotification, + hostFingerprint: string, + coalescedCount: number +): string { + if (coalescedCount > 1 || notification.notificationId === undefined) { + return hostCollapseId(hostFingerprint) + } + return truncateUtf8(notification.notificationId, PUSH_LIMITS.apnsCollapseIdMaxBytes) +} + +export function buildPushDelivery(input: { + registrationId: string + hostFingerprint: string + notification: PushNotification + title: string + body: string + coalescedCount: number +}): PushDelivery { + const { notification, hostFingerprint, coalescedCount } = input + return { + ...(notification.sound === false ? { sound: false } : {}), + registrationId: input.registrationId, + hostFingerprint, + title: input.title, + body: input.body, + collapseId: collapseIdFor(notification, hostFingerprint, coalescedCount), + orca: { + hostFingerprint, + ...(notification.worktreeId === undefined ? {} : { worktreeId: notification.worktreeId }), + ...(notification.notificationId === undefined + ? {} + : { notificationId: notification.notificationId }), + notificationSeq: notification.notificationSeq, + notificationEpoch: notification.notificationEpoch, + source: notification.source, + agentState: notification.agentState, + coalescedCount + } + } +} + +export function orcaDataStrings(orca: PushOrcaData): Record { + return Object.fromEntries( + Object.entries(orca) + .filter(([, value]) => value !== undefined && value !== null) + .map(([key, value]) => [key, String(value)]) + ) +} diff --git a/cloud/apps/push/src/push-dispatcher.ts b/cloud/apps/push/src/push-dispatcher.ts new file mode 100644 index 00000000000..39d17c92d11 --- /dev/null +++ b/cloud/apps/push/src/push-dispatcher.ts @@ -0,0 +1,74 @@ +import type { ApnsClient } from './apns-client.js' +import type { PushDeviceRegistryStore } from './device-registry-store.js' +import type { FcmClient } from './fcm-client.js' +import { fingerprintLogPrefix } from './host-fingerprint.js' +import type { PushDelivery } from './push-delivery-message.js' +import type { PushProviderOutcome } from './push-provider-outcome.js' + +export type PushDispatcherOptions = { + devices: PushDeviceRegistryStore + apns?: ApnsClient + fcm?: FcmClient + wait?: (ms: number) => Promise + now?: () => number + onRetry?: () => void + onOutcome?: (outcome: PushProviderOutcome['status']) => void +} + +// Sends one coalesced delivery through the provider the registration belongs +// to, and retires the registration when the provider says the token is gone. +export class PushDispatcher { + constructor(private readonly options: PushDispatcherOptions) {} + + async deliver(delivery: PushDelivery): Promise { + const now = this.options.now ?? Date.now + const deadline = now() + 120_000 + for (let attempt = 0; attempt < 3; attempt++) { + if (now() >= deadline) return + const retry = await this.deliverAttempt(delivery) + if (!retry || attempt === 2) return + const delay = Math.max(retry.delayMs, 1000 * 2 ** attempt) + Math.floor(Math.random() * 250) + if (now() + delay >= deadline) return + this.options.onRetry?.() + await (this.options.wait ?? ((ms) => new Promise((resolve) => setTimeout(resolve, ms))))( + delay + ) + } + } + + private async deliverAttempt(delivery: PushDelivery): Promise<{ delayMs: number } | undefined> { + const device = await this.options.devices.findById(delivery.registrationId) + if (!device || device.dead) return + let outcome: PushProviderOutcome + if (device.platform === 'ios') { + outcome = this.options.apns + ? await this.options.apns.send(delivery, { + token: device.token, + apnsEnvironment: device.apnsEnvironment ?? 'production' + }) + : { status: 'error', reason: 'apns_not_configured' } + } else { + outcome = this.options.fcm + ? await this.options.fcm.send(delivery, { token: device.token }) + : { status: 'error', reason: 'fcm_not_configured' } + } + this.options.onOutcome?.(outcome.status) + if (outcome.status === 'dead') { + await this.options.devices.markDead(delivery.registrationId, device) + } + if (outcome.status !== 'sent') { + console.warn( + JSON.stringify({ + event: 'orca_push_delivery_failed', + platform: device.platform, + status: outcome.status, + reason: outcome.reason, + host: fingerprintLogPrefix(delivery.hostFingerprint) + }) + ) + } + if (outcome.status === 'error' && outcome.retryable) + return { delayMs: outcome.retryAfterMs ?? 0 } + return undefined + } +} diff --git a/cloud/apps/push/src/push-notification-sound.test.ts b/cloud/apps/push/src/push-notification-sound.test.ts new file mode 100644 index 00000000000..17e30661fb0 --- /dev/null +++ b/cloud/apps/push/src/push-notification-sound.test.ts @@ -0,0 +1,31 @@ +import { expect, it } from 'vitest' +import { apnsBody } from './apns-client.js' +import { fcmMessageBody } from './fcm-client.js' +import { buildPushDelivery } from './push-delivery-message.js' +import { PushNotificationSchema } from '@orca-cloud/push-contract' + +it('carries a silent preference through validation to APNs and Android payloads', () => { + const notification = PushNotificationSchema.parse({ + notificationSeq: 1, + notificationEpoch: 'epoch', + source: 'terminal-bell', + agentState: null, + title: 'Bell', + body: '', + sound: false + }) + const delivery = buildPushDelivery({ + registrationId: 'reg', + hostFingerprint: 'host', + notification, + title: 'Bell', + body: '', + coalescedCount: 1 + }) + expect(JSON.parse(apnsBody(delivery)).aps).not.toHaveProperty('sound') + expect( + JSON.parse(fcmMessageBody({ delivery, token: 'test-token', channelId: 'orca-desktop' })).message + .android.notification.channel_id + ).toBe('orca-desktop-silent') + expect(JSON.parse(apnsBody({ ...delivery, sound: undefined })).aps.sound).toBe('default') +}) diff --git a/cloud/apps/push/src/push-observability.ts b/cloud/apps/push/src/push-observability.ts new file mode 100644 index 00000000000..4840723b7ec --- /dev/null +++ b/cloud/apps/push/src/push-observability.ts @@ -0,0 +1,73 @@ +type PushCounterName = + | 'ip_rate_limited' + | 'request_error' + | 'challenge_issued' + | 'challenge_rejected' + | 'session_issued' + | 'session_rejected' + | 'device_registered' + | 'device_rejected' + | 'device_deleted' + | 'send_queued' + | 'send_dead' + | 'send_rate_limited' + | 'send_error' + | 'delivery_sent' + | 'delivery_dead' + | 'delivery_error' + | 'delivery_retry' + +const COUNTER_NAMES: PushCounterName[] = [ + 'ip_rate_limited', + 'request_error', + 'challenge_issued', + 'challenge_rejected', + 'session_issued', + 'session_rejected', + 'device_registered', + 'device_rejected', + 'device_deleted', + 'send_queued', + 'send_dead', + 'send_rate_limited', + 'send_error', + 'delivery_sent', + 'delivery_dead', + 'delivery_error', + 'delivery_retry' +] + +// Aggregate counters only. Nothing here may accept a token, a title, a body, +// or more than the first four characters of a host fingerprint. +export class PushObservability { + private counters = new Map() + private timer: NodeJS.Timeout | null = null + + record(name: PushCounterName, delta = 1): void { + this.counters.set(name, (this.counters.get(name) ?? 0) + delta) + } + + consume(): Record { + const snapshot = Object.fromEntries( + COUNTER_NAMES.map((name) => [name, this.counters.get(name) ?? 0]) + ) as Record + this.counters = new Map() + return snapshot + } + + start(intervalMs = 60_000): void { + if (this.timer) return + this.timer = setInterval(() => { + const counters = this.consume() + if (Object.values(counters).every((value) => value === 0)) return + console.warn(JSON.stringify({ event: 'orca_push_counters', ...counters })) + }, intervalMs) + this.timer.unref() + } + + stop(): void { + if (!this.timer) return + clearInterval(this.timer) + this.timer = null + } +} diff --git a/cloud/apps/push/src/push-provider-outcome.ts b/cloud/apps/push/src/push-provider-outcome.ts new file mode 100644 index 00000000000..bc65d10c175 --- /dev/null +++ b/cloud/apps/push/src/push-provider-outcome.ts @@ -0,0 +1,6 @@ +// What a provider send resolved to, before the send route maps it onto the +// contract's queued / dead / rate_limited / error statuses. +export type PushProviderOutcome = + | { status: 'sent' } + | { status: 'dead'; reason: string } + | { status: 'error'; reason: string; retryable?: boolean; retryAfterMs?: number } diff --git a/cloud/apps/push/src/push-readiness.ts b/cloud/apps/push/src/push-readiness.ts new file mode 100644 index 00000000000..d652fbca1c1 --- /dev/null +++ b/cloud/apps/push/src/push-readiness.ts @@ -0,0 +1,33 @@ +import type { PushDatabase } from './push-database.js' + +export type PushReadinessOptions = { + cacheMs?: number + now?: () => number + observe?: (observation: { ready: boolean; sqlLatencyMs: number }) => void +} + +// The gateway holds no JWKS dependency, so readiness is exactly "can we reach +// the database": /health stays unconditional for the container probe. +export function createPushReadiness( + database: PushDatabase, + options: PushReadinessOptions = {} +): () => Promise { + const cacheMs = options.cacheMs ?? 10_000 + const now = options.now ?? Date.now + let cachedAt = Number.NEGATIVE_INFINITY + let cached = false + + return async () => { + if (now() - cachedAt < cacheMs) return cached + const startedAt = now() + try { + await database.query('SELECT 1 AS ready') + cached = true + } catch { + cached = false + } + cachedAt = now() + options.observe?.({ ready: cached, sqlLatencyMs: Math.max(0, cachedAt - startedAt) }) + return cached + } +} diff --git a/cloud/apps/push/src/push-request-drain.ts b/cloud/apps/push/src/push-request-drain.ts new file mode 100644 index 00000000000..4acaf09ca67 --- /dev/null +++ b/cloud/apps/push/src/push-request-drain.ts @@ -0,0 +1,28 @@ +import type { MiddlewareHandler } from 'hono' + +export class PushRequestDrain { + private draining = false + private active = 0 + private readonly waiters = new Set<() => void>() + + readonly middleware: MiddlewareHandler = async (context, next) => { + if (this.draining) return context.json({ error: 'shutting_down' }, 503) + this.active++ + try { + await next() + } finally { + this.active-- + if (this.active === 0) { + for (const resolve of this.waiters) resolve() + this.waiters.clear() + } + } + } + + begin(): Promise { + this.draining = true + return this.active === 0 + ? Promise.resolve() + : new Promise((resolve) => this.waiters.add(resolve)) + } +} diff --git a/cloud/apps/push/src/push-schema.ts b/cloud/apps/push/src/push-schema.ts new file mode 100644 index 00000000000..1be71bc97bd --- /dev/null +++ b/cloud/apps/push/src/push-schema.ts @@ -0,0 +1,71 @@ +// The five tables the gateway spec names. Applied at startup for both dialects, +// so every column type has to read the same in SQLite and PostgreSQL. +const PUSH_SCHEMA = ` +CREATE TABLE IF NOT EXISTS push_hosts ( + host_fingerprint TEXT PRIMARY KEY, + host_public_key TEXT NOT NULL, + created_at BIGINT NOT NULL, + last_seen_at BIGINT NOT NULL +); + +CREATE TABLE IF NOT EXISTS push_challenges ( + challenge_id TEXT PRIMARY KEY, + host_fingerprint TEXT NOT NULL, + -- Carried here so a host row is only written once a proof succeeds; an + -- unauthenticated challenge must not be able to create one. + host_public_key TEXT NOT NULL, + secret_hash TEXT NOT NULL, + transcript TEXT NOT NULL, + expires_at BIGINT NOT NULL, + consumed_at BIGINT +); +CREATE INDEX IF NOT EXISTS push_challenges_expires_at ON push_challenges(expires_at); + +CREATE TABLE IF NOT EXISTS push_sessions ( + token_hash TEXT PRIMARY KEY, + host_fingerprint TEXT NOT NULL, + expires_at BIGINT NOT NULL, + created_at BIGINT NOT NULL +); +CREATE INDEX IF NOT EXISTS push_sessions_expires_at ON push_sessions(expires_at); + +CREATE TABLE IF NOT EXISTS push_devices ( + registration_id TEXT PRIMARY KEY, + host_fingerprint TEXT NOT NULL, + device_id TEXT NOT NULL, + platform TEXT NOT NULL, + token TEXT NOT NULL, + apns_environment TEXT, + filter_json TEXT NOT NULL, + dead_at BIGINT, + created_at BIGINT NOT NULL, + updated_at BIGINT NOT NULL +); +CREATE UNIQUE INDEX IF NOT EXISTS push_devices_host_device + ON push_devices(host_fingerprint, device_id); + +CREATE TABLE IF NOT EXISTS push_send_log ( + send_id TEXT PRIMARY KEY, + host_fingerprint TEXT NOT NULL, + registration_id TEXT NOT NULL, + sent_at BIGINT NOT NULL +); +-- Both quota windows scan by identity and time, and the pruner scans by time alone. +CREATE INDEX IF NOT EXISTS push_send_log_host_sent_at ON push_send_log(host_fingerprint, sent_at); +CREATE INDEX IF NOT EXISTS push_send_log_registration_sent_at + ON push_send_log(registration_id, sent_at); +CREATE INDEX IF NOT EXISTS push_send_log_sent_at ON push_send_log(sent_at); + +-- The stale-host pruner scans by last contact. Its owning-host subquery rides +-- the push_devices_host_device index. +CREATE INDEX IF NOT EXISTS push_hosts_last_seen_at ON push_hosts(last_seen_at); +` + +export function pushSchemaStatements(): string[] { + // Comments are stripped before the split so a ';' inside one cannot cut a + // statement in half and hand SQLite an "incomplete input" fragment. + return PUSH_SCHEMA.replace(/--[^\n]*/g, '') + .split(';') + .map((statement) => statement.trim()) + .filter((statement) => statement.length > 0) +} diff --git a/cloud/apps/push/src/push-send-idempotency.test.ts b/cloud/apps/push/src/push-send-idempotency.test.ts new file mode 100644 index 00000000000..ec79512f70e --- /dev/null +++ b/cloud/apps/push/src/push-send-idempotency.test.ts @@ -0,0 +1,34 @@ +import { afterEach, expect, it } from 'vitest' +import { createPushServerHarness, notification } from './push-server-harness.test-fixture.js' +import { createPushHostKeypair } from './host-challenge-answering.test-fixture.js' +const harnesses: Awaited>[] = [] +afterEach(async () => { + await Promise.all(harnesses.splice(0).map((h) => h.close())) +}) + +it('returns queued for concurrent retries without double quota or a false summary', async () => { + const h = await createPushServerHarness() + harnesses.push(h) + const token = await h.signIn(createPushHostKeypair(2)) + const registrationId = await h.registerAndroid(token) + const body = { v: 1, registrationIds: [registrationId], notification: notification() } + const responses = await Promise.all( + Array.from({ length: 10 }, () => h.post('/v1/send', body, token)) + ) + for (const response of responses) + expect(await response.json()).toEqual({ results: [{ registrationId, status: 'queued' }] }) + expect(h.server.coalescer.pendingCount(registrationId)).toBe(1) + await h.server.coalescer.flushAll() + await h.post('/v1/send', body, token) + await h.server.coalescer.flushAll() + expect(h.fcmRequests).toHaveLength(1) + expect(JSON.parse(h.fcmRequests[0]!.body).message.data.coalescedCount).toBe('1') + expect((await h.database.query('SELECT COUNT(*) AS count FROM push_send_log'))[0]?.count).toBe(1) + await h.post( + '/v1/send', + { ...body, notification: notification({ notificationEpoch: 'new-epoch' }) }, + token + ) + await h.server.coalescer.flushAll() + expect(h.fcmRequests).toHaveLength(2) +}) diff --git a/cloud/apps/push/src/push-server-auth.test.ts b/cloud/apps/push/src/push-server-auth.test.ts new file mode 100644 index 00000000000..e15bd64aba8 --- /dev/null +++ b/cloud/apps/push/src/push-server-auth.test.ts @@ -0,0 +1,162 @@ +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import { afterEach, beforeEach, describe, expect, it } from 'vitest' +import { createPushHostKeypair } from './host-challenge-answering.test-fixture.js' +import type { PushDatabase } from './push-database.js' +import { createPushServer } from './push-server.js' +import { + createPushServerHarness, + FILTER, + testPushConfig +} from './push-server-harness.test-fixture.js' + +describe('push gateway authentication and device routes', () => { + let harness: Awaited> + + beforeEach(async () => { + harness = await createPushServerHarness() + }) + + afterEach(async () => { + await harness.close() + }) + + it('answers health unconditionally and ready from the database', async () => { + expect((await harness.server.app.request('/health')).status).toBe(200) + expect((await harness.server.app.request('/ready')).status).toBe(200) + }) + + it('reports not ready when the database is unreachable', async () => { + const unreachable: PushDatabase = { + dialect: 'sqlite', + query: async () => { + throw new Error('no connection') + }, + transaction: async (operation) => await operation(unreachable), + lockQuotaScope: async () => undefined, + close: async () => undefined + } + const broken = createPushServer(testPushConfig(), unreachable, { + fcmAccessToken: async () => 'token', + fcmTransport: async () => ({ status: 200, body: '{}' }) + }) + expect((await broken.app.request('/health')).status).toBe(200) + expect((await broken.app.request('/ready')).status).toBe(503) + broken.coalescer.stop() + }) + + it('completes challenge, session, register, list, delete', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(11)) + const registrationId = await harness.registerAndroid(sessionToken) + + const list = await harness.authorized('/v1/devices', {}, sessionToken) + expect(await list.json()).toEqual({ + devices: [{ registrationId, deviceId: 'device-1', platform: 'android', dead: false }] + }) + + const deleted = await harness.authorized( + `/v1/devices/${registrationId}`, + { method: 'DELETE' }, + sessionToken + ) + expect(deleted.status).toBe(204) + expect(await harness.server.devices.findById(registrationId)).toBeNull() + }) + + it('refuses a request with no bearer, a bogus bearer, and an expired session', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(12)) + expect((await harness.server.app.request('/v1/devices')).status).toBe(401) + const bogus = await harness.authorized('/v1/devices', {}, 'nonsense') + expect(bogus.status).toBe(401) + expect(await bogus.json()).toEqual({ error: 'invalid_token' }) + + harness.advanceClock(PUSH_LIMITS.sessionTtlMs + 1) + const expired = await harness.authorized('/v1/devices', {}, sessionToken) + expect(expired.status).toBe(401) + expect(await expired.json()).toEqual({ error: 'session_expired' }) + }) + + it('refuses a replayed proof and an unknown challenge', async () => { + const host = createPushHostKeypair(13) + const challenge = await harness.issueChallenge(host) + const proof = harness.answer(challenge, host) + expect( + (await harness.post('/v1/host/session', { + v: 1, + challengeId: challenge.challengeId, + proofB64: proof + })).status + ).toBe(200) + + const replay = await harness.post('/v1/host/session', { + v: 1, + challengeId: challenge.challengeId, + proofB64: proof + }) + expect(replay.status).toBe(401) + expect(await replay.json()).toEqual({ error: 'invalid_proof' }) + + const unknown = await harness.post('/v1/host/session', { + v: 1, + challengeId: 'no-such-challenge', + proofB64: proof + }) + expect(await unknown.json()).toEqual({ error: 'invalid_challenge' }) + }) + + it('never returns the host fingerprint on the challenge itself', async () => { + const challenge = await harness.issueChallenge(createPushHostKeypair(22)) + expect(Object.keys(challenge).sort()).toEqual([ + 'challengeId', + 'ciphertextB64', + 'expiresAt', + 'gatewayEphemeralPublicKeyB64', + 'nonceB64' + ]) + }) + + it('lets only the owning host delete a registration', async () => { + const ownerToken = await harness.signIn(createPushHostKeypair(14)) + const intruderToken = await harness.signIn(createPushHostKeypair(15)) + const registrationId = await harness.registerAndroid(ownerToken) + + const forbidden = await harness.authorized( + `/v1/devices/${registrationId}`, + { method: 'DELETE' }, + intruderToken + ) + expect(forbidden.status).toBe(404) + expect(await forbidden.json()).toEqual({ error: 'not_found' }) + expect(await harness.server.devices.findById(registrationId)).not.toBeNull() + }) + + it('replaces the token on a re-registration and keeps one registration id', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(23)) + const first = await harness.registerAndroid(sessionToken) + const again = await harness.post( + '/v1/devices', + { + v: 1, + deviceId: 'device-1', + platform: 'android', + token: 'rotated_token:APA91b-newnewnewnewnewnewnewnewnewnew', + filter: FILTER + }, + sessionToken + ) + expect(await again.json()).toEqual({ registrationId: first }) + expect(await harness.server.devices.findById(first)).toMatchObject({ + token: 'rotated_token:APA91b-newnewnewnewnewnewnewnewnewnew' + }) + }) + + it('rejects a malformed registration body', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(16)) + const bad = await harness.post( + '/v1/devices', + { v: 1, deviceId: 'device-1', platform: 'ios', token: 'not-hex', filter: FILTER }, + sessionToken + ) + expect(bad.status).toBe(400) + expect(await bad.json()).toEqual({ error: 'invalid_request' }) + }) +}) diff --git a/cloud/apps/push/src/push-server-harness.test-fixture.ts b/cloud/apps/push/src/push-server-harness.test-fixture.ts new file mode 100644 index 00000000000..4b955fcf68a --- /dev/null +++ b/cloud/apps/push/src/push-server-harness.test-fixture.ts @@ -0,0 +1,165 @@ +import { generateKeyPairSync } from 'node:crypto' +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import { expect } from 'vitest' +import type { ApnsRequest, ApnsResponse } from './apns-http2-transport.js' +import type { PushConfig } from './config.js' +import type { FcmRequest, FcmResponse } from './fcm-client.js' +import { + answerPushHostChallenge, + hostPublicKeyB64, + type PushHostKeypair +} from './host-challenge-answering.test-fixture.js' +import { openInMemoryPushDatabase, type PushDatabase } from './push-database.js' +import { createPushServer } from './push-server.js' + +export const GATEWAY_ORIGIN = 'https://push.onorca.dev' +export const APNS_TOKEN = 'a'.repeat(64) +export const FCM_TOKEN = 'cQ1abcDEF_gh:APA91bZZ-zz0123456789abcdefghijklmnopqrstuvwxyz' +export const FILTER = { sources: ['agent-task-complete'], agentStates: ['needs-input'] } + +export function notification(overrides: Record = {}): Record { + return { + notificationId: 'note-1', + notificationSeq: 1, + notificationEpoch: 'epoch-1', + source: 'agent-task-complete', + agentState: 'needs-input', + title: 'Agent needs input', + body: 'Waiting on your answer', + worktreeId: 'wt-1', + ...overrides + } +} + +export function testPushConfig(): PushConfig { + const { privateKey } = generateKeyPairSync('ec', { + namedCurve: 'P-256', + privateKeyEncoding: { type: 'pkcs8', format: 'pem' }, + publicKeyEncoding: { type: 'spki', format: 'pem' } + }) + return { + port: 0, + publicUrl: GATEWAY_ORIGIN, + dataDir: './data/push-test', + databasePoolMax: 10, + apns: { keyPem: privateKey, keyId: 'ABCDE12345', teamId: 'TEAM123456' }, + apnsTopic: 'com.stably.orca.mobile', + fcmProjectId: 'onorca-cloud', + coalesceMs: PUSH_LIMITS.coalesceWindowMs, + trustedProxyHops: 0 + } +} + +type ChallengeWire = { + challengeId: string + gatewayEphemeralPublicKeyB64: string + nonceB64: string + ciphertextB64: string + expiresAt: number +} + +export async function createPushServerHarness() { + const database: PushDatabase = await openInMemoryPushDatabase() + let clock = 1_700_000_000_000 + const apnsRequests: ApnsRequest[] = [] + const fcmRequests: FcmRequest[] = [] + let apnsResponse: ApnsResponse = { status: 200, body: '' } + let fcmResponse: FcmResponse = { status: 200, body: '{}' } + const server = createPushServer(testPushConfig(), database, { + now: () => clock, + providerRetryWait: async () => undefined, + apnsTransport: async (request) => { + apnsRequests.push(request) + return apnsResponse + }, + fcmTransport: async (request) => { + fcmRequests.push(request) + return fcmResponse + }, + fcmAccessToken: async () => 'access-token', + // Windows are flushed explicitly so the 3s timer never gates a test. + setTimer: () => ({ handle: null }), + clearTimer: () => undefined + }) + + const post = async (path: string, body: unknown, token?: string): Promise => + await server.app.request(path, { + method: 'POST', + headers: { + 'content-type': 'application/json', + ...(token ? { authorization: `Bearer ${token}` } : {}) + }, + body: JSON.stringify(body) + }) + + const issueChallenge = async (keypair: PushHostKeypair): Promise => { + const response = await post('/v1/host/challenge', { + v: 1, + hostPublicKeyB64: hostPublicKeyB64(keypair) + }) + expect(response.status).toBe(200) + return (await response.json()) as ChallengeWire + } + + const answer = (challenge: ChallengeWire, keypair: PushHostKeypair): string => { + const proof = answerPushHostChallenge(challenge, { + gatewayOrigin: GATEWAY_ORIGIN, + keypair, + now: () => clock + }) + expect(proof).not.toBeNull() + return proof! + } + + return { + server, + database, + apnsRequests, + fcmRequests, + post, + issueChallenge, + answer, + now: () => clock, + advanceClock: (deltaMs: number): void => { + clock += deltaMs + }, + setApnsResponse: (response: ApnsResponse): void => { + apnsResponse = response + }, + setFcmResponse: (response: FcmResponse): void => { + fcmResponse = response + }, + authorized: async (path: string, init: RequestInit = {}, token?: string): Promise => + await server.app.request(path, { + ...init, + headers: { + ...(init.headers as Record | undefined), + ...(token ? { authorization: `Bearer ${token}` } : {}) + } + }), + signIn: async (keypair: PushHostKeypair): Promise => { + const challenge = await issueChallenge(keypair) + const response = await post('/v1/host/session', { + v: 1, + challengeId: challenge.challengeId, + proofB64: answer(challenge, keypair) + }) + expect(response.status).toBe(200) + return ((await response.json()) as { sessionToken: string }).sessionToken + }, + registerAndroid: async (token: string, deviceId = 'device-1'): Promise => { + const response = await post( + '/v1/devices', + { v: 1, deviceId, platform: 'android', token: FCM_TOKEN, filter: FILTER }, + token + ) + expect(response.status).toBe(200) + return ((await response.json()) as { registrationId: string }).registrationId + }, + close: async (): Promise => { + server.coalescer.stop() + // A test may close the database itself to provoke a route failure. + await database.close().catch(() => undefined) + } + } +} diff --git a/cloud/apps/push/src/push-server-limits.test.ts b/cloud/apps/push/src/push-server-limits.test.ts new file mode 100644 index 00000000000..9423e4022a7 --- /dev/null +++ b/cloud/apps/push/src/push-server-limits.test.ts @@ -0,0 +1,270 @@ +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { + createPushHostKeypair, + hostPublicKeyB64 +} from './host-challenge-answering.test-fixture.js' +import { + createPushServerHarness, + FCM_TOKEN, + FILTER, + notification +} from './push-server-harness.test-fixture.js' + +const CLIENT_IP = '203.0.113.7' +const OTHER_CLIENT_IP = '198.51.100.9' + +function oversizedChallengeBody(): string { + return JSON.stringify({ v: 1, filler: 'x'.repeat(PUSH_LIMITS.maxHttpBodyBytes) }) +} + +function chunkedRequest(path: string, body: string): Request { + const stream = new ReadableStream({ + start(controller) { + controller.enqueue(new TextEncoder().encode(body)) + controller.close() + } + }) + return new Request(`http://push.test${path}`, { + method: 'POST', + headers: { 'content-type': 'application/json' }, + body: stream, + duplex: 'half' + } as RequestInit) +} + +describe('push gateway request limits', () => { + let harness: Awaited> + + beforeEach(async () => { + harness = await createPushServerHarness() + }) + + afterEach(async () => { + await harness.close() + }) + + it('refuses an oversized chunked body that declares no content length', async () => { + const request = chunkedRequest('/v1/host/challenge', oversizedChallengeBody()) + expect(request.headers.get('content-length')).toBeNull() + + const response = await harness.server.app.request(request) + expect(response.status).toBe(413) + expect(await response.json()).toEqual({ error: 'request_too_large' }) + }) + + it('still refuses an oversized body that declares a content length', async () => { + const body = oversizedChallengeBody() + const response = await harness.server.app.request('/v1/host/challenge', { + method: 'POST', + headers: { + 'content-type': 'application/json', + 'content-length': String(Buffer.byteLength(body)) + }, + body + }) + expect(response.status).toBe(413) + expect(await response.json()).toEqual({ error: 'request_too_large' }) + }) + + it('lets a chunked body under the cap through to schema validation', async () => { + const response = await harness.server.app.request( + chunkedRequest( + '/v1/host/challenge', + JSON.stringify({ v: 1, hostPublicKeyB64: hostPublicKeyB64(createPushHostKeypair(60)) }) + ) + ) + expect(response.status).toBe(200) + }) + + it('caps an authenticated oversized send as well', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(61)) + const response = await harness.server.app.request( + new Request('http://push.test/v1/send', { + method: 'POST', + headers: { + 'content-type': 'application/json', + authorization: `Bearer ${sessionToken}` + }, + body: new ReadableStream({ + start(controller) { + controller.enqueue(new TextEncoder().encode(oversizedChallengeBody())) + controller.close() + } + }), + duplex: 'half' + } as RequestInit) + ) + expect(response.status).toBe(413) + expect(await response.json()).toEqual({ error: 'request_too_large' }) + }) + + it('rate limits one client ip across both unauthenticated routes', async () => { + const body = JSON.stringify({ + v: 1, + hostPublicKeyB64: hostPublicKeyB64(createPushHostKeypair(62)) + }) + // Cloud Run appends the peer, so the caller's own IP is the last value. + const headers = { + 'content-type': 'application/json', + 'x-forwarded-for': `10.0.0.1, ${CLIENT_IP}` + } + for (let index = 0; index < PUSH_LIMITS.unauthenticatedRequestsPerMinutePerIp; index++) { + const allowed = await harness.server.app.request('/v1/host/challenge', { + method: 'POST', + headers, + body + }) + expect(allowed.status).toBe(200) + } + + const limited = await harness.server.app.request('/v1/host/challenge', { + method: 'POST', + headers, + body + }) + expect(limited.status).toBe(429) + expect(await limited.json()).toEqual({ error: 'rate_limited' }) + + // The session route draws on the same bucket, so a flood cannot simply move. + const session = await harness.server.app.request('/v1/host/session', { + method: 'POST', + headers, + body: JSON.stringify({ v: 1, challengeId: 'anything', proofB64: 'x'.repeat(44) }) + }) + expect(session.status).toBe(429) + + const other = await harness.server.app.request('/v1/host/challenge', { + method: 'POST', + headers: { ...headers, 'x-forwarded-for': `10.0.0.1, ${OTHER_CLIENT_IP}` }, + body + }) + expect(other.status).toBe(200) + + // A caller rewriting the left of the chain lands in its own bucket anyway. + const spoofed = await harness.server.app.request('/v1/host/challenge', { + method: 'POST', + headers: { ...headers, 'x-forwarded-for': `198.51.100.250, ${CLIENT_IP}` }, + body + }) + expect(spoofed.status).toBe(429) + }) + + it('lets a throttled client back in once the window refills', async () => { + const body = JSON.stringify({ + v: 1, + hostPublicKeyB64: hostPublicKeyB64(createPushHostKeypair(63)) + }) + const headers = { 'content-type': 'application/json', 'x-forwarded-for': CLIENT_IP } + for (let index = 0; index < PUSH_LIMITS.unauthenticatedRequestsPerMinutePerIp; index++) { + await harness.server.app.request('/v1/host/challenge', { method: 'POST', headers, body }) + } + expect( + (await harness.server.app.request('/v1/host/challenge', { method: 'POST', headers, body })) + .status + ).toBe(429) + + harness.advanceClock(60_000) + expect( + (await harness.server.app.request('/v1/host/challenge', { method: 'POST', headers, body })) + .status + ).toBe(200) + }) + + it('gives the authenticated routes their own, wider bucket per client ip', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(64)) + const headers = { 'x-forwarded-for': CLIENT_IP } + for (let index = 0; index < PUSH_LIMITS.authenticatedRequestsPerMinutePerIp; index++) { + const listed = await harness.authorized('/v1/devices', { headers }, sessionToken) + expect(listed.status).toBe(200) + } + const limited = await harness.authorized('/v1/devices', { headers }, sessionToken) + expect(limited.status).toBe(429) + // The handshake bucket is untouched by any of that. + const challenge = await harness.server.app.request('/v1/host/challenge', { + method: 'POST', + headers: { ...headers, 'content-type': 'application/json' }, + body: JSON.stringify({ v: 1, hostPublicKeyB64: hostPublicKeyB64(createPushHostKeypair(67)) }) + }) + expect(challenge.status).toBe(200) + }) + + it('caps a flood of forged bearers before any of them reaches the session lookup', async () => { + const headers = { 'x-forwarded-for': CLIENT_IP } + const [before] = await harness.database.query('SELECT COUNT(*) AS sessions FROM push_sessions') + for (let index = 0; index < PUSH_LIMITS.authenticatedRequestsPerMinutePerIp; index++) { + const refused = await harness.authorized('/v1/send', { method: 'POST', headers }, 'forged') + expect(refused.status).toBe(401) + } + const limited = await harness.authorized('/v1/send', { method: 'POST', headers }, 'forged') + expect(limited.status).toBe(429) + expect(await limited.json()).toEqual({ error: 'rate_limited' }) + expect(harness.server.unauthenticatedIps.trackedIpCount()).toBe(0) + const [after] = await harness.database.query('SELECT COUNT(*) AS sessions FROM push_sessions') + expect(Number(after?.sessions)).toBe(Number(before?.sessions)) + }) + + it('answers 409 once a host has registered its device allowance', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(66)) + for (let index = 0; index < PUSH_LIMITS.maxDevicesPerHost; index++) { + const accepted = await harness.post( + '/v1/devices', + { v: 1, deviceId: `device-${index}`, platform: 'android', token: FCM_TOKEN, filter: FILTER }, + sessionToken + ) + expect(accepted.status).toBe(200) + } + + const refused = await harness.post( + '/v1/devices', + { v: 1, deviceId: 'one-too-many', platform: 'android', token: FCM_TOKEN, filter: FILTER }, + sessionToken + ) + expect(refused.status).toBe(409) + expect(await refused.json()).toEqual({ error: 'too_many_devices' }) + + const listed = await harness.authorized('/v1/devices', {}, sessionToken) + expect(((await listed.json()) as { devices: unknown[] }).devices).toHaveLength( + PUSH_LIMITS.maxDevicesPerHost + ) + }) + + // Why: a database error carries the failing row in its message. The response + // and the log must both stop at the error's name. + it('answers an unexpected route failure with a bare 500 and logs only the name', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(66)) + const warn = vi.spyOn(console, 'warn').mockImplementation(() => undefined) + try { + await harness.database.close() + const response = await harness.authorized('/v1/devices', {}, sessionToken) + expect(response.status).toBe(500) + expect(await response.json()).toEqual({ error: 'internal' }) + const logged = warn.mock.calls.map((call) => String(call[0])).join('\n') + expect(logged).toContain('"event":"orca_push_request_failed"') + expect(logged).not.toContain('SELECT') + expect(logged).not.toContain('push_devices') + expect(harness.server.observability.consume().request_error).toBe(1) + } finally { + warn.mockRestore() + } + }) + + it('charges a repeated registration id once and returns one result', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(65)) + const registrationId = await harness.registerAndroid(sessionToken) + + const response = await harness.post( + '/v1/send', + { + v: 1, + registrationIds: [registrationId, registrationId, registrationId], + notification: notification() + }, + sessionToken + ) + expect(await response.json()).toEqual({ results: [{ registrationId, status: 'queued' }] }) + expect(harness.server.coalescer.pendingCount(registrationId)).toBe(1) + const [row] = await harness.database.query('SELECT COUNT(*) AS sends FROM push_send_log') + expect(Number(row?.sends)).toBe(1) + }) +}) diff --git a/cloud/apps/push/src/push-server-send.test.ts b/cloud/apps/push/src/push-server-send.test.ts new file mode 100644 index 00000000000..35d0c60c89e --- /dev/null +++ b/cloud/apps/push/src/push-server-send.test.ts @@ -0,0 +1,182 @@ +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import { afterEach, beforeEach, describe, expect, it } from 'vitest' +import { createPushHostKeypair } from './host-challenge-answering.test-fixture.js' +import { + APNS_TOKEN, + createPushServerHarness, + FCM_TOKEN, + FILTER, + notification +} from './push-server-harness.test-fixture.js' + +describe('push gateway send route', () => { + let harness: Awaited> + + beforeEach(async () => { + harness = await createPushServerHarness() + }) + + afterEach(async () => { + await harness.close() + }) + + it('rejects a batch over the registration cap', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(16)) + const oversized = await harness.post( + '/v1/send', + { + v: 1, + registrationIds: Array.from( + { length: PUSH_LIMITS.maxRegistrationIdsPerSend + 1 }, + (_, index) => `reg-${index}` + ), + notification: notification() + }, + sessionToken + ) + expect(oversized.status).toBe(400) + expect(await oversized.json()).toEqual({ error: 'invalid_request' }) + }) + + it('queues a send, delivers it to fcm, and reports a dead token on the next send', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(17)) + const registrationId = await harness.registerAndroid(sessionToken) + + const queued = await harness.post( + '/v1/send', + { v: 1, registrationIds: [registrationId], notification: notification() }, + sessionToken + ) + expect(await queued.json()).toEqual({ results: [{ registrationId, status: 'queued' }] }) + + harness.setFcmResponse({ + status: 404, + body: JSON.stringify({ error: { status: 'UNREGISTERED', message: 'gone' } }) + }) + await harness.server.coalescer.flushAll() + expect(harness.fcmRequests).toHaveLength(1) + expect(JSON.parse(harness.fcmRequests[0]!.body)).toMatchObject({ + message: { token: FCM_TOKEN, notification: { title: 'Agent needs input' } } + }) + + const afterDeath = await harness.post( + '/v1/send', + { v: 1, registrationIds: [registrationId], notification: notification() }, + sessionToken + ) + expect(await afterDeath.json()).toEqual({ results: [{ registrationId, status: 'dead' }] }) + + const listed = await harness.authorized('/v1/devices', {}, sessionToken) + expect(await listed.json()).toEqual({ + devices: [{ registrationId, deviceId: 'device-1', platform: 'android', dead: true }] + }) + }) + + it('leaves a live registration alone when the provider reports a transient failure', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(24)) + const registrationId = await harness.registerAndroid(sessionToken) + await harness.post( + '/v1/send', + { v: 1, registrationIds: [registrationId], notification: notification() }, + sessionToken + ) + harness.setFcmResponse({ + status: 503, + body: JSON.stringify({ error: { status: 'UNAVAILABLE', message: 'backend busy' } }) + }) + await harness.server.coalescer.flushAll() + expect(await harness.server.devices.findById(registrationId)).toMatchObject({ dead: false }) + }) + + it('coalesces a burst into one apns summary under the host collapse id', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(18)) + const registration = await harness.post( + '/v1/devices', + { + v: 1, + deviceId: 'iphone-1', + platform: 'ios', + token: APNS_TOKEN, + apnsEnvironment: 'sandbox', + filter: FILTER + }, + sessionToken + ) + const { registrationId } = (await registration.json()) as { registrationId: string } + for (const seq of [1, 2, 3]) { + await harness.post( + '/v1/send', + { + v: 1, + registrationIds: [registrationId], + notification: notification({ notificationId: `note-${seq}`, notificationSeq: seq }) + }, + sessionToken + ) + } + await harness.server.coalescer.flushAll() + expect(harness.apnsRequests).toHaveLength(1) + const request = harness.apnsRequests[0]! + expect(request.host).toBe('api.sandbox.push.apple.com') + const body = JSON.parse(request.body) as { + aps: { alert: { title: string; body: string } } + orca: { coalescedCount: number; notificationSeq: number } + } + expect(body.aps.alert).toEqual({ title: 'Orca', body: '3 agents need attention' }) + expect(body.orca.coalescedCount).toBe(3) + expect(body.orca.notificationSeq).toBe(3) + expect(request.headers['apns-collapse-id']).toMatch(/^host:/) + }) + + it('sends a lone event through unchanged with its own collapse id', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(25)) + const registrationId = await harness.registerAndroid(sessionToken) + await harness.post( + '/v1/send', + { v: 1, registrationIds: [registrationId], notification: notification() }, + sessionToken + ) + await harness.server.coalescer.flushAll() + const message = JSON.parse(harness.fcmRequests[0]!.body) as { + message: { android: { notification: { tag: string } }; data: Record } + } + expect(message.message.android.notification.tag).toBe('note-1') + expect(message.message.data.coalescedCount).toBe('1') + }) + + it('reports an error for a registration the host does not own', async () => { + const ownerToken = await harness.signIn(createPushHostKeypair(19)) + const intruderToken = await harness.signIn(createPushHostKeypair(20)) + const registrationId = await harness.registerAndroid(ownerToken) + + const foreign = await harness.post( + '/v1/send', + { v: 1, registrationIds: [registrationId, 'made-up'], notification: notification() }, + intruderToken + ) + expect(await foreign.json()).toEqual({ + results: [ + { registrationId, status: 'error' }, + { registrationId: 'made-up', status: 'error' } + ] + }) + expect(harness.server.coalescer.pendingCount(registrationId)).toBe(0) + }) + + it('rate limits a host that exhausted its hourly allowance', async () => { + const sessionToken = await harness.signIn(createPushHostKeypair(21)) + const registrationId = await harness.registerAndroid(sessionToken) + const hostFingerprint = (await harness.server.devices.findById(registrationId))!.hostFingerprint + for (let index = 0; index < PUSH_LIMITS.hostSendsPerRollingHour; index++) { + expect(await harness.server.quota.reserve(hostFingerprint, registrationId)).toBe('allowed') + } + const limited = await harness.post( + '/v1/send', + { v: 1, registrationIds: [registrationId], notification: notification() }, + sessionToken + ) + expect(limited.status).toBe(200) + expect(await limited.json()).toEqual({ results: [{ registrationId, status: 'rate_limited' }] }) + expect(harness.server.coalescer.pendingCount(registrationId)).toBe(0) + }) +}) diff --git a/cloud/apps/push/src/push-server.ts b/cloud/apps/push/src/push-server.ts new file mode 100644 index 00000000000..1201748095b --- /dev/null +++ b/cloud/apps/push/src/push-server.ts @@ -0,0 +1,289 @@ +import { createAdaptorServer } from '@hono/node-server' +import { + PUSH_LIMITS, + PushDeviceRegistrationRequestSchema, + PushHostChallengeRequestSchema, + PushHostSessionRequestSchema, + PushSendRequestSchema, + type PushSendResult +} from '@orca-cloud/push-contract' +import { Hono, type MiddlewareHandler } from 'hono' +import { bodyLimit } from 'hono/body-limit' +import { ApnsClient } from './apns-client.js' +import { createApnsHttp2Transport, type ApnsTransport } from './apns-http2-transport.js' +import { clientIpRateLimit, ClientIpRateLimiter } from './client-ip-rate-limit.js' +import { PushCoalescer } from './coalescer.js' +import type { PushConfig } from './config.js' +import { PushDeviceRegistryStore } from './device-registry-store.js' +import { createFcmAccessTokenProvider } from './fcm-access-token.js' +import { createFcmFetchTransport, FcmClient, type FcmTransport } from './fcm-client.js' +import { PushHostChallengeStore } from './host-challenge-store.js' +import { PushHostSessionStore } from './host-session-store.js' +import type { PushDatabase } from './push-database.js' +import { PushDispatcher } from './push-dispatcher.js' +import { PushObservability } from './push-observability.js' +import { createPushReadiness } from './push-readiness.js' +import { PushRequestDrain } from './push-request-drain.js' +import { PushSendQuota } from './send-quota.js' + +export type PushServerOptions = { + now?: () => number + providerRetryWait?: (ms: number) => Promise + apnsTransport?: ApnsTransport + fcmTransport?: FcmTransport + fcmAccessToken?: () => Promise + setTimer?: PushCoalescerTimerFactory + clearTimer?: (timer: { readonly handle: unknown }) => void +} + +type PushCoalescerTimerFactory = ( + callback: () => void, + delayMs: number +) => { readonly handle: unknown } + +type PushVariables = { hostFingerprint: string } + +export function readBearer(header: string | undefined): string | null { + if (!header) return null + const [scheme, ...rest] = header.split(' ') + const token = rest.join(' ').trim() + return scheme?.toLowerCase() === 'bearer' && token.length > 0 ? token : null +} + +// Hono's body limit, not a Content-Length check: a chunked body declares no +// length, and req.json() would buffer all of it before any handler ran. +const limitBody = bodyLimit({ + maxSize: PUSH_LIMITS.maxHttpBodyBytes, + onError: (context) => context.json({ error: 'request_too_large' }, 413) +}) + +export function createPushServer( + config: PushConfig, + database: PushDatabase, + options: PushServerOptions = {} +) { + const now = options.now ?? Date.now + const observability = new PushObservability() + const challenges = new PushHostChallengeStore(database, config.publicUrl, now) + const sessions = new PushHostSessionStore(database, now) + const devices = new PushDeviceRegistryStore(database, now) + const quota = new PushSendQuota(database, now) + const apnsTransport = options.apnsTransport ?? (config.apns ? createApnsHttp2Transport() : null) + const dispatcher = new PushDispatcher({ + devices, + now, + ...(options.providerRetryWait ? { wait: options.providerRetryWait } : {}), + onRetry: () => observability.record('delivery_retry'), + ...(config.apns && apnsTransport + ? { + apns: new ApnsClient({ + topic: config.apnsTopic, + credentials: config.apns, + transport: apnsTransport, + now + }) + } + : {}), + fcm: new FcmClient({ + projectId: config.fcmProjectId, + accessToken: options.fcmAccessToken ?? createFcmAccessTokenProvider(), + transport: options.fcmTransport ?? createFcmFetchTransport() + }), + onOutcome: (status) => + observability.record( + status === 'sent' ? 'delivery_sent' : status === 'dead' ? 'delivery_dead' : 'delivery_error' + ) + }) + const coalescer = new PushCoalescer({ + windowMs: config.coalesceMs, + deliver: (delivery) => dispatcher.deliver(delivery), + ...(options.setTimer ? { setTimer: options.setTimer } : {}), + ...(options.clearTimer ? { clearTimer: options.clearTimer } : {}), + onDeliveryFailed: () => observability.record('delivery_error') + }) + const ready = createPushReadiness(database, { now }) + const unauthenticatedIps = new ClientIpRateLimiter({ now }) + const limitUnauthenticatedIp = clientIpRateLimit(unauthenticatedIps, { + trustedProxyHops: config.trustedProxyHops, + onLimited: () => observability.record('ip_rate_limited') + }) + // Why a second bucket: a bearer has to be looked up before it can be refused, + // and that lookup takes one of very few pool connections. Capping the caller + // first keeps a flood of forged bearers from starving real hosts of the pool. + const authenticatedIps = new ClientIpRateLimiter({ + now, + capacity: PUSH_LIMITS.authenticatedRequestsPerMinutePerIp + }) + const limitAuthenticatedIp = clientIpRateLimit(authenticatedIps, { + trustedProxyHops: config.trustedProxyHops, + onLimited: () => observability.record('ip_rate_limited') + }) + const app = new Hono<{ Variables: PushVariables }>() + const requestDrain = new PushRequestDrain() + app.use('*', requestDrain.middleware) + // Hono's default handler prints the whole error, and a pg error carries the + // offending row in `detail`. Only the error's name may reach the logs. + app.onError((error, context) => { + observability.record('request_error') + console.warn( + JSON.stringify({ + event: 'orca_push_request_failed', + error: error instanceof Error ? error.name : 'unknown' + }) + ) + return context.json({ error: 'internal' }, 500) + }) + + app.get('/health', (context) => context.json({ ok: true, pushProtocol: 1 })) + app.get('/ready', async (context) => + (await ready()) + ? context.json({ ok: true }) + : context.json({ error: 'dependency_unavailable' }, 503) + ) + + const bearerSession: MiddlewareHandler<{ Variables: PushVariables }> = async (context, next) => { + const bearer = readBearer(context.req.header('authorization')) + if (!bearer) return context.json({ error: 'invalid_token' }, 401) + const session = await sessions.resolve(bearer) + if (!session.ok) { + return context.json( + { error: session.reason === 'session_expired' ? 'session_expired' : 'invalid_token' }, + 401 + ) + } + context.set('hostFingerprint', session.hostFingerprint) + await next() + return + } + // `/v1/devices/*` matches `/v1/devices` itself; a second registration for the + // bare path would run both middlewares twice on it. + app.use('/v1/devices/*', limitAuthenticatedIp, bearerSession) + app.use('/v1/send', limitAuthenticatedIp, bearerSession) + + app.post('/v1/host/challenge', limitUnauthenticatedIp, limitBody, async (context) => { + const body = PushHostChallengeRequestSchema.safeParse( + await context.req.json().catch(() => null) + ) + if (!body.success) return context.json({ error: 'invalid_request' }, 400) + const issued = await challenges.issue(body.data.hostPublicKeyB64) + if (!issued) { + observability.record('challenge_rejected') + return context.json({ error: 'invalid_request' }, 400) + } + observability.record('challenge_issued') + const { hostFingerprint: _bound, ...response } = issued + return context.json(response) + }) + + app.post('/v1/host/session', limitUnauthenticatedIp, limitBody, async (context) => { + const body = PushHostSessionRequestSchema.safeParse(await context.req.json().catch(() => null)) + if (!body.success) return context.json({ error: 'invalid_request' }, 400) + const verification = await challenges.verify(body.data.challengeId, body.data.proofB64) + if (!verification.ok) { + observability.record('session_rejected') + return context.json( + { + error: verification.reason === 'unknown_challenge' ? 'invalid_challenge' : 'invalid_proof' + }, + 401 + ) + } + observability.record('session_issued') + return context.json(await sessions.create(verification.hostFingerprint)) + }) + + app.post('/v1/devices', limitBody, async (context) => { + const body = PushDeviceRegistrationRequestSchema.safeParse( + await context.req.json().catch(() => null) + ) + if (!body.success) return context.json({ error: 'invalid_request' }, 400) + const registered = await devices.upsert({ + hostFingerprint: context.get('hostFingerprint'), + deviceId: body.data.deviceId, + platform: body.data.platform, + token: body.data.token, + ...(body.data.apnsEnvironment === undefined + ? {} + : { apnsEnvironment: body.data.apnsEnvironment }), + filter: body.data.filter + }) + if (!registered.ok) { + observability.record('device_rejected') + return context.json({ error: 'too_many_devices' }, 409) + } + observability.record('device_registered') + return context.json({ registrationId: registered.registrationId }) + }) + + app.delete('/v1/devices/:registrationId', async (context) => { + const deleted = await devices.deleteOwned( + context.get('hostFingerprint'), + context.req.param('registrationId') + ) + if (!deleted) return context.json({ error: 'not_found' }, 404) + observability.record('device_deleted') + return context.body(null, 204) + }) + + app.get('/v1/devices', async (context) => + context.json({ devices: await devices.list(context.get('hostFingerprint')) }) + ) + + app.post('/v1/send', limitBody, async (context) => { + const body = PushSendRequestSchema.safeParse(await context.req.json().catch(() => null)) + if (!body.success) return context.json({ error: 'invalid_request' }, 400) + const hostFingerprint = context.get('hostFingerprint') + const owned = await devices.findOwned(hostFingerprint, body.data.registrationIds) + const results: PushSendResult[] = [] + for (const registrationId of body.data.registrationIds) { + const device = owned.get(registrationId) + if (!device) { + observability.record('send_error') + results.push({ registrationId, status: 'error' }) + continue + } + if (device.dead) { + observability.record('send_dead') + results.push({ registrationId, status: 'dead' }) + continue + } + const reservation = await quota.reserve( + hostFingerprint, + registrationId, + body.data.notification + ) + if (reservation === 'duplicate') { + results.push({ registrationId, status: 'queued' }) + continue + } + if (reservation === 'rate_limited') { + observability.record('send_rate_limited') + results.push({ registrationId, status: 'rate_limited' }) + continue + } + coalescer.enqueue({ registrationId, hostFingerprint, notification: body.data.notification }) + observability.record('send_queued') + results.push({ registrationId, status: 'queued' }) + } + return context.json({ results }) + }) + + return { + app, + requestDrain, + server: createAdaptorServer(app), + challenges, + sessions, + devices, + quota, + unauthenticatedIps, + coalescer, + observability, + ready, + closeTransports: (): void => { + if (apnsTransport && 'close' in apnsTransport) { + ;(apnsTransport as { close: () => void }).close() + } + } + } +} diff --git a/cloud/apps/push/src/push-session-concurrency.test.ts b/cloud/apps/push/src/push-session-concurrency.test.ts new file mode 100644 index 00000000000..a43daf0f07b --- /dev/null +++ b/cloud/apps/push/src/push-session-concurrency.test.ts @@ -0,0 +1,73 @@ +import { randomUUID } from 'node:crypto' +import { tmpdir } from 'node:os' +import { afterEach, describe, expect, it } from 'vitest' +import { openInMemoryPushDatabase, openPushDatabase, type PushDatabase } from './push-database.js' +import { PushHostSessionStore } from './host-session-store.js' +import { ensurePushSessionIndex } from './push-session-schema.js' +const databases: PushDatabase[] = [] +afterEach(async () => { + await Promise.all(databases.splice(0).map((db) => db.close())) +}) + +async function concurrentSessions(db: PushDatabase) { + databases.push(db) + const host = randomUUID() + const store = new PushHostSessionStore(db) + try { + const sessions = await Promise.all(Array.from({ length: 20 }, () => store.create(host))) + const decisions = await Promise.all( + sessions.map((session) => store.resolve(session.sessionToken)) + ) + expect(decisions.filter((decision) => decision.ok)).toHaveLength(1) + const [row] = await db.query( + 'SELECT COUNT(*) AS count FROM push_sessions WHERE host_fingerprint = ?', + [host] + ) + expect(Number(row?.count)).toBe(1) + } finally { + await db.query('DELETE FROM push_sessions WHERE host_fingerprint = ?', [host]) + } +} +it('serializes sessions on SQLite', async () => { + await concurrentSessions(await openInMemoryPushDatabase()) +}) + +it('migrates existing duplicate hosts to the newest session and enforces uniqueness', async () => { + const db = await openInMemoryPushDatabase() + databases.push(db) + await db.query('DROP INDEX push_sessions_host') + for (const [token, created] of [ + ['old', 1], + ['new', 2] + ] as const) { + await db.query('INSERT INTO push_sessions VALUES (?, ?, ?, ?)', [token, 'host', 100, created]) + } + await ensurePushSessionIndex(db) + expect(await db.query('SELECT token_hash FROM push_sessions')).toEqual([{ token_hash: 'new' }]) + await expect( + db.query('INSERT INTO push_sessions VALUES (?, ?, ?, ?)', ['third', 'host', 100, 3]) + ).rejects.toThrow() +}) + +describe.skipIf(!process.env.ORCA_PUSH_TEST_DATABASE_URL)('PostgreSQL push sessions', () => { + it('leaves exactly one live token after concurrent creates', async () => { + await concurrentSessions( + await openPushDatabase({ + databaseUrl: process.env.ORCA_PUSH_TEST_DATABASE_URL!, + dataDir: tmpdir() + }) + ) + }) + it('allows concurrent schema startup', async () => { + const opened = await Promise.all( + Array.from({ length: 4 }, () => + openPushDatabase({ + databaseUrl: process.env.ORCA_PUSH_TEST_DATABASE_URL!, + dataDir: tmpdir() + }) + ) + ) + databases.push(...opened) + for (const db of opened) expect(await db.query('SELECT 1 AS ok')).toEqual([{ ok: 1 }]) + }) +}) diff --git a/cloud/apps/push/src/push-session-schema.ts b/cloud/apps/push/src/push-session-schema.ts new file mode 100644 index 00000000000..aeb690ce048 --- /dev/null +++ b/cloud/apps/push/src/push-session-schema.ts @@ -0,0 +1,23 @@ +import type { PushDatabase } from './push-database.js' + +export async function ensurePushSessionIndex(database: PushDatabase): Promise { + await database.transaction(async (transaction) => { + await transaction.lockQuotaScope('orca-push-session-schema') + const indexQuery = + database.dialect === 'postgres' + ? "SELECT indexname FROM pg_indexes WHERE schemaname = current_schema() AND tablename = 'push_sessions' AND indexname = 'push_sessions_host'" + : "SELECT name FROM sqlite_master WHERE type = 'index' AND name = 'push_sessions_host'" + if ((await transaction.query(indexQuery)).length) return + // Retain the newest session when upgrading a database with duplicate hosts. + await transaction.query(`DELETE FROM push_sessions WHERE token_hash IN ( + SELECT token_hash FROM ( + SELECT token_hash, ROW_NUMBER() OVER ( + PARTITION BY host_fingerprint ORDER BY created_at DESC, token_hash DESC + ) AS position FROM push_sessions + ) AS ranked WHERE position > 1 + )`) + await transaction.query( + 'CREATE UNIQUE INDEX IF NOT EXISTS push_sessions_host ON push_sessions(host_fingerprint)' + ) + }) +} diff --git a/cloud/apps/push/src/send-quota-postgres.test.ts b/cloud/apps/push/src/send-quota-postgres.test.ts new file mode 100644 index 00000000000..9ccdf176f46 --- /dev/null +++ b/cloud/apps/push/src/send-quota-postgres.test.ts @@ -0,0 +1,100 @@ +import { randomUUID } from 'node:crypto' +import { tmpdir } from 'node:os' +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import { afterEach, beforeEach, describe, expect, it } from 'vitest' +import { PushDeviceRegistryStore } from './device-registry-store.js' +import { openPushDatabase, type PushDatabase } from './push-database.js' +import { PushSendQuota } from './send-quota.js' + +// Cloud Verify supplies a disposable PostgreSQL; SQLite cannot expose these races. +const DATABASE_URL = process.env.ORCA_PUSH_TEST_DATABASE_URL +const CONCURRENT_RESERVES = 80 + +describe.skipIf(!DATABASE_URL)('push send quota on postgres', () => { + let database: PushDatabase + let hostFingerprint: string + + beforeEach(async () => { + database = await openPushDatabase({ + databaseUrl: DATABASE_URL!, + dataDir: tmpdir(), + applicationName: 'orca-push-test' + }) + // Every run owns a fresh identity, so a shared database needs no truncation. + hostFingerprint = randomUUID().replaceAll('-', '').slice(0, 16) + }) + + afterEach(async () => { + await database.query('DELETE FROM push_send_log WHERE host_fingerprint = ?', [hostFingerprint]) + await database.query('DELETE FROM push_devices WHERE host_fingerprint = ?', [hostFingerprint]) + await database.close() + }) + + it('admits exactly the hourly allowance when every reserve races at once', async () => { + const quota = new PushSendQuota(database) + const decisions = await Promise.all( + Array.from({ length: CONCURRENT_RESERVES }, () => quota.reserve(hostFingerprint, 'reg-1')) + ) + expect(decisions.filter((decision) => decision === 'allowed')).toHaveLength( + PUSH_LIMITS.hostSendsPerRollingHour + ) + expect(decisions.filter((decision) => decision === 'rate_limited')).toHaveLength( + CONCURRENT_RESERVES - PUSH_LIMITS.hostSendsPerRollingHour + ) + + const [row] = await database.query( + 'SELECT COUNT(*) AS sends FROM push_send_log WHERE host_fingerprint = ?', + [hostFingerprint] + ) + expect(Number(row?.sends)).toBe(PUSH_LIMITS.hostSendsPerRollingHour) + }) + + it('holds the per-host device cap when every registration races at once', async () => { + const devices = new PushDeviceRegistryStore(database) + const attempts = PUSH_LIMITS.maxDevicesPerHost + 20 + const results = await Promise.all( + Array.from({ length: attempts }, (_, index) => + devices.upsert({ + hostFingerprint, + deviceId: `device-${index}`, + platform: 'android', + token: `token-${index}`, + filter: { sources: ['agent-task-complete'], agentStates: ['needs-input'] } + }) + ) + ) + expect(results.filter((result) => result.ok)).toHaveLength(PUSH_LIMITS.maxDevicesPerHost) + + const [row] = await database.query( + 'SELECT COUNT(*) AS devices FROM push_devices WHERE host_fingerprint = ?', + [hostFingerprint] + ) + expect(Number(row?.devices)).toBe(PUSH_LIMITS.maxDevicesPerHost) + }) + + it('does not let one host lock block another host reserving at the same time', async () => { + const quota = new PushSendQuota(database) + const otherHost = randomUUID().replaceAll('-', '').slice(0, 16) + try { + const decisions = await Promise.all([ + ...Array.from({ length: 40 }, () => quota.reserve(hostFingerprint, 'reg-1')), + ...Array.from({ length: 40 }, () => quota.reserve(otherHost, 'reg-2')) + ]) + expect(decisions.every((decision) => decision === 'allowed')).toBe(true) + } finally { + await database.query('DELETE FROM push_send_log WHERE host_fingerprint = ?', [otherHost]) + } + }) + it('reserves a retried event once under concurrent PostgreSQL transactions', async () => { + const quota = new PushSendQuota(database) + const event = { notificationEpoch: 'epoch', notificationSeq: 1 } + const results = await Promise.all( + Array.from({ length: 40 }, () => quota.reserve(hostFingerprint, 'reg-dedupe', event)) + ) + expect(results.filter((result) => result === 'allowed')).toHaveLength(1) + expect(results.filter((result) => result === 'duplicate')).toHaveLength(39) + expect( + await quota.reserve(hostFingerprint, 'reg-dedupe', { ...event, notificationEpoch: 'next' }) + ).toBe('allowed') + }) +}) diff --git a/cloud/apps/push/src/send-quota.test.ts b/cloud/apps/push/src/send-quota.test.ts new file mode 100644 index 00000000000..dc5b1260020 --- /dev/null +++ b/cloud/apps/push/src/send-quota.test.ts @@ -0,0 +1,70 @@ +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import { afterEach, beforeEach, describe, expect, it } from 'vitest' +import { openInMemoryPushDatabase, type PushDatabase } from './push-database.js' +import { PushSendQuota } from './send-quota.js' + +const HOST = 'abcdefghijklmnop' +const HOUR_MS = 60 * 60 * 1000 +const DAY_MS = 24 * HOUR_MS + +describe('push send quota', () => { + let database: PushDatabase + let clock = 1_700_000_000_000 + let quota: PushSendQuota + + beforeEach(async () => { + database = await openInMemoryPushDatabase() + clock = 1_700_000_000_000 + quota = new PushSendQuota(database, () => clock) + }) + + afterEach(async () => { + await database.close() + }) + + async function reserveMany(count: number, registrationId: string): Promise { + const decisions: string[] = [] + for (let index = 0; index < count; index++) { + decisions.push(await quota.reserve(HOST, registrationId)) + } + return decisions + } + + it('admits exactly the hourly host allowance and refuses the next send', async () => { + const decisions = await reserveMany(PUSH_LIMITS.hostSendsPerRollingHour, 'reg-1') + expect(decisions.every((decision) => decision === 'allowed')).toBe(true) + await expect(quota.reserve(HOST, 'reg-1')).resolves.toBe('rate_limited') + }) + + it('lets the host window roll forward', async () => { + await reserveMany(PUSH_LIMITS.hostSendsPerRollingHour, 'reg-1') + clock += HOUR_MS + await expect(quota.reserve(HOST, 'reg-1')).resolves.toBe('allowed') + }) + + it('limits a single registration across a rolling day even as hosts rotate', async () => { + // Spread the day allowance across hours so the hourly host cap never binds. + for (let index = 0; index < PUSH_LIMITS.registrationSendsPerRollingDay; index++) { + expect(await quota.reserve(HOST, 'reg-1')).toBe('allowed') + if ((index + 1) % PUSH_LIMITS.hostSendsPerRollingHour === 0) clock += HOUR_MS + 1 + } + await expect(quota.reserve(HOST, 'reg-1')).resolves.toBe('rate_limited') + await expect(quota.reserve(HOST, 'reg-2')).resolves.toBe('allowed') + clock += DAY_MS + await expect(quota.reserve(HOST, 'reg-1')).resolves.toBe('allowed') + }) + + it('never logs a send it refused', async () => { + await reserveMany(PUSH_LIMITS.hostSendsPerRollingHour + 5, 'reg-1') + const [row] = await database.query('SELECT COUNT(*) AS sends FROM push_send_log') + expect(Number(row?.sends)).toBe(PUSH_LIMITS.hostSendsPerRollingHour) + }) + + it('prunes the log past the retention window only', async () => { + await quota.reserve(HOST, 'reg-1') + clock += PUSH_LIMITS.sendLogRetentionMs + expect(await quota.prune()).toBe(0) + clock += 1 + expect(await quota.prune()).toBe(1) + }) +}) diff --git a/cloud/apps/push/src/send-quota.ts b/cloud/apps/push/src/send-quota.ts new file mode 100644 index 00000000000..3049cb312b1 --- /dev/null +++ b/cloud/apps/push/src/send-quota.ts @@ -0,0 +1,75 @@ +import { createHash, randomUUID } from 'node:crypto' +import { PUSH_LIMITS } from '@orca-cloud/push-contract' +import type { PushDatabase } from './push-database.js' + +const QUOTA_LOCK_PREFIX = 'orca-push-send-quota:' +const ROLLING_HOUR_MS = 60 * 60 * 1000 +const ROLLING_DAY_MS = 24 * ROLLING_HOUR_MS + +export type PushQuotaDecision = 'allowed' | 'rate_limited' | 'duplicate' + +export class PushSendQuota { + constructor( + private readonly database: PushDatabase, + private readonly now: () => number = Date.now + ) {} + + // One transaction is not enough on its own: PostgreSQL reads at READ + // COMMITTED, so concurrent reserves would each see the same under-quota count + // and all be admitted. The host lock serializes them. The registration count + // rides the same lock because a registration belongs to exactly one host. + async reserve( + hostFingerprint: string, + registrationId: string, + event?: { notificationEpoch: string; notificationSeq: number } + ): Promise { + const now = this.now() + const sendId = event + ? createHash('sha256') + .update( + JSON.stringify([ + hostFingerprint, + registrationId, + event.notificationEpoch, + event.notificationSeq + ]) + ) + .digest('hex') + : randomUUID() + return await this.database.transaction(async (transaction) => { + await transaction.lockQuotaScope(`${QUOTA_LOCK_PREFIX}${hostFingerprint}`) + if ( + event && + (await transaction.query('SELECT send_id FROM push_send_log WHERE send_id = ?', [sendId])) + .length + ) { + return 'duplicate' + } + const [hostRow] = await transaction.query( + 'SELECT COUNT(*) AS sends FROM push_send_log WHERE host_fingerprint = ? AND sent_at > ?', + [hostFingerprint, now - ROLLING_HOUR_MS] + ) + if (Number(hostRow?.sends ?? 0) >= PUSH_LIMITS.hostSendsPerRollingHour) return 'rate_limited' + const [registrationRow] = await transaction.query( + 'SELECT COUNT(*) AS sends FROM push_send_log WHERE registration_id = ? AND sent_at > ?', + [registrationId, now - ROLLING_DAY_MS] + ) + if (Number(registrationRow?.sends ?? 0) >= PUSH_LIMITS.registrationSendsPerRollingDay) { + return 'rate_limited' + } + await transaction.query( + `INSERT INTO push_send_log (send_id, host_fingerprint, registration_id, sent_at) + VALUES (?, ?, ?, ?)`, + [sendId, hostFingerprint, registrationId, now] + ) + return 'allowed' + }) + } + + async prune(): Promise { + const [result] = await this.database.query('DELETE FROM push_send_log WHERE sent_at < ?', [ + this.now() - PUSH_LIMITS.sendLogRetentionMs + ]) + return Number(result?.changes ?? 0) + } +} diff --git a/cloud/apps/push/tsconfig.build.json b/cloud/apps/push/tsconfig.build.json new file mode 100644 index 00000000000..5e71eb0f951 --- /dev/null +++ b/cloud/apps/push/tsconfig.build.json @@ -0,0 +1,10 @@ +{ + "extends": "./tsconfig.json", + "compilerOptions": { + "declaration": true, + "noEmit": false, + "outDir": "dist", + "rootDir": "src" + }, + "exclude": ["src/**/*.test.ts", "src/**/*.test-fixture.ts"] +} diff --git a/cloud/apps/push/tsconfig.json b/cloud/apps/push/tsconfig.json new file mode 100644 index 00000000000..a552e34dbe9 --- /dev/null +++ b/cloud/apps/push/tsconfig.json @@ -0,0 +1,5 @@ +{ + "extends": "../../tsconfig.base.json", + "compilerOptions": { "noEmit": true }, + "include": ["src/**/*.ts"] +} diff --git a/cloud/apps/push/vitest.config.ts b/cloud/apps/push/vitest.config.ts new file mode 100644 index 00000000000..bffcc30e39e --- /dev/null +++ b/cloud/apps/push/vitest.config.ts @@ -0,0 +1,5 @@ +import { defineConfig } from 'vitest/config' + +export default defineConfig({ + test: { name: 'push', include: ['src/**/*.test.ts'], testTimeout: 15_000, hookTimeout: 15_000 } +}) diff --git a/cloud/apps/relay/Dockerfile b/cloud/apps/relay/Dockerfile index 12516cbf749..f0abcf9f5b3 100644 --- a/cloud/apps/relay/Dockerfile +++ b/cloud/apps/relay/Dockerfile @@ -3,11 +3,13 @@ WORKDIR /app RUN corepack enable COPY package.json pnpm-lock.yaml pnpm-workspace.yaml tsconfig.base.json ./ COPY packages/relay-contract/package.json packages/relay-contract/package.json +COPY packages/postgres-schema/package.json packages/postgres-schema/package.json COPY apps/relay/package.json apps/relay/package.json RUN pnpm install --frozen-lockfile COPY packages/relay-contract packages/relay-contract COPY apps/relay apps/relay -RUN pnpm --filter @orca-cloud/relay-contract build && pnpm --filter @orca-cloud/relay build +COPY packages/postgres-schema packages/postgres-schema +RUN pnpm --filter @orca-cloud/postgres-schema build && pnpm --filter @orca-cloud/relay-contract build && pnpm --filter @orca-cloud/relay build FROM node:24-alpine AS runtime ENV NODE_ENV=production @@ -16,8 +18,10 @@ WORKDIR /app RUN corepack enable COPY package.json pnpm-lock.yaml pnpm-workspace.yaml ./ COPY packages/relay-contract/package.json packages/relay-contract/package.json +COPY packages/postgres-schema/package.json packages/postgres-schema/package.json COPY apps/relay/package.json apps/relay/package.json COPY --from=build /app/packages/relay-contract/dist packages/relay-contract/dist +COPY --from=build /app/packages/postgres-schema/dist packages/postgres-schema/dist COPY --from=build /app/apps/relay/dist apps/relay/dist RUN pnpm install --prod --frozen-lockfile --filter @orca-cloud/relay... USER node diff --git a/cloud/apps/relay/package.json b/cloud/apps/relay/package.json index 4c2b2e4269c..ea69572b4f6 100644 --- a/cloud/apps/relay/package.json +++ b/cloud/apps/relay/package.json @@ -9,13 +9,14 @@ "clean": "node -e \"require('fs').rmSync('dist', { recursive: true, force: true })\"", "dev": "tsx watch src/index.ts", "lint": "tsc -p tsconfig.json --noEmit", - "pretest": "pnpm --filter @orca-cloud/relay-contract build", + "pretest": "pnpm --filter @orca-cloud/postgres-schema build && pnpm --filter @orca-cloud/relay-contract build", "start": "node dist/index.js", "test": "vitest run", "typecheck": "tsc -p tsconfig.json --noEmit" }, "dependencies": { "@hono/node-server": "^1.19.14", + "@orca-cloud/postgres-schema": "workspace:*", "@orca-cloud/relay-contract": "workspace:*", "hono": "^4.12.27", "jose": "^6.1.3", diff --git a/cloud/apps/relay/src/postgres-schema-startup.ts b/cloud/apps/relay/src/postgres-schema-startup.ts index ba9efc6a792..3a3428eda32 100644 --- a/cloud/apps/relay/src/postgres-schema-startup.ts +++ b/cloud/apps/relay/src/postgres-schema-startup.ts @@ -1,105 +1 @@ -const RETRYABLE_SCHEMA_CODES = new Set(['55P03', '57014']) -const DEFAULT_RETRY_DEADLINE_MS = 30_000 -const RETRY_BASE_DELAY_MS = 250 -const RETRY_MAX_DELAY_MS = 2_000 - -type SchemaStartupOptions = { - now?: () => number - random?: () => number - retryDeadlineMs?: number - wait?: (delayMs: number) => Promise -} - -function retryDelayMs(attempt: number, random: () => number): number { - const ceiling = Math.min( - RETRY_BASE_DELAY_MS * 2 ** (attempt - 1), - RETRY_MAX_DELAY_MS - ) - return Math.ceil(ceiling * (0.5 + random() * 0.5)) -} - -function wait(delayMs: number): Promise { - return new Promise((resolve) => setTimeout(resolve, delayMs)) -} - -const CREATE_TABLE_IF_NOT_EXISTS = /^\s*CREATE\s+TABLE\s+IF\s+NOT\s+EXISTS\b/i -const CREATE_INDEX_IF_NOT_EXISTS = /^\s*CREATE\s+(?:UNIQUE\s+)?INDEX\s+IF\s+NOT\s+EXISTS\b/i - -// `IF NOT EXISTS` only checks the name before the catalog inserts, so the loser of a concurrent -// CREATE can fail on the catalog unique index (23505) or, when the winner has already committed by -// the time the loser reaches TypeCreate/heap_create_with_catalog, on the name check those routines -// repeat (42710 duplicate type, 42P07 duplicate relation). Each is a no-op on the next attempt. -function concurrentCreateCollision( - value: { code?: unknown; constraint?: unknown }, - statement: string -): boolean { - if (CREATE_TABLE_IF_NOT_EXISTS.test(statement)) { - return ( - (value.code === '23505' && value.constraint === 'pg_type_typname_nsp_index') || - value.code === '42710' || - value.code === '42P07' - ) - } - if (CREATE_INDEX_IF_NOT_EXISTS.test(statement)) { - return ( - (value.code === '23505' && value.constraint === 'pg_class_relname_nsp_index') || - value.code === '42P07' - ) - } - return false -} - -function retryableSchemaError(error: unknown, statement: string): boolean { - const value = error as { code?: unknown; constraint?: unknown } - return ( - RETRYABLE_SCHEMA_CODES.has(String(value.code)) || concurrentCreateCollision(value, statement) - ) -} - -export async function applyPostgresSchema( - statements: string[], - query: (statement: string) => Promise, - options: SchemaStartupOptions = {} -): Promise { - const now = options.now ?? Date.now - const random = options.random ?? Math.random - const pause = options.wait ?? wait - const deadlineAt = now() + (options.retryDeadlineMs ?? DEFAULT_RETRY_DEADLINE_MS) - - for (const statement of statements) { - let attempt = 1 - while (true) { - try { - await query(statement) - break - } catch (error) { - const code = String((error as { code?: unknown }).code) - const remainingMs = deadlineAt - now() - const retryable = retryableSchemaError(error, statement) - if (!retryable || remainingMs <= 0) { - if (retryable) { - console.warn( - JSON.stringify({ - event: 'orca_relay_postgres_schema_retry_exhausted', - code, - attempts: attempt - }) - ) - } - throw error - } - const delayMs = Math.min(remainingMs, retryDelayMs(attempt, random)) - console.warn( - JSON.stringify({ - event: 'orca_relay_postgres_schema_retry', - code, - attempt, - delayMs - }) - ) - await pause(delayMs) - attempt += 1 - } - } - } -} +export { applyPostgresSchema } from '@orca-cloud/postgres-schema' diff --git a/cloud/dev/fixtures/terraform-root-partition/families.json b/cloud/dev/fixtures/terraform-root-partition/families.json index dfe100fd2dd..18ca2c8df4b 100644 --- a/cloud/dev/fixtures/terraform-root-partition/families.json +++ b/cloud/dev/fixtures/terraform-root-partition/families.json @@ -92,11 +92,14 @@ "google_certificate_manager_certificate_map.relay_gce", "google_certificate_manager_certificate_map_entry.relay_gce", "google_certificate_manager_dns_authorization.relay_gce", + "google_cloud_run_domain_mapping.push", "google_cloud_run_domain_mapping.relay", "google_cloud_run_domain_mapping.relay_cell", + "google_cloud_run_v2_service.push", "google_cloud_run_v2_service.relay", "google_cloud_run_v2_service.relay_cell", "google_cloud_run_v2_service.relay_fence_broker", + "google_cloud_run_v2_service_iam_member.github_production_push_developer", "google_cloud_run_v2_service_iam_member.github_production_relay_director_developer", "google_cloud_run_v2_service_iam_member.github_production_relay_fence_broker_developer", "google_cloud_run_v2_service_iam_member.github_staging_relay_capacity_developer", @@ -169,6 +172,9 @@ "google_project_iam_member.github_staging_relay_capacity_viewer", "google_project_iam_member.github_staging_relay_deploy_compute_viewer", "google_project_iam_member.github_staging_relay_power", + "google_project_iam_member.push_runtime_cloudsql_client", + "google_project_iam_member.push_runtime_fcm_admin", + "google_project_iam_member.push_runtime_service_usage_consumer", "google_project_iam_member.relay_director_runtime_cloudsql_client", "google_project_iam_member.relay_fence_broker_artifact_reader", "google_project_iam_member.relay_fence_broker_compute_viewer", @@ -177,9 +183,13 @@ "google_project_iam_member.relay_runtime_artifact_reader", "google_project_iam_member.relay_runtime_cloudsql_client", "google_project_iam_member.relay_runtime_log_writer", + "google_secret_manager_secret.push_database_url", + "google_secret_manager_secret.push_provider", "google_secret_manager_secret.relay_assignment_signing_key", "google_secret_manager_secret.relay_database_url", "google_secret_manager_secret.relay_regional_placement_enabled", + "google_secret_manager_secret_iam_member.push_database_url_runtime_accessor", + "google_secret_manager_secret_iam_member.push_provider_runtime_accessor", "google_secret_manager_secret_iam_member.relay_assignment_signing_key_accessor", "google_secret_manager_secret_iam_member.relay_assignment_signing_key_director_accessor", "google_secret_manager_secret_iam_member.relay_database_url_accessor", @@ -189,6 +199,7 @@ "google_secret_manager_secret_iam_member.relay_regional_placement_deploy_viewer", "google_secret_manager_secret_iam_member.relay_regional_placement_director_accessor", "google_secret_manager_secret_iam_member.relay_regional_placement_runtime_accessor", + "google_secret_manager_secret_version.push_database_url", "google_secret_manager_secret_version.relay_assignment_signing_key", "google_secret_manager_secret_version.relay_database_url", "google_secret_manager_secret_version.relay_regional_placement_enabled", @@ -199,12 +210,15 @@ "google_service_account.github_relay_asia_topology", "google_service_account.github_staging_relay_capacity", "google_service_account.github_staging_relay_deploy", + "google_service_account.push_runtime", "google_service_account.relay_director_runtime", "google_service_account.relay_fence_broker", "google_service_account.relay_runtime", "google_service_account_iam_member.github_accepted_repository_workload_identity_user", "google_service_account_iam_member.github_fence_workload_identity_user", "google_service_account_iam_member.github_monitor_workload_identity_user", + "google_service_account_iam_member.github_production_push_runtime_token_creator", + "google_service_account_iam_member.github_production_push_runtime_user", "google_service_account_iam_member.github_production_relay_capacity_runtime_user", "google_service_account_iam_member.github_production_relay_capacity_workload_identity_user", "google_service_account_iam_member.github_relay_asia_proof_workload_identity_user", @@ -218,7 +232,9 @@ "google_service_account_iam_member.github_staging_relay_deploy_auth_runtime_user", "google_service_account_iam_member.github_staging_relay_deploy_workload_identity_user", "google_service_account_iam_member.relay_fence_broker_requester_token_creator", + "google_sql_database.push", "google_sql_database.relay", + "google_sql_user.push", "google_sql_user.relay", "google_storage_bucket_iam_member.github_production_relay_capacity_state", "google_storage_bucket_iam_member.github_relay_asia_topology_state", @@ -228,6 +244,7 @@ "google_storage_bucket_iam_member.github_staging_relay_deploy_state_list", "google_storage_bucket_iam_member.relay_fence_broker_bucket_reader", "google_storage_bucket_iam_member.relay_fence_broker_state_objects", + "random_password.push_database", "random_password.relay_assignment_signing_key", "random_password.relay_database" ], diff --git a/cloud/dev/scripts/cloud-sql-rollout-lock-census.mjs b/cloud/dev/scripts/cloud-sql-rollout-lock-census.mjs index 76193746f2c..2f7157d823c 100644 --- a/cloud/dev/scripts/cloud-sql-rollout-lock-census.mjs +++ b/cloud/dev/scripts/cloud-sql-rollout-lock-census.mjs @@ -283,6 +283,8 @@ export const LEASED_WORKFLOWS = named([ 'operate-relay-production-rehome.yml', production({ leaseFiles: ['operate-relay-production-rehome-job.yml'] }) ], + // The gateway applies its schema at startup, so its deploy revision is the schema step. + ['push-deploy.yml', production()], ['deploy-relay-asia-topology.yml', eitherEnvironment()], ['operate-relay-asia-admission.yml', eitherEnvironment()], ['deploy-relay-staging.yml', staging()], diff --git a/cloud/dev/scripts/push-gateway-recovery.test.mjs b/cloud/dev/scripts/push-gateway-recovery.test.mjs new file mode 100644 index 00000000000..abed4bc6885 --- /dev/null +++ b/cloud/dev/scripts/push-gateway-recovery.test.mjs @@ -0,0 +1,93 @@ +import assert from 'node:assert/strict' +import { mkdtempSync, rmSync } from 'node:fs' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { spawnSync } from 'node:child_process' +import test from 'node:test' +import { readRelayWorkflow } from './relay-repository.mjs' + +const workflow = readRelayWorkflow('push-deploy.yml') +function step(name) { + const start = workflow.indexOf(` - name: ${name}\n`) + assert.notEqual(start, -1) + const end = workflow.indexOf('\n - name:', start + 1) + const block = workflow.slice(start, end === -1 ? undefined : end) + return block.slice(block.indexOf(' run: |\n') + ' run: |\n'.length) + .split('\n').filter((line) => line.startsWith(' ')).map((line) => line.slice(10)).join('\n') +} +const candidate = step('Deploy the candidate revision with no traffic') +const shift = step('Shift all traffic to the verified candidate') +const rollback = step('Roll traffic back to the previous revision') +const cleanup = step('Delete the rejected candidate revision') +const env = { SERVICE_NAME: 'push-test', GCP_PROJECT_ID: 'test', GCP_REGION: 'test', + GITHUB_RUN_ID: '123', GITHUB_RUN_ATTEMPT: '1', IMAGE: 'synthetic-image', + CANDIDATE_REVISION: 'push-test-c123-1', ROLLBACK_REVISION: 'push-test-old' } + +function exercise(body) { + const dir = mkdtempSync(join(tmpdir(), 'push-workflow-')) + try { + const run = spawnSync('bash', ['-c', body], { encoding: 'utf8', timeout: 10000, + env: { ...process.env, ...env, GITHUB_ENV: join(dir, 'env'), GITHUB_STEP_SUMMARY: join(dir, 'summary'), + TRACE: join(dir, 'trace'), STATE: join(dir, 'state') } }) + assert.equal(run.status, 0, run.stderr) + } finally { rmSync(dir, { recursive: true, force: true }) } +} + +// Workflow shell behavior is Linux-specific; these tests never call a real cloud CLI. +test('failed candidate discovery retains enough state to remove tag and revision', { skip: process.platform === 'win32' }, () => { + exercise(` + gcloud() { + case "$*" in + 'run deploy '*) echo deployed > "$STATE" ;; + 'run services describe '*) return 1 ;; + *) echo "$*" >> "$TRACE" ;; + esac + } + jq() { return 1; } + ( ${candidate} ) + test "$?" != 0 || exit 1 + source "$GITHUB_ENV" + test "$CANDIDATE_TAG" = c123-1 || exit 1 + test "$CANDIDATE_REVISION" = push-test-c123-1 || exit 1 + ( ${cleanup} ) || exit 1 + grep -q -- '--remove-tags c123-1' "$TRACE" || exit 1 + grep -q 'run revisions delete push-test-c123-1' "$TRACE" || exit 1 + `) +}) + +test('failed post-promotion read retains intent and restores previous traffic', { skip: process.platform === 'win32' }, () => { + exercise(` + gcloud() { + case "$*" in + 'run services update-traffic '*) echo "$*" >> "$TRACE" ;; + 'run services describe '*) return 1 ;; + esac + } + jq() { return 1; } + ( ${shift} ) + test "$?" != 0 || exit 1 + source "$GITHUB_ENV" + test "$TRAFFIC_SHIFT_ATTEMPTED" = true || exit 1 + gcloud() { + case "$*" in + 'run services update-traffic '*) echo "$*" >> "$TRACE" ;; + 'run services describe '*) echo '{}' ;; + esac + } + jq() { echo "$ROLLBACK_REVISION"; } + ( ${rollback} ) || exit 1 + source "$GITHUB_ENV" + test "$TRAFFIC_ROLLED_BACK" = true || exit 1 + grep -q -- '--to-revisions push-test-old=100' "$TRACE" || exit 1 + `) +}) + +test('ambiguous promotion failure also leaves rollback intent', { skip: process.platform === 'win32' }, () => { + exercise(` + gcloud() { return 1; } + ( ${shift} ) + test "$?" != 0 || exit 1 + source "$GITHUB_ENV" + test "$TRAFFIC_SHIFT_ATTEMPTED" = true + `) +}) diff --git a/cloud/dev/scripts/push-gateway-workflow.test.mjs b/cloud/dev/scripts/push-gateway-workflow.test.mjs new file mode 100644 index 00000000000..b7c8c7db3fe --- /dev/null +++ b/cloud/dev/scripts/push-gateway-workflow.test.mjs @@ -0,0 +1,299 @@ +import assert from 'node:assert/strict' +import { readFileSync } from 'node:fs' +import test from 'node:test' +import { + concurrencyBlocks, + jobIf, + jobs, + LEASE_ACTION, + leaseSteps +} from './cloud-sql-rollout-lock-census.mjs' +import { readRelayWorkflow, relayWorkflowFile } from './relay-repository.mjs' + +// Why: the push gateway holds the APNs key and is the only thing standing between a paired +// phone and a silent notification pipeline. Its deploy is a blue/green rollout against the +// shared Cloud SQL instance, and each of the guarantees below is one careless edit from gone. +const WORKFLOW = 'push-deploy.yml' +const workflow = readRelayWorkflow(WORKFLOW) +const deploy = () => { + const job = jobs(workflow).find((entry) => entry.id === 'deploy') + assert.ok(job, 'the workflow no longer declares a deploy job') + return job +} + +function terraform(file) { + return readFileSync(new URL(`../../infra/terraform/${file}`, import.meta.url), 'utf8') +} + +// The ordered step names; every assertion below reads positions out of this list rather than +// restating them, so a reordering that breaks the no-traffic guarantee fails here. +const stepNames = () => [...workflow.matchAll(/^ {6}- name: (.+)$/gm)].map((match) => match[1]) + +const indexOfStep = (name) => { + const index = stepNames().indexOf(name) + assert.notEqual(index, -1, `the workflow no longer has a "${name}" step`) + return index +} + +test('the whole surface stays inert until the owner enables cloud operations', () => { + const guard = jobIf(deploy().text) + assert.ok(guard.includes("vars.ORCA_CLOUD_OPERATIONS_ENABLED == 'true'"), guard) + assert.ok(guard.includes("github.ref == 'refs/heads/main'"), guard) + assert.equal(jobs(workflow).length, 1, 'a second job would need its own gate') +}) + +test('it authenticates through Workload Identity and holds no repository secret', () => { + assert.match(workflow, /uses: google-github-actions\/auth@v2/) + assert.match(workflow, /workload_identity_provider: \$\{\{ vars\.PRODUCTION_GCP_RELAY_DEPLOY_WORKLOAD_IDENTITY_PROVIDER \}\}/) + assert.match(workflow, /service_account: \$\{\{ vars\.PRODUCTION_GCP_RELAY_DEPLOY_SERVICE_ACCOUNT \}\}/) + assert.match(workflow, /environment: production/) + for (const [, name] of workflow.matchAll(/secrets\.([A-Za-z_][A-Za-z0-9_]*)/g)) { + assert.equal(name, 'GITHUB_TOKEN', `the workflow reads secrets.${name}`) + } +}) + +// Why: Terraform trusts exact workflow filenames, not a prefix. A rename here without the +// matching tfvars-independent list entry would fail authentication at dispatch time only. +test('Terraform trusts this exact workflow file on the production deploy provider', () => { + assert.match(terraform('relay-github-actions.tf'), /^\s*"push-deploy\.yml"$/m) + assert.equal(relayWorkflowFile(WORKFLOW), 'cloud-push-deploy.yml') +}) + +test('the rollout is serialized and leases the production Cloud SQL rollout lock', () => { + const blocks = concurrencyBlocks(workflow) + assert.equal(blocks.length, 1) + assert.equal(blocks[0].group, 'production-cloud-sql-rollout') + assert.equal(blocks[0].cancelInProgress, 'false') + const steps = leaseSteps(workflow) + assert.equal(steps.length, 1, 'exactly one lease step, held for the whole run') + assert.equal(steps[0].bucket, 'onorca-cloud-terraform-state') + assert.equal(steps[0].object, 'terraform/state/cloud-sql-rollout/production.lock') + assert.equal(steps[0].release, undefined, 'release stays at its default for a single-job run') +}) + +// Why: the ops guardrail is that a piped command only fails the step when pipefail is set, and +// pipefail only applies under an explicit bash shell. Every multi-line body here opts in. +test('every multi-line command runs under bash with pipefail', () => { + const bodies = [...workflow.matchAll(/^ {8}(shell: bash\n {8})?run: \|\n((?: {10}.*\n|\n)+)/gm)] + assert.ok(bodies.length >= 8, `only ${bodies.length} multi-line commands were found`) + for (const match of bodies) { + assert.ok(match[1], `a multi-line command does not declare shell: bash:\n${match[2].slice(0, 120)}`) + assert.match(match[2], /^ {10}set -euo pipefail$/m) + } +}) + +test('the candidate revision takes no traffic and is addressed by its own tag', () => { + assert.match(workflow, /gcloud run deploy "\$\{SERVICE_NAME\}"/) + assert.match(workflow, /^ {12}--no-traffic \\$/m) + assert.match(workflow, /--tag "\$\{tag\}"/) + assert.match(workflow, /test "\$\{CANDIDATE_REVISION\}" != "\$\{ROLLBACK_REVISION\}"/) + assert.ok( + indexOfStep('Record the serving revision and require its Terraform-owned scaling') < + indexOfStep('Deploy the candidate revision with no traffic'), + 'the rollback target must be captured before the candidate exists' + ) +}) + +// Why: scaling is a Terraform-owned field that `lifecycle.ignore_changes` does not cover, so a +// deploy that passed --max-instances would revert a later push_max_instances raise on every run. +// The workflow asserts the shape instead of writing it, on the serving revision before the +// candidate exists and on the candidate that inherits it. +test('the deploy asserts the Terraform-owned scaling instead of mutating it', () => { + assert.doesNotMatch(workflow, /--max-instances/, 'the deploy must not write a scaling field') + assert.doesNotMatch(workflow, /--min-instances "/, 'the deploy must not write a scaling field') + // The floor is the variables.tf default; production.tfvars overrides only the ceiling, down to + // the two instances the Cloud SQL connection budget leaves room for. + assert.match(workflow, /PUSH_MIN_INSTANCES: 1$/m) + assert.match(workflow, /PUSH_MAX_INSTANCES: 2$/m) + assert.match(terraform('variables.tf'), /variable "push_min_instances"[\s\S]*?default {5}= 1/) + assert.match(terraform('environments/production.tfvars'), /^push_max_instances {9}= 2$/m) + const gate = indexOfStep('Record the serving revision and require its Terraform-owned scaling') + assert.ok(gate < indexOfStep('Deploy the candidate revision with no traffic')) + assert.match(workflow, /autoscaling\.knative\.dev\/minScale/) + assert.match(workflow, /\[\[ "\$\{floor:-0\}" -lt "\$\{PUSH_MIN_INSTANCES\}" \]\]/) + assert.match(workflow, /test "\$\{ceiling\}" = "\$\{PUSH_MAX_INSTANCES\}"/) + assert.match(workflow, /test "\$\{candidate_ceiling\}" = "\$\{PUSH_MAX_INSTANCES\}"/) +}) + +// Why: the image build is not a Cloud SQL operation, and the lease is a global serialization +// point. A build inside it blocks every relay deploy and rehome for its duration. +test('the image is built before the rollout lease is taken', () => { + const lease = workflow.indexOf(`- uses: ${LEASE_ACTION}`) + assert.notEqual(lease, -1) + const build = workflow.indexOf('- name: Build and publish the immutable gateway image') + const deployCandidate = workflow.indexOf('- name: Deploy the candidate revision with no traffic') + assert.ok(build < lease, 'the build must finish before the run takes the lease') + assert.ok(lease < deployCandidate, 'the lease must still cover the deploy, probe, and shift') +}) + +// Why: the gateway's Cloud SQL draw is instances x pool, and the root that takes the rollout +// lease can only account for a pool it declares. Leaving it at the application default hid it. +test('the database pool size is Terraform-owned and bounded at plan time', () => { + const source = terraform('push-gateway.tf') + assert.match(source, /name {2}= "ORCA_PUSH_DATABASE_POOL_MAX"/) + assert.match(source, /value = tostring\(var\.push_database_pool_max\)/) + assert.match(terraform('variables.tf'), /variable "push_database_pool_max"[\s\S]*?default {5}= 2/) + const block = /resource "google_cloud_run_v2_service" "push"[\s\S]*?\n lifecycle \{([\s\S]*?)\n \}/.exec(source) + assert.ok(block, 'the push service no longer declares a lifecycle block') + assert.match( + block[1], + /var\.push_max_instances \* var\.push_database_pool_max <= 4/, + 'instances x pool must be bounded at plan time' + ) + assert.match( + readFileSync(new URL('../../apps/push/src/config.ts', import.meta.url), 'utf8'), + /ORCA_PUSH_DATABASE_POOL_MAX/, + 'the gateway must read the variable Terraform sets' + ) +}) + +test('the candidate is probed on its own URL before any traffic moves', () => { + const probe = indexOfStep('Probe the candidate readiness endpoint') + assert.ok(probe > indexOfStep('Deploy the candidate revision with no traffic')) + assert.ok(probe < indexOfStep('Shift all traffic to the verified candidate')) + assert.match(workflow, /"\$\{CANDIDATE_URL\}\/ready"/) + assert.match(workflow, /test "\$\{code\}" = 200/) + assert.doesNotMatch(workflow, /\$\{CANDIDATE_URL\}\/health/, 'liveness is not readiness') +}) + +// Why: a gateway that answers /ready can still hold no usable FCM credential. The probe must be +// validate-only, must use a token that cannot exist, and must treat a denied credential as the +// failure. Accepting PERMISSION_DENIED would make the whole step decorative. +test('the FCM probe is validate-only and separates a bad token from a bad credential', () => { + const fcm = indexOfStep('Prove the runtime identity can reach FCM') + assert.ok(fcm > indexOfStep('Probe the candidate readiness endpoint')) + assert.ok(fcm < indexOfStep('Shift all traffic to the verified candidate')) + assert.match(workflow, /"validate_only":true/) + assert.match(workflow, /https:\/\/fcm\.googleapis\.com\/v1\/projects\/\$\{GCP_PROJECT_ID\}\/messages:send/) + assert.match(workflow, /GCP_PROJECT_ID: onorca-cloud$/m) + assert.match(workflow, /orca-push-deploy-probe-invalid-token/) + assert.match(workflow, /test "\$\{status\}" = INVALID_ARGUMENT/) + assert.match(workflow, /test "\$\{status\}" = PERMISSION_DENIED/) + // Only those four answers are conclusive; a 429 or a 5xx says nothing about the credential, so + // it is retried rather than read as either verdict. A denied credential still fails at once. + assert.match(workflow, /for attempt in \$\(seq 1 5\); do/) + const probe = workflow.slice( + workflow.indexOf('- name: Prove the runtime identity can reach FCM'), + workflow.indexOf('- name: Shift all traffic to the verified candidate') + ) + assert.match(probe, /for attempt in \$\(seq 1 5\); do/) + assert.match(probe, /test "\$\{code\}" = 401 \|\| test "\$\{code\}" = 403; then\n {14}break/) + assert.match( + workflow, + /--impersonate-service-account "\$\{PUSH_RUNTIME_SERVICE_ACCOUNT\}"/, + 'the probe must exercise the runtime credential, not the deploy identity' + ) + // Why: that token reads the Apple signing key. Masking it means a later `set -x` or a + // debug re-run cannot print it into a public log. + assert.match( + probe, + /test -n "\$\{token\}"\n {10}echo "::add-mask::\$\{token\}"/, + 'the impersonated token must be masked before anything else runs' + ) + assert.match(workflow, /PUSH_RUNTIME_SERVICE_ACCOUNT: orca-cloud-push@onorca-cloud\.iam\.gserviceaccount\.com/) +}) + +// Why: a deploy ends with traffic pinned to an exact revision, and a rollback pins it to the +// previous one. Terraform reverting the service to 100% LATEST would undo either silently. +test('Terraform does not own the image or the traffic split', () => { + const source = terraform('push-gateway.tf') + const block = /resource "google_cloud_run_v2_service" "push"[\s\S]*?\n lifecycle \{([\s\S]*?)\n \}/.exec(source) + assert.ok(block, 'the push service no longer declares a lifecycle block') + assert.match(block[1], /template\[0\]\.containers\[0\]\.image/) + assert.match(block[1], /^\s*traffic$/m) +}) + +test('impersonating the runtime identity is a Terraform-declared grant', () => { + const source = terraform('push-gateway.tf') + assert.match(source, /resource "google_service_account_iam_member" "github_production_push_runtime_token_creator"/) + assert.match(source, /role\s+= "roles\/iam\.serviceAccountTokenCreator"/) + assert.match(source, /resource "google_cloud_run_v2_service_iam_member" "github_production_push_developer"/) +}) + +test('the traffic shift is all-or-nothing and is verified after the fact', () => { + const shift = indexOfStep('Shift all traffic to the verified candidate') + assert.match(workflow, /gcloud run services update-traffic "\$\{SERVICE_NAME\}"/) + assert.match(workflow, /--to-revisions "\$\{CANDIDATE_REVISION\}=100"/) + assert.match(workflow, /test "\$\{serving\}" = "\$\{CANDIDATE_REVISION\}"/) + assert.ok(shift < indexOfStep('Verify the public origin after the shift')) + assert.match(workflow, /PUSH_ORIGIN: https:\/\/push\.onorca\.dev/) + assert.match(workflow, /"\$\{PUSH_ORIGIN\}\/ready"/) +}) + +// Why: the origin can lag the traffic move by seconds, and a single unlucky curl would otherwise +// roll a healthy deploy back. It retries on the same schedule as the candidate probe. +test('the post-shift origin check retries like the candidate probe', () => { + const check = workflow.slice( + workflow.indexOf('- name: Verify the public origin after the shift'), + workflow.indexOf('- name: Roll traffic back to the previous revision') + ) + assert.match(check, /for attempt in \$\(seq 1 30\); do/) + assert.match(check, /sleep 5/) + assert.match(check, /test "\$\{code\}" = 200/) +}) + +// Why: the summary carries the rollback target. Writing it after the origin check meant the one +// run that needed it, the run whose check failed, was the one run that never got it. +test('the summary is written before anything that can fail after the shift', () => { + const summary = indexOfStep('Publish the rollout summary') + assert.ok(summary > indexOfStep('Shift all traffic to the verified candidate')) + assert.ok(summary < indexOfStep('Verify the public origin after the shift')) + assert.match(workflow, /--to-revisions \$\{ROLLBACK_REVISION\}=100/) + assert.match(workflow, /GITHUB_STEP_SUMMARY/) +}) + +// Why: everything after the shift runs with production on the candidate, so a failure there is a +// live gateway that has to go back. The marker is what separates that case from a failure before +// the shift, where production never moved and the candidate is the thing to clean up. +test('a failure after the shift rolls production back automatically', () => { + const rollback = indexOfStep('Roll traffic back to the previous revision') + assert.ok(rollback > indexOfStep('Verify the public origin after the shift')) + assert.match(workflow, /echo "TRAFFIC_SHIFTED=true" >> "\$\{GITHUB_ENV\}"/) + const shift = workflow.indexOf('- name: Shift all traffic to the verified candidate') + assert.ok( + workflow.indexOf('echo "TRAFFIC_SHIFTED=true"') > shift, + 'the success marker follows the shift step' + ) + const body = workflow.slice( + workflow.indexOf('- name: Roll traffic back to the previous revision'), + workflow.indexOf('- name: Delete the rejected candidate revision') + ) + assert.match( + body, + /if: \$\{\{ \(failure\(\) \|\| cancelled\(\)\) && env\.TRAFFIC_SHIFT_ATTEMPTED == 'true' \}\}/, + 'the rollback must be conditioned on both failure and the shift marker' + ) + assert.match(body, /test -n "\$\{ROLLBACK_REVISION:-\}"/) + assert.match(body, /--to-revisions "\$\{ROLLBACK_REVISION\}=100"/) + assert.match(body, /test "\$\{serving\}" = "\$\{ROLLBACK_REVISION\}"/) + assert.match(body, /GITHUB_STEP_SUMMARY/, 'the rollback must be reported in the summary') +}) + +// Why: a candidate that never took traffic still holds a warm instance and a Cloud SQL pool. Its +// tag comes off first, because Cloud Run refuses to delete a revision a traffic target names. +test('a failure before the shift deletes the candidate it created', () => { + const body = workflow.slice( + workflow.indexOf('- name: Delete the rejected candidate revision'), + workflow.indexOf('- name: Drop the candidate traffic tag') + ) + assert.match( + body, + /env\.TRAFFIC_SHIFT_ATTEMPTED != 'true' \|\| env\.TRAFFIC_ROLLED_BACK == 'true'/, + 'the cleanup must be conditioned on both failure and the absence of the shift marker' + ) + assert.match(body, /test -n "\$\{CANDIDATE_REVISION:-\}" \|\| exit 0/) + assert.ok( + body.indexOf('--remove-tags') < body.indexOf('gcloud run revisions delete'), + 'the tag must come off before the revision is deleted' + ) + assert.match(body, /echo "CANDIDATE_TAG=" >> "\$\{GITHUB_ENV\}"/) +}) + +test('the run always drops its traffic tag', () => { + const cleanup = indexOfStep('Drop the candidate traffic tag') + assert.equal(cleanup, stepNames().length - 1, 'tag cleanup must be the last step') + assert.match(workflow, /--remove-tags "\$\{CANDIDATE_TAG\}"/) + const body = workflow.slice(workflow.indexOf('- name: Drop the candidate traffic tag')) + assert.match(body, /if: always\(\)/) + assert.match(body, /test -n "\$\{CANDIDATE_TAG:-\}" \|\| exit 0/) +}) diff --git a/cloud/dev/scripts/relay-cloud-sql-connection-budget.mjs b/cloud/dev/scripts/relay-cloud-sql-connection-budget.mjs index 79036918f23..6965986845c 100644 --- a/cloud/dev/scripts/relay-cloud-sql-connection-budget.mjs +++ b/cloud/dev/scripts/relay-cloud-sql-connection-budget.mjs @@ -33,6 +33,13 @@ function requiredInteger(source, pattern, label) { return value } +// A tfvars file states only what it overrides, so an absent key means the variable default holds. +// Reading the default as the fallback keeps this honest either way. +function overriddenInteger(override, overridePattern, source, pattern, label) { + if (!overridePattern.test(override)) return requiredInteger(source, pattern, label) + return requiredInteger(override, overridePattern, label) +} + function productionCells(source, defaultPoolMax) { const fencedMatch = source.match(/relay_gce_fenced_cells\s*=\s*\[([^\]]*)\]/) if (!fencedMatch) throw new Error('could not read fenced Relay cells') @@ -52,11 +59,13 @@ function productionCells(source, defaultPoolMax) { } export function calculateRelayCloudSqlConnectionBudget(inputs) { + const pushDraw = inputs.pushInstances * inputs.pushPoolMax const consumers = { cells: inputs.cellPoolTotal + inputs.asiaCellCount * inputs.asiaPoolMax, directors: inputs.directorInstances * inputs.directorPoolMax, auth: inputs.authInstances * inputs.authPoolMax, - api: inputs.apiInstances * inputs.apiPoolMax + api: inputs.apiInstances * inputs.apiPoolMax, + push: pushDraw } const configuredMaximum = Object.values(consumers).reduce((total, value) => total + value, 0) const retainedDirectorRollback = inputs.directorInstances * inputs.directorPoolMax @@ -64,6 +73,11 @@ export function calculateRelayCloudSqlConnectionBudget(inputs) { relayDirectorCandidate: retainedDirectorRollback * 2, apiCandidate: retainedDirectorRollback + inputs.apiInstances * inputs.apiPoolMax, authCandidate: retainedDirectorRollback + inputs.authInstances * inputs.authPoolMax, + // The push candidate doubles rather than adding one copy, like the director candidate and + // unlike the API and auth ones: cloud-push-deploy.yml probes a *tagged* revision, which is + // directly addressable and so sits outside the service-wide instance cap, letting the + // candidate and the serving revision each reach push_max_instances at the same time. + pushCandidate: retainedDirectorRollback + pushDraw * 2, relayCells: retainedDirectorRollback } const rolloutOverlap = Math.max(...Object.values(candidateOverlap)) @@ -131,6 +145,20 @@ export function readRelayCloudSqlConnectionBudget({ /variable\s+"relay_director_database_pool_max"[\s\S]*?default\s*=\s*(\d+)/, 'director pool maximum' ), + // The mobile push gateway shares this instance. Its draw was invisible here until Terraform + // declared the pool: docs/push-gateway.md, "Shape". + pushInstances: overriddenInteger( + productionTfvars, + /^\s*push_max_instances\s*=\s*(\d+)/m, + terraformVariables, + /variable\s+"push_max_instances"[\s\S]*?default\s*=\s*(\d+)/, + 'push gateway instances' + ), + pushPoolMax: requiredInteger( + terraformVariables, + /variable\s+"push_database_pool_max"[\s\S]*?default\s*=\s*(\d+)/, + 'push gateway pool maximum' + ), authInstances: apps.authInstances, authPoolMax: apps.authPoolMax, apiInstances: apps.apiInstances, diff --git a/cloud/dev/scripts/relay-cloud-sql-connection-budget.test.mjs b/cloud/dev/scripts/relay-cloud-sql-connection-budget.test.mjs index a26d24c274d..4e3536c0e2b 100644 --- a/cloud/dev/scripts/relay-cloud-sql-connection-budget.test.mjs +++ b/cloud/dev/scripts/relay-cloud-sql-connection-budget.test.mjs @@ -6,28 +6,91 @@ import { readRelayCloudSqlConnectionBudget } from './relay-cloud-sql-connection-budget.mjs' -test('production plus three Asia pools preserves allowance and reserve below the ceiling', () => { +// Why these numbers are this tight: the shared instance's 400 connections were already spoken +// for, and the relay shape below leaves exactly five. The gateway is sized to fit in four, two +// instances times a two-connection pool, and its rollout overlap of 23 stays under the API +// candidate's 65, so the Math.max is the API candidate rather than the gateway. +// +// `Deploy Relay Asia Topology` gates on `withinBudget == true`, so the single remaining +// connection is the whole margin. Anything that raises a pool or an instance count moves it. +test('production plus the push gateway keeps allowance and reserve below the ceiling', () => { const report = readRelayCloudSqlConnectionBudget() - assert.deepEqual(report.consumers, { cells: 230, directors: 15, auth: 20, api: 50 }) + assert.deepEqual(report.consumers, { cells: 230, directors: 15, auth: 20, api: 50, push: 4 }) assert.deepEqual(report.asia, { cells: 3, poolMax: 10 }) - assert.equal(report.configuredMaximum, 315) + assert.equal(report.configuredMaximum, 319) assert.equal(report.rolloutOverlap.relayDirectorCandidate, 30) assert.equal(report.rolloutOverlap.apiCandidate, 65) assert.equal(report.rolloutOverlap.authCandidate, 35) + assert.equal(report.rolloutOverlap.pushCandidate, 23) assert.equal(report.rolloutOverlap.relayCells, 15) assert.equal(report.rolloutOverlap.retainedDirectorRollback, 15) + // The gateway does not set the maximum; the API candidate does, as it did before it existed. assert.equal(report.rolloutOverlap.maximum, 65) assert.equal(report.maintenanceAdminAllowance, 5) assert.equal(report.explicitReserve, 10) assert.equal(report.usableCeiling, 390) + assert.equal(report.operatingMaximum, 389) + assert.equal(report.remainingWithinUsableCeiling, 1) + assert.equal(report.budgetedTotal, 399) + assert.equal(report.unallocated, 1) + assert.equal(report.withinBudget, true) +}) + +// Why: the same relay shape without a push gateway is the before picture, and it stood at five +// connections clear. Holding it here keeps the gateway's cost visible as the four it takes, +// rather than letting drift elsewhere in the budget hide inside the same margin. +test('the same relay shape without the gateway stays inside the ceiling', () => { + const report = calculateRelayCloudSqlConnectionBudget({ + cellPoolTotal: 200, + asiaCellCount: 3, + asiaPoolMax: 10, + directorInstances: 5, + directorPoolMax: 3, + authInstances: 2, + authPoolMax: 10, + apiInstances: 10, + apiPoolMax: 5, + pushInstances: 0, + pushPoolMax: 0, + maxConnections: 400, + maintenanceAdminAllowance: 5, + explicitReserve: 10 + }) + + assert.equal(report.consumers.push, 0) + assert.equal(report.rolloutOverlap.maximum, 65) assert.equal(report.operatingMaximum, 385) assert.equal(report.remainingWithinUsableCeiling, 5) - assert.equal(report.budgetedTotal, 395) - assert.equal(report.unallocated, 5) assert.equal(report.withinBudget, true) }) +// Why: a tagged candidate is directly addressable and sits outside the service-wide cap, so both +// push revisions can reach the ceiling at once. The API and auth candidates add one copy; this +// one adds two, like the director candidate. +test('the push rollout scenario doubles the gateway draw over the retained director', () => { + const report = calculateRelayCloudSqlConnectionBudget({ + cellPoolTotal: 0, + asiaCellCount: 0, + asiaPoolMax: 0, + directorInstances: 5, + directorPoolMax: 3, + authInstances: 0, + authPoolMax: 0, + apiInstances: 0, + apiPoolMax: 0, + pushInstances: 2, + pushPoolMax: 2, + maxConnections: 400, + maintenanceAdminAllowance: 5, + explicitReserve: 10 + }) + + assert.equal(report.consumers.push, 4) + // 15 retained director rollback, plus the 4-connection draw counted twice. + assert.equal(report.rolloutOverlap.pushCandidate, 23) +}) + test('fails closed when pool growth consumes the explicit reserve', () => { const report = calculateRelayCloudSqlConnectionBudget({ cellPoolTotal: 200, @@ -39,12 +102,14 @@ test('fails closed when pool growth consumes the explicit reserve', () => { authPoolMax: 10, apiInstances: 20, apiPoolMax: 5, + pushInstances: 4, + pushPoolMax: 10, maxConnections: 400, maintenanceAdminAllowance: 5, explicitReserve: 10 }) - assert.equal(report.operatingMaximum, 515) + assert.equal(report.operatingMaximum, 555) assert.equal(report.withinBudget, false) }) @@ -63,7 +128,11 @@ test('excludes fenced cell pools and reads per-cell pool overrides', () => { } } `, - terraformVariables: 'variable "relay_director_database_pool_max" { default = 3 }', + terraformVariables: [ + 'variable "relay_director_database_pool_max" { default = 3 }', + 'variable "push_max_instances" { default = 1 }', + 'variable "push_database_pool_max" { default = 2 }' + ].join('\n'), relayConfig: 'export const RELAY_DATABASE_POOL_MAX = 10' }, maxConnections: 100, @@ -72,8 +141,42 @@ test('excludes fenced cell pools and reads per-cell pool overrides', () => { }) assert.equal(report.consumers.cells, 14) - assert.equal(report.operatingMaximum, 46) - assert.equal(report.budgetedTotal, 47) + // No push_max_instances in this tfvars, so the variable default of one instance holds. + assert.equal(report.consumers.push, 2) + assert.equal(report.operatingMaximum, 48) + assert.equal(report.budgetedTotal, 49) +}) + +// Why: production.tfvars overrides push_max_instances down to 2 while variables.tf still defaults +// to 4, so reading the default instead of the override would overstate the live draw by half. +test('a tfvars push_max_instances override wins over the variable default', () => { + const report = readRelayCloudSqlConnectionBudget({ + proposedAsiaCellCount: 1, + appConsumers: { authInstances: 1, authPoolMax: 10, apiInstances: 1, apiPoolMax: 5, maxConnections: 100 }, + sources: { + productionTfvars: ` + relay_max_instances = 1 + push_max_instances = 3 + relay_gce_fenced_cells = [] + relay_gce_cells = { + "production-gce-c2" = { database_pool_max = 4 + } + } + `, + terraformVariables: [ + 'variable "relay_director_database_pool_max" { default = 3 }', + 'variable "push_max_instances" { default = 1 }', + 'variable "push_database_pool_max" { default = 2 }' + ].join('\n'), + relayConfig: 'export const RELAY_DATABASE_POOL_MAX = 10' + }, + maxConnections: 100, + maintenanceAdminAllowance: 1, + explicitReserve: 1 + }) + + assert.equal(report.consumers.push, 6) + assert.equal(report.rolloutOverlap.pushCandidate, 15) }) test('requires strict headroom below the physical ceiling', () => { @@ -87,12 +190,14 @@ test('requires strict headroom below the physical ceiling', () => { authPoolMax: 10, apiInstances: 1, apiPoolMax: 5, + pushInstances: 1, + pushPoolMax: 2, maxConnections: 50, maintenanceAdminAllowance: 9, explicitReserve: 3 }) - assert.equal(report.budgetedTotal, 63) + assert.equal(report.budgetedTotal, 65) assert.equal(report.withinBudget, false) }) diff --git a/cloud/dev/scripts/relay-production-identity-boundaries.test.mjs b/cloud/dev/scripts/relay-production-identity-boundaries.test.mjs index 7e8ea2a05c1..f97e742215b 100644 --- a/cloud/dev/scripts/relay-production-identity-boundaries.test.mjs +++ b/cloud/dev/scripts/relay-production-identity-boundaries.test.mjs @@ -32,7 +32,8 @@ test('no workflow names the retired generic production deploy identity', async ( 'deploy-relay-production.yml', 'operate-relay-asia-admission.yml', 'operate-relay-production-rehome-job.yml', - 'publish-relay-production.yml' + 'publish-relay-production.yml', + 'push-deploy.yml' ].map((name) => relayWorkflowFile(name)).sort()) }) diff --git a/cloud/dev/scripts/relay-public-workflow-contract.test.mjs b/cloud/dev/scripts/relay-public-workflow-contract.test.mjs index 56393d07bd1..d25ffb221f4 100644 --- a/cloud/dev/scripts/relay-public-workflow-contract.test.mjs +++ b/cloud/dev/scripts/relay-public-workflow-contract.test.mjs @@ -20,7 +20,7 @@ const UNGATED = relayWorkflowFile('verify.yml') const relayWorkflows = () => workflowFiles().filter((file) => file !== UNGATED) test('the copy carries every relay workflow', () => { - assert.equal(relayWorkflows().length, 24) + assert.equal(relayWorkflows().length, 25) }) // Why: workflow_run chains match by display name, not filename. Renaming a file is safe; renaming diff --git a/cloud/dev/scripts/workload-identity-attribute-conditions.test.mjs b/cloud/dev/scripts/workload-identity-attribute-conditions.test.mjs index 1d3f3ce4d79..2dd65e1b6f4 100644 --- a/cloud/dev/scripts/workload-identity-attribute-conditions.test.mjs +++ b/cloud/dev/scripts/workload-identity-attribute-conditions.test.mjs @@ -31,7 +31,7 @@ const EXPECTED_CONDITIONS = { production: { relay: { github: - "assertion.repository == 'stablyai/orca' && assertion.repository_id == '1183888342' && assertion.repository_owner_id == '127256420' && assertion.ref == 'refs/heads/main' && assertion.environment == 'production' && ((assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-fence-broker.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-capacity.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-director.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-multi-target.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-operate-relay-asia-admission.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-publish-relay-production.yml@refs/heads/main') || (assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-operate-relay-production-rehome.yml@refs/heads/main' && assertion.job_workflow_ref == 'stablyai/orca/.github/workflows/cloud-operate-relay-production-rehome-job.yml@refs/heads/main') || (assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-same-cap.yml@refs/heads/main' && (assertion.job_workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-same-cap-job.yml@refs/heads/main' || assertion.job_workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-same-cap.yml@refs/heads/main')))", + "assertion.repository == 'stablyai/orca' && assertion.repository_id == '1183888342' && assertion.repository_owner_id == '127256420' && assertion.ref == 'refs/heads/main' && assertion.environment == 'production' && ((assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-fence-broker.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-capacity.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-director.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-multi-target.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-operate-relay-asia-admission.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-publish-relay-production.yml@refs/heads/main' || assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-push-deploy.yml@refs/heads/main') || (assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-operate-relay-production-rehome.yml@refs/heads/main' && assertion.job_workflow_ref == 'stablyai/orca/.github/workflows/cloud-operate-relay-production-rehome-job.yml@refs/heads/main') || (assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-same-cap.yml@refs/heads/main' && (assertion.job_workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-same-cap-job.yml@refs/heads/main' || assertion.job_workflow_ref == 'stablyai/orca/.github/workflows/cloud-deploy-relay-production-same-cap.yml@refs/heads/main')))", github_monitor: "assertion.repository == 'stablyai/orca' && assertion.repository_id == '1183888342' && assertion.repository_owner_id == '127256420' && assertion.ref == 'refs/heads/main' && assertion.environment == 'production' && assertion.workflow_ref == 'stablyai/orca/.github/workflows/cloud-monitor-relay-production.yml@refs/heads/main' && assertion.job_workflow_ref == 'stablyai/orca/.github/workflows/cloud-monitor-relay-production-job.yml@refs/heads/main'", github_fence: diff --git a/cloud/docs/push-gateway.md b/cloud/docs/push-gateway.md new file mode 100644 index 00000000000..f373c7a2bfb --- /dev/null +++ b/cloud/docs/push-gateway.md @@ -0,0 +1,337 @@ +# Orca mobile push gateway + +`orca-cloud-push` is a public Cloud Run service in `onorca-cloud` that turns a desktop +notification into an APNs or FCM push for a paired phone. The desktop registers each phone's +native token with it and calls `POST /v1/send` after the socket fan-out it already does; the +phone dedupes by `notificationId#notificationSeq`. The service is the only place the Apple +`.p8` signing key is readable, which is the reason it exists as a service at all. + +The contract every lane builds against is `docs/reference/mobile-push-contract.md` in the +repository root. This document covers only the deploy surface: what Terraform owns, how the +credentials rotate, and what the other repository still has to publish. + +**There is no staging push gateway.** That is a decision, not an omission. `push_gateway_enabled` +is false in `environments/staging.tfvars` and true in `environments/production.tfvars`, and every +resource in `infra/terraform/push-gateway.tf` is behind it. A staging gateway would be a tfvars +edit plus a second set of Apple credentials. + +## Shape + +| Setting | Value | Where | +| --- | --- | --- | +| Cloud Run service | `orca-cloud-push` | `push_cloud_run_service_name` | +| Region | `us-central1` | `region` | +| Instances | min 1, max 2 | `push_min_instances`, `push_max_instances` | +| Database pool | 2 per instance | `push_database_pool_max` | +| Concurrency | 80 | `push_concurrency` | +| Ingress | all | `INGRESS_TRAFFIC_ALL` | +| Invoker | IAM disabled | `invoker_iam_disabled = true` on the service | +| Runtime identity | `orca-cloud-push@onorca-cloud.iam.gserviceaccount.com` | `google_service_account.push_runtime` | +| Database | `orca_push` on the shared Cloud SQL instance | `google_sql_database.push` | +| Hostname | `push.onorca.dev` | `push_base_url` | + +The minimum of one instance is deliberate and did not move when the ceiling came down to two. A +cold start delays a notification past the point where it is worth showing, and the three-second +coalescing window lives in instance memory, so the floor is what keeps a notification prompt. The +ceiling is a different question, answered below. + +The maximum and the pool are set by the connection budget, not by the gateway's own appetite. Two +instances times a two-connection pool is a draw of 4, and a rollout doubles it to 8, because the +tagged candidate is directly addressable and sits outside the service-wide cap. The shared Cloud +SQL instance's 400 connections were already spoken for by the relay cells, the directors, auth, +and the API, which left five. Four is the whole of the room there was, and the gateway fits in +it. + +Two connections per instance is enough for the work. A send runs two or three short queries, so +at concurrency 80 requests queue against the pool for microseconds rather than holding it. A +`lifecycle` precondition refuses a plan whose instances times pool exceeds 4, because a fifth +connection puts the checked budget over its ceiling and blocks `Deploy Relay Asia Topology`, +which gates on it. `dev/scripts/relay-cloud-sql-connection-budget.mjs` counts the gateway and +prints the whole picture. + +Authentication is the host proof in `POST /v1/host/challenge`, not Cloud Run IAM, so the service +opts out of invoker IAM with `invoker_iam_disabled = true`, exactly as the relay director does. +The project's domain-restricted-sharing policy refuses an `allUsers` invoker binding, so that is +the only way to reach an open service here. + +## Environment + +Set on the container by Terraform: + +| Variable | Source | +| --- | --- | +| `PORT` | Cloud Run, container port 8080 | +| `ORCA_PUSH_PUBLIC_URL` | `push_base_url` | +| `ORCA_PUSH_FCM_PROJECT_ID` | `push_fcm_project_id`, empty means `project_id` | +| `ORCA_PUSH_DATABASE_URL` | Secret `orca-cloud-push-database-url`, version `latest` | +| `ORCA_PUSH_DATABASE_POOL_MAX` | `push_database_pool_max`, 2 per instance | +| `ORCA_PUSH_APNS_KEY` | Secret `orca-cloud-push-apns-key`, version `latest` | +| `ORCA_PUSH_APNS_KEY_ID` | Secret `orca-cloud-push-apns-key-id`, version `latest` | +| `ORCA_PUSH_APPLE_TEAM_ID` | Secret `orca-cloud-push-apple-team-id`, version `latest` | + +`ORCA_PUSH_APNS_TOPIC` and `ORCA_PUSH_COALESCE_MS` are left to their application defaults +(`com.stably.orca.mobile` and `3000`). Add them here only when one of them has to differ from +the code default, so that a code-side change stays visible rather than silently overridden. + +Terraform owns the three Apple secret **names, labels, and replication, and never a version.** +The `.p8` is issued by the Apple developer portal, so a Terraform-managed version would put the +private key in state and would fight the rotation below. The database URL secret is different: +Terraform generates that password, so it owns that version, exactly as `relay-database.tf` does. +That puts the generated password and the full database URL in the state bucket, which the shared +deploy identity can read; the Apple key never appears there. The three Apple secrets and the +`orca_push` database carry `prevent_destroy`, so disabling the gateway fails the plan instead +of deleting the only copy of the signing key or every live device token. + +## Importing what already exists + +The runtime account, the three Apple secrets, and their accessor bindings were created out of +band alongside the Apple credentials. They are declared so a plan is clean, and imported once. +Run these from `cloud/` after `pnpm infra:init --env production`, review the resulting plan, and +expect the imported resources to show no changes. + +```sh +terraform -chdir=infra/terraform import -var-file=environments/production.tfvars \ + 'google_service_account.push_runtime[0]' \ + projects/onorca-cloud/serviceAccounts/orca-cloud-push@onorca-cloud.iam.gserviceaccount.com + +terraform -chdir=infra/terraform import -var-file=environments/production.tfvars \ + 'google_project_iam_member.push_runtime_fcm_admin[0]' \ + 'onorca-cloud roles/firebasecloudmessaging.admin serviceAccount:orca-cloud-push@onorca-cloud.iam.gserviceaccount.com' + +terraform -chdir=infra/terraform import -var-file=environments/production.tfvars \ + 'google_project_iam_member.push_runtime_service_usage_consumer[0]' \ + 'onorca-cloud roles/serviceusage.serviceUsageConsumer serviceAccount:orca-cloud-push@onorca-cloud.iam.gserviceaccount.com' + +terraform -chdir=infra/terraform import -var-file=environments/production.tfvars \ + 'google_secret_manager_secret.push_provider["orca-cloud-push-apns-key"]' \ + projects/onorca-cloud/secrets/orca-cloud-push-apns-key + +terraform -chdir=infra/terraform import -var-file=environments/production.tfvars \ + 'google_secret_manager_secret.push_provider["orca-cloud-push-apns-key-id"]' \ + projects/onorca-cloud/secrets/orca-cloud-push-apns-key-id + +terraform -chdir=infra/terraform import -var-file=environments/production.tfvars \ + 'google_secret_manager_secret.push_provider["orca-cloud-push-apple-team-id"]' \ + projects/onorca-cloud/secrets/orca-cloud-push-apple-team-id + +terraform -chdir=infra/terraform import -var-file=environments/production.tfvars \ + 'google_secret_manager_secret_iam_member.push_provider_runtime_accessor["orca-cloud-push-apns-key"]' \ + 'projects/onorca-cloud/secrets/orca-cloud-push-apns-key roles/secretmanager.secretAccessor serviceAccount:orca-cloud-push@onorca-cloud.iam.gserviceaccount.com' + +terraform -chdir=infra/terraform import -var-file=environments/production.tfvars \ + 'google_secret_manager_secret_iam_member.push_provider_runtime_accessor["orca-cloud-push-apns-key-id"]' \ + 'projects/onorca-cloud/secrets/orca-cloud-push-apns-key-id roles/secretmanager.secretAccessor serviceAccount:orca-cloud-push@onorca-cloud.iam.gserviceaccount.com' + +terraform -chdir=infra/terraform import -var-file=environments/production.tfvars \ + 'google_secret_manager_secret_iam_member.push_provider_runtime_accessor["orca-cloud-push-apple-team-id"]' \ + 'projects/onorca-cloud/secrets/orca-cloud-push-apple-team-id roles/secretmanager.secretAccessor serviceAccount:orca-cloud-push@onorca-cloud.iam.gserviceaccount.com' +``` + +Everything else in `push-gateway.tf` is new and is created by the apply: the `orca_push` +database and user, the database-URL secret and its accessor, the `roles/cloudsql.client` binding +on the runtime account, the Cloud Run service, the domain mapping, and the +three deploy-identity bindings. Save that plan and review it before applying; this root carries +unrelated standing drift, so an untargeted apply is never automatic. + +Two things this root does **not** declare, because the carve assigns them elsewhere. Neither +affects whether this root's plan is clean, since an undeclared resource is invisible to it. + +- `firebase.googleapis.com` and `fcm.googleapis.com` are project service enablement, which is + `google_project_service.required` in the foundation root. They are already enabled; add them + to the foundation root's list so a foundation plan stays clean. +- The Firebase attachment on `onorca-cloud` is project-level and belongs with foundation for the + same reason. It exists already. + +## Deploying + +`Deploy Push Gateway Production` (`.github/workflows/cloud-push-deploy.yml`) is the only +supported path. Like every `cloud-*` workflow it does nothing until `ORCA_CLOUD_OPERATIONS_ENABLED` +is `true`, it runs only on `main`, and it needs the confirmation string `DEPLOY_PUSH_GATEWAY`. + +It authenticates as the shared production deploy identity through +`PRODUCTION_GCP_RELAY_DEPLOY_WORKLOAD_IDENTITY_PROVIDER` and +`PRODUCTION_GCP_RELAY_DEPLOY_SERVICE_ACCOUNT`, which are already published. No new GitHub +variable is required. That account was chosen because the Cloud SQL rollout lease grant is +foundation-owned and names only that account; a dedicated identity could not take that lease from +this root, and the gateway's schema rollout has to serialize against the relay's. + +**That choice widens what this workflow can reach, and the widening is deliberate.** Adding +`push-deploy.yml` to the provider allowlist gives the run the account's whole existing authority, +not only the push bindings: Artifact Registry writer on `orca-cloud`, `roles/run.developer` on +the relay director and the fence broker, accessor and version-adder on the relay +regional-placement secret, and service-account user on the relay runtime identities. It was +accepted as the price of the lease. What `push-gateway.tf` adds on top is three bindings scoped +to the gateway alone: Cloud Run developer on this one service, and service-account user plus +token creator on the runtime account. The bound on the rest is the provider condition, which +admits this exact workflow file on `main` in the `production` environment only, and the workflow +itself, which is dispatch-only behind a typed confirmation. + +The run, in order: + +1. Builds `apps/push/Dockerfile` with the `cloud/` build context and pushes to the existing + `orca-cloud` Artifact Registry repository as `push:sha-`, then resolves the digest. + This happens **before** the lease is taken. Artifact Registry is not the Cloud SQL instance, + and a multi-minute build inside the lease would block every relay deploy and rehome for its + duration. +2. Takes the production Cloud SQL rollout lease and holds it from here to the end. The gateway + applies its schema while the new revision starts, so the revision **is** the schema step + (on a one-connection pool with no statement timeout, closed before the serving pool opens, + exactly as the relay does since #18722); + there is no separate migration command to wrap. The lease therefore covers exactly the + connection-budget window: deploy, probe, shift. +3. Records the currently serving revision as the rollback target, and requires it to still hold + the Terraform-owned floor and ceiling. The candidate inherits that scaling, so a drifted + serving revision would be latched rather than corrected. +4. `gcloud run deploy --no-traffic` with a per-run traffic tag, so the candidate boots and + applies schema while every phone still reaches the previous revision. The deploy passes no + scaling flag: the shape is Terraform's, and the candidate's inherited ceiling is asserted + instead. +5. Probes the tagged candidate's own `/ready`, up to 30 times at five-second intervals. +6. Sends a validate-only FCM message as the runtime identity, by impersonation. See below. +7. Shifts 100% of traffic to the candidate and verifies it is the only revision serving. +8. Writes the run summary, including the rollback command, before checking the public origin, so + the summary exists even when the check that follows does not pass. +9. Checks `https://push.onorca.dev/ready`, up to 30 times at five-second intervals, since the + origin can lag the traffic move by a few seconds. +10. Always removes the traffic tag, so tags do not accumulate across runs. + +**Failure after the shift rolls itself back.** Everything from step 8 on runs with production +already on the candidate, so a failure there is not a failed deploy, it is a live gateway that +has to go back. The run returns traffic to the recorded rollback revision, verifies the move, and +reports it in the summary. A failure *before* the shift leaves production untouched and deletes +the candidate revision, which otherwise sits holding a warm instance and a Cloud SQL pool for +nothing. + +To move traffic by hand, from the revision named in the run summary: + +```sh +gcloud run services update-traffic orca-cloud-push \ + --project onorca-cloud --region us-central1 \ + --to-revisions =100 +``` + +### Why the FCM probe impersonates the runtime account + +A gateway that boots and answers `/ready` can still be unable to send: the FCM grant lives on +the runtime service account, not on anything the readiness check touches. The probe therefore +mints an access token for `orca-cloud-push@onorca-cloud.iam.gserviceaccount.com` and posts +`validate_only: true` with a token that cannot exist. `validate_only` stops Google before any +delivery, and a healthy credential answers `INVALID_ARGUMENT` because the device token is +garbage. `PERMISSION_DENIED`, `401`, and `403` are the failures the step exists to catch, and +they fail the run immediately, before traffic moves. Those four answers are the only conclusive +ones: a `429`, a `5xx`, or a transport failure says nothing about the credential, so the send is +retried up to five times at five-second intervals rather than read as either verdict. Probing as the deploy identity instead would prove +something true about the wrong account. + +## Rotating the APNs key + +Apple keys do not expire, so this is for a suspected compromise or a routine rotation. Order +matters: the new key must be serving before the old one is revoked, or every iOS push fails in +the window between. + +1. In the Apple developer portal, create a **new** APNs authentication key. Download the `.p8` + once; Apple will not show it again. Note the new key ID. A team may hold two APNs keys at a + time, which is what makes this overlap possible. +2. Add a version to each changed secret, without printing the value: + + ```sh + gcloud secrets versions add orca-cloud-push-apns-key \ + --project onorca-cloud --data-file /path/to/AuthKey_NEW.p8 + printf '%s' '' | gcloud secrets versions add orca-cloud-push-apns-key-id \ + --project onorca-cloud --data-file=- + ``` + + The team ID does not change, so `orca-cloud-push-apple-team-id` is untouched. +3. Dispatch `Deploy Push Gateway Production`. The container reads `latest` at start, so only a + new revision picks the key up; there is no in-place reload. +4. Verify from a real device that an iOS notification still arrives. The workflow's FCM probe + covers Android only, and APNs has no validate-only equivalent. +5. Only then revoke the old key in the Apple portal, and disable the superseded secret versions: + + ```sh + gcloud secrets versions disable \ + --project onorca-cloud --secret orca-cloud-push-apns-key + ``` + + Disable rather than destroy, so a rollback to the previous revision still works. Destroy + after the next clean deploy. + +Delete the downloaded `.p8` from disk when you are done. It is the whole credential. + +## Dead tokens + +A push token stops working when the app is uninstalled, when the user restores to a new device, +or when iOS reissues it. Both providers report this, and the shapes differ: + +- APNs: HTTP 410, or 400 with `BadDeviceToken`, `Unregistered`, or `DeviceTokenNotForTopic`. + `DeviceTokenNotForTopic` also fires when a sandbox token is sent to the production host, which + is a configuration bug rather than a dead token; check `apns_environment` on the registration + before concluding the device is gone. +- FCM: `UNREGISTERED`, or `INVALID_ARGUMENT` whose message names the token. + +The gateway marks the registration `dead_at` and returns `status: "dead"` for it, and the +desktop drops the registration when it sees that. Nothing here retries a dead token. A phone +that comes back registers again and gets a fresh `registrationId`, so a rising dead count is +normal churn; a dead count that spikes across many hosts at once is a credential or topic +problem, not device churn. + +## Quotas + +Two independent limits, both enforced in the gateway and both returning HTTP 200 with +`status: "rate_limited"` per result rather than failing the request: + +| Limit | Scope | +| --- | --- | +| 60 sends per rolling hour | per `hostFingerprint` | +| 200 sends per rolling day | per `registrationId` | +| 20 `registrationIds` | per request, hard cap, HTTP 400 over it | + +Ahead of all three sit two per-client-IP token buckets that answer HTTP 429: 30 requests per +minute on the two unauthenticated handshake routes, and 240 per minute on every other `/v1` +route, applied before the bearer is looked up so that a flood of forged bearers cannot spend +the two-connection pool on session lookups. Both are per instance and in memory. + +`push_send_log` backs the two rolling counts and is pruned after 25 hours. Upstream of all +three, FCM V1 bills project quota against `ORCA_PUSH_FCM_PROJECT_ID`, which is why the runtime +account holds `roles/serviceusage.serviceUsageConsumer`; a project-level FCM quota exhaustion +surfaces as `RESOURCE_EXHAUSTED` and is not something the per-host limits can prevent. + +Logging is aggregate counters only. Never log a token, a title, a body, or a full fingerprint; +the first four characters of a fingerprint are the most that may appear. + +## DNS: one hand-managed record + +The Cloud Run domain mapping is created here, and Google issues and renews the certificate. The +`onorca.dev` zone is not in this root: it is a Cloudflare zone whose Terraform-managed records +live in the apps root in `stablyai/orca-cloud`, and whose relay and auth records are managed by +hand. The push record follows the relay's precedent and was created by hand on 2026-09-04: + +```text +push.onorca.dev. CNAME ghs.googlehosted.com. (DNS only, not proxied) +``` + +`terraform -chdir=infra/terraform output push_dns_record` prints the same three fields. If the +record is ever lost, recreate it exactly like that; Cloudflare proxying blocks certificate +issuance and breaks Cloud Run host routing. + + +### Recovery and delivery guarantees + +Candidate tags and deterministic revision names are recorded before deployment. Promotion intent is +recorded before changing traffic, so a failed verification or ambiguous mutation result still triggers +rollback. Failed candidates are deleted only before attempted promotion or after verified rollback. +The summary runs even if candidate discovery or traffic verification fails. + +Push uses the relay's schema-startup retry implementation through `@orca-cloud/postgres-schema`. +Session replacement is serialized per host and a unique host index upgrades older databases by +retaining their newest session. Cloud Verify runs push concurrency tests against PostgreSQL. + +Accepted sends deduplicate by host, registration, epoch, and sequence for the quota ledger's 25-hour +retention period. Provider failures retry at most three times within two minutes, respecting provider +retry delays. Queues remain in memory; a crash or the nine-second shutdown deadline can still lose work. +Graceful shutdown first refuses new requests, waits for admitted handlers, and drains pending and active +deliveries before closing transports and SQL. `delivery_retry` counters accompany existing outcomes. + +Notification and worktree IDs allow 2048 characters each, subject to a combined notification JSON +budget of 3000 UTF-8 bytes. This preserves normal long and Unicode paths without exceeding provider +envelope space. No identity is truncated to meet this budget. diff --git a/cloud/docs/relay-workflows.md b/cloud/docs/relay-workflows.md index 14bb2a25d7c..88c574f3206 100644 --- a/cloud/docs/relay-workflows.md +++ b/cloud/docs/relay-workflows.md @@ -400,3 +400,42 @@ after checkout and authentication, before package installation, revision checks, Their typed confirmations are `PAUSE_REGIONAL_REHOMING` and `DISABLE_REGIONAL_REHOMING`. Keep the default 3,600,000 ms drain grace so existing splices can finish. The job summary contains only fresh aggregate active, receipt, registration, completion, and abort counts. + +## Mobile push gateway + +`Deploy Push Gateway Production` (`.github/workflows/cloud-push-deploy.yml`) is the deploy path +for `orca-cloud-push`, the mobile push gateway. It is the one `cloud-*` workflow that is not a +relay operation, and it is here because it shares this repository's Cloud SQL instance, its +Artifact Registry repository, and its rollout lease. + +It needs **no new GitHub environment variable.** It authenticates as the shared production deploy +identity through the already-published `PRODUCTION_GCP_RELAY_DEPLOY_WORKLOAD_IDENTITY_PROVIDER` +and `PRODUCTION_GCP_RELAY_DEPLOY_SERVICE_ACCOUNT`, and reads `PRODUCTION_GCP_REGION` like the +rest. That account holds the foundation-owned Cloud SQL rollout lease grant, which names it and nothing +else, so a dedicated identity could not be given that lease from this root. + +`infra/terraform/push-gateway.tf` adds three bindings scoped to the gateway: Cloud Run developer +on that one service, and service-account user plus token creator on the gateway's runtime +account. Those three are not the workflow's whole authority. Running as the shared account gives +the run every role that account already holds for the relay: Artifact Registry writer on +`orca-cloud`, `roles/run.developer` on the relay director and the fence broker, accessor and +version-adder on the relay regional-placement secret, and service-account user on the relay +runtime identities. That widening was accepted as the price of the lease, and it is bounded by +the provider condition and by the workflow being dispatch-only behind a typed confirmation. + +The provider's workflow allowlist gained exactly one entry, `cloud-push-deploy.yml`, on `main` in +the `production` environment. That entry is required: the allowlist compares complete workflow +refs by equality, so the `cloud-` filename prefix alone does not admit a new file. + +The run builds `apps/push/Dockerfile` **before** taking the lease, so an image build never blocks +a relay deploy or rehome, then holds the production rollout lease across the deploy itself, +because the gateway applies its schema while the new revision starts. Under the lease it checks +the serving revision's Terraform-owned scaling, deploys with `--no-traffic` behind a per-run +traffic tag and no scaling flag of its own, probes the candidate's own `/ready`, proves the +runtime identity can reach FCM with a validate-only send, and only then shifts 100% of traffic. A +failure after the shift returns traffic to the recorded rollback revision; a failure before it +deletes the candidate. There is no staging gateway, so there is no staging counterpart to run +first. + +Full runbook, including the APNs key rotation and the DNS record the `stablyai/orca-cloud` apps +root still owes, is in `docs/push-gateway.md`. diff --git a/cloud/infra/terraform/environments/production.tfvars b/cloud/infra/terraform/environments/production.tfvars index 8e442c75900..79db1904ee6 100644 --- a/cloud/infra/terraform/environments/production.tfvars +++ b/cloud/infra/terraform/environments/production.tfvars @@ -408,3 +408,13 @@ relay_region_rehome_source_cell_ids = [ # Slack #orca-relay-alerts, created out of band on 2026-08-05. Declared here because an apply # was otherwise going to strip it from every policy, leaving the alerts firing at nobody. relay_alert_notification_channels = ["projects/onorca-cloud/notificationChannels/4879431412695417284"] + +# Mobile push gateway. Production is the only environment that runs one; the runtime account, +# the three Apple secrets, and their accessor bindings already exist and are imported once +# (see docs/push-gateway.md). +push_gateway_enabled = true +push_base_url = "https://push.onorca.dev" +# Sized so the gateway's rollout overlap, the retained director rollback plus its doubled draw, +# stays under the API candidate's, which keeps the checked Cloud SQL connection budget green. +push_max_instances = 2 +manage_push_domain_mapping = true diff --git a/cloud/infra/terraform/environments/staging.tfvars b/cloud/infra/terraform/environments/staging.tfvars index 4a32458fcd5..72b5306336b 100644 --- a/cloud/infra/terraform/environments/staging.tfvars +++ b/cloud/infra/terraform/environments/staging.tfvars @@ -81,3 +81,7 @@ relay_gce_cells = { } relay_region_rehome_source_cell_ids = ["staging-gce-c2", "staging-gce-c3"] + +# No staging push gateway by decision (mobile-push-contract.md, "Non-goals"). Stated rather than +# left to the default so a future staging gateway is one obvious edit. +push_gateway_enabled = false diff --git a/cloud/infra/terraform/outputs.tf b/cloud/infra/terraform/outputs.tf index 220aa5cf94f..184b3be61f7 100644 --- a/cloud/infra/terraform/outputs.tf +++ b/cloud/infra/terraform/outputs.tf @@ -189,3 +189,27 @@ output "relay_gce_cell_deployments" { error_message = "relay_gce_fenced_cells may contain only configured relay_gce_cells keys." } } + +output "push_cloud_run_service_uri" { + value = try(google_cloud_run_v2_service.push[0].uri, null) + description = "Default push gateway service URI for pre-domain smoke tests." +} + +output "push_runtime_service_account" { + value = try(google_service_account.push_runtime[0].email, null) + description = "Runtime identity that holds the APNs key and sends through FCM." +} + +output "push_database_name" { + value = try(google_sql_database.push[0].name, null) + description = "Database isolated for durable push gateway state." +} + +output "push_dns_record" { + value = var.push_gateway_enabled ? { + name = local.push_fqdn + type = "CNAME" + data = "ghs.googlehosted.com." + } : null + description = "Record the stablyai/orca-cloud apps root must publish in the onorca.dev zone." +} diff --git a/cloud/infra/terraform/push-gateway.tf b/cloud/infra/terraform/push-gateway.tf new file mode 100644 index 00000000000..87d12ae2693 --- /dev/null +++ b/cloud/infra/terraform/push-gateway.tf @@ -0,0 +1,405 @@ +# Orca mobile push gateway (`cloud/apps/push`). +# +# One public Cloud Run service that holds the APNs key and sends through APNs and FCM V1 on +# behalf of paired phones. Contract: `docs/reference/mobile-push-contract.md`, "Infra" and +# "Gateway env". Operations: `docs/push-gateway.md`. +# +# There is no staging push gateway by decision, so every resource here is behind +# `var.push_gateway_enabled`, which only `environments/production.tfvars` sets true. The file +# still reads every environment-shaped value from a variable, like the rest of this root, so a +# future staging gateway is a tfvars edit rather than a rewrite. +# +# Several resources below already exist in `onorca-cloud`; they are declared so a plan is clean +# and imported once. `docs/push-gateway.md` carries the exact `terraform import` commands. + +locals { + push_gateway_count = var.push_gateway_enabled ? 1 : 0 + + # The runtime account, the three provider secrets, and their accessor bindings already exist in + # production and were created out of band with the Apple credentials. + push_runtime_service_account_id = "${var.name_prefix}-push" + + # Secret Manager holds the Apple credentials. Terraform owns the secret names, labels, and + # replication; it never owns a version. The `.p8` is issued by the Apple developer portal and + # rotated by `docs/push-gateway.md`, so a Terraform-managed version would either put the key in + # state or fight the rotation. `ignore_changes` on the whole resource is not available, so the + # versions are simply not declared and every consumer reads `latest`. + push_provider_secret_ids = var.push_gateway_enabled ? toset([ + "${var.name_prefix}-push-apns-key", + "${var.name_prefix}-push-apns-key-id", + "${var.name_prefix}-push-apple-team-id" + ]) : toset([]) + + push_provider_secret_env = { + "${var.name_prefix}-push-apns-key" = "ORCA_PUSH_APNS_KEY" + "${var.name_prefix}-push-apns-key-id" = "ORCA_PUSH_APNS_KEY_ID" + "${var.name_prefix}-push-apple-team-id" = "ORCA_PUSH_APPLE_TEAM_ID" + } + + push_fcm_project_id = var.push_fcm_project_id == "" ? var.project_id : var.push_fcm_project_id + + push_fqdn = replace(replace(var.push_base_url, "https://", ""), "http://", "") + + # The shared production deploy identity runs `cloud-push-deploy.yml`. The grants this file adds + # are scoped to this service and its runtime account alone, but the workflow inherits every + # other grant that account already holds for the relay; see the deploy-identity section below. + # The account itself is declared in relay-github-actions.tf and is production-only. + push_gateway_deploy_count = ( + var.push_gateway_enabled && local.relay_create_production_ops_identity ? 1 : 0 + ) +} + +# --- Runtime identity --------------------------------------------------------------------- + +resource "google_service_account" "push_runtime" { + count = local.push_gateway_count + + project = var.project_id + account_id = local.push_runtime_service_account_id + display_name = "Orca mobile push gateway" + description = "Runtime identity for the Orca mobile push gateway; sends through FCM V1." +} + +# FCM V1 sends are authorized by the runtime account's own metadata-server token. +resource "google_project_iam_member" "push_runtime_fcm_admin" { + count = local.push_gateway_count + + project = var.project_id + role = "roles/firebasecloudmessaging.admin" + member = google_service_account.push_runtime[0].member +} + +# The FCM V1 endpoint bills against the caller's project quota, which the caller must consume. +resource "google_project_iam_member" "push_runtime_service_usage_consumer" { + count = local.push_gateway_count + + project = var.project_id + role = "roles/serviceusage.serviceUsageConsumer" + member = google_service_account.push_runtime[0].member +} + +resource "google_project_iam_member" "push_runtime_cloudsql_client" { + count = local.push_gateway_count + + project = var.project_id + role = "roles/cloudsql.client" + member = google_service_account.push_runtime[0].member +} + +# --- Database ----------------------------------------------------------------------------- +# Gateway state shares the foundation-owned Cloud SQL instance with auth and the relay, and uses +# an isolated database and principal, exactly as relay-database.tf does. The application applies +# its own schema at startup. + +resource "google_sql_database" "push" { + count = local.push_gateway_count + + project = var.project_id + name = "orca_push" + instance = local.relay_database_instance_name + + # Why: this database holds every live device token. Disabling the gateway must not drop it. + lifecycle { + prevent_destroy = true + } +} + +resource "random_password" "push_database" { + count = local.push_gateway_count + + length = 32 + special = false +} + +resource "google_sql_user" "push" { + count = local.push_gateway_count + + project = var.project_id + name = "orca_push" + instance = local.relay_database_instance_name + password = random_password.push_database[0].result +} + +resource "google_secret_manager_secret" "push_database_url" { + count = local.push_gateway_count + + project = var.project_id + secret_id = "${var.name_prefix}-push-database-url" + labels = local.relay_shared_labels + + replication { + auto {} + } +} + +resource "google_secret_manager_secret_version" "push_database_url" { + count = local.push_gateway_count + + secret = google_secret_manager_secret.push_database_url[0].id + secret_data = format( + "postgresql://%s:%s@/%s?host=/cloudsql/%s", + google_sql_user.push[0].name, + random_password.push_database[0].result, + google_sql_database.push[0].name, + local.relay_database_connection_name + ) +} + +resource "google_secret_manager_secret_iam_member" "push_database_url_runtime_accessor" { + count = local.push_gateway_count + + project = var.project_id + secret_id = google_secret_manager_secret.push_database_url[0].secret_id + role = "roles/secretmanager.secretAccessor" + member = google_service_account.push_runtime[0].member +} + +# --- Apple credentials ---------------------------------------------------------------------- + +resource "google_secret_manager_secret" "push_provider" { + for_each = local.push_provider_secret_ids + + project = var.project_id + secret_id = each.value + labels = local.relay_shared_labels + + replication { + auto {} + } + + # Why: Apple issues a `.p8` once and Secret Manager has no undelete. Turning the gateway off + # must fail the plan rather than destroy the only copy of the signing key. + lifecycle { + prevent_destroy = true + } +} + +resource "google_secret_manager_secret_iam_member" "push_provider_runtime_accessor" { + for_each = local.push_provider_secret_ids + + project = var.project_id + secret_id = google_secret_manager_secret.push_provider[each.value].secret_id + role = "roles/secretmanager.secretAccessor" + member = google_service_account.push_runtime[0].member +} + +# --- Service -------------------------------------------------------------------------------- + +resource "google_cloud_run_v2_service" "push" { + count = local.push_gateway_count + + project = var.project_id + name = var.push_cloud_run_service_name + location = var.region + ingress = "INGRESS_TRAFFIC_ALL" + # Why: the host proof in `POST /v1/host/challenge` is the authentication, not Cloud Run IAM. + # The project's domain-restricted-sharing policy refuses an `allUsers` invoker binding, so the + # service opts out of invoker IAM exactly as the relay director does. + invoker_iam_disabled = true + deletion_protection = var.environment == "production" + labels = local.relay_shared_labels + + template { + service_account = google_service_account.push_runtime[0].email + timeout = "${var.push_request_timeout_seconds}s" + max_instance_request_concurrency = var.push_concurrency + + scaling { + min_instance_count = var.push_min_instances + max_instance_count = var.push_max_instances + } + + volumes { + name = "cloudsql" + + cloud_sql_instance { + instances = [local.relay_database_connection_name] + } + } + + containers { + image = var.push_cloud_run_image + + ports { + container_port = 8080 + } + + volume_mounts { + name = "cloudsql" + mount_path = "/cloudsql" + } + + env { + name = "ORCA_PUSH_PUBLIC_URL" + value = var.push_base_url + } + + env { + name = "ORCA_PUSH_FCM_PROJECT_ID" + value = local.push_fcm_project_id + } + + # Declared rather than left to the application default, so the gateway's share of the + # shared Cloud SQL connection budget is a value this root states and the precondition + # below can bound. + env { + name = "ORCA_PUSH_DATABASE_POOL_MAX" + value = tostring(var.push_database_pool_max) + } + + env { + name = "ORCA_PUSH_DATABASE_URL" + + value_source { + secret_key_ref { + secret = google_secret_manager_secret.push_database_url[0].secret_id + version = "latest" + } + } + } + + # Rotation adds a new version and redeploys; `latest` is what the redeploy picks up. + dynamic "env" { + for_each = local.push_provider_secret_env + + content { + name = env.value + + value_source { + secret_key_ref { + secret = google_secret_manager_secret.push_provider[env.key].secret_id + version = "latest" + } + } + } + } + + resources { + limits = { + cpu = var.push_cloud_run_cpu + memory = var.push_cloud_run_memory + } + + cpu_idle = false + } + + startup_probe { + failure_threshold = 12 + initial_delay_seconds = 0 + period_seconds = 5 + timeout_seconds = 2 + + http_get { + path = "/health" + port = 8080 + } + } + } + } + + # Deploys update the immutable image and shift traffic; Terraform owns the shape and IAM. + # + # `traffic` is ignored as well as the image. A deploy ends with traffic pinned to an exact + # revision and a rollback pins it to the previous one; an apply that reset the service to + # 100% LATEST would silently undo either, and this root carries unrelated standing drift, so + # that apply need not be a push change at all. + lifecycle { + # Why: the gateway draws instances x pool from the shared Cloud SQL instance, and a rollout + # doubles it, because the tagged candidate is directly addressable and sits outside the + # service-wide cap. The instance's 400 connections were already spoken for by the relay + # cells, directors, auth, and API, which left five: 4 is the whole of the gateway's share and + # it fits, with the doubled 8 still under the API candidate's rollout overlap, the term + # dev/scripts/relay-cloud-sql-connection-budget.mjs maximizes over. A fifth connection here + # puts the checked budget over its ceiling and blocks Deploy Relay Asia Topology, which gates + # on it, so catch a raise at plan time rather than in someone else's rollout. + precondition { + condition = var.push_max_instances * var.push_database_pool_max <= 4 + error_message = "Push gateway instances x database pool must stay within its 4-connection share of the shared Cloud SQL instance." + } + + ignore_changes = [ + client, + client_version, + template[0].containers[0].image, + traffic + ] + } + + depends_on = [ + data.google_artifact_registry_repository.relay_images, + google_project_iam_member.push_runtime_cloudsql_client, + google_secret_manager_secret_iam_member.push_database_url_runtime_accessor, + google_secret_manager_secret_iam_member.push_provider_runtime_accessor, + google_secret_manager_secret_version.push_database_url + ] +} + +# Google issues and renews the certificate for the mapping. The DNS record itself is a +# hand-managed Cloudflare CNAME to ghs.googlehosted.com, like relay.onorca.dev; this root has no +# Cloudflare surface by design. `terraform output push_dns_record` prints the record. +resource "google_cloud_run_domain_mapping" "push" { + count = var.push_gateway_enabled && var.manage_push_domain_mapping ? 1 : 0 + + location = var.region + name = local.push_fqdn + + metadata { + namespace = var.project_id + } + + spec { + route_name = google_cloud_run_v2_service.push[0].name + } + + # Same reason as relay-dns.tf: a gcloud-created mapping reports an empty legacy + # certificate_mode, and replacing it would reset issuance for no behavioral change. + lifecycle { + ignore_changes = [spec[0].certificate_mode] + } +} + +# --- Deploy identity grants ------------------------------------------------------------------- +# `cloud-push-deploy.yml` authenticates as the shared production deploy account, because that +# account is the one the foundation root grants the Cloud SQL rollout lease to; the grant names +# that account and nothing else, so a dedicated push identity could not take the lease from this +# root and the gateway's schema rollout could not be serialized against the relay's. +# +# The three bindings below are the whole of that account's authority over the *push gateway*, but +# they are not the whole of what the workflow can do. Adding `push-deploy.yml` to the provider's +# allowlist in relay-github-actions.tf gives the run the account's entire existing authority: +# Artifact Registry writer on `orca-cloud`, `roles/run.developer` on the relay director and the +# fence broker, accessor and version-adder on the relay regional-placement secret, and +# service-account user on the relay runtime identities. That widening was accepted deliberately +# as the price of the lease. It is bounded by the provider condition, which admits this exact +# workflow file on `main` in the `production` environment only, and by the workflow itself, which +# is dispatch-only behind a typed confirmation. + +resource "google_cloud_run_v2_service_iam_member" "github_production_push_developer" { + count = local.push_gateway_deploy_count + + project = var.project_id + location = var.region + name = google_cloud_run_v2_service.push[0].name + role = "roles/run.developer" + member = local.relay_github_deploy_service_account_member +} + +resource "google_service_account_iam_member" "github_production_push_runtime_user" { + count = local.push_gateway_deploy_count + + service_account_id = google_service_account.push_runtime[0].name + role = "roles/iam.serviceAccountUser" + member = local.relay_github_deploy_service_account_member +} + +# Why: the deploy workflow's validate-only FCM send has to exercise the credential the gateway +# will actually use. Impersonating the runtime account proves its firebasecloudmessaging grant; +# granting the deploy account FCM admin outright would prove nothing about the runtime account +# and would widen a project-level role on the shared identity. +resource "google_service_account_iam_member" "github_production_push_runtime_token_creator" { + count = local.push_gateway_deploy_count + + service_account_id = google_service_account.push_runtime[0].name + role = "roles/iam.serviceAccountTokenCreator" + member = local.relay_github_deploy_service_account_member +} diff --git a/cloud/infra/terraform/relay-github-actions.tf b/cloud/infra/terraform/relay-github-actions.tf index 450ea64cc0a..a73e8f511e5 100644 --- a/cloud/infra/terraform/relay-github-actions.tf +++ b/cloud/infra/terraform/relay-github-actions.tf @@ -19,7 +19,16 @@ locals { "deploy-relay-production-multi-target.yml", "deploy-relay-production.yml", "operate-relay-asia-admission.yml", - "publish-relay-production.yml" + "publish-relay-production.yml", + # The push gateway deploy runs as this account because the Cloud SQL rollout lease grant is + # foundation-owned and names only this account; a dedicated identity could not take that + # lease, and the gateway's schema rollout has to serialize against the relay's. + # + # This entry therefore grants that workflow every role the account already holds, not just + # the three push bindings in push-gateway.tf: Artifact Registry writer, run.developer on the + # relay director and fence broker, relay secret accessor and version-adder, and + # serviceAccountUser on the relay runtime identities. Accepted as the price of the lease. + "push-deploy.yml" ] github_production_relay_capacity_workflow_file = "deploy-relay-production-capacity.yml" github_production_relay_capacity_job_workflow_file = "deploy-relay-production-capacity-job.yml" diff --git a/cloud/infra/terraform/variables.tf b/cloud/infra/terraform/variables.tf index 91f67e8ebe0..1ef74bbc40f 100644 --- a/cloud/infra/terraform/variables.tf +++ b/cloud/infra/terraform/variables.tf @@ -484,3 +484,108 @@ variable "relay_gce_cloud_sql_proxy_image" { error_message = "relay_gce_cloud_sql_proxy_image must be pinned by sha256 digest." } } + +# --- Mobile push gateway --------------------------------------------------------------------- +# There is no staging push gateway by decision, so this defaults false and only +# environments/production.tfvars turns it on. Everything in push-gateway.tf is behind it. +variable "push_gateway_enabled" { + type = bool + description = "Create the Orca mobile push gateway, its database, secrets, and identity." + default = false +} + +variable "push_base_url" { + type = string + description = "Public TLS origin of the mobile push gateway." + default = "https://push.onorca.dev" + + validation { + condition = can(regex("^https://[^/]+$", var.push_base_url)) + error_message = "push_base_url must be an HTTPS origin with no path." + } +} + +variable "push_cloud_run_service_name" { + type = string + description = "Cloud Run service name for the mobile push gateway." + default = "orca-cloud-push" +} + +variable "push_cloud_run_image" { + type = string + description = "Initial image for the Terraform-created push gateway service; deploys own it after." + default = "us-docker.pkg.dev/cloudrun/container/hello" +} + +variable "push_cloud_run_cpu" { + type = string + description = "CPU limit for the push gateway container." + default = "1" +} + +variable "push_cloud_run_memory" { + type = string + description = "Memory limit for the push gateway container." + default = "512Mi" +} + +# Why: a cold start would delay a notification past the point where it is worth showing, and the +# 3 s coalescing window lives in instance memory, so the floor is one warm instance. +variable "push_min_instances" { + type = number + description = "Minimum instances for the push gateway." + default = 1 +} + +variable "push_max_instances" { + type = number + description = "Maximum instances for the push gateway." + default = 4 + + validation { + condition = var.push_max_instances >= 1 + error_message = "The push gateway needs at least one instance." + } +} + +# Why: the gateway's draw on the shared Cloud SQL instance is instances x pool, and the rollout +# lease is taken for twice that, because a tagged candidate is directly addressable and sits +# outside the service-wide cap. Leaving the pool at its application default made that draw +# invisible to this root, so it is declared here and set on the container. +# +# Two is sized to the work, not to the default: a send runs two or three short queries, and at +# concurrency 80 those queue against the pool for microseconds rather than holding it. +variable "push_database_pool_max" { + type = number + description = "Push gateway database pool size per instance; instances x pool is its Cloud SQL draw." + default = 2 + + validation { + condition = var.push_database_pool_max >= 1 && var.push_database_pool_max <= 100 + error_message = "The push gateway pool must hold at least one connection and stay under the per-service bound." + } +} + +variable "push_concurrency" { + type = number + description = "Cloud Run concurrency for short-lived push gateway HTTP requests." + default = 80 +} + +variable "push_request_timeout_seconds" { + type = number + description = "Cloud Run timeout for push gateway requests; every route is short-lived." + default = 30 +} + +variable "push_fcm_project_id" { + type = string + description = "Firebase project for FCM V1 sends; empty uses project_id." + default = "" +} + +variable "manage_push_domain_mapping" { + type = bool + description = "Manage the push gateway Cloud Run domain mapping; the DNS record stays in the apps root." + default = false +} diff --git a/cloud/package.json b/cloud/package.json index 62dbadc7455..3e33f245527 100644 --- a/cloud/package.json +++ b/cloud/package.json @@ -21,7 +21,7 @@ "load:relay:recovery-gate": "node dev/scripts/run-relay-recovery-wave-gate.mjs", "ops:relay": "pnpm --filter @orca-cloud/relay-ops dev", "pretest": "node --test dev/scripts/capture-terraform-plan-baseline.test.mjs dev/scripts/operate-relay-asia-admission.test.mjs dev/scripts/prepare-relay-asia-director-cells.test.mjs dev/scripts/prepare-relay-asia-topology-input.test.mjs dev/scripts/production-cloud-sql-rollout-lock.test.mjs dev/scripts/read-relay-serving-regional-placement-version.test.mjs dev/scripts/relay-asia-admission-workflow.test.mjs dev/scripts/relay-asia-rollout-evidence.test.mjs dev/scripts/relay-asia-topology-workflow.test.mjs dev/scripts/relay-cloud-sql-connection-budget.test.mjs dev/scripts/relay-load-reader-evidence.test.mjs dev/scripts/relay-staging-deploy-identity.test.mjs dev/scripts/sanitize-relay-asia-admission-result.test.mjs dev/scripts/terraform-root-partition.test.mjs dev/scripts/validate-relay-asia-topology-plan.test.mjs ../.github/actions/cloud-sql-rollout-lease/action-contract.test.mjs ../.github/actions/cloud-sql-rollout-lease/storage-lease.test.mjs", - "test": "pnpm -r test && node --test dev/scripts/classify-relay-production-capacity-director.test.mjs dev/scripts/classify-relay-staging-bootstrap.test.mjs dev/scripts/deploy-relay-blue-green.test.mjs dev/scripts/deploy-relay-gce-candidate.test.mjs dev/scripts/deploy-relay-gce-multi-target.test.mjs dev/scripts/github-smoke-token.test.mjs dev/scripts/infra.test.mjs dev/scripts/operate-relay-regional-rehome.test.mjs dev/scripts/power-staging-relay.test.mjs dev/scripts/prepare-relay-capacity-canary.test.mjs dev/scripts/prepare-relay-production-capacity-canary.test.mjs dev/scripts/probe-relay-legacy-admission.test.mjs dev/scripts/probe-relay-rehome-trust.test.mjs dev/scripts/production-cell-image-digest-consistency.test.mjs dev/scripts/read-relay-production-capacity-identity.test.mjs dev/scripts/relay-admin-endpoint-retry-workflow.test.mjs dev/scripts/relay-admin-transient-retry.test.mjs dev/scripts/relay-admission-selector.test.mjs dev/scripts/relay-gce-terraform-fence.test.mjs dev/scripts/relay-load-connection-failure.test.mjs dev/scripts/relay-load-control-peer.test.mjs dev/scripts/relay-load-director-capacity-gate.test.mjs dev/scripts/relay-load-model.test.mjs dev/scripts/relay-load-phase-barrier.test.mjs dev/scripts/relay-load-placement-boundary.test.mjs dev/scripts/relay-load-profile.test.mjs dev/scripts/relay-load-rebind-boundary.test.mjs dev/scripts/relay-load-region-behavior.test.mjs dev/scripts/relay-load-request-unit-boundary.test.mjs dev/scripts/relay-load-run-lifecycle.test.mjs dev/scripts/relay-monitor-evidence.test.mjs dev/scripts/relay-production-capacity-wave.test.mjs dev/scripts/relay-production-capacity-workflow.test.mjs dev/scripts/relay-production-identity-boundaries.test.mjs dev/scripts/relay-production-same-cap-wave.test.mjs dev/scripts/relay-public-workflow-contract.test.mjs dev/scripts/relay-recovery-wave-gate.test.mjs dev/scripts/relay-region-observation-evidence.test.mjs dev/scripts/relay-regional-rehome-workflow.test.mjs dev/scripts/relay-rehome-aggregate-evidence.test.mjs dev/scripts/relay-repository.test.mjs dev/scripts/relay-same-cap-script-census.test.mjs dev/scripts/relay-staging-c4-refresh-workflow.test.mjs dev/scripts/relay-staging-capacity-identity.test.mjs dev/scripts/staging-relay-apply-guard.test.mjs dev/scripts/validate-relay-capacity-plan.test.mjs dev/scripts/verify-relay-capacity-transition.test.mjs dev/scripts/verify-relay-legacy-bootstrap.test.mjs dev/scripts/workload-identity-attribute-conditions.test.mjs", + "test": "pnpm -r test && node --test dev/scripts/classify-relay-production-capacity-director.test.mjs dev/scripts/classify-relay-staging-bootstrap.test.mjs dev/scripts/deploy-relay-blue-green.test.mjs dev/scripts/deploy-relay-gce-candidate.test.mjs dev/scripts/deploy-relay-gce-multi-target.test.mjs dev/scripts/github-smoke-token.test.mjs dev/scripts/infra.test.mjs dev/scripts/operate-relay-regional-rehome.test.mjs dev/scripts/power-staging-relay.test.mjs dev/scripts/prepare-relay-capacity-canary.test.mjs dev/scripts/prepare-relay-production-capacity-canary.test.mjs dev/scripts/probe-relay-legacy-admission.test.mjs dev/scripts/probe-relay-rehome-trust.test.mjs dev/scripts/production-cell-image-digest-consistency.test.mjs dev/scripts/push-gateway-workflow.test.mjs dev/scripts/push-gateway-recovery.test.mjs dev/scripts/read-relay-production-capacity-identity.test.mjs dev/scripts/relay-admin-endpoint-retry-workflow.test.mjs dev/scripts/relay-admin-transient-retry.test.mjs dev/scripts/relay-admission-selector.test.mjs dev/scripts/relay-gce-terraform-fence.test.mjs dev/scripts/relay-load-connection-failure.test.mjs dev/scripts/relay-load-control-peer.test.mjs dev/scripts/relay-load-director-capacity-gate.test.mjs dev/scripts/relay-load-model.test.mjs dev/scripts/relay-load-phase-barrier.test.mjs dev/scripts/relay-load-placement-boundary.test.mjs dev/scripts/relay-load-profile.test.mjs dev/scripts/relay-load-rebind-boundary.test.mjs dev/scripts/relay-load-region-behavior.test.mjs dev/scripts/relay-load-request-unit-boundary.test.mjs dev/scripts/relay-load-run-lifecycle.test.mjs dev/scripts/relay-monitor-evidence.test.mjs dev/scripts/relay-production-capacity-wave.test.mjs dev/scripts/relay-production-capacity-workflow.test.mjs dev/scripts/relay-production-identity-boundaries.test.mjs dev/scripts/relay-production-same-cap-wave.test.mjs dev/scripts/relay-public-workflow-contract.test.mjs dev/scripts/relay-recovery-wave-gate.test.mjs dev/scripts/relay-region-observation-evidence.test.mjs dev/scripts/relay-regional-rehome-workflow.test.mjs dev/scripts/relay-rehome-aggregate-evidence.test.mjs dev/scripts/relay-repository.test.mjs dev/scripts/relay-same-cap-script-census.test.mjs dev/scripts/relay-staging-c4-refresh-workflow.test.mjs dev/scripts/relay-staging-capacity-identity.test.mjs dev/scripts/staging-relay-apply-guard.test.mjs dev/scripts/validate-relay-capacity-plan.test.mjs dev/scripts/verify-relay-capacity-transition.test.mjs dev/scripts/verify-relay-legacy-bootstrap.test.mjs dev/scripts/workload-identity-attribute-conditions.test.mjs", "typecheck": "pnpm -r typecheck" }, "devDependencies": { diff --git a/cloud/packages/postgres-schema/package.json b/cloud/packages/postgres-schema/package.json new file mode 100644 index 00000000000..e170973cf2b --- /dev/null +++ b/cloud/packages/postgres-schema/package.json @@ -0,0 +1,20 @@ +{ + "name": "@orca-cloud/postgres-schema", + "version": "0.0.0", + "private": true, + "type": "module", + "main": "dist/index.js", + "types": "dist/index.d.ts", + "scripts": { + "build": "tsc -p tsconfig.build.json", + "clean": "node -e \"require('fs').rmSync('dist', { recursive: true, force: true })\"", + "lint": "tsc -p tsconfig.json --noEmit", + "test": "pnpm build", + "typecheck": "tsc -p tsconfig.json --noEmit" + }, + "devDependencies": { + "@types/node": "^24.10.0", + "typescript": "^5.9.3", + "vitest": "^4.0.8" + } +} diff --git a/cloud/packages/postgres-schema/src/index.ts b/cloud/packages/postgres-schema/src/index.ts new file mode 100644 index 00000000000..10c144b0ad3 --- /dev/null +++ b/cloud/packages/postgres-schema/src/index.ts @@ -0,0 +1,103 @@ +const RETRYABLE_SCHEMA_CODES = new Set(['55P03', '57014']) +const DEFAULT_RETRY_DEADLINE_MS = 30_000 +const RETRY_BASE_DELAY_MS = 250 +const RETRY_MAX_DELAY_MS = 2_000 + +type SchemaStartupOptions = { + eventPrefix?: string + now?: () => number + random?: () => number + retryDeadlineMs?: number + wait?: (delayMs: number) => Promise +} + +function retryDelayMs(attempt: number, random: () => number): number { + const ceiling = Math.min(RETRY_BASE_DELAY_MS * 2 ** (attempt - 1), RETRY_MAX_DELAY_MS) + return Math.ceil(ceiling * (0.5 + random() * 0.5)) +} + +function wait(delayMs: number): Promise { + return new Promise((resolve) => setTimeout(resolve, delayMs)) +} + +const CREATE_TABLE_IF_NOT_EXISTS = /^\s*CREATE\s+TABLE\s+IF\s+NOT\s+EXISTS\b/i +const CREATE_INDEX_IF_NOT_EXISTS = /^\s*CREATE\s+(?:UNIQUE\s+)?INDEX\s+IF\s+NOT\s+EXISTS\b/i + +// `IF NOT EXISTS` only checks the name before the catalog inserts, so the loser of a concurrent +// CREATE can fail on the catalog unique index (23505) or, when the winner has already committed by +// the time the loser reaches TypeCreate/heap_create_with_catalog, on the name check those routines +// repeat (42710 duplicate type, 42P07 duplicate relation). Each is a no-op on the next attempt. +function concurrentCreateCollision( + value: { code?: unknown; constraint?: unknown }, + statement: string +): boolean { + if (CREATE_TABLE_IF_NOT_EXISTS.test(statement)) { + return ( + (value.code === '23505' && value.constraint === 'pg_type_typname_nsp_index') || + value.code === '42710' || + value.code === '42P07' + ) + } + if (CREATE_INDEX_IF_NOT_EXISTS.test(statement)) { + return ( + (value.code === '23505' && value.constraint === 'pg_class_relname_nsp_index') || + value.code === '42P07' + ) + } + return false +} + +function retryableSchemaError(error: unknown, statement: string): boolean { + const value = error as { code?: unknown; constraint?: unknown } + return ( + RETRYABLE_SCHEMA_CODES.has(String(value.code)) || concurrentCreateCollision(value, statement) + ) +} + +export async function applyPostgresSchema( + statements: string[], + query: (statement: string) => Promise, + options: SchemaStartupOptions = {} +): Promise { + const now = options.now ?? Date.now + const random = options.random ?? Math.random + const pause = options.wait ?? wait + const deadlineAt = now() + (options.retryDeadlineMs ?? DEFAULT_RETRY_DEADLINE_MS) + + for (const statement of statements) { + let attempt = 1 + while (true) { + try { + await query(statement) + break + } catch (error) { + const code = String((error as { code?: unknown }).code) + const remainingMs = deadlineAt - now() + const retryable = retryableSchemaError(error, statement) + if (!retryable || remainingMs <= 0) { + if (retryable) { + console.warn( + JSON.stringify({ + event: `${options.eventPrefix ?? 'orca_relay_postgres_schema'}_retry_exhausted`, + code, + attempts: attempt + }) + ) + } + throw error + } + const delayMs = Math.min(remainingMs, retryDelayMs(attempt, random)) + console.warn( + JSON.stringify({ + event: `${options.eventPrefix ?? 'orca_relay_postgres_schema'}_retry`, + code, + attempt, + delayMs + }) + ) + await pause(delayMs) + attempt += 1 + } + } + } +} diff --git a/cloud/packages/postgres-schema/tsconfig.build.json b/cloud/packages/postgres-schema/tsconfig.build.json new file mode 100644 index 00000000000..94c84b60803 --- /dev/null +++ b/cloud/packages/postgres-schema/tsconfig.build.json @@ -0,0 +1,11 @@ +{ + "extends": "./tsconfig.json", + "compilerOptions": { + "declaration": true, + "emitDeclarationOnly": false, + "noEmit": false, + "outDir": "dist", + "rootDir": "src" + }, + "exclude": ["src/**/*.test.ts"] +} diff --git a/cloud/packages/postgres-schema/tsconfig.json b/cloud/packages/postgres-schema/tsconfig.json new file mode 100644 index 00000000000..a552e34dbe9 --- /dev/null +++ b/cloud/packages/postgres-schema/tsconfig.json @@ -0,0 +1,5 @@ +{ + "extends": "../../tsconfig.base.json", + "compilerOptions": { "noEmit": true }, + "include": ["src/**/*.ts"] +} diff --git a/cloud/packages/push-contract/package.json b/cloud/packages/push-contract/package.json new file mode 100644 index 00000000000..072b5e7193f --- /dev/null +++ b/cloud/packages/push-contract/package.json @@ -0,0 +1,23 @@ +{ + "name": "@orca-cloud/push-contract", + "private": true, + "version": "0.0.0", + "type": "module", + "main": "dist/index.js", + "types": "dist/index.d.ts", + "scripts": { + "build": "pnpm clean && tsc -p tsconfig.build.json", + "clean": "node -e \"require('fs').rmSync('dist', { recursive: true, force: true })\"", + "lint": "tsc -p tsconfig.json --noEmit", + "test": "vitest run", + "typecheck": "tsc -p tsconfig.json --noEmit" + }, + "dependencies": { + "zod": "^3.25.76" + }, + "devDependencies": { + "@types/node": "^24.10.0", + "typescript": "^5.9.3", + "vitest": "^4.0.8" + } +} diff --git a/cloud/packages/push-contract/src/apns-token-length.test.ts b/cloud/packages/push-contract/src/apns-token-length.test.ts new file mode 100644 index 00000000000..ec67383fefe --- /dev/null +++ b/cloud/packages/push-contract/src/apns-token-length.test.ts @@ -0,0 +1,27 @@ +import { expect, it } from 'vitest' +import { PushDeviceRegistrationRequestSchema } from './device-registration-messages.js' + +const registration = (token: string) => ({ + v: 1, + deviceId: 'qa-device', + platform: 'ios', + token, + apnsEnvironment: 'sandbox', + filter: { sources: ['agent-task-complete'], agentStates: ['finished'] } +}) + +it.each([32, 64, 160, 256])( + 'accepts variable-length APNs device tokens (%i hex characters)', + (length) => { + expect( + PushDeviceRegistrationRequestSchema.safeParse(registration('aB'.repeat(length / 2))).success + ).toBe(true) + } +) + +it.each(['', 'abc', 'not-hex', 'ab cd', 'ab'.repeat(2049)])( + 'rejects malformed or oversized APNs tokens', + (token) => { + expect(PushDeviceRegistrationRequestSchema.safeParse(registration(token)).success).toBe(false) + } +) diff --git a/cloud/packages/push-contract/src/contract.test.ts b/cloud/packages/push-contract/src/contract.test.ts new file mode 100644 index 00000000000..e81ac2ad02f --- /dev/null +++ b/cloud/packages/push-contract/src/contract.test.ts @@ -0,0 +1,216 @@ +import { describe, expect, it } from 'vitest' +import { + ApnsEnvironmentSchema, + PushDeviceListResponseSchema, + PushDeviceRegistrationRequestSchema, + PushDeviceRegistrationResponseSchema, + PushNotificationFilterSchema +} from './device-registration-messages.js' +import { + PushErrorResponseSchema, + PushHostChallengeRequestSchema, + PushHostChallengeResponseSchema, + PushHostSessionRequestSchema, + PushHostSessionResponseSchema +} from './host-auth-messages.js' +import { PUSH_DEFAULTS, PUSH_LIMITS } from './push-limits.js' + +const KEY_B64 = Buffer.alloc(32, 1).toString('base64') +const NONCE_B64 = Buffer.alloc(24, 2).toString('base64') +const SESSION_TOKEN = Buffer.alloc(32, 3).toString('base64url') +const FINGERPRINT = 'abcdefghijklmnop' +const APNS_TOKEN = 'a'.repeat(64) +const FCM_TOKEN = 'cQ1abcDEF_gh:APA91bZZ-zz0123456789abcdefghijklmnopqrstuvwxyz' + +function notification(): Record { + return { + notificationId: 'note-1', + notificationSeq: 4, + notificationEpoch: '5c9e9a1e-0000-4000-8000-000000000000', + source: 'agent-task-complete', + agentState: 'needs-input', + title: 'Agent needs input', + body: 'Waiting on your answer', + worktreeId: 'wt-1' + } +} + +describe('push contract limits', () => { + it('locks the normative limits the desktop and gateway both assume', () => { + expect(PUSH_LIMITS).toMatchObject({ + titleMaxChars: 80, + bodyMaxChars: 180, + maxRegistrationIdsPerSend: 20, + maxDevicesPerHost: 64, + maxDevicesPerListResponse: 1_024, + hostSendsPerRollingHour: 60, + registrationSendsPerRollingDay: 200, + coalesceWindowMs: 3_000, + challengeTtlMs: 10_000, + clockSkewToleranceMs: 30_000, + sessionTtlMs: 86_400_000, + sendLogRetentionMs: 90_000_000, + notificationTtlSeconds: 14_400, + apnsCollapseIdMaxBytes: 64, + hostRetentionMs: 3_600_000, + unauthenticatedRequestsPerMinutePerIp: 30, + authenticatedRequestsPerMinutePerIp: 240 + }) + expect(PUSH_DEFAULTS.apnsTopic).toBe('com.stably.orca.mobile') + expect(PUSH_DEFAULTS.fcmProjectId).toBe('onorca-cloud') + expect(PUSH_DEFAULTS.androidChannelId).toBe('orca-desktop') + }) +}) + +describe('host authentication schemas', () => { + it('accepts a well formed challenge round trip', () => { + expect( + PushHostChallengeRequestSchema.safeParse({ v: 1, hostPublicKeyB64: KEY_B64 }).success + ).toBe(true) + expect( + PushHostChallengeResponseSchema.safeParse({ + challengeId: 'challenge-1', + gatewayEphemeralPublicKeyB64: KEY_B64, + nonceB64: NONCE_B64, + ciphertextB64: Buffer.alloc(96, 5).toString('base64'), + expiresAt: 1_700_000_010_000 + }).success + ).toBe(true) + expect( + PushHostSessionRequestSchema.safeParse({ + v: 1, + challengeId: 'challenge-1', + proofB64: KEY_B64 + }).success + ).toBe(true) + expect( + PushHostSessionResponseSchema.safeParse({ + sessionToken: SESSION_TOKEN, + expiresAt: 1_700_086_400_000, + hostFingerprint: FINGERPRINT + }).success + ).toBe(true) + }) + + it('rejects unknown keys, wrong versions, and mis-sized keys', () => { + expect( + PushHostChallengeRequestSchema.safeParse({ + v: 1, + hostPublicKeyB64: KEY_B64, + extra: true + }).success + ).toBe(false) + expect(PushHostChallengeRequestSchema.safeParse({ v: 2, hostPublicKeyB64: KEY_B64 }).success) + .toBe(false) + expect( + PushHostChallengeRequestSchema.safeParse({ + v: 1, + hostPublicKeyB64: Buffer.alloc(31, 1).toString('base64') + }).success + ).toBe(false) + expect( + PushHostSessionResponseSchema.safeParse({ + sessionToken: SESSION_TOKEN, + expiresAt: 1_700_086_400_000, + hostFingerprint: 'short' + }).success + ).toBe(false) + }) + + it('names only the error codes the gateway may return', () => { + expect(PushErrorResponseSchema.safeParse({ error: 'session_expired' }).success).toBe(true) + expect(PushErrorResponseSchema.safeParse({ error: 'too_many_devices' }).success).toBe(true) + expect(PushErrorResponseSchema.safeParse({ error: 'rate_limited' }).success).toBe(true) + expect(PushErrorResponseSchema.safeParse({ error: 'teapot' }).success).toBe(false) + }) +}) + +describe('device registration schemas', () => { + it('requires an apns environment and a hex token for ios', () => { + expect( + PushDeviceRegistrationRequestSchema.safeParse({ + v: 1, + deviceId: 'device-1', + platform: 'ios', + token: APNS_TOKEN, + apnsEnvironment: 'sandbox', + filter: { sources: ['agent-task-complete'], agentStates: ['needs-input'] } + }).success + ).toBe(true) + expect( + PushDeviceRegistrationRequestSchema.safeParse({ + v: 1, + deviceId: 'device-1', + platform: 'ios', + token: APNS_TOKEN, + filter: { sources: [], agentStates: [] } + }).success + ).toBe(false) + expect( + PushDeviceRegistrationRequestSchema.safeParse({ + v: 1, + deviceId: 'device-1', + platform: 'ios', + token: 'not-hex', + apnsEnvironment: 'production', + filter: { sources: [], agentStates: [] } + }).success + ).toBe(false) + }) + + it('rejects an apns environment on android and accepts an fcm token', () => { + expect( + PushDeviceRegistrationRequestSchema.safeParse({ + v: 1, + deviceId: 'device-2', + platform: 'android', + token: FCM_TOKEN, + filter: { sources: ['plugin', 'terminal-bell'], agentStates: [] } + }).success + ).toBe(true) + expect( + PushDeviceRegistrationRequestSchema.safeParse({ + v: 1, + deviceId: 'device-2', + platform: 'android', + token: FCM_TOKEN, + apnsEnvironment: 'sandbox', + filter: { sources: [], agentStates: [] } + }).success + ).toBe(false) + }) + + it('rejects duplicate filter entries and unknown filter keys', () => { + expect( + PushNotificationFilterSchema.safeParse({ + sources: ['plugin', 'plugin'], + agentStates: [] + }).success + ).toBe(false) + expect( + PushNotificationFilterSchema.safeParse({ + sources: [], + agentStates: ['finished'], + worktrees: [] + }).success + ).toBe(false) + expect(ApnsEnvironmentSchema.safeParse('adhoc').success).toBe(false) + }) + + it('shapes the registration and list responses', () => { + expect(PushDeviceRegistrationResponseSchema.safeParse({ registrationId: 'reg-1' }).success) + .toBe(true) + expect( + PushDeviceListResponseSchema.safeParse({ + devices: [ + { registrationId: 'reg-1', deviceId: 'device-1', platform: 'ios', dead: false } + ] + }).success + ).toBe(true) + expect( + PushDeviceListResponseSchema.safeParse({ + devices: [{ registrationId: 'reg-1', deviceId: 'device-1', platform: 'ios' }] + }).success + ).toBe(false) + }) +}) diff --git a/cloud/packages/push-contract/src/device-registration-messages.ts b/cloud/packages/push-contract/src/device-registration-messages.ts new file mode 100644 index 00000000000..d13e5094861 --- /dev/null +++ b/cloud/packages/push-contract/src/device-registration-messages.ts @@ -0,0 +1,104 @@ +import { z } from 'zod' +import { PUSH_LIMITS } from './push-limits.js' +import { OpaqueIdSchema } from './wire-scalars.js' + +export const PushPlatformSchema = z.enum(['ios', 'android']) +export const ApnsEnvironmentSchema = z.enum(['sandbox', 'production']) +export const PushNotificationSourceSchema = z.enum([ + 'agent-task-complete', + 'terminal-bell', + 'plugin' +]) +export const PushAgentStateSchema = z.enum(['needs-input', 'finished']) + +// APNs tokens are variable-length byte strings, including longer simulator tokens. +const APNS_TOKEN_PATTERN = /^(?:[0-9a-fA-F]{2})+$/ +const FCM_TOKEN_PATTERN = /^[A-Za-z0-9_:.\-]{32,4096}$/ + +export const PushNotificationFilterSchema = z + .object({ + sources: z.array(PushNotificationSourceSchema).max(3), + agentStates: z.array(PushAgentStateSchema).max(2) + }) + .strict() + .superRefine((value, context) => { + if (new Set(value.sources).size !== value.sources.length) { + context.addIssue({ code: 'custom', path: ['sources'], message: 'sources must be unique' }) + } + if (new Set(value.agentStates).size !== value.agentStates.length) { + context.addIssue({ + code: 'custom', + path: ['agentStates'], + message: 'agentStates must be unique' + }) + } + }) + +export const PushDeviceRegistrationRequestSchema = z + .object({ + v: z.literal(1), + deviceId: OpaqueIdSchema, + platform: PushPlatformSchema, + token: z.string().min(1).max(4096), + apnsEnvironment: ApnsEnvironmentSchema.optional(), + filter: PushNotificationFilterSchema + }) + .strict() + .superRefine((value, context) => { + if (value.platform === 'ios') { + if (value.apnsEnvironment === undefined) { + context.addIssue({ + code: 'custom', + path: ['apnsEnvironment'], + message: 'apnsEnvironment is required for ios' + }) + } + if (!APNS_TOKEN_PATTERN.test(value.token)) { + context.addIssue({ + code: 'custom', + path: ['token'], + message: 'ios token must be hex-encoded bytes' + }) + } + return + } + if (value.apnsEnvironment !== undefined) { + context.addIssue({ + code: 'custom', + path: ['apnsEnvironment'], + message: 'apnsEnvironment is ios only' + }) + } + if (!FCM_TOKEN_PATTERN.test(value.token)) { + context.addIssue({ + code: 'custom', + path: ['token'], + message: 'android token must be an FCM registration string' + }) + } + }) + +export const PushDeviceRegistrationResponseSchema = z + .object({ registrationId: OpaqueIdSchema }) + .strict() + +export const PushDeviceSummarySchema = z + .object({ + registrationId: OpaqueIdSchema, + deviceId: OpaqueIdSchema, + platform: PushPlatformSchema, + dead: z.boolean() + }) + .strict() + +export const PushDeviceListResponseSchema = z + .object({ devices: z.array(PushDeviceSummarySchema).max(PUSH_LIMITS.maxDevicesPerListResponse) }) + .strict() + +export type PushPlatform = z.infer +export type ApnsEnvironment = z.infer +export type PushNotificationSource = z.infer +export type PushAgentState = z.infer +export type PushNotificationFilter = z.infer +export type PushDeviceRegistrationRequest = z.infer +export type PushDeviceSummary = z.infer diff --git a/cloud/packages/push-contract/src/host-auth-messages.ts b/cloud/packages/push-contract/src/host-auth-messages.ts new file mode 100644 index 00000000000..01085af543c --- /dev/null +++ b/cloud/packages/push-contract/src/host-auth-messages.ts @@ -0,0 +1,59 @@ +import { z } from 'zod' +import { + Base6432ByteSchema, + Base64Raw24ByteSchema, + Base64Url32ByteSchema, + BoundedCiphertextSchema, + EpochMsSchema, + OpaqueIdSchema, + PushHostFingerprintSchema +} from './wire-scalars.js' + +export const PushHostChallengeRequestSchema = z + .object({ v: z.literal(1), hostPublicKeyB64: Base6432ByteSchema }) + .strict() + +export const PushHostChallengeResponseSchema = z + .object({ + challengeId: OpaqueIdSchema, + gatewayEphemeralPublicKeyB64: Base6432ByteSchema, + nonceB64: Base64Raw24ByteSchema, + ciphertextB64: BoundedCiphertextSchema, + expiresAt: EpochMsSchema + }) + .strict() + +export const PushHostSessionRequestSchema = z + .object({ v: z.literal(1), challengeId: OpaqueIdSchema, proofB64: Base6432ByteSchema }) + .strict() + +export const PushHostSessionResponseSchema = z + .object({ + sessionToken: Base64Url32ByteSchema, + expiresAt: EpochMsSchema, + hostFingerprint: PushHostFingerprintSchema + }) + .strict() + +export const PUSH_ERROR_CODES = [ + 'invalid_request', + 'invalid_challenge', + 'invalid_proof', + 'invalid_token', + 'session_expired', + 'not_found', + 'too_many_devices', + 'request_too_large', + 'rate_limited', + 'dependency_unavailable' +] as const + +export const PushErrorResponseSchema = z + .object({ error: z.enum(PUSH_ERROR_CODES) }) + .strict() + +export type PushHostChallengeRequest = z.infer +export type PushHostChallengeResponse = z.infer +export type PushHostSessionRequest = z.infer +export type PushHostSessionResponse = z.infer +export type PushErrorCode = (typeof PUSH_ERROR_CODES)[number] diff --git a/cloud/packages/push-contract/src/index.ts b/cloud/packages/push-contract/src/index.ts new file mode 100644 index 00000000000..3bd8a871f28 --- /dev/null +++ b/cloud/packages/push-contract/src/index.ts @@ -0,0 +1,6 @@ +export * from './device-registration-messages.js' +export * from './host-auth-messages.js' +export * from './push-host-proof-transcript.js' +export * from './push-limits.js' +export * from './send-messages.js' +export * from './wire-scalars.js' diff --git a/cloud/packages/push-contract/src/notification-identity-limits.test.ts b/cloud/packages/push-contract/src/notification-identity-limits.test.ts new file mode 100644 index 00000000000..e19fd93140a --- /dev/null +++ b/cloud/packages/push-contract/src/notification-identity-limits.test.ts @@ -0,0 +1,32 @@ +import { expect, it } from 'vitest' +import { PushNotificationSchema } from './send-messages.js' +const base = { + source: 'agent-task-complete', + agentState: 'finished', + notificationSeq: 1, + notificationEpoch: 'epoch', + title: 'Done', + body: '' +} +it.each([ + 'repo::/Users/developer/orca/workspaces/monorepo/packages/desktop/integrations/feature-mobile-background-notifications', + 'repo::C:\\Users\\developer\\Documents\\projects\\monorepo\\packages\\desktop\\feature-mobile-notifications', + 'folder::/home/developer/projects/通知/作業ディレクトリ/機能', + 'ssh:host::/home/developer/workspaces/monorepo/packages/desktop/feature-mobile-background-notifications' +])('preserves long desktop identities: %s', (path) => { + const worktreeId = `12345678-1234-1234-1234-123456789012::${path}` + const notificationId = [ + 'agent', + encodeURIComponent(worktreeId), + encodeURIComponent('12345678-1234-1234-1234-123456789012:87654321-4321-4321-4321-210987654321'), + '1780000000123' + ].join(':') + const result = PushNotificationSchema.parse({ ...base, worktreeId, notificationId }) + expect(result.worktreeId).toBe(worktreeId) + expect(result.notificationId).toBe(notificationId) +}) +it('rejects oversized provider data by UTF-8 bytes instead of truncating identities', () => { + expect(PushNotificationSchema.safeParse({ ...base, worktreeId: '界'.repeat(1100) }).success).toBe( + false + ) +}) diff --git a/cloud/packages/push-contract/src/push-host-proof-transcript.test.ts b/cloud/packages/push-contract/src/push-host-proof-transcript.test.ts new file mode 100644 index 00000000000..34423beaf6d --- /dev/null +++ b/cloud/packages/push-contract/src/push-host-proof-transcript.test.ts @@ -0,0 +1,106 @@ +import { describe, expect, it } from 'vitest' +import { + buildPushHostChallengePlaintext, + buildPushHostProofMacInput, + buildPushHostProofTranscript, + PUSH_HOST_CHALLENGE_PLAINTEXT_DOMAIN, + PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN, + PUSH_HOST_PROOF_TRANSCRIPT_FIELD_COUNT +} from './push-host-proof-transcript.js' +import { PUSH_LIMITS } from './push-limits.js' + +const transcriptInput = { + gatewayOrigin: 'https://push.onorca.dev', + gatewayEphemeralPublicKey: new Uint8Array(32).fill(7), + challengeNonce: new Uint8Array(24).fill(9), + challengeId: 'challenge-1', + issuedAt: 1_700_000_000_000, + expiresAt: 1_700_000_000_000 + PUSH_LIMITS.challengeTtlMs, + hostFingerprint: 'abcdefghijklmnop', + hostPublicKey: new Uint8Array(32).fill(4) +} + +describe('push host proof transcript', () => { + it('is deterministic and order dependent', () => { + const first = buildPushHostProofTranscript(transcriptInput) + const second = buildPushHostProofTranscript({ ...transcriptInput }) + expect(Buffer.from(first).equals(Buffer.from(second))).toBe(true) + const different = buildPushHostProofTranscript({ + ...transcriptInput, + challengeId: 'challenge-2' + }) + expect(Buffer.from(first).equals(Buffer.from(different))).toBe(false) + }) + + it('encodes exactly the ten specified fields in order', () => { + const transcript = buildPushHostProofTranscript(transcriptInput) + const view = new DataView(transcript.buffer, transcript.byteOffset, transcript.byteLength) + const names: string[] = [] + let offset = 0 + while (offset < transcript.byteLength) { + const nameLength = view.getUint32(offset, false) + offset += 4 + names.push(Buffer.from(transcript.slice(offset, offset + nameLength)).toString('utf8')) + offset += nameLength + offset += 4 + view.getUint32(offset, false) + } + expect(names).toEqual([ + 'protocol', + 'version', + 'gatewayOrigin', + 'gatewayEphemeralPublicKey', + 'challengeNonce', + 'challengeId', + 'issuedAt', + 'expiresAt', + 'hostFingerprint', + 'hostPublicKey' + ]) + expect(names).toHaveLength(PUSH_HOST_PROOF_TRANSCRIPT_FIELD_COUNT) + expect(offset).toBe(transcript.byteLength) + }) + + it('rejects mis-sized key material', () => { + expect(() => + buildPushHostProofTranscript({ + ...transcriptInput, + hostPublicKey: new Uint8Array(31) + }) + ).toThrow('hostPublicKey must be 32 bytes') + expect(() => + buildPushHostProofTranscript({ ...transcriptInput, challengeNonce: new Uint8Array(23) }) + ).toThrow('challengeNonce must be 24 bytes') + }) + + it('frames the challenge plaintext as domain, length, transcript, secret', () => { + const transcript = buildPushHostProofTranscript(transcriptInput) + const secret = new Uint8Array(32).fill(11) + const plaintext = buildPushHostChallengePlaintext(transcript, secret) + const domain = Buffer.from(`${PUSH_HOST_CHALLENGE_PLAINTEXT_DOMAIN}\0`, 'utf8') + expect(Buffer.from(plaintext.slice(0, domain.byteLength)).equals(domain)).toBe(true) + const declared = new DataView( + plaintext.buffer, + plaintext.byteOffset + domain.byteLength, + 4 + ).getUint32(0, false) + expect(declared).toBe(transcript.byteLength) + expect(plaintext.byteLength).toBe(domain.byteLength + 4 + transcript.byteLength + 32) + expect( + Buffer.from(plaintext.slice(plaintext.byteLength - 32)).equals(Buffer.from(secret)) + ).toBe(true) + expect(() => buildPushHostChallengePlaintext(transcript, new Uint8Array(16))).toThrow( + 'challengeSecret must be 32 bytes' + ) + }) + + it('separates the ack mac input from the challenge domain', () => { + const transcript = buildPushHostProofTranscript(transcriptInput) + const macInput = buildPushHostProofMacInput(transcript) + expect(Buffer.from(macInput).toString('utf8')).toContain( + `${PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN}\0ack\0` + ) + expect(macInput.byteLength).toBe( + Buffer.byteLength(`${PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN}\0ack\0`) + transcript.byteLength + ) + }) +}) diff --git a/cloud/packages/push-contract/src/push-host-proof-transcript.ts b/cloud/packages/push-contract/src/push-host-proof-transcript.ts new file mode 100644 index 00000000000..a375b18ca76 --- /dev/null +++ b/cloud/packages/push-contract/src/push-host-proof-transcript.ts @@ -0,0 +1,90 @@ +const textEncoder = new TextEncoder() + +export const PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN = 'orca-push-host-proof/v1' +export const PUSH_HOST_CHALLENGE_PLAINTEXT_DOMAIN = 'orca-push-host-challenge/v1' +export const PUSH_HOST_CHALLENGE_BOX_ALGORITHM = 'Curve25519-XSalsa20-Poly1305' +export const PUSH_HOST_PROOF_ALGORITHM = 'HMAC-SHA-256' + +export interface PushHostProofTranscriptInput { + gatewayOrigin: string + gatewayEphemeralPublicKey: Uint8Array + challengeNonce: Uint8Array + challengeId: string + issuedAt: number + expiresAt: number + hostFingerprint: string + hostPublicKey: Uint8Array +} + +export const PUSH_HOST_PROOF_TRANSCRIPT_FIELD_COUNT = 10 + +function uint32(value: number): Uint8Array { + const bytes = new Uint8Array(4) + new DataView(bytes.buffer).setUint32(0, value, false) + return bytes +} + +function uint64(value: number): Uint8Array { + const bytes = new Uint8Array(8) + new DataView(bytes.buffer).setBigUint64(0, BigInt(value), false) + return bytes +} + +function concat(parts: readonly Uint8Array[]): Uint8Array { + const output = new Uint8Array(parts.reduce((total, part) => total + part.byteLength, 0)) + let offset = 0 + for (const part of parts) { + output.set(part, offset) + offset += part.byteLength + } + return output +} + +function field(name: string, value: Uint8Array): Uint8Array { + const encodedName = textEncoder.encode(name) + return concat([uint32(encodedName.byteLength), encodedName, uint32(value.byteLength), value]) +} + +function text(value: string): Uint8Array { + return textEncoder.encode(value) +} + +function requireByteLength(value: Uint8Array, expected: number, name: string): void { + if (value.byteLength !== expected) throw new Error(`${name} must be ${expected} bytes`) +} + +export function buildPushHostProofTranscript(input: PushHostProofTranscriptInput): Uint8Array { + requireByteLength(input.gatewayEphemeralPublicKey, 32, 'gatewayEphemeralPublicKey') + requireByteLength(input.challengeNonce, 24, 'challengeNonce') + requireByteLength(input.hostPublicKey, 32, 'hostPublicKey') + return concat([ + field('protocol', text(PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN)), + field('version', new Uint8Array([1])), + field('gatewayOrigin', text(input.gatewayOrigin)), + field('gatewayEphemeralPublicKey', input.gatewayEphemeralPublicKey), + field('challengeNonce', input.challengeNonce), + field('challengeId', text(input.challengeId)), + field('issuedAt', uint64(input.issuedAt)), + field('expiresAt', uint64(input.expiresAt)), + field('hostFingerprint', text(input.hostFingerprint)), + field('hostPublicKey', input.hostPublicKey) + ]) +} + +export function buildPushHostChallengePlaintext( + transcript: Uint8Array, + challengeSecret: Uint8Array +): Uint8Array { + if (challengeSecret.byteLength !== 32) throw new Error('challengeSecret must be 32 bytes') + // Why: the encrypted random secret makes the public transcript insufficient to forge the ack. + return concat([ + text(`${PUSH_HOST_CHALLENGE_PLAINTEXT_DOMAIN}\0`), + uint32(transcript.byteLength), + transcript, + challengeSecret + ]) +} + +export function buildPushHostProofMacInput(transcript: Uint8Array): Uint8Array { + return concat([text(`${PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN}\0ack\0`), transcript]) +} diff --git a/cloud/packages/push-contract/src/push-host-proof-vector.json b/cloud/packages/push-contract/src/push-host-proof-vector.json new file mode 100644 index 00000000000..128eba46980 --- /dev/null +++ b/cloud/packages/push-contract/src/push-host-proof-vector.json @@ -0,0 +1,16 @@ +{ + "hostSecretKeyB64": "BwcHBwcHBwcHBwcHBwcHBwcHBwcHBwcHBwcHBwcHBwc=", + "hostPublicKeyB64": "E75P6uryBMf9M1j8nAByGIHRdCeBKCJ+xnTzf3/pe20=", + "hostFingerprint": "D20lU_8MD0R64gLt", + "gatewayOrigin": "https://push.onorca.dev", + "challenge": { + "challengeId": "vector-challenge-1", + "gatewayEphemeralPublicKeyB64": "V9tLNZ8jrl4Ubk4lEgVnBHIlBjSMFQwUdT0Mkz0E1CE=", + "nonceB64": "AwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMD", + "ciphertextB64": "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", + "expiresAt": 1800000010000 + }, + "issuedAt": 1800000000000, + "challengeSecretB64": "BQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQU=", + "transcriptB64": "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" +} diff --git a/cloud/packages/push-contract/src/push-limits.ts b/cloud/packages/push-contract/src/push-limits.ts new file mode 100644 index 00000000000..5d46b994d06 --- /dev/null +++ b/cloud/packages/push-contract/src/push-limits.ts @@ -0,0 +1,43 @@ +export const PUSH_LIMITS = { + titleMaxChars: 80, + bodyMaxChars: 180, + maxRegistrationIdsPerSend: 20, + // A host pairs phones, not a fleet. The cap bounds what one session can write + // through a caller-chosen deviceId. + maxDevicesPerHost: 64, + // The list response is bounded well above the per-host cap so the query LIMIT + // and the response schema can never disagree. + maxDevicesPerListResponse: 1024, + maxHttpBodyBytes: 16 * 1024, + hostSendsPerRollingHour: 60, + registrationSendsPerRollingDay: 200, + coalesceWindowMs: 3_000, + challengeTtlMs: 10_000, + // Covers routine NTP drift without extending the signed challenge window. + clockSkewToleranceMs: 30_000, + sessionTtlMs: 24 * 60 * 60 * 1000, + // One hour past the widest quota window so a rolling day never reads a pruned row. + sendLogRetentionMs: 25 * 60 * 60 * 1000, + notificationTtlSeconds: 4 * 60 * 60, + apnsCollapseIdMaxBytes: 64, + // Nothing reads a host row, and any keypair mints one for free, so a host + // with no registration left is kept only long enough to survive a phone swap. + hostRetentionMs: 60 * 60 * 1000, + // The challenge and session routes are the only unauthenticated writes, so + // they are capped per client IP before any key material is generated. + unauthenticatedRequestsPerMinutePerIp: 30, + // Every other route looks its bearer up in the database before it can refuse + // it, so a flood of forged bearers is capped per client IP ahead of that. + // Wide enough for an office NAT full of hosts, each of which sends at most + // its hourly quota plus a registration per connect. + authenticatedRequestsPerMinutePerIp: 240 +} as const + +export const PUSH_DEFAULTS = { + apnsTopic: 'com.stably.orca.mobile', + fcmProjectId: 'onorca-cloud', + androidChannelId: 'orca-desktop', + gatewayUrl: 'https://push.onorca.dev' +} as const + +export const PUSH_HOST_FINGERPRINT_LENGTH = 16 diff --git a/cloud/packages/push-contract/src/send-messages.test.ts b/cloud/packages/push-contract/src/send-messages.test.ts new file mode 100644 index 00000000000..8a261938c43 --- /dev/null +++ b/cloud/packages/push-contract/src/send-messages.test.ts @@ -0,0 +1,126 @@ +import { describe, expect, it } from 'vitest' +import { PUSH_LIMITS } from './push-limits.js' +import { + PushSendRequestSchema, + PushSendResponseSchema, + PushSendStatusSchema +} from './send-messages.js' + +function notification(): Record { + return { + notificationId: 'note-1', + notificationSeq: 4, + notificationEpoch: '5c9e9a1e-0000-4000-8000-000000000000', + source: 'agent-task-complete', + agentState: 'needs-input', + title: 'Agent needs input', + body: 'Waiting on your answer', + worktreeId: 'wt-1' + } +} + +describe('send schemas', () => { + it('accepts a batch at the registration cap and a terminal bell without an id', () => { + const ids = Array.from({ length: PUSH_LIMITS.maxRegistrationIdsPerSend }, (_, i) => `reg-${i}`) + expect( + PushSendRequestSchema.safeParse({ v: 1, registrationIds: ids, notification: notification() }) + .success + ).toBe(true) + const { notificationId: _dropped, ...bell } = notification() + expect( + PushSendRequestSchema.safeParse({ + v: 1, + registrationIds: ['reg-1'], + notification: { ...bell, source: 'terminal-bell', agentState: null } + }).success + ).toBe(true) + }) + + it('rejects an oversized batch, over-long copy, and unknown notification keys', () => { + const ids = Array.from( + { length: PUSH_LIMITS.maxRegistrationIdsPerSend + 1 }, + (_, i) => `reg-${i}` + ) + expect( + PushSendRequestSchema.safeParse({ v: 1, registrationIds: ids, notification: notification() }) + .success + ).toBe(false) + expect( + PushSendRequestSchema.safeParse({ + v: 1, + registrationIds: ['reg-1'], + notification: { ...notification(), title: 'x'.repeat(PUSH_LIMITS.titleMaxChars + 1) } + }).success + ).toBe(false) + expect( + PushSendRequestSchema.safeParse({ + v: 1, + registrationIds: ['reg-1'], + notification: { ...notification(), body: 'x'.repeat(PUSH_LIMITS.bodyMaxChars + 1) } + }).success + ).toBe(false) + expect( + PushSendRequestSchema.safeParse({ + v: 1, + registrationIds: ['reg-1'], + notification: { ...notification(), coalescedCount: 2 } + }).success + ).toBe(false) + expect(PushSendRequestSchema.safeParse({ v: 1, registrationIds: [], notification: notification() }).success) + .toBe(false) + }) + + it('rejects a notification id that could not be sent as a collapse header', () => { + for (const notificationId of ['line\nbreak', 'nul\0byte', 'émoji', '\t']) { + expect( + PushSendRequestSchema.safeParse({ + v: 1, + registrationIds: ['reg-1'], + notification: { ...notification(), notificationId } + }).success + ).toBe(false) + } + expect( + PushSendRequestSchema.safeParse({ + v: 1, + registrationIds: ['reg-1'], + notification: { + ...notification(), + notificationId: 'agent:repo%3A%3A%2FUsers%2Fme:pane-1:1700000000000' + } + }).success + ).toBe(true) + }) + + it('dedupes repeated registration ids and keeps the first-seen order', () => { + const parsed = PushSendRequestSchema.safeParse({ + v: 1, + registrationIds: ['reg-b', 'reg-a', 'reg-b', 'reg-c', 'reg-a'], + notification: notification() + }) + expect(parsed.success).toBe(true) + expect(parsed.success && parsed.data.registrationIds).toEqual(['reg-b', 'reg-a', 'reg-c']) + }) + + it('counts duplicates against the batch cap before deduping them', () => { + const ids = Array.from({ length: PUSH_LIMITS.maxRegistrationIdsPerSend + 1 }, () => 'reg-1') + expect( + PushSendRequestSchema.safeParse({ v: 1, registrationIds: ids, notification: notification() }) + .success + ).toBe(false) + }) + + it('locks the send result statuses', () => { + expect(PushSendStatusSchema.options).toEqual(['queued', 'dead', 'rate_limited', 'error']) + expect( + PushSendResponseSchema.safeParse({ + results: [{ registrationId: 'reg-1', status: 'queued' }] + }).success + ).toBe(true) + expect( + PushSendResponseSchema.safeParse({ + results: [{ registrationId: 'reg-1', status: 'sent' }] + }).success + ).toBe(false) + }) +}) diff --git a/cloud/packages/push-contract/src/send-messages.ts b/cloud/packages/push-contract/src/send-messages.ts new file mode 100644 index 00000000000..a088248d935 --- /dev/null +++ b/cloud/packages/push-contract/src/send-messages.ts @@ -0,0 +1,67 @@ +import { z } from 'zod' +import { + PushAgentStateSchema, + PushNotificationSourceSchema +} from './device-registration-messages.js' +import { PUSH_LIMITS } from './push-limits.js' +import { OpaqueIdSchema, SequenceSchema } from './wire-scalars.js' + +export const PushNotificationSchema = z + .object({ + // Absent for terminal-bell, which the desktop raises without a notification record. + // Printable ASCII only: the id becomes the APNs collapse header, and the + // desktop builds it from URL-encoded parts, so anything else is not Orca's. + notificationId: z + .string() + .min(1) + .max(2048) + .regex(/^[\x20-\x7e]+$/) + .optional(), + notificationSeq: SequenceSchema, + notificationEpoch: OpaqueIdSchema, + source: PushNotificationSourceSchema, + sound: z.boolean().optional(), + agentState: PushAgentStateSchema.nullable(), + title: z.string().min(1).max(PUSH_LIMITS.titleMaxChars), + body: z.string().max(PUSH_LIMITS.bodyMaxChars), + worktreeId: z.string().min(1).max(2048).optional() + }) + .strict() + .refine( + (notification) => new TextEncoder().encode(JSON.stringify(notification)).byteLength <= 3000, + { + message: 'notification exceeds provider payload budget' + } + ) + +export const PushSendRequestSchema = z + .object({ + v: z.literal(1), + // Deduped before the gateway sees it: a repeated id would otherwise reserve + // quota twice and inflate the coalesced count for one banner. + registrationIds: z + .array(OpaqueIdSchema) + .min(1) + .max(PUSH_LIMITS.maxRegistrationIdsPerSend) + .transform((ids) => [...new Set(ids)]), + notification: PushNotificationSchema + }) + .strict() + +export const PushSendStatusSchema = z.enum(['queued', 'dead', 'rate_limited', 'error']) + +export const PushSendResultSchema = z + .object({ registrationId: OpaqueIdSchema, status: PushSendStatusSchema }) + .strict() + +export const PushSendResponseSchema = z + .object({ + results: z.array(PushSendResultSchema).max(PUSH_LIMITS.maxRegistrationIdsPerSend) + }) + .strict() + +export type PushNotification = z.infer +export type PushSendRequest = z.infer +export type PushSendStatus = z.infer +export type PushSendResult = z.infer +export type PushSendResponse = z.infer diff --git a/cloud/packages/push-contract/src/wire-scalars.ts b/cloud/packages/push-contract/src/wire-scalars.ts new file mode 100644 index 00000000000..10e8effb69f --- /dev/null +++ b/cloud/packages/push-contract/src/wire-scalars.ts @@ -0,0 +1,25 @@ +import { z } from 'zod' + +// Copied from relay-contract rather than imported: the push gateway ships as a +// standalone image and must not pull the relay wire contract into its closure. +export const Base64Url32ByteSchema = z.string().regex(/^[A-Za-z0-9_-]{43}$/) +export const Base6432ByteSchema = z.string().regex(/^(?:[A-Za-z0-9+/]{4}){10}[A-Za-z0-9+/]{3}=$/) +export const Base64Raw24ByteSchema = z.string().regex(/^(?:[A-Za-z0-9+/]{4}){8}$/) +export const PushHostFingerprintSchema = z.string().regex(/^[A-Za-z0-9_-]{16}$/) +export const OpaqueIdSchema = z.string().min(1).max(128) +export const EpochMsSchema = z.number().int().nonnegative().max(Number.MAX_SAFE_INTEGER) +export const SequenceSchema = z.number().int().nonnegative().max(Number.MAX_SAFE_INTEGER) +export const BoundedCiphertextSchema = z + .string() + .min(1) + .max(16 * 1024) + .regex(/^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/) + +export const CanonicalHttpsOriginSchema = z.string().max(2048).refine((value) => { + try { + const url = new URL(value) + return url.protocol === 'https:' && url.origin === value && url.pathname === '/' + } catch { + return false + } +}, 'must be a canonical HTTPS origin') diff --git a/cloud/packages/push-contract/tsconfig.build.json b/cloud/packages/push-contract/tsconfig.build.json new file mode 100644 index 00000000000..94c84b60803 --- /dev/null +++ b/cloud/packages/push-contract/tsconfig.build.json @@ -0,0 +1,11 @@ +{ + "extends": "./tsconfig.json", + "compilerOptions": { + "declaration": true, + "emitDeclarationOnly": false, + "noEmit": false, + "outDir": "dist", + "rootDir": "src" + }, + "exclude": ["src/**/*.test.ts"] +} diff --git a/cloud/packages/push-contract/tsconfig.json b/cloud/packages/push-contract/tsconfig.json new file mode 100644 index 00000000000..a552e34dbe9 --- /dev/null +++ b/cloud/packages/push-contract/tsconfig.json @@ -0,0 +1,5 @@ +{ + "extends": "../../tsconfig.base.json", + "compilerOptions": { "noEmit": true }, + "include": ["src/**/*.ts"] +} diff --git a/cloud/pnpm-lock.yaml b/cloud/pnpm-lock.yaml index 27fdd29071a..6011b2f62d5 100644 --- a/cloud/pnpm-lock.yaml +++ b/cloud/pnpm-lock.yaml @@ -21,11 +21,57 @@ importers: specifier: ^4.0.8 version: 4.1.9(@types/node@24.13.2)(vite@8.0.16(@types/node@24.13.2)(esbuild@0.28.1)(tsx@4.22.4)) + apps/push: + dependencies: + '@hono/node-server': + specifier: ^1.19.14 + version: 1.19.14(hono@4.12.27) + '@orca-cloud/postgres-schema': + specifier: workspace:* + version: link:../../packages/postgres-schema + '@orca-cloud/push-contract': + specifier: workspace:* + version: link:../../packages/push-contract + google-auth-library: + specifier: ^10.5.0 + version: 10.9.1 + hono: + specifier: ^4.12.27 + version: 4.12.27 + pg: + specifier: ^8.22.0 + version: 8.22.0 + tweetnacl: + specifier: ^1.0.3 + version: 1.0.3 + zod: + specifier: ^3.25.76 + version: 3.25.76 + devDependencies: + '@types/node': + specifier: ^24.10.0 + version: 24.13.2 + '@types/pg': + specifier: ^8.20.0 + version: 8.20.0 + tsx: + specifier: ^4.21.0 + version: 4.22.4 + typescript: + specifier: ^5.9.3 + version: 5.9.3 + vitest: + specifier: ^4.0.8 + version: 4.1.9(@types/node@24.13.2)(vite@8.0.16(@types/node@24.13.2)(esbuild@0.28.1)(tsx@4.22.4)) + apps/relay: dependencies: '@hono/node-server': specifier: ^1.19.14 version: 1.19.14(hono@4.12.27) + '@orca-cloud/postgres-schema': + specifier: workspace:* + version: link:../../packages/postgres-schema '@orca-cloud/relay-contract': specifier: workspace:* version: link:../../packages/relay-contract @@ -117,6 +163,34 @@ importers: specifier: ^4.0.8 version: 4.1.9(@types/node@24.13.2)(vite@8.0.16(@types/node@24.13.2)(esbuild@0.28.1)(tsx@4.22.4)) + packages/postgres-schema: + devDependencies: + '@types/node': + specifier: ^24.10.0 + version: 24.13.2 + typescript: + specifier: ^5.9.3 + version: 5.9.3 + vitest: + specifier: ^4.0.8 + version: 4.1.9(@types/node@24.13.2)(vite@8.0.16(@types/node@24.13.2)(esbuild@0.28.1)(tsx@4.22.4)) + + packages/push-contract: + dependencies: + zod: + specifier: ^3.25.76 + version: 3.25.76 + devDependencies: + '@types/node': + specifier: ^24.10.0 + version: 24.13.2 + typescript: + specifier: ^5.9.3 + version: 5.9.3 + vitest: + specifier: ^4.0.8 + version: 4.1.9(@types/node@24.13.2)(vite@8.0.16(@types/node@24.13.2)(esbuild@0.28.1)(tsx@4.22.4)) + packages/relay-contract: dependencies: zod: @@ -463,10 +537,23 @@ packages: '@vitest/utils@4.1.9': resolution: {integrity: sha512-A51o8ymO5PpqlWNnBP9ZHPXDIpuMtTLlGSjN7la4US+LJzoUMyhwjA5QXlm39JexgwHKW4Xjs8Z2d3dLCXOeuA==} + agent-base@7.1.4: + resolution: {integrity: sha512-MnA+YT8fwfJPgBx3m60MNqakm30XOkyIoH1y6huTQvC0PwZG7ki8NacLBcrPbNoo8vEZy7Jpuk7+jMO+CUovTQ==} + engines: {node: '>= 14'} + assertion-error@2.0.1: resolution: {integrity: sha512-Izi8RQcffqCeNVgFigKli1ssklIbpHnCYc6AknXGYoB6grJqyeby7jv12JUQgmTAnIDnbck1uxksT4dzN3PWBA==} engines: {node: '>=12'} + base64-js@1.5.1: + resolution: {integrity: sha512-AKpaYlHn8t4SVbOHCy+b5+KKgvR4vrsD8vbvrbiQJps7fKDTkjkDry6ji0rUJjC0kzbNePLwzxq8iypo41qeWA==} + + bignumber.js@9.3.1: + resolution: {integrity: sha512-Ko0uX15oIUS7wJ3Rb30Fs6SkVbLmPBAKdlm7q9+ak9bbIeFf0MwuBsQV6z7+X768/cHsfg+WlysDWJcmthjsjQ==} + + buffer-equal-constant-time@1.0.1: + resolution: {integrity: sha512-zRpUiDwd/xk6ADqPMATG8vc9VPrkck7T07OIx0gnjmJAnHnTVXNQG3vfvWNuiZIkwu9KrKdA1iJKfsfTVxE6NA==} + chai@6.2.2: resolution: {integrity: sha512-NUPRluOfOiTKBKvWPtSD4PhFvWCqOi0BGStNWs57X9js7XGTprSmFoz5F0tWhR4WPjNeR9jXqdC7/UpSJTnlRg==} engines: {node: '>=18'} @@ -474,10 +561,26 @@ packages: convert-source-map@2.0.0: resolution: {integrity: sha512-Kvp459HrV2FEJ1CAsi1Ku+MY3kasH19TFykTz2xWmMeq6bk2NU3XXvfJ+Q61m0xktWwt+1HSYf3JZsTms3aRJg==} + data-uri-to-buffer@4.0.1: + resolution: {integrity: sha512-0R9ikRb668HB7QDxT1vkpuUBtqc53YyAwMwGeUFKRojY/NWKvdZ+9UYtRfGmhqNbRkTSVpMbmyhXipFFv2cb/A==} + engines: {node: '>= 12'} + + debug@4.4.3: + resolution: {integrity: sha512-RGwwWnwQvkVfavKVt22FGLw+xYSdzARwm0ru6DhTVA3umU5hZc28V3kO4stgYryrTlLpuvgI9GiijltAjNbcqA==} + engines: {node: '>=6.0'} + peerDependencies: + supports-color: '*' + peerDependenciesMeta: + supports-color: + optional: true + detect-libc@2.1.2: resolution: {integrity: sha512-Btj2BOOO83o3WyH59e8MgXsxEQVcarkUOpEYrubB0urwnN10yQ364rsiByU11nZlqWYZm05i/of7io4mzihBtQ==} engines: {node: '>=8'} + ecdsa-sig-formatter@1.0.11: + resolution: {integrity: sha512-nagl3RYrbNv6kQkeJIpt6NJZy8twLB/2vtz6yN9Z4vRKHN4/QZJIEbqohALSgwKdnksuY3k5Addp5lg8sVoVcQ==} + es-module-lexer@2.1.0: resolution: {integrity: sha512-n27zTYMjYu1aj4MjCWzSP7G9r75utsaoc8m61weK+W8JMBGGQybd43GstCXZ3WNmSFtGT9wi59qQTW6mhTR5LQ==} @@ -493,6 +596,9 @@ packages: resolution: {integrity: sha512-knvyeauYhqjOYvQ66MznSMs83wmHrCycNEN6Ao+2AeYEfxUIkuiVxdEa1qlGEPK+We3n0THiDciYSsCcgW/DoA==} engines: {node: '>=12.0.0'} + extend@3.0.2: + resolution: {integrity: sha512-fjquC59cD7CyW6urNXK0FBufkZcoiGG80wTuPujX590cB5Ttln20E2UB4S/WARVqhXffZl2LNgS+gQdPIIim/g==} + fdir@6.5.0: resolution: {integrity: sha512-tIbYtZbucOs0BRGqPJkshJUYdL+SDH7dVM8gjy+ERp3WAUjLEFJE+02kanyHtwjWOnwrKYBiwAmM0p4kLJAnXg==} engines: {node: '>=12.0.0'} @@ -502,18 +608,55 @@ packages: picomatch: optional: true + fetch-blob@3.2.0: + resolution: {integrity: sha512-7yAQpD2UMJzLi1Dqv7qFYnPbaPx7ZfFK6PiIxQ4PfkGPyNyl2Ugx+a/umUonmKqjhM4DnfbMvdX6otXq83soQQ==} + engines: {node: ^12.20 || >= 14.13} + + formdata-polyfill@4.0.10: + resolution: {integrity: sha512-buewHzMvYL29jdeQTVILecSaZKnt/RJWjoZCF5OW60Z67/GmSLBkOFM7qh1PI3zFNtJbaZL5eQu1vLfazOwj4g==} + engines: {node: '>=12.20.0'} + fsevents@2.3.3: resolution: {integrity: sha512-5xoDfX+fL7faATnagmWPpbFtwh/R77WmMMqqHGS65C3vvB0YHrgF+B1YmZ3441tMj5n63k0212XNoJwzlhffQw==} engines: {node: ^8.16.0 || ^10.6.0 || >=11.0.0} os: [darwin] + gaxios@7.3.1: + resolution: {integrity: sha512-kB3rzJV7d9juLZh8/56QTXCwQfxyhdOMdyYk1HdQKFtF8TJTDTZQJtixWIwXdE9Jji91mC41DUNpjleo4L4eAQ==} + engines: {node: '>=18'} + + gcp-metadata@8.1.2: + resolution: {integrity: sha512-zV/5HKTfCeKWnxG0Dmrw51hEWFGfcF2xiXqcA3+J90WDuP0SvoiSO5ORvcBsifmx/FoIjgQN3oNOGaQ5PhLFkg==} + engines: {node: '>=18'} + + google-auth-library@10.9.1: + resolution: {integrity: sha512-i1ydyHrqcIxXkWh/uBmVkzCvIuq5yiK2ATndIe5XxKholrG/MTYP9xGYka4sQhrbIAgGjL2B6NOE7rFaiF3fXw==} + engines: {node: '>=18'} + + google-logging-utils@1.1.3: + resolution: {integrity: sha512-eAmLkjDjAFCVXg7A1unxHsLf961m6y17QFqXqAXGj/gVkKFrEICfStRfwUlGNfeCEjNRa32JEWOUTlYXPyyKvA==} + engines: {node: '>=14'} + hono@4.12.27: resolution: {integrity: sha512-1yrb/+w6HWQJrUCLkJ2IF5jNIPvvFkblV5RNOYl6bV+OA6p9GLcMpHFFGTosSvHvcAUibuUukRqhlYI4z32C7Q==} engines: {node: '>=16.9.0'} + https-proxy-agent@7.0.6: + resolution: {integrity: sha512-vK9P5/iUfdl95AI+JVyUuIcVtd4ofvtrOr3HNtM2yxC9bnMbEdp3x01OhQNnjb8IJYi38VlTE3mBXwcfvywuSw==} + engines: {node: '>= 14'} + jose@6.2.3: resolution: {integrity: sha512-YYVDInQKFJfR/xa3ojUTl8c2KoTwiL1R5Wg9YCydwH0x0B9grbzlg5HC7mMjCtUJjbQ/YnGEZIhI5tCgfTb4Hw==} + json-bigint@1.0.0: + resolution: {integrity: sha512-SiPv/8VpZuWbvLSMtTDU8hEfrZWg/mH/nV/b4o0CYbSxu1UIQPLdwKOCIyLQX+VIPO5vrLX3i8qtqFyhdPSUSQ==} + + jwa@2.0.1: + resolution: {integrity: sha512-hRF04fqJIP8Abbkq5NKGN0Bbr3JxlQ+qhZufXVr0DvujKy93ZCbXZMHDL4EOtodSbCWxOqR8MS1tXA5hwqCXDg==} + + jws@4.0.1: + resolution: {integrity: sha512-EKI/M/yqPncGUUh44xz0PxSidXFr/+r0pA70+gIYhjv+et7yxM+s29Y+VGDkovRofQem0fs7Uvf4+YmAdyRduA==} + lightningcss-android-arm64@1.32.0: resolution: {integrity: sha512-YK7/ClTt4kAK0vo6w3X+Pnm0D2cf2vPHbhOXdoNti1Ga0al1P4TBZhwjATvjNwLEBCnKvjJc2jQgHXH0NEwlAg==} engines: {node: '>= 12.0.0'} @@ -587,11 +730,23 @@ packages: magic-string@0.30.21: resolution: {integrity: sha512-vd2F4YUyEXKGcLHoq+TEyCjxueSeHnFxyyjNp80yg0XV4vUhnDer/lvvlqM/arB5bXQN5K2/3oinyCRyx8T2CQ==} + ms@2.1.3: + resolution: {integrity: sha512-6FlzubTLZG3J2a/NVCAleEhjzq5oxgHyaCU9yYXvcLsvoVaHJq/s5xXI6/XXP6tz7R9xAOtHnSO/tXtF3WRTlA==} + nanoid@3.3.13: resolution: {integrity: sha512-sPdqC6ByMVVGvF1ynvvMo0/o+oD1VX7DaHhijt1bFgjvBkHBib4t49GoNDhf2NDta4oeUNlaGbSt5K7qjZ955Q==} engines: {node: ^10 || ^12 || ^13.7 || ^14 || >=15.0.1} hasBin: true + node-domexception@1.0.0: + resolution: {integrity: sha512-/jKZoMpw0F8GRwl4/eLROPA3cfcXtLApP0QzLmUT/HuPCZWyB7IY9ZrMeKw2O/nFIqPQB3PVM9aYm0F312AXDQ==} + engines: {node: '>=10.5.0'} + deprecated: Use your platform's native DOMException instead + + node-fetch@3.3.2: + resolution: {integrity: sha512-dRB78srN/l6gqWulah9SrxeYnxeddIG30+GOqK/9OlLVyLg3HPnr6SqOWTWOXKRwC2eGYCkZ59NNuSgvSrpgOA==} + engines: {node: ^12.20.0 || ^14.13.1 || >=16.0.0} + obug@2.1.3: resolution: {integrity: sha512-9miFgM2OFba7hB+pRgvtV84pYTBaoTHohvmIgiRt6dRIzbwEOIaNaP+dIlGs2fNFoB0SeISs0Jz5WFVRid6Xyg==} engines: {node: '>=12.20.0'} @@ -665,6 +820,9 @@ packages: engines: {node: ^20.19.0 || >=22.12.0} hasBin: true + safe-buffer@5.2.1: + resolution: {integrity: sha512-rp3So07KcdmmKbGvgaNxQSJr7bGVSVk5S9Eq1F+ppbRo70+YeaDxkw5Dd8NPN+GD6bjnYm2VuPuCXmpuYvmCXQ==} + siginfo@2.0.0: resolution: {integrity: sha512-ybx0WO1/8bSBLEWXZvEd7gMW3Sn3JFlW3TvX1nREbDLRNQNaeNN8WK0meBwPdAaOI7TtRRRJn/Es1zhrrCHu7g==} @@ -800,6 +958,10 @@ packages: jsdom: optional: true + web-streams-polyfill@3.3.3: + resolution: {integrity: sha512-d2JWLCivmZYTSIoge9MsgFCZrt571BikcWGYkjC1khllbTeDlGqZ2D8vD8E/lJa8WGWbb7Plm8/XJYV7IJHZZw==} + engines: {node: '>= 8'} + why-is-node-running@2.3.0: resolution: {integrity: sha512-hUrmaWBdVDcxvYqnyh09zunKzROWjbZTiNy8dBEjkS7ehEDQibXJ7XvlmtbwuTclUiIyN+CyXQD4Vmko8fNm8w==} engines: {node: '>=8'} @@ -1057,14 +1219,32 @@ snapshots: convert-source-map: 2.0.0 tinyrainbow: 3.1.0 + agent-base@7.1.4: {} + assertion-error@2.0.1: {} + base64-js@1.5.1: {} + + bignumber.js@9.3.1: {} + + buffer-equal-constant-time@1.0.1: {} + chai@6.2.2: {} convert-source-map@2.0.0: {} + data-uri-to-buffer@4.0.1: {} + + debug@4.4.3: + dependencies: + ms: 2.1.3 + detect-libc@2.1.2: {} + ecdsa-sig-formatter@1.0.11: + dependencies: + safe-buffer: 5.2.1 + es-module-lexer@2.1.0: {} esbuild@0.28.1: @@ -1102,17 +1282,79 @@ snapshots: expect-type@1.3.0: {} + extend@3.0.2: {} + fdir@6.5.0(picomatch@4.0.4): optionalDependencies: picomatch: 4.0.4 + fetch-blob@3.2.0: + dependencies: + node-domexception: 1.0.0 + web-streams-polyfill: 3.3.3 + + formdata-polyfill@4.0.10: + dependencies: + fetch-blob: 3.2.0 + fsevents@2.3.3: optional: true + gaxios@7.3.1: + dependencies: + extend: 3.0.2 + https-proxy-agent: 7.0.6 + node-fetch: 3.3.2 + transitivePeerDependencies: + - supports-color + + gcp-metadata@8.1.2: + dependencies: + gaxios: 7.3.1 + google-logging-utils: 1.1.3 + json-bigint: 1.0.0 + transitivePeerDependencies: + - supports-color + + google-auth-library@10.9.1: + dependencies: + base64-js: 1.5.1 + ecdsa-sig-formatter: 1.0.11 + gaxios: 7.3.1 + gcp-metadata: 8.1.2 + google-logging-utils: 1.1.3 + jws: 4.0.1 + transitivePeerDependencies: + - supports-color + + google-logging-utils@1.1.3: {} + hono@4.12.27: {} + https-proxy-agent@7.0.6: + dependencies: + agent-base: 7.1.4 + debug: 4.4.3 + transitivePeerDependencies: + - supports-color + jose@6.2.3: {} + json-bigint@1.0.0: + dependencies: + bignumber.js: 9.3.1 + + jwa@2.0.1: + dependencies: + buffer-equal-constant-time: 1.0.1 + ecdsa-sig-formatter: 1.0.11 + safe-buffer: 5.2.1 + + jws@4.0.1: + dependencies: + jwa: 2.0.1 + safe-buffer: 5.2.1 + lightningcss-android-arm64@1.32.0: optional: true @@ -1166,8 +1408,18 @@ snapshots: dependencies: '@jridgewell/sourcemap-codec': 1.5.5 + ms@2.1.3: {} + nanoid@3.3.13: {} + node-domexception@1.0.0: {} + + node-fetch@3.3.2: + dependencies: + data-uri-to-buffer: 4.0.1 + fetch-blob: 3.2.0 + formdata-polyfill: 4.0.10 + obug@2.1.3: {} pathe@2.0.3: {} @@ -1248,6 +1500,8 @@ snapshots: '@rolldown/binding-win32-arm64-msvc': 1.0.3 '@rolldown/binding-win32-x64-msvc': 1.0.3 + safe-buffer@5.2.1: {} + siginfo@2.0.0: {} source-map-js@1.2.1: {} @@ -1324,6 +1578,8 @@ snapshots: transitivePeerDependencies: - msw + web-streams-polyfill@3.3.3: {} + why-is-node-running@2.3.0: dependencies: siginfo: 2.0.0 diff --git a/config/oxlint-performance-audit.json b/config/oxlint-performance-audit.json index 2912c6b8e03..15d3fcd0f68 100644 --- a/config/oxlint-performance-audit.json +++ b/config/oxlint-performance-audit.json @@ -28,6 +28,7 @@ "app-store-performance/require-selector": "warn", "app-store-performance/no-identity-selector": "warn", "app-store-performance/no-fresh-selector-result": "warn", + "app-store-performance/no-nested-fresh-under-shallow": "warn", "quadratic-buffer-concat/no-loop-carried-concat": "warn", "sort-comparator-performance/no-repeated-collator": "warn" }, diff --git a/config/oxlint-plugins/app-store-performance.mjs b/config/oxlint-plugins/app-store-performance.mjs index 9da732f5825..d8bfe4131d9 100644 --- a/config/oxlint-plugins/app-store-performance.mjs +++ b/config/oxlint-plugins/app-store-performance.mjs @@ -8,6 +8,19 @@ const ALLOCATING_METHODS = new Set([ 'toSpliced', 'with' ]) +const ALLOCATING_OBJECT_STATICS = new Set([ + 'assign', + 'create', + 'entries', + 'fromEntries', + 'keys', + 'values' +]) +const FUNCTION_NODES = new Set([ + 'ArrowFunctionExpression', + 'FunctionDeclaration', + 'FunctionExpression' +]) function identifierName(node) { return node?.type === 'Identifier' ? node.name : null @@ -25,8 +38,12 @@ function propertyName(node) { : null } +function functionNode(node) { + return FUNCTION_NODES.has(node?.type) ? node : null +} + function returnedExpressions(selector) { - if (selector?.type !== 'ArrowFunctionExpression' && selector?.type !== 'FunctionExpression') { + if (!functionNode(selector)) { return [] } if (selector.body.type !== 'BlockStatement') { @@ -37,10 +54,7 @@ function returnedExpressions(selector) { if (!node || typeof node !== 'object') { return } - if ( - node !== selector.body && - ['ArrowFunctionExpression', 'FunctionDeclaration', 'FunctionExpression'].includes(node.type) - ) { + if (node !== selector.body && FUNCTION_NODES.has(node.type)) { return } if (node.type === 'ReturnStatement') { @@ -76,10 +90,7 @@ function unwrapShallowSelector(selector, shallowHooks) { } function isIdentitySelector(selector) { - if (selector?.type !== 'ArrowFunctionExpression' && selector?.type !== 'FunctionExpression') { - return false - } - const parameter = selector.params[0] + const parameter = functionNode(selector)?.params[0] if (parameter?.type !== 'Identifier') { return false } @@ -88,14 +99,23 @@ function isIdentitySelector(selector) { ) } -function isAllocatingExpression(expression) { - if (expression?.type === 'ConditionalExpression') { - return ( - isAllocatingExpression(expression.consequent) || isAllocatingExpression(expression.alternate) - ) - } - if (expression?.type === 'LogicalExpression') { - return isAllocatingExpression(expression.left) || isAllocatingExpression(expression.right) +/** + * `everyBranch` decides how a conditional counts. An inline selector is flagged + * when ANY branch allocates; a helper the selector delegates to must allocate on + * EVERY branch, so the `cache.get(k) ?? build(state)` identity-caching shape is + * not a false positive. + */ +function allocates(expression, everyBranch) { + const branches = + expression?.type === 'ConditionalExpression' + ? [expression.consequent, expression.alternate] + : expression?.type === 'LogicalExpression' + ? [expression.left, expression.right] + : null + if (branches) { + return everyBranch + ? branches.every((branch) => allocates(branch, true)) + : branches.some((branch) => allocates(branch, false)) } if ( expression?.type === 'ArrayExpression' || @@ -107,44 +127,104 @@ function isAllocatingExpression(expression) { if (expression?.type !== 'CallExpression') { return false } - const method = propertyName(expression.callee) - if (method && ALLOCATING_METHODS.has(method)) { - return true - } const callee = expression.callee + const method = propertyName(callee) return ( - callee.type === 'MemberExpression' && - identifierName(callee.object) === 'Object' && - ['assign', 'create', 'entries', 'fromEntries', 'keys', 'values'].includes(propertyName(callee)) + ALLOCATING_METHODS.has(method) || + (identifierName(callee.object) === 'Object' && ALLOCATING_OBJECT_STATICS.has(method)) ) } -function importedLocalName(specifier, importedName) { - if (specifier.type !== 'ImportSpecifier' || identifierName(specifier.imported) !== importedName) { - return null +function isAllocatingExpression(expression) { + return allocates(expression, false) +} + +// Project-local zustand hooks follow the useStore convention; React's +// useSyncExternalStore matches that shape but is not a store subscription. +const STORE_HOOK_NAME = /^use[A-Z][A-Za-z0-9]*Store$/ +const NON_STORE_HOOKS = new Set(['useSyncExternalStore']) + +function isLocalModuleSource(source) { + return typeof source === 'string' && (source.startsWith('.') || source.startsWith('@/')) +} + +/** Module scope only: a component-local helper must not shadow a same-named import. */ +function isModuleScope(node) { + const parent = node.parent + return ( + parent?.type === 'Program' || + (parent?.type === 'ExportNamedDeclaration' && parent.parent?.type === 'Program') + ) +} + +/** Records module-scope `const selectX = (state) => ...` so identifier selectors resolve. */ +function recordNamedSelector(node, state) { + if (!isModuleScope(node)) { + return } - return identifierName(specifier.local) + const declared = + node.type === 'FunctionDeclaration' + ? [[node.id, node]] + : node.declarations.map((declarator) => [declarator.id, declarator.init]) + for (const [id, initializer] of declared) { + const name = identifierName(id) + if (name && functionNode(initializer)) { + state.namedSelectors.set(name, initializer) + } + } +} + +/** Inline function, or a module-scope selector referenced by name. */ +function resolveSelector(argument, state) { + return functionNode(argument) ?? state.namedSelectors.get(identifierName(argument)) ?? null +} + +/** + * One hop: a selector that delegates to a module-scope helper is the idiomatic + * shape here, and neither the inline-body check nor a reviewer reading the call + * site can see what that helper returns. An unresolvable helper is left alone. + */ +function expandThroughNamedHelper(expression, state) { + const helper = + expression?.type === 'CallExpression' + ? state.namedSelectors.get(identifierName(expression.callee)) + : undefined + const returned = helper ? returnedExpressions(helper) : [] + return returned.length > 0 && returned.every((entry) => allocates(entry, true)) + ? returned + : [expression] } function createRuleState() { return { appStoreHooks: new Set(), - shallowHooks: new Set() + shallowHooks: new Set(), + namedSelectors: new Map(), + deferredCalls: [] } } function recordImports(node, state) { - if (node.source?.value === 'zustand/react/shallow') { - for (const specifier of node.specifiers) { - const localName = importedLocalName(specifier, 'useShallow') - if (localName) { - state.shallowHooks.add(localName) - } - } - } + const source = node.source?.value for (const specifier of node.specifiers) { - const localName = importedLocalName(specifier, 'useAppStore') - if (localName) { + if (specifier.type !== 'ImportSpecifier') { + continue + } + const imported = identifierName(specifier.imported) + const localName = identifierName(specifier.local) + if (!imported || !localName) { + continue + } + if (source === 'zustand/react/shallow' && imported === 'useShallow') { + state.shallowHooks.add(localName) + } + // useAppStore is the app store wherever it is re-exported from; sibling + // stores are trusted by naming convention only when they come from this codebase. + if ( + STORE_HOOK_NAME.test(imported) && + !NON_STORE_HOOKS.has(imported) && + (imported === 'useAppStore' || isLocalModuleSource(source)) + ) { state.appStoreHooks.add(localName) } } @@ -176,52 +256,107 @@ function requireSelectorRule() { } } -function noIdentitySelectorRule() { +/** + * Selector arguments are collected during traversal and judged at Program:exit so a + * selector hoisted below its call site still resolves. + */ +function deferredSelectorRule(inspect) { const state = createRuleState() return { ImportDeclaration(node) { recordImports(node, state) }, + FunctionDeclaration(node) { + recordNamedSelector(node, state) + }, + VariableDeclaration(node) { + recordNamedSelector(node, state) + }, CallExpression(node) { - if (!isAppStoreCall(node, state)) { - return + if (isAppStoreCall(node, state)) { + state.deferredCalls.push(node) } - const { selector } = unwrapShallowSelector(node.arguments[0], state.shallowHooks) - if (isIdentitySelector(selector)) { - this.report({ - node: selector, - message: - 'Select the smallest required fields instead of subscribing to the entire app store.' + }, + 'Program:exit'() { + for (const node of state.deferredCalls) { + const { selector: argument, shallow } = unwrapShallowSelector( + node.arguments[0], + state.shallowHooks + ) + const report = inspect({ + selector: resolveSelector(argument, state), + shallow, + state }) + if (report) { + this.report(report) + } } } } } +function noIdentitySelectorRule() { + return deferredSelectorRule(({ selector }) => + isIdentitySelector(selector) + ? { + node: selector, + message: + 'Select the smallest required fields instead of subscribing to the entire app store.' + } + : null + ) +} + function noFreshSelectorResultRule() { - const state = createRuleState() - return { - ImportDeclaration(node) { - recordImports(node, state) - }, - CallExpression(node) { - if (!isAppStoreCall(node, state)) { - return - } - const { selector, shallow } = unwrapShallowSelector(node.arguments[0], state.shallowHooks) - if (shallow) { - return - } - const freshResult = returnedExpressions(selector).find(isAllocatingExpression) - if (freshResult) { - this.report({ + return deferredSelectorRule(({ selector, shallow, state }) => { + if (shallow || !selector) { + return null + } + const freshResult = returnedExpressions(selector) + .flatMap((expression) => expandThroughNamedHelper(expression, state)) + .find(isAllocatingExpression) + return freshResult + ? { node: freshResult, message: 'This selector returns a fresh reference on every store write; select a stable field, cache the result, or use useShallow.' - }) - } - } + } + : null + }) +} + +/** useShallow compares one level deep, so a fresh reference nested inside its result never matches. */ +function nestedFreshValues(expression) { + if (expression?.type === 'ObjectExpression') { + return expression.properties + .map((property) => (property.type === 'Property' ? property.value : null)) + .filter(Boolean) } + if (expression?.type === 'ArrayExpression') { + return expression.elements.filter(Boolean) + } + return [] +} + +function noNestedFreshUnderShallowRule() { + return deferredSelectorRule(({ selector, shallow, state }) => { + if (!shallow || !selector) { + return null + } + const nestedFresh = returnedExpressions(selector) + .flatMap((expression) => expandThroughNamedHelper(expression, state)) + .flatMap(nestedFreshValues) + .flatMap((expression) => expandThroughNamedHelper(expression, state)) + .find(isAllocatingExpression) + return nestedFresh + ? { + node: nestedFresh, + message: + 'useShallow compares only one level deep, so this nested fresh reference changes on every store write and defeats the memo; project the primitives the component actually renders.' + } + : null + }) } function bindContext(createVisitors) { @@ -239,6 +374,7 @@ export default { rules: { 'require-selector': { create: bindContext(requireSelectorRule) }, 'no-identity-selector': { create: bindContext(noIdentitySelectorRule) }, - 'no-fresh-selector-result': { create: bindContext(noFreshSelectorResultRule) } + 'no-fresh-selector-result': { create: bindContext(noFreshSelectorResultRule) }, + 'no-nested-fresh-under-shallow': { create: bindContext(noNestedFreshUnderShallowRule) } } } diff --git a/config/patches/@vscode__windows-process-tree@0.8.0.patch b/config/patches/@vscode__windows-process-tree@0.8.0.patch index fe5e4be44b1..7c930a5fca5 100644 --- a/config/patches/@vscode__windows-process-tree@0.8.0.patch +++ b/config/patches/@vscode__windows-process-tree@0.8.0.patch @@ -1,5 +1,5 @@ diff --git a/binding.gyp b/binding.gyp -index 855bd4b86f0a3c18c7594212c0e42b6e35bc4001..33774e7ae296f0de39dd94156673c9e773638bf4 100644 +index 855bd4b86f0a3c18c7594212c0e42b6e35bc4001..0bb2af7923b6e6f1f0da40cae8067304cd1fea14 100644 --- a/binding.gyp +++ b/binding.gyp @@ -3,7 +3,6 @@ @@ -10,7 +10,8 @@ index 855bd4b86f0a3c18c7594212c0e42b6e35bc4001..33774e7ae296f0de39dd94156673c9e7 ], "conditions": [ ['OS=="win"', { -@@ -15,12 +14,11 @@ +@@ -14,13 +13,12 @@ + "src/process_worker.cc", "src/process_commandline.cc" ], - "include_dirs": [], @@ -26,314 +27,207 @@ index 855bd4b86f0a3c18c7594212c0e42b6e35bc4001..33774e7ae296f0de39dd94156673c9e7 "AdditionalOptions": [ "/guard:cf", "/sdl", +diff --git a/lib/index.js b/lib/index.js +index 9747a7402600cd252859144d32580ed45c8c93f7..001e81fa8bc89091971d06aaf9d051ba20906615 100644 +--- a/lib/index.js ++++ b/lib/index.js +@@ -7,11 +7,13 @@ Object.defineProperty(exports, "__esModule", { value: true }); + exports.getAllProcesses = exports.getProcessTree = exports.getProcessCpuUsage = exports.getProcessList = exports.filterProcessList = exports.buildProcessTree = exports.ProcessDataFlag = void 0; + const util_1 = require("util"); + const native = process.platform === 'win32' ? require('../build/Release/windows_process_tree.node') : undefined; ++exports.supportedProcessDataFlags = native === undefined ? undefined : native.supportedProcessDataFlags; + var ProcessDataFlag; + (function (ProcessDataFlag) { + ProcessDataFlag[ProcessDataFlag["None"] = 0] = "None"; + ProcessDataFlag[ProcessDataFlag["Memory"] = 1] = "Memory"; + ProcessDataFlag[ProcessDataFlag["CommandLine"] = 2] = "CommandLine"; ++ ProcessDataFlag[ProcessDataFlag["CreationTime"] = 4] = "CreationTime"; + })(ProcessDataFlag = exports.ProcessDataFlag || (exports.ProcessDataFlag = {})); + // requestInProgress is used for any function that uses CreateToolhelp32Snapshot, as multiple calls + // to this cannot be done at the same time. +@@ -66,11 +68,12 @@ function buildProcessTree(rootPid, processList, maxDepth = MAX_FILTER_DEPTH) { + // • the properties are inlined/splatted + // • the 'ppid' field is omitted + // • the depth of the tree is limited by `maxDepth` +- const buildNode = ({ info: { pid, name, memory, commandLine }, children }, depth) => ({ ++ const buildNode = ({ info: { pid, name, memory, commandLine, creationTimeMs }, children }, depth) => ({ + pid, + name, + memory, + commandLine, ++ creationTimeMs, + children: depth > 0 ? children.map(c => buildNode(c, depth - 1)) : [], + }); + return buildNode(root, maxDepth); +diff --git a/lib/index.ts b/lib/index.ts +index f9aa005d9ced9e42885b8a976de5eb5bd61899ee..1b509af0b9065918bcb5cb75f2d7f23821d4a56a 100644 +--- a/lib/index.ts ++++ b/lib/index.ts +@@ -6,12 +6,15 @@ + import { promisify } from 'util'; + + const native = process.platform === 'win32' ? require('../build/Release/windows_process_tree.node') : undefined; ++/** The flag bits this compiled addon reports; undefined off win32. */ ++export const supportedProcessDataFlags: number | undefined = native?.supportedProcessDataFlags; + import { IProcessInfo, IProcessTreeNode, IProcessCpuInfo } from '@vscode/windows-process-tree'; + + export enum ProcessDataFlag { + None = 0, + Memory = 1, +- CommandLine = 2 ++ CommandLine = 2, ++ CreationTime = 4 + } + + type RequestCallback = (processList: IProcessInfo[]) => void; +@@ -81,11 +84,12 @@ export function buildProcessTree(rootPid: number, processList: Iterable ({ ++ const buildNode = ({ info: { pid, name, memory, commandLine, creationTimeMs }, children }: IProcessInfoNode, depth: number): IProcessTreeNode => ({ + pid, + name, + memory, + commandLine, ++ creationTimeMs, + children: depth > 0 ? children.map(c => buildNode(c, depth - 1)) : [], + }); + +diff --git a/src/addon.cc b/src/addon.cc +index 9214aff281251e797a70ecb9f6e0b52932a0503f..722edd42ddb4740296bfc47582a181bd6d00c464 100644 +--- a/src/addon.cc ++++ b/src/addon.cc +@@ -53,6 +53,10 @@ void GetProcessCpuUsage(const Napi::CallbackInfo& args) { + Napi::Object Init(Napi::Env env, Napi::Object exports) { + exports.Set("getProcessList", Napi::Function::New(env, GetProcessList)); + exports.Set("getProcessCpuUsage", Napi::Function::New(env, GetProcessCpuUsage)); ++ // Lets a caller prove THIS BINARY understands CREATIONTIME. The JS enum is ++ // patched source and says nothing about what the .node was compiled from. ++ exports.Set("supportedProcessDataFlags", ++ Napi::Number::New(env, MEMORY | COMMANDLINE | CREATIONTIME)); + return exports; + } + diff --git a/src/process.cc b/src/process.cc -index 3eea92077c4d1d433119361d5c432881859131e9..738775f6fcdfb676054386fe34c0380327ed1863 100644 +index 3eea92077c4d1d433119361d5c432881859131e9..22a47421da919c76e2194280974d39c2287b098d 100644 --- a/src/process.cc +++ b/src/process.cc -@@ -1,108 +1,112 @@ --/*--------------------------------------------------------------------------------------------- -- * Copyright (c) Microsoft Corporation. All rights reserved. -- * Licensed under the MIT License. See License.txt in the project root for license information. -- *--------------------------------------------------------------------------------------------*/ -- --#include "process.h" --#include "process_commandline.h" -- --#include --#include --#include -- --uint32_t GetRawProcessList(std::vector& process_info, -- DWORD process_data_flags) { -- // Fetch the PID and PPIDs -- PROCESSENTRY32 process_entry = { 0 }; -- DWORD parent_pid = 0; -- uint32_t process_count = 0; -- HANDLE snapshot_handle = CreateToolhelp32Snapshot(TH32CS_SNAPPROCESS, 0); -- process_entry.dwSize = sizeof(PROCESSENTRY32); -- if (Process32First(snapshot_handle, &process_entry)) { -- do { -- if (process_entry.th32ProcessID != 0) { +@@ -21,7 +21,8 @@ uint32_t GetRawProcessList(std::vector& process_info, + if (Process32First(snapshot_handle, &process_entry)) { + do { + if (process_entry.th32ProcessID != 0) { - ProcessInfo pinfo; -- pinfo.pid = process_entry.th32ProcessID; -- pinfo.ppid = process_entry.th32ParentProcessID; -- -- if (MEMORY & process_data_flags) { -- GetProcessMemoryUsage(pinfo); -- } -- -- if (COMMANDLINE & process_data_flags) { -- GetProcessCommandLine(pinfo); -- } -- -- strcpy(pinfo.name, process_entry.szExeFile); -- process_info.push_back(std::move(pinfo)); -- process_count++; -- } -- } while (process_count < 1024 && Process32Next(snapshot_handle, &process_entry)); -- } -- -- CloseHandle(snapshot_handle); -- return process_count; --} -- --void GetProcessMemoryUsage(ProcessInfo& process_info) { -- DWORD pid = process_info.pid; -- HANDLE hProcess; -- PROCESS_MEMORY_COUNTERS pmc; -- -- hProcess = OpenProcess(PROCESS_QUERY_INFORMATION | PROCESS_VM_READ, false, pid); -- -- if (hProcess == NULL) { -- return; -- } -- -- if (GetProcessMemoryInfo(hProcess, &pmc, sizeof(pmc))) { -- process_info.memory = (DWORD)pmc.WorkingSetSize; -- } -- -- CloseHandle(hProcess); --} -- --// Per documentation, it is not recommended to add or subtract values from the FILETIME --// structure, or to cast it to ULARGE_INTEGER as this can cause alignment faults on 64-bit Windows. --// Copy the high and low part to a ULARGE_INTEGER and peform arithmetic on that instead. --// See https://msdn.microsoft.com/en-us/library/windows/desktop/ms724284(v=vs.85).aspx --ULONGLONG GetTotalTime(const FILETIME* kernelTime, const FILETIME* userTime) { -- ULARGE_INTEGER kt, ut; -- kt.LowPart = (*kernelTime).dwLowDateTime; -- kt.HighPart = (*kernelTime).dwHighDateTime; -- -- ut.LowPart = (*userTime).dwLowDateTime; -- ut.HighPart = (*userTime).dwHighDateTime; -- -- return kt.QuadPart + ut.QuadPart; --} -- --void GetCpuUsage(Cpu& cpu_info, bool first_pass) { -- DWORD pid = cpu_info.pid; -- HANDLE hProcess; -- -- hProcess = OpenProcess(PROCESS_QUERY_INFORMATION | PROCESS_VM_READ, false, pid); -- -- if (hProcess == NULL) { -- return; -- } -- -- FILETIME creationTime, exitTime, kernelTime, userTime; -- FILETIME sysIdleTime, sysKernelTime, sysUserTime; -- if (GetProcessTimes(hProcess, &creationTime, &exitTime, &kernelTime, &userTime) -- && GetSystemTimes(&sysIdleTime, &sysKernelTime, &sysUserTime)) { -- if (first_pass) { -- cpu_info.initialProcRunTime = GetTotalTime(&kernelTime, &userTime); -- cpu_info.initialSystemTime = GetTotalTime(&sysKernelTime, &sysUserTime); -- } else { -- ULONGLONG endProcTime = GetTotalTime(&kernelTime, &userTime); -- ULONGLONG endSysTime = GetTotalTime(&sysKernelTime, &sysUserTime); -- -- cpu_info.cpu = 100.0 * (endProcTime - cpu_info.initialProcRunTime) / (endSysTime - cpu_info.initialSystemTime); -- } -- } else { -- cpu_info.cpu = std::numeric_limits::quiet_NaN(); -- } -- -- CloseHandle(hProcess); -+/*--------------------------------------------------------------------------------------------- -+ * Copyright (c) Microsoft Corporation. All rights reserved. -+ * Licensed under the MIT License. See License.txt in the project root for license information. -+ *--------------------------------------------------------------------------------------------*/ -+ -+#include "process.h" -+#include "process_commandline.h" -+ -+#include -+#include -+#include -+ -+uint32_t GetRawProcessList(std::vector& process_info, -+ DWORD process_data_flags) { -+ // Fetch the PID and PPIDs -+ PROCESSENTRY32 process_entry = { 0 }; -+ DWORD parent_pid = 0; -+ uint32_t process_count = 0; -+ HANDLE snapshot_handle = CreateToolhelp32Snapshot(TH32CS_SNAPPROCESS, 0); -+ process_entry.dwSize = sizeof(PROCESSENTRY32); -+ if (Process32First(snapshot_handle, &process_entry)) { -+ do { -+ if (process_entry.th32ProcessID != 0) { + // Value-initialize: `memory` is otherwise stack garbage when the flag is unset. + ProcessInfo pinfo{}; -+ pinfo.pid = process_entry.th32ProcessID; -+ pinfo.ppid = process_entry.th32ParentProcessID; -+ -+ if (MEMORY & process_data_flags) { -+ GetProcessMemoryUsage(pinfo); + pinfo.pid = process_entry.th32ProcessID; + pinfo.ppid = process_entry.th32ParentProcessID; + +@@ -33,23 +34,51 @@ uint32_t GetRawProcessList(std::vector& process_info, + GetProcessCommandLine(pinfo); + } + ++ if (CREATIONTIME & process_data_flags) { ++ GetProcessCreationTime(pinfo); + } + -+ if (COMMANDLINE & process_data_flags) { -+ GetProcessCommandLine(pinfo); -+ } -+ -+ strcpy(pinfo.name, process_entry.szExeFile); -+ process_info.push_back(std::move(pinfo)); -+ process_count++; -+ } + strcpy(pinfo.name, process_entry.szExeFile); + process_info.push_back(std::move(pinfo)); + process_count++; + } +- } while (process_count < 1024 && Process32Next(snapshot_handle, &process_entry)); + } while (Process32Next(snapshot_handle, &process_entry)); -+ } -+ -+ CloseHandle(snapshot_handle); -+ return process_count; -+} -+ -+void GetProcessMemoryUsage(ProcessInfo& process_info) { -+ DWORD pid = process_info.pid; -+ HANDLE hProcess; -+ PROCESS_MEMORY_COUNTERS pmc; -+ -+ // PROCESS_VM_READ is never used here -- GetProcessMemoryInfo reads counters the -+ // kernel keeps, not the address space -- and acquiring it is what EDR scores. -+ hProcess = OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, false, pid); -+ -+ if (hProcess == NULL) { -+ return; -+ } -+ -+ if (GetProcessMemoryInfo(hProcess, &pmc, sizeof(pmc))) { -+ process_info.memory = (DWORD)pmc.WorkingSetSize; -+ } -+ -+ CloseHandle(hProcess); -+} -+ -+// Per documentation, it is not recommended to add or subtract values from the FILETIME -+// structure, or to cast it to ULARGE_INTEGER as this can cause alignment faults on 64-bit Windows. -+// Copy the high and low part to a ULARGE_INTEGER and peform arithmetic on that instead. -+// See https://msdn.microsoft.com/en-us/library/windows/desktop/ms724284(v=vs.85).aspx -+ULONGLONG GetTotalTime(const FILETIME* kernelTime, const FILETIME* userTime) { -+ ULARGE_INTEGER kt, ut; -+ kt.LowPart = (*kernelTime).dwLowDateTime; -+ kt.HighPart = (*kernelTime).dwHighDateTime; -+ -+ ut.LowPart = (*userTime).dwLowDateTime; -+ ut.HighPart = (*userTime).dwHighDateTime; -+ -+ return kt.QuadPart + ut.QuadPart; -+} -+ -+void GetCpuUsage(Cpu& cpu_info, bool first_pass) { -+ DWORD pid = cpu_info.pid; -+ HANDLE hProcess; -+ -+ // GetProcessTimes needs no more than PROCESS_QUERY_LIMITED_INFORMATION. -+ hProcess = OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, false, pid); -+ + } + + CloseHandle(snapshot_handle); + return process_count; + } + ++void GetProcessCreationTime(ProcessInfo& process_info) { ++ HANDLE hProcess = OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, false, process_info.pid); + if (hProcess == NULL) { + return; + } + + FILETIME creationTime, exitTime, kernelTime, userTime; -+ FILETIME sysIdleTime, sysKernelTime, sysUserTime; -+ if (GetProcessTimes(hProcess, &creationTime, &exitTime, &kernelTime, &userTime) -+ && GetSystemTimes(&sysIdleTime, &sysKernelTime, &sysUserTime)) { -+ if (first_pass) { -+ cpu_info.initialProcRunTime = GetTotalTime(&kernelTime, &userTime); -+ cpu_info.initialSystemTime = GetTotalTime(&sysKernelTime, &sysUserTime); -+ } else { -+ ULONGLONG endProcTime = GetTotalTime(&kernelTime, &userTime); -+ ULONGLONG endSysTime = GetTotalTime(&sysKernelTime, &sysUserTime); -+ -+ cpu_info.cpu = 100.0 * (endProcTime - cpu_info.initialProcRunTime) / (endSysTime - cpu_info.initialSystemTime); ++ if (GetProcessTimes(hProcess, &creationTime, &exitTime, &kernelTime, &userTime)) { ++ ULARGE_INTEGER timestamp; ++ timestamp.LowPart = creationTime.dwLowDateTime; ++ timestamp.HighPart = creationTime.dwHighDateTime; ++ constexpr ULONGLONG WINDOWS_EPOCH_OFFSET_100NS = 116444736000000000ULL; ++ constexpr ULONGLONG HUNDRED_NS_PER_MILLISECOND = 10000ULL; ++ if (timestamp.QuadPart >= WINDOWS_EPOCH_OFFSET_100NS) { ++ process_info.creationTimeMs = ++ (timestamp.QuadPart - WINDOWS_EPOCH_OFFSET_100NS) / HUNDRED_NS_PER_MILLISECOND; + } -+ } else { -+ cpu_info.cpu = std::numeric_limits::quiet_NaN(); + } + + CloseHandle(hProcess); - } -\ No newline at end of file ++} ++ + void GetProcessMemoryUsage(ProcessInfo& process_info) { + DWORD pid = process_info.pid; + HANDLE hProcess; + PROCESS_MEMORY_COUNTERS pmc; + +- hProcess = OpenProcess(PROCESS_QUERY_INFORMATION | PROCESS_VM_READ, false, pid); ++ // PROCESS_VM_READ is never used here -- GetProcessMemoryInfo reads counters the ++ // kernel keeps, not the address space -- and acquiring it is what EDR scores. ++ hProcess = OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, false, pid); + + if (hProcess == NULL) { + return; +@@ -81,7 +110,8 @@ void GetCpuUsage(Cpu& cpu_info, bool first_pass) { + DWORD pid = cpu_info.pid; + HANDLE hProcess; + +- hProcess = OpenProcess(PROCESS_QUERY_INFORMATION | PROCESS_VM_READ, false, pid); ++ // GetProcessTimes needs no more than PROCESS_QUERY_LIMITED_INFORMATION. ++ hProcess = OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, false, pid); + + if (hProcess == NULL) { + return; +diff --git a/src/process.h b/src/process.h +index 82f8e4bcfa742551e5d874a7632736a7611d7aa7..78d1d2c3b2360ed06fd624b4cb2f5042510f7a77 100644 +--- a/src/process.h ++++ b/src/process.h +@@ -22,18 +22,22 @@ struct ProcessInfo { + DWORD ppid; + DWORD memory; // Reported in bytes + std::string commandLine; ++ ULONGLONG creationTimeMs; + }; + + enum ProcessDataFlags { + NONE = 0, + MEMORY = 1, +- COMMANDLINE = 2 ++ COMMANDLINE = 2, ++ CREATIONTIME = 4 + }; + + uint32_t GetRawProcessList(std::vector& process_info, DWORD flags); + + void GetProcessMemoryUsage(ProcessInfo& process_info); + ++void GetProcessCreationTime(ProcessInfo& process_info); ++ + void GetCpuUsage(Cpu& cpu_info, bool first_run); + + #endif // SRC_PROCESS_H_ diff --git a/src/process_commandline.cc b/src/process_commandline.cc index ea822b120e8038a4803e34647042f08f4aaf5ca1..25907c0bf542bed6c72b1b462b19bcf3210c3cfd 100644 --- a/src/process_commandline.cc +++ b/src/process_commandline.cc -@@ -1,67 +1,125 @@ --/*--------------------------------------------------------------------------------------------- -- * Copyright (c) Microsoft Corporation. All rights reserved. -- * Licensed under the MIT License. See License.txt in the project root for license information. -- *--------------------------------------------------------------------------------------------*/ -- --#include "process.h" --#include "process_commandline.h" --#include --#include +@@ -7,61 +7,119 @@ + #include "process_commandline.h" + #include + #include -#include -- ++#include + -bool GetProcessCommandLine(ProcessInfo& process_info) { - HINSTANCE ntdll = GetModuleHandleW(L"ntdll.dll"); -- if (!ntdll) { -- return false; -- } -- -- decltype(NtQueryInformationProcess)* nt_query_information_process = -- reinterpret_cast( -- GetProcAddress(ntdll, "NtQueryInformationProcess")); -- -- if (!nt_query_information_process) { -- return false; -- } -- -- PROCESS_BASIC_INFORMATION pbi{}; -- PEB peb = {NULL}; -- RTL_USER_PROCESS_PARAMETERS process_parameters = {NULL}; -- -- // Get process handle -- DWORD pid = process_info.pid; -- HANDLE hProcess = OpenProcess(PROCESS_QUERY_INFORMATION | PROCESS_VM_READ, FALSE, pid); -- if (hProcess == INVALID_HANDLE_VALUE) { -- return false; -- } -- -- // Get Process Environment Block (PEB) -- NTSTATUS status = nt_query_information_process(hProcess, ProcessBasicInformation, &pbi, sizeof(pbi), nullptr); -- if (NT_SUCCESS(status) && pbi.PebBaseAddress) { -- // Read PEB -- if (ReadProcessMemory(hProcess, pbi.PebBaseAddress, &peb, sizeof(peb), nullptr)) { -- // Read the processs parameters -- if (ReadProcessMemory(hProcess, peb.ProcessParameters, &process_parameters, sizeof(RTL_USER_PROCESS_PARAMETERS), nullptr)) { -- if (process_parameters.CommandLine.Length > 0) { -- std::wstring buffer; -- buffer.resize(process_parameters.CommandLine.Length / sizeof(wchar_t)); -- if (ReadProcessMemory(hProcess, process_parameters.CommandLine.Buffer, &buffer[0], process_parameters.CommandLine.Length, nullptr)) { -- int wide_length = static_cast(buffer.length()); -- int charcount = WideCharToMultiByte(CP_UTF8, 0, buffer.data(), wide_length, -- NULL, 0, NULL, NULL); -- if (charcount) { -- process_info.commandLine.resize(static_cast(charcount)); -- WideCharToMultiByte(CP_UTF8, 0, buffer.data(), wide_length, -- &process_info.commandLine[0], charcount, -- NULL, NULL); -- } -- CloseHandle(hProcess); -- return true; -- } -- } -- } -- } -- } -- -- CloseHandle(hProcess); -- return false; --} -+/*--------------------------------------------------------------------------------------------- -+ * Copyright (c) Microsoft Corporation. All rights reserved. -+ * Licensed under the MIT License. See License.txt in the project root for license information. -+ *--------------------------------------------------------------------------------------------*/ -+ -+#include "process.h" -+#include "process_commandline.h" -+#include -+#include -+#include -+ +namespace { + +// Windows 8.1 and later hand back a process's command line as a UNICODE_STRING @@ -366,7 +260,7 @@ index ea822b120e8038a4803e34647042f08f4aaf5ca1..25907c0bf542bed6c72b1b462b19bcf3 +// ntdll ships no import library for this entry point; it has to be resolved. +NtQueryInformationProcessFn ResolveNtQueryInformationProcess() { + HMODULE ntdll = GetModuleHandleW(L"ntdll.dll"); -+ if (!ntdll) { + if (!ntdll) { + return nullptr; + } + return reinterpret_cast( @@ -385,8 +279,8 @@ index ea822b120e8038a4803e34647042f08f4aaf5ca1..25907c0bf542bed6c72b1b462b19bcf3 + int length = static_cast(wide_length); + int charcount = WideCharToMultiByte(CP_UTF8, 0, data, length, NULL, 0, NULL, NULL); + if (!charcount) { -+ return false; -+ } + return false; + } + process_info.commandLine.resize(static_cast(charcount)); + WideCharToMultiByte(CP_UTF8, 0, data, length, &process_info.commandLine[0], charcount, NULL, + NULL); @@ -394,18 +288,25 @@ index ea822b120e8038a4803e34647042f08f4aaf5ca1..25907c0bf542bed6c72b1b462b19bcf3 +} + +} // namespace -+ + +- decltype(NtQueryInformationProcess)* nt_query_information_process = +- reinterpret_cast( +- GetProcAddress(ntdll, "NtQueryInformationProcess")); +bool GetProcessCommandLine(ProcessInfo& process_info) { + NtQueryInformationProcessFn query = NtQueryInformationProcessEntry(); + if (!query) { + return false; + } -+ + +- if (!nt_query_information_process) { + HANDLE process = OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, FALSE, process_info.pid); + if (process == NULL) { -+ return false; -+ } -+ + return false; + } + +- PROCESS_BASIC_INFORMATION pbi{}; +- PEB peb = {NULL}; +- RTL_USER_PROCESS_PARAMETERS process_parameters = {NULL}; + ULONG size = 0; + NTSTATUS status = query(process, kProcessCommandLineInformation, nullptr, 0, &size); + if (NT_SUCCESS(status)) { @@ -421,14 +322,44 @@ index ea822b120e8038a4803e34647042f08f4aaf5ca1..25907c0bf542bed6c72b1b462b19bcf3 + CloseHandle(process); + return false; + } -+ + +- // Get process handle +- DWORD pid = process_info.pid; +- HANDLE hProcess = OpenProcess(PROCESS_QUERY_INFORMATION | PROCESS_VM_READ, FALSE, pid); +- if (hProcess == INVALID_HANDLE_VALUE) { + std::vector buffer(size); + status = query(process, kProcessCommandLineInformation, &buffer[0], size, &size); + CloseHandle(process); + if (!NT_SUCCESS(status)) { -+ return false; -+ } -+ + return false; + } + +- // Get Process Environment Block (PEB) +- NTSTATUS status = nt_query_information_process(hProcess, ProcessBasicInformation, &pbi, sizeof(pbi), nullptr); +- if (NT_SUCCESS(status) && pbi.PebBaseAddress) { +- // Read PEB +- if (ReadProcessMemory(hProcess, pbi.PebBaseAddress, &peb, sizeof(peb), nullptr)) { +- // Read the processs parameters +- if (ReadProcessMemory(hProcess, peb.ProcessParameters, &process_parameters, sizeof(RTL_USER_PROCESS_PARAMETERS), nullptr)) { +- if (process_parameters.CommandLine.Length > 0) { +- std::wstring buffer; +- buffer.resize(process_parameters.CommandLine.Length / sizeof(wchar_t)); +- if (ReadProcessMemory(hProcess, process_parameters.CommandLine.Buffer, &buffer[0], process_parameters.CommandLine.Length, nullptr)) { +- int wide_length = static_cast(buffer.length()); +- int charcount = WideCharToMultiByte(CP_UTF8, 0, buffer.data(), wide_length, +- NULL, 0, NULL, NULL); +- if (charcount) { +- process_info.commandLine.resize(static_cast(charcount)); +- WideCharToMultiByte(CP_UTF8, 0, buffer.data(), wide_length, +- &process_info.commandLine[0], charcount, +- NULL, NULL); +- } +- CloseHandle(hProcess); +- return true; +- } +- } +- } +- } + // Header and characters arrive in one allocation, but treat the header as + // untrusted: a hooked ntdll is the case this reader is written for, and an + // unchecked Buffer/Length here would be an over-read encoded straight into JS. @@ -440,11 +371,70 @@ index ea822b120e8038a4803e34647042f08f4aaf5ca1..25907c0bf542bed6c72b1b462b19bcf3 + if (chars == nullptr || chars < begin + sizeof(UNICODE_STRING) || chars > end || + command_line->Length > static_cast(end - chars)) { + return false; -+ } -+ + } + +- CloseHandle(hProcess); +- return false; + // True only when a command line was actually stored, so "empty" and "not + // recovered" stay the same answer they were before this reader replaced the + // PEB read. `src/process.cc` discards the result either way. + return StoreCommandLineUtf8(process_info, command_line->Buffer, + command_line->Length / sizeof(wchar_t)); -+} + } +diff --git a/src/process_worker.cc b/src/process_worker.cc +index c9e3457a759c1acaa2644231a4917d45aed951f8..3f26a354477f062b34bd31fbd17be529e6a2fd7a 100644 +--- a/src/process_worker.cc ++++ b/src/process_worker.cc +@@ -43,6 +43,11 @@ void GetProcessesWorker::OnOK() { + Napi::String::New(env, pinfo.commandLine)); + } + ++ if ((CREATIONTIME & process_data_flags_) && pinfo.creationTimeMs != 0) { ++ object.Set("creationTimeMs", ++ Napi::Number::New(env, static_cast(pinfo.creationTimeMs))); ++ } ++ + result.Set(i, object); + } + +diff --git a/typings/windows-process-tree.d.ts b/typings/windows-process-tree.d.ts +index 08bdac2fdc5ead6f0fcfb5ee5a021e2298c7d523..458981566fc45c0084badff566b1e3791ec1b629 100644 +--- a/typings/windows-process-tree.d.ts ++++ b/typings/windows-process-tree.d.ts +@@ -7,9 +7,17 @@ declare module '@vscode/windows-process-tree' { + export enum ProcessDataFlag { + None = 0, + Memory = 1, +- CommandLine = 2 ++ CommandLine = 2, ++ CreationTime = 4 + } + ++ /** ++ * The flag bits the compiled addon actually understands, or undefined off ++ * win32. `ProcessDataFlag` above is source; this is what the binary reports, ++ * so it is the only way to tell a patched build from a stale prebuilt. ++ */ ++ export const supportedProcessDataFlags: number | undefined; ++ + export interface IProcessInfo { + pid: number; + ppid: number; +@@ -24,6 +32,9 @@ declare module '@vscode/windows-process-tree' { + * The string returned is at most 512 chars, strings exceeding this length are truncated. + */ + commandLine?: string; ++ ++ /** Process creation time in Unix milliseconds. */ ++ creationTimeMs?: number; + } + + export interface IProcessCpuInfo extends IProcessInfo { +@@ -35,6 +46,7 @@ declare module '@vscode/windows-process-tree' { + name: string; + memory?: number; + commandLine?: string; ++ creationTimeMs?: number; + children: IProcessTreeNode[]; + } + diff --git a/config/reliability-gates.jsonc b/config/reliability-gates.jsonc index a6ac6b1fe23..2bcba7cb737 100644 --- a/config/reliability-gates.jsonc +++ b/config/reliability-gates.jsonc @@ -3063,7 +3063,7 @@ "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/browser-preview-tool-authorization.test.ts src/main/browser/doc-preview-guest-policy.test.ts src/main/ipc/browser.test.ts", "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/browser-preview-tool-authorization.test.ts src/main/browser/browser-manager-annotation-bridge.test.ts src/main/browser/browser-manager-guest-lifecycle.test.ts src/shared/doc-preview-scheme.test.ts src/renderer/src/components/browser-pane/workspace-doc/doc-preview-document-actions.test.ts src/renderer/src/components/browser-pane/workspace-doc/use-doc-preview-guest-tools.test.ts src/renderer/src/components/editor/EditorPanelShell.header.test.tsx src/renderer/src/components/browser-pane/workspace-doc/HtmlDocPreview.toolbar.test.tsx", "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/browser-preview-tool-authorization.test.ts src/main/browser/browser-manager-annotation-bridge.test.ts src/main/browser/browser-manager-guest-lifecycle.test.ts src/main/browser/offscreen-browser-backend-lifecycle.test.ts src/main/browser/doc-preview-guest-policy.test.ts src/shared/doc-preview-scheme.test.ts src/renderer/src/components/browser-pane/workspace-doc/doc-preview-document-actions.test.ts src/renderer/src/components/browser-pane/workspace-doc/use-doc-preview-guest-tools.test.ts src/renderer/src/components/editor/EditorPanelShell.header.test.tsx src/renderer/src/components/browser-pane/workspace-doc/HtmlDocPreview.toolbar.test.tsx", - "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/doc-preview-grant-ipc.test.ts src/renderer/src/components/terminal-pane/TerminalLinkActionPopover.test.tsx src/renderer/src/runtime/sync-runtime-graph-editor-diff-tabs.test.ts src/renderer/src/store/slices/tabs-hydration.test.ts", + "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/doc-preview-grant-ipc.test.ts src/renderer/src/components/link-actions/LinkActionPopover.test.tsx src/renderer/src/runtime/sync-runtime-graph-editor-diff-tabs.test.ts src/renderer/src/store/slices/tabs-hydration.test.ts", "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/browser-preview-tool-authorization.test.ts src/main/ipc/browser-tab-registration-wait.test.ts src/main/ipc/doc-preview-grant-ipc.test.ts src/main/browser/browser-manager-guest-lifecycle.test.ts src/main/browser/browser-manager-guest-policy-profile.test.ts src/main/browser/browser-manager-annotation-bridge.test.ts src/main/browser/offscreen-browser-backend-lifecycle.test.ts src/main/browser/doc-preview-guest-policy.test.ts src/shared/doc-preview-scheme.test.ts src/renderer/src/components/browser-pane/workspace-doc/use-doc-preview-guest-tools.test.ts src/renderer/src/components/browser-pane/workspace-doc/HtmlDocPreview.toolbar.test.tsx", // STA-5681 address-bar convergence: conversion is page replacement (fresh id, one store // commit flips page + mirror + mobile observables), typed workspace paths convert via the @@ -3127,7 +3127,7 @@ "src/renderer/src/components/terminal-pane/terminal-file-link-actions.test.ts", "src/main/ipc/doc-preview-grant-ipc.test.ts", "src/renderer/src/store/slices/tabs-hydration.test.ts", - "src/renderer/src/components/terminal-pane/TerminalLinkActionPopover.test.tsx", + "src/renderer/src/components/link-actions/LinkActionPopover.test.tsx", "src/renderer/src/runtime/sync-runtime-graph-editor-diff-tabs.test.ts", "src/renderer/src/store/slices/browser-page-conversion.test.ts", "src/renderer/src/runtime/sync-runtime-graph-conversion-publish.test.ts", @@ -3559,13 +3559,13 @@ "summary": "29/29 on the candidate that makes the preview a browser tab. The preview action now creates a page located by the document; reopening the same document activates the tab it is already in rather than minting a second grant on one file; and closing that tab revokes its grant, which nothing else does now that the editor tab's close hook is gone. Red-green with each mutant as the sole delta: dropping the reuse lookup opens a second tab for a document already on screen, and dropping the release on close leaves the document readable through a grant nothing revokes until the process ends. Both are paired with presence preconditions in the same runs — a second, different document still gets its own tab, and a URL tab closed beside the document tab revokes nothing, so a release fired for every close would fail rather than pass." }, { - "date": "2026-08-27", + "date": "2026-09-06", "runner": "local", "platform": "macos", - "command": "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/doc-preview-grant-ipc.test.ts src/renderer/src/components/terminal-pane/TerminalLinkActionPopover.test.tsx src/renderer/src/runtime/sync-runtime-graph-editor-diff-tabs.test.ts src/renderer/src/store/slices/tabs-hydration.test.ts", + "command": "pnpm exec vitest run --config config/vitest.config.ts src/main/ipc/doc-preview-grant-ipc.test.ts src/renderer/src/components/link-actions/LinkActionPopover.test.tsx src/renderer/src/runtime/sync-runtime-graph-editor-diff-tabs.test.ts src/renderer/src/store/slices/tabs-hydration.test.ts", "result": "passed", - "durationSeconds": 4.77, - "summary": "40/40 on the candidate that removes the editor preview species and moves its two boundary duties onto the browser tab. The publish filter moved with the tab: it used to exclude an editor file by mode, and now excludes a browser workspace by whether it is located by a document — asserted at both the group projection and the tab loop, because a mutant that filters only one of them still publishes. Migration is the other half: sessions written by builds that made previews editor tabs carry chrome whose entity id encodes the document, and whose document was never persisted, so it has always come back naming a surface no restore produces; that chrome is now dropped. Each mutant as the sole delta: filtering neither place publishes the document tab to mobile, filtering only the projection still publishes it, and removing the migration leaves the stale strip entry. Presence preconditions in the same runs: an ordinary browser tab beside the document tab does publish, and the ordinary editor tab for the very same document survives hydration." + "durationSeconds": 7.12, + "summary": "44/44 across all four files after PR #19130 moved the terminal popover suite to the shared LinkActionPopover path; all eight popover cases remain. This replaces the 2026-08-27 command that named the removed test path. Historical evidence from that run (4.77 seconds; mutation checks were not repeated in this rerun): 40/40 on the candidate that removes the editor preview species and moves its two boundary duties onto the browser tab. The publish filter moved with the tab: it used to exclude an editor file by mode, and now excludes a browser workspace by whether it is located by a document — asserted at both the group projection and the tab loop, because a mutant that filters only one of them still publishes. Migration is the other half: sessions written by builds that made previews editor tabs carry chrome whose entity id encodes the document, and whose document was never persisted, so it has always come back naming a surface no restore produces; that chrome is now dropped. Each mutant as the sole delta: filtering neither place publishes the document tab to mobile, filtering only the projection still publishes it, and removing the migration leaves the stale strip entry. Presence preconditions in the same runs: an ordinary browser tab beside the document tab does publish, and the ordinary editor tab for the very same document survives hydration." }, { "date": "2026-08-27", @@ -18472,6 +18472,198 @@ "The new PR lane is outside verify until reliability is established." ], "demotionRule": "Keep experimental if provisioning or an execution flakes; never promote by skipping a case, raising timeouts, or retrying until green." + }, + { + "id": "browser.packaged-mixed-version-placement", + "title": "Packaged browser placement across versions", + "maturity": "experimental", + "protection": "partial", + "owner": "browser-runtime", + "layer": "electron-packaged", + "surfaces": [ + "paired browser placement" + ], + "platforms": [ + "linux", + "macos", + "windows" + ], + "providers": [ + "paired-runtime" + ], + "coveredPlatforms": [ + "linux" + ], + "coveredProviders": [ + "paired-runtime" + ], + "coverageNotes": "Published Linux 1.4.188 desktop against current source in both directions; scheduled weekly and manually runnable. No required PR check.", + "motivatingLinks": [ + "https://github.com/stablyai/orca/actions/runs/34069063016" + ], + "invariant": "A paired client and host without client-hosted browser capabilities retain server-hosted browser placement across supported version skew.", + "oracle": "Require both existing named browser placement scenarios to pass three times with one attempt, zero skips, zero failures, and no report errors.", + "commands": [ + "gh workflow run packaged-browser-e2e.yml", + "pnpm exec playwright test tests/e2e/packaged-mixed-version-browser-placement.spec.ts --config tests/playwright.config.ts --project=electron-headless --workers=1 --repeat-each=3 --retries=0", + "node_modules/.bin/vitest run --config config/vitest.config.ts config/scripts/packaged-browser-lane-contract.test.mjs config/scripts/verify-packaged-browser-participation.test.mjs", + "gh run view 34069063016 --log" + ], + "testFiles": [ + "tests/e2e/packaged-mixed-version-browser-placement.spec.ts", + "config/scripts/packaged-browser-lane-contract.test.mjs", + "config/scripts/verify-packaged-browser-participation.test.mjs" + ], + "assertionRefs": [ + { + "file": "tests/e2e/packaged-mixed-version-browser-placement.spec.ts", + "assertions": [ + "old client and old host lack client-host and browser-tunnel capabilities", + "browser contents remain owned by the server and the expected snapshot marker is readable" + ] + }, + { + "file": "config/scripts/verify-packaged-browser-participation.test.mjs", + "assertions": [ + "reject missing, substituted, skipped and retried scenarios" + ] + }, + { + "file": "config/scripts/packaged-browser-lane-contract.test.mjs", + "assertions": [ + "verify pinned package checksum before extraction", + "require both directions three times and run report verification even on failure" + ] + } + ], + "evidenceRuns": [ + { + "date": "2026-09-07", + "runner": "ci", + "platform": "linux", + "result": "passed", + "command": "gh run view 34069063016 --log", + "durationSeconds": 120, + "summary": "Both unmodified compatibility cases passed three times at 5a99f935 with published1.4.188 and main f7d52160162; retries0. Final workflow34069429156 also passed6/6; its downloaded JSON passed the same participation verifier." + } + ], + "runtimeBudget": { + "p95Seconds": 1500, + "scope": "CI job timeout; not a measured p95" + }, + "flakeHistory": { + "status": "soaking", + "evidence": "Initial executable discovery matched CLI and desktop and was corrected before any tests ran. Corrected baseline2/2 and repeat6/6 pass." + }, + "redGreenEvidence": { + "status": "partial", + "evidence": "Participation unit tests reject missing and retried scenarios; no application mutation proof." + }, + "performanceBudget": { + "required": false, + "evidence": "Compatibility assertions, not a performance benchmark." + }, + "promotionCriteria": [ + "Final workflow JSON report proves all six executions.", + "Collect repeated scheduled history before making this required." + ], + "knownGaps": [ + "Linux1.4.188 only; no macOS or Windows packaged coverage.", + "No folder workspace, SSH execution host or live-service coverage.", + "Other released version pairs remain untested; not a required PR check." + ], + "demotionRule": "Keep experimental if any direction skips or fails; do not extend timeouts or retry to green." + }, + { + "id": "terminal-input.native-wayland-hangul-digit", + "title": "Native Wayland Hangul terminating digits reach the PTY exactly once", + "maturity": "experimental", + "protection": "partial", + "owner": "terminal-input", + "layer": "electron-native-ime-e2e", + "surfaces": [ + "native Hangul composition", + "Wayland terminal input" + ], + "platforms": [ + "linux" + ], + "providers": [ + "local" + ], + "coveredPlatforms": [ + "linux" + ], + "coveredProviders": [ + "local" + ], + "coverageNotes": "Ubuntu 22.04 nested GNOME and IBus Hangul drive three complete native executions in GitHub Actions. GNOME owns IBus; daemon and CLI share its default config discovery path.", + "motivatingLinks": [ + "https://github.com/stablyai/orca/pull/19174" + ], + "invariant": "Typing d k 1 Return through native IBus Hangul delivers exactly 아1 followed by newline without missing, duplicate, or reordered characters.", + "oracle": "Three executions each assert three exact UTF-8 PTY lines. Verify the exact Playwright title, zero skips/retries, each individual native composition receipt, and the nested launch Wayland flag.", + "commands": [ + "gh workflow run terminal-ime-e2e.yml", + "gh run view 34074017928 --log", + "pnpm exec playwright test --config tests/playwright.config.ts tests/e2e/terminal-hangul-terminating-digit-native.spec.ts --project=electron-headful --workers=1 --repeat-each=3 --retries=0 --reporter=list,json", + "ORCA_BACKGROUND_LAUNCH=1 node_modules/.bin/vitest run --config config/vitest.config.ts config/scripts/terminal-ime-e2e-workflow.test.mjs" + ], + "testFiles": [ + "tests/e2e/terminal-hangul-terminating-digit-native.spec.ts", + "config/scripts/terminal-ime-e2e-workflow.test.mjs" + ], + "assertionRefs": [ + { + "file": "tests/e2e/terminal-hangul-terminating-digit-native.spec.ts", + "assertions": [ + "a digit typed right after a Hangul syllable reaches the pty" + ] + }, + { + "file": "config/scripts/terminal-ime-e2e-workflow.test.mjs", + "assertions": [ + "runs native Wayland independently with CJK fonts and retained evidence" + ] + } + ], + "evidenceRuns": [ + { + "date": "2026-09-07", + "runner": "ci", + "platform": "linux", + "command": "gh run view 34074017928 --log", + "result": "passed", + "summary": "Permanent runner passed three native executions, nine exact lines and zero skips/retries. Downloaded participation report and all three engagement receipts verified; compositor cleanup reported no remaining group members. Independent X11 job passed.", + "durationSeconds": 76.61 + } + ], + "runtimeBudget": { + "p95Seconds": 1500, + "scope": "CI job timeout including installation/build; measured p95 not established" + }, + "flakeHistory": { + "status": "soaking", + "evidence": "Earlier diagnostic repetition had one unexplained missing Hangul commit. GNOME-owned diagnostic and corrected permanent runner each passed 3/3. Long-term soak is missing." + }, + "redGreenEvidence": { + "status": "partial", + "evidence": "Original exact-byte assertions retained. Permanent startup failed until the GNOME config-discovery mismatch was corrected. No intentional production regression was introduced." + }, + "performanceBudget": { + "required": false, + "evidence": "CI-only harness; no application runtime changes." + }, + "promotionCriteria": [ + "Collect 100 soak runs across 14 days with no unexplained flakes.", + "Exercise native Wayland desktops beyond nested GNOME before broadening the claim." + ], + "knownGaps": [ + "Only Hangul terminating digits; no native candidate-selection or other input-method coverage claim.", + "No macOS, Windows, SSH terminal, packaged build, or mixed-version claim.", + "Default config paths are shared with GNOME on a disposable hosted CI runner; nested mode refuses non-GitHub-Actions execution." + ], + "demotionRule": "Keep experimental on unexplained failures; retain exact bytes and participation checks without retries, skips, or longer deadlines." } ] } diff --git a/config/scripts/app-store-performance-plugin.test.mjs b/config/scripts/app-store-performance-plugin.test.mjs index bb2f305ba92..d8e2568165f 100644 --- a/config/scripts/app-store-performance-plugin.test.mjs +++ b/config/scripts/app-store-performance-plugin.test.mjs @@ -12,7 +12,8 @@ function lintSource(source) { rules: { 'app-store-performance/require-selector': 'warn', 'app-store-performance/no-identity-selector': 'warn', - 'app-store-performance/no-fresh-selector-result': 'warn' + 'app-store-performance/no-fresh-selector-result': 'warn', + 'app-store-performance/no-nested-fresh-under-shallow': 'warn' } }) } @@ -52,4 +53,92 @@ describe('app store performance Oxlint plugin', () => { expect(diagnostics).toEqual([]) }) + + it('resolves selectors referenced by name, including ones hoisted below the call', () => { + const diagnostics = lintSource(` + import { useAppStore } from '@/store' + const EarlyFresh = () => useAppStore(selectFreshRows) + const selectFreshRows = (state) => state.rows.filter(Boolean) + const Stable = () => useAppStore(selectActiveId) + const selectActiveId = (state) => state.activeId + `) + + expect(diagnostics.map((diagnostic) => diagnostic.code)).toEqual([ + 'app-store-performance(no-fresh-selector-result)' + ]) + }) + + it('does not let a component-local helper resolve a same-named imported selector', () => { + const diagnostics = lintSource(` + import { useAppStore } from '@/store' + import { selectRows } from './selectors' + const Other = () => { + const selectRows = (state) => state.rows.map((row) => row.id) + return selectRows + } + const Imported = () => useAppStore(selectRows) + `) + + expect(diagnostics).toEqual([]) + }) + + it('covers sibling store hooks but not useSyncExternalStore', () => { + const diagnostics = lintSource(` + import { usePluginPanelsStore } from '@/store/plugin-panels' + import { useSyncExternalStore } from 'react' + const WholePanels = () => usePluginPanelsStore() + const FreshPanels = () => usePluginPanelsStore((state) => ({ open: state.open })) + const External = () => useSyncExternalStore(subscribe, () => ({ open: true })) + `) + + expect(diagnostics.map((diagnostic) => diagnostic.code)).toEqual([ + 'app-store-performance(require-selector)', + 'app-store-performance(no-fresh-selector-result)' + ]) + }) + + it('reports fresh references nested inside a useShallow projection', () => { + const diagnostics = lintSource(` + import { useAppStore } from '@/store' + import { useShallow } from 'zustand/react/shallow' + const NestedObject = () => useAppStore(useShallow((state) => ({ ids: state.rows.map((row) => row.id) }))) + const NestedArray = () => useAppStore(useShallow((state) => [state.activeId, state.rows.filter(Boolean)])) + const Flat = () => useAppStore(useShallow((state) => ({ activeId: state.activeId, rows: state.rows }))) + `) + + expect(diagnostics.map((diagnostic) => diagnostic.code)).toEqual([ + 'app-store-performance(no-nested-fresh-under-shallow)', + 'app-store-performance(no-nested-fresh-under-shallow)' + ]) + }) + + it('follows a selector one hop into a module-scope helper', () => { + const diagnostics = lintSource(` + import { useAppStore } from '@/store' + import { useShallow } from 'zustand/react/shallow' + const buildRows = (state) => state.rows.map((row) => row.id) + const Delegating = () => useAppStore((state) => buildRows(state)) + const NestedDelegating = () => useAppStore(useShallow((state) => ({ ids: buildRows(state) }))) + `) + + expect(diagnostics.map((diagnostic) => diagnostic.code)).toEqual([ + 'app-store-performance(no-fresh-selector-result)', + 'app-store-performance(no-nested-fresh-under-shallow)' + ]) + }) + + it('does not flag a helper that returns a cached reference on some branch', () => { + const diagnostics = lintSource(` + import { useAppStore } from '@/store' + import { useShallow } from 'zustand/react/shallow' + // The identity-caching shape: fresh only on a miss, cached otherwise. + const selectCachedRows = (state) => cache.get(state.key) ?? state.rows.filter(Boolean) + const Cached = () => useAppStore((state) => selectCachedRows(state)) + const CachedNested = () => useAppStore(useShallow((state) => ({ rows: selectCachedRows(state) }))) + // An unknown helper cannot be resolved, so it must not be guessed at. + const External = () => useAppStore((state) => externalBuild(state)) + `) + + expect(diagnostics).toEqual([]) + }) }) diff --git a/config/scripts/build-windows-process-tree-relay-addon.mjs b/config/scripts/build-windows-process-tree-relay-addon.mjs index 9243f5a5b78..912bbd3c174 100644 --- a/config/scripts/build-windows-process-tree-relay-addon.mjs +++ b/config/scripts/build-windows-process-tree-relay-addon.mjs @@ -98,6 +98,210 @@ function assertPatchApplied() { 'config/patches/@vscode__windows-process-tree@0.8.0.patch; run pnpm install.' ) } + // Every string the repair below can write, so a repaired tree cannot be + // declared patched while one of the pieces is silently missing. + const requiredCreationTimeSources = [ + ['src/process.h', 'CREATIONTIME = 4'], + ['src/process.h', 'ULONGLONG creationTimeMs'], + ['src/process.cc', 'GetProcessCreationTime(pinfo)'], + ['src/process.cc', 'GetProcessTimes(hProcess, &creationTime'], + ['src/process_worker.cc', 'object.Set("creationTimeMs"'], + ['src/addon.cc', 'exports.Set("supportedProcessDataFlags"'], + ['lib/index.js', '["CreationTime"] = 4'], + ['lib/index.js', 'exports.supportedProcessDataFlags'], + ['lib/index.js', 'creationTimeMs,'], + ['lib/index.ts', 'CreationTime = 4'], + ['lib/index.ts', 'export const supportedProcessDataFlags'], + ['lib/index.ts', 'creationTimeMs,'], + ['typings/windows-process-tree.d.ts', 'creationTimeMs?: number'], + // A regex because IProcessInfo declares the same field: only the tree node + // is followed by `children`, and that is the one buildNode fills. + ['typings/windows-process-tree.d.ts', /creationTimeMs\?: number;\r?\n\s*children:/], + ['typings/windows-process-tree.d.ts', 'export const supportedProcessDataFlags'] + ] + for (const [relativePath, expected] of requiredCreationTimeSources) { + const source = readFileSync(join(PACKAGE_DIR, relativePath), 'utf8') + const present = typeof expected === 'string' ? source.includes(expected) : expected.test(source) + if (!present) { + throw new Error( + `${relativePath} does not contain the process creation-time patch (${expected}). ` + + 'Run pnpm install before building the relay addon.' + ) + } + } +} + +function repairCreationTimeSources() { + let repaired = false + const rewrite = (relativePath, transform) => { + const filePath = join(PACKAGE_DIR, relativePath) + const source = readFileSync(filePath, 'utf8') + const next = transform(source, source.includes('\r\n') ? '\r\n' : '\n') + if (next !== source) { + writeFileSync(filePath, next) + repaired = true + } + } + + rewrite('src/process.h', (source, eol) => { + let next = source + if (!next.includes('ULONGLONG creationTimeMs')) { + next = next.replace( + / std::string commandLine;\r?\n/, + ` std::string commandLine;${eol} ULONGLONG creationTimeMs;${eol}` + ) + } + if (!next.includes('CREATIONTIME = 4')) { + next = next.replace( + / COMMANDLINE = 2\r?\n/, + ` COMMANDLINE = 2,${eol} CREATIONTIME = 4${eol}` + ) + } + if (!next.includes('void GetProcessCreationTime')) { + next = next.replace( + /void GetProcessMemoryUsage\(ProcessInfo& process_info\);\r?\n/, + `void GetProcessMemoryUsage(ProcessInfo& process_info);${eol}${eol}` + + `void GetProcessCreationTime(ProcessInfo& process_info);${eol}` + ) + } + return next + }) + + rewrite('src/process.cc', (source, eol) => { + let next = source.replace('ProcessInfo pinfo;', 'ProcessInfo pinfo{};') + if (!next.includes('GetProcessCreationTime(pinfo)')) { + next = next.replace( + /( if \(COMMANDLINE & process_data_flags\) \{\r?\n GetProcessCommandLine\(pinfo\);\r?\n \})/, + `$1${eol}${eol} if (CREATIONTIME & process_data_flags) {${eol}` + + ` GetProcessCreationTime(pinfo);${eol} }` + ) + } + if (!next.includes('void GetProcessCreationTime(ProcessInfo& process_info) {')) { + const producer = [ + 'void GetProcessCreationTime(ProcessInfo& process_info) {', + ' HANDLE hProcess = OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, false, process_info.pid);', + ' if (hProcess == NULL) {', + ' return;', + ' }', + '', + ' FILETIME creationTime, exitTime, kernelTime, userTime;', + ' if (GetProcessTimes(hProcess, &creationTime, &exitTime, &kernelTime, &userTime)) {', + ' ULARGE_INTEGER timestamp;', + ' timestamp.LowPart = creationTime.dwLowDateTime;', + ' timestamp.HighPart = creationTime.dwHighDateTime;', + ' constexpr ULONGLONG WINDOWS_EPOCH_OFFSET_100NS = 116444736000000000ULL;', + ' constexpr ULONGLONG HUNDRED_NS_PER_MILLISECOND = 10000ULL;', + ' if (timestamp.QuadPart >= WINDOWS_EPOCH_OFFSET_100NS) {', + ' process_info.creationTimeMs =', + ' (timestamp.QuadPart - WINDOWS_EPOCH_OFFSET_100NS) / HUNDRED_NS_PER_MILLISECOND;', + ' }', + ' }', + '', + ' CloseHandle(hProcess);', + '}', + '' + ].join(eol) + next = next.replace( + 'void GetProcessMemoryUsage', + `${producer}${eol}void GetProcessMemoryUsage` + ) + } + return next + }) + + rewrite('src/process_worker.cc', (source, eol) => { + if (source.includes('object.Set("creationTimeMs"')) { + return source + } + const emission = [ + ' if ((CREATIONTIME & process_data_flags_) && pinfo.creationTimeMs != 0) {', + ' object.Set("creationTimeMs",', + ' Napi::Number::New(env, static_cast(pinfo.creationTimeMs)));', + ' }', + '' + ].join(eol) + return source.replace( + ' result.Set(i, object);', + `${emission}${eol} result.Set(i, object);` + ) + }) + + rewrite('src/addon.cc', (source, eol) => { + if (source.includes('exports.Set("supportedProcessDataFlags"')) { + return source + } + return source.replace( + /( exports\.Set\("getProcessCpuUsage", Napi::Function::New\(env, GetProcessCpuUsage\)\);\r?\n)/, + `$1 exports.Set("supportedProcessDataFlags",${eol}` + + ` Napi::Number::New(env, MEMORY | COMMANDLINE | CREATIONTIME));${eol}` + ) + }) + + // Each piece is guarded on its own: an early-out on the enum alone would let a + // tree with the enum but no buildNode splat pass as repaired. + const NATIVE_CONST = + "const native = process.platform === 'win32' ? require('../build/Release/windows_process_tree.node') : undefined;" + for (const relativePath of ['lib/index.ts', 'lib/index.js']) { + const isTs = relativePath.endsWith('.ts') + rewrite(relativePath, (source, eol) => { + let next = source + if (!next.includes('CreationTime')) { + next = isTs + ? next.replace(' CommandLine = 2', ` CommandLine = 2,${eol} CreationTime = 4`) + : next.replace( + ' ProcessDataFlag[ProcessDataFlag["CommandLine"] = 2] = "CommandLine";', + ' ProcessDataFlag[ProcessDataFlag["CommandLine"] = 2] = "CommandLine";' + + `${eol} ProcessDataFlag[ProcessDataFlag["CreationTime"] = 4] = "CreationTime";` + ) + } + if (!next.includes('supportedProcessDataFlags')) { + const reExport = isTs + ? `/** The flag bits this compiled addon reports; undefined off win32. */${eol}` + + 'export const supportedProcessDataFlags: number | undefined = native?.supportedProcessDataFlags;' + : 'exports.supportedProcessDataFlags = native === undefined ? undefined : native.supportedProcessDataFlags;' + next = next.replace(NATIVE_CONST, `${NATIVE_CONST}${eol}${reExport}`) + } + // buildNode drops any field it does not name, so the destructure and the + // splat have to move together. + next = next.replace(/(memory, commandLine)( \}, children \})/, '$1, creationTimeMs$2') + if (!/\bcreationTimeMs,/.test(next)) { + next = next.replace( + /(\r?\n)(\s*)commandLine,(\r?\n\s*children:)/, + `$1$2commandLine,$1$2creationTimeMs,$3` + ) + } + return next + }) + } + + rewrite('typings/windows-process-tree.d.ts', (source, eol) => { + let next = source + if (!next.includes('CreationTime = 4')) { + next = next.replace(' CommandLine = 2', ` CommandLine = 2,${eol} CreationTime = 4`) + } + if (!next.includes('supportedProcessDataFlags')) { + next = next.replace( + /( CreationTime = 4\r?\n \}\r?\n)/, + `$1${eol} /** The flag bits the compiled addon reports; undefined off win32. */${eol}` + + ` export const supportedProcessDataFlags: number | undefined;${eol}` + ) + } + if (!next.includes('creationTimeMs?: number')) { + next = next.replace( + / commandLine\?: string;\r?\n/, + ` commandLine?: string;${eol}${eol}` + + ` /** Process creation time in Unix milliseconds. */${eol}` + + ` creationTimeMs?: number;${eol}` + ) + } + // IProcessTreeNode is the second declaration; only it is followed by children. + next = next.replace( + /( commandLine\?: string;\r?\n)( children:)/, + `$1 creationTimeMs?: number;${eol}$2` + ) + return next + }) + return repaired } // pnpm can materialize this CRLF package without applying its patch. Repair the @@ -146,9 +350,15 @@ function applyWindowsProcessTreeBuildFixes() { if (processCc !== originalProcess) { writeFileSync(processPath, processCc) } + const repairedCreationTime = repairCreationTimeSources() stageWindowsProcessTreeNodeAddonApiHeaders(PACKAGE_DIR) const repairedCommandLine = ensureWindowsProcessTreeCommandLinePatch(PACKAGE_DIR) - if (bindingGyp !== originalBinding || processCc !== originalProcess || repairedCommandLine) { + if ( + bindingGyp !== originalBinding || + processCc !== originalProcess || + repairedCommandLine || + repairedCreationTime + ) { console.warn('[windows-process-tree] Repaired un-applied pnpm patch hunks before build.') } } diff --git a/config/scripts/ensure-native-runtime.mjs b/config/scripts/ensure-native-runtime.mjs index b2a47b99d5b..10e8426c2a5 100644 --- a/config/scripts/ensure-native-runtime.mjs +++ b/config/scripts/ensure-native-runtime.mjs @@ -2,7 +2,7 @@ import { spawnSync } from 'node:child_process' import { createRequire } from 'node:module' -import { existsSync, readFileSync } from 'node:fs' +import { existsSync, readFileSync, realpathSync } from 'node:fs' import { release } from 'node:os' import { basename, dirname, resolve } from 'node:path' import { @@ -14,6 +14,7 @@ import { const require = createRequire(import.meta.url) const { assertNodePtyJobOwnership } = require('./node-pty-job-ownership.cjs') +const { assertWindowsProcessTreeCreationTime } = require('./windows-process-tree-creation-time.cjs') const scriptPath = import.meta.filename const projectDir = resolve(import.meta.dirname, '../..') const runtime = readRuntimeArg() @@ -262,9 +263,10 @@ function loadNativeModule(moduleName) { // A bare require loads the .node addon on win32, so it catches an ABI // mismatch on its own. What it cannot catch is *which* addon loaded: the // published tarball ships a prebuilt built from unpatched source that is - // node-addon-api, so it requires cleanly and then reads every process's - // command line out of its address space. Check the binary, not the load. - require(moduleName) + // node-addon-api, so it requires cleanly, reads every process's command + // line out of its address space, and ignores the CreationTime flag. Check + // the binary on both counts, not the load. + assertWindowsProcessTreeCreationTime({ module: require(moduleName) }) if (inspectWindowsProcessTreeAddon(windowsProcessTreeAddonPath()) === 'unpatched') { throw new Error( 'the loaded addon still calls ReadProcessMemory, so it was not built from the patched ' + @@ -380,14 +382,18 @@ function getWindowsBuildNumber() { function rebuildNodeRuntimeModules(moduleNames) { for (const moduleName of moduleNames) { - const moduleDir = dirname(require.resolve(`${moduleName}/package.json`)) + let moduleDir = dirname(require.resolve(`${moduleName}/package.json`)) if (moduleName === '@vscode/windows-process-tree') { // Why before node-gyp: this module is rebuilt precisely because the // binary was the unpatched one, and pnpm materializes it unpatched often // enough that compiling the source as-is would just rebuild the same - // reader and fail the verify pass. + // reader and fail the verify pass. The patched binding.gyp then includes + // deps/node-addon-api, which the tarball does not ship, and node-gyp must + // run from the physical dir -- both reasons live in + // windows-process-tree-gyp-rebuild.mjs. ensureWindowsProcessTreeCommandLinePatch(moduleDir) stageWindowsProcessTreeNodeAddonApiHeaders(moduleDir) + moduleDir = realpathSync(moduleDir) } console.warn(`[native-runtime] Rebuilding ${moduleName} with node-gyp.`) runPnpm(['exec', 'node-gyp', 'rebuild'], { cwd: moduleDir }) diff --git a/config/scripts/ensure-native-runtime.test.mjs b/config/scripts/ensure-native-runtime.test.mjs index 973e2f6852d..1e7d888d2e2 100644 --- a/config/scripts/ensure-native-runtime.test.mjs +++ b/config/scripts/ensure-native-runtime.test.mjs @@ -15,9 +15,12 @@ import { describe, expect, it } from 'vitest' import { copyScriptWithLocalModules } from './script-module-dependencies.mjs' const sourceScriptPath = fileURLToPath(new URL('./ensure-native-runtime.mjs', import.meta.url)) -const sourceNodePtyJobOwnershipPath = fileURLToPath( - new URL('./node-pty-job-ownership.cjs', import.meta.url) -) +// The import walk sees `from './x.mjs'` only, so the createRequire'd CJS +// siblings have to be named. Without them the temp project cannot even load. +const REQUIRED_CJS_SIBLINGS = [ + 'node-pty-job-ownership.cjs', + 'windows-process-tree-creation-time.cjs' +] describe('ensure-native-runtime', () => { it('rechecks Node native modules in fresh child processes after rebuilding', () => { @@ -197,10 +200,12 @@ function mkTempProject() { // Walked, not listed: the script imports windows-process-tree-gyp-rebuild.mjs, and a fixture // missing it fails every case with a module-resolution error instead of the defect under test. copyScriptWithLocalModules(sourceScriptPath, join(projectDir, 'config', 'scripts')) - copyFileSync( - sourceNodePtyJobOwnershipPath, - join(projectDir, 'config', 'scripts', 'node-pty-job-ownership.cjs') - ) + for (const name of REQUIRED_CJS_SIBLINGS) { + copyFileSync( + fileURLToPath(new URL(`./${name}`, import.meta.url)), + join(projectDir, 'config', 'scripts', name) + ) + } return projectDir } diff --git a/config/scripts/focus-nested-wayland-terminal.sh b/config/scripts/focus-nested-wayland-terminal.sh new file mode 100755 index 00000000000..6d75699c54c --- /dev/null +++ b/config/scripts/focus-nested-wayland-terminal.sh @@ -0,0 +1,11 @@ +#!/usr/bin/env bash +set -euo pipefail +[[ "${GITHUB_ACTIONS:-}" == true ]] +# The isolated X server owns exactly one nested compositor window. +mapfile -t windows < <(xwininfo -root -tree | awk '$2 == "\"gnome-shell\":" {print $1}') +[[ ${#windows[@]} -eq 1 ]] +xdotool windowmap --sync "${windows[0]}" +xdotool windowfocus --sync "${windows[0]}" +read -r width height < <(xwininfo -id "${windows[0]}" | awk '$1 == "Width:" {w=$2} $1 == "Height:" {print w,$2}') +# The native spec opens a single terminal; a seat click activates its Wayland client. +xdotool mousemove --window "${windows[0]}" "$((width / 2))" "$((height / 2))" click 1 diff --git a/config/scripts/package-electron-runtime-contract.test.mjs b/config/scripts/package-electron-runtime-contract.test.mjs index 950d5ed258a..aa34e043268 100644 --- a/config/scripts/package-electron-runtime-contract.test.mjs +++ b/config/scripts/package-electron-runtime-contract.test.mjs @@ -579,6 +579,9 @@ describe('Electron runtime package contract', () => { expect(packageScripts['test:e2e:terminal-rendering-golden']).not.toContain( 'terminal-long-table-scroll-restore.spec.ts' ) + const goldenCommand = packageScripts['test:e2e:terminal-rendering-golden'] + expect(goldenCommand).toContain('--project electron-headless') + expect(goldenCommand).toContain('--project electron-headful') expect(packageScripts['test:e2e:windows-fresh-startup-golden']).toContain( 'golden-windows-fresh-startup.spec.ts' ) diff --git a/config/scripts/packaged-browser-lane-contract.test.mjs b/config/scripts/packaged-browser-lane-contract.test.mjs new file mode 100644 index 00000000000..bac077d57d4 --- /dev/null +++ b/config/scripts/packaged-browser-lane-contract.test.mjs @@ -0,0 +1,45 @@ +import { readFileSync } from 'node:fs' +import { describe, expect, it } from 'vitest' +import { parse } from 'yaml' + +const workflow = parse( + readFileSync(new URL('../../.github/workflows/packaged-browser-e2e.yml', import.meta.url), 'utf8') +) +const steps = workflow.jobs.compatibility.steps + +describe('packaged browser compatibility lane', () => { + it('runs weekly and supports immutable manual or reusable revisions', () => { + expect(workflow.on.schedule).toHaveLength(1) + for (const trigger of ['workflow_dispatch', 'workflow_call']) { + expect(workflow.on[trigger].inputs.ref).toMatchObject({ type: 'string', required: false }) + } + expect(steps[0].with.ref).toBe('${{ inputs.ref || github.sha }}') + expect(workflow.permissions).toEqual({ contents: 'read' }) + }) + + it('verifies the pinned package before selecting the desktop executable', () => { + const download = steps.find((step) => step.name === 'Download pinned old release').run + expect(download).toContain('gh release download v1.4.188') + expect(download).toContain('hashlib.sha512(package.read_bytes())') + expect(download).toContain("extracted/'opt'/'Orca'/'orca-ide'") + expect(download).toContain('assert base64.') + expect(download).toContain('decode()==expected') + expect(download).toContain("['dpkg-deb'") + expect(download.indexOf('assert base64.')).toBeLessThan(download.indexOf("['dpkg-deb'")) + }) + + it('requires both directions three times and rejects silent skips', () => { + const run = steps.find((step) => step.name === 'Run both mixed-version directions') + expect(run.run).toContain('tests/e2e/packaged-mixed-version-browser-placement.spec.ts') + expect(run.run).toContain('--repeat-each=3') + expect(run.run).toContain('--retries=0') + expect(run.run).toContain('--reporter=list,json') + const verify = steps.find((step) => step.name === 'Require all six compatibility executions') + expect(verify.if).toBe('always()') + expect(verify.run).toBe( + `node config/scripts/verify-packaged-browser-participation.mjs ${run.env.PLAYWRIGHT_JSON_OUTPUT_FILE}` + ) + expect(steps.at(-1).if).toBe('always()') + expect(steps.at(-1).with.path).toBe('test-results/') + }) +}) diff --git a/config/scripts/pr-code-change-scope.mjs b/config/scripts/pr-code-change-scope.mjs index 3089d376b2a..befcb06fe1f 100644 --- a/config/scripts/pr-code-change-scope.mjs +++ b/config/scripts/pr-code-change-scope.mjs @@ -140,6 +140,8 @@ const NATIVE_RUNTIME_PREFIXES = [ 'config/scripts/ensure-native-runtime', 'config/scripts/rebuild-native-deps', 'config/scripts/node-pty-job-ownership', + 'config/scripts/windows-process-tree-creation-time', + 'config/scripts/windows-process-tree-gyp-rebuild', 'config/scripts/electron-builder-native-rebuild', 'config/patches/node-pty@', 'config/patches/@vscode__windows-process-tree' @@ -224,6 +226,7 @@ const WINDOWS_PACKAGE_TESTS = [ 'src/main/windows/windows-pty-job.win32.test.ts', 'src/main/windows/windows-host-job.win32.test.ts', 'src/main/windows/windows-process-tree-command-line-patch.test.ts', + 'src/main/windows/windows-process-table-native-addon.win32.test.ts', 'src/main/windows-live-tree-kill.win32.test.ts', 'src/main/wsl/wsl-runner.test.ts', 'src/main/wsl/wsl-guest-environment.test.ts', diff --git a/config/scripts/pr-e2e-source-routing.mjs b/config/scripts/pr-e2e-source-routing.mjs index bda022159d4..3b8f2e90afb 100644 --- a/config/scripts/pr-e2e-source-routing.mjs +++ b/config/scripts/pr-e2e-source-routing.mjs @@ -10,7 +10,7 @@ const NATIVE_IME_PRODUCT_SOURCE = /** The harness itself: the session runner, the boundary probes, and the native specs. */ const NATIVE_IME_HARNESS = - /^(?:config\/scripts\/(?:run-terminal-ibus-hangul-e2e|terminal-ime-engagement-receipt)\.mjs$|tests\/e2e\/terminal-ime-(?:boundary-probe|byte-reader|engagement-receipt)\.ts$|tests\/e2e\/terminal-(?:ibus-hangul|hangul-terminating-digit|macos-2set-korean)-native\.spec\.ts$)/ + /^(?:config\/scripts\/focus-nested-wayland-terminal\.sh$|config\/scripts\/(?:run-terminal-ibus-hangul-e2e|terminal-ime-engagement-receipt)\.mjs$|tests\/e2e\/terminal-ime-(?:boundary-probe|byte-reader|engagement-receipt)\.ts$|tests\/e2e\/terminal-(?:ibus-hangul|hangul-terminating-digit|macos-2set-korean)-native\.spec\.ts$)/ export const PR_E2E_SOURCE_ROUTES = [ { @@ -41,7 +41,7 @@ export const PR_E2E_SOURCE_ROUTES = [ ], matches: (file) => isProductSource(file) && - /^(?:config\/scripts\/verify-wsl-e2e-participation\.mjs$|src\/main\/(?:wsl[/-]|pty\/.*wsl|providers\/wsl)|src\/shared\/(?:wsl-|windows-terminal-shell)|src\/renderer\/src\/.*(?:terminal-paste|pty-paste)|tests\/e2e\/(?:golden-tab-bar-agent-launch\.spec|terminal-windows-shell-paste-ownership\.spec|helpers\/(?:wsl-golden-stub-agent|golden-stub-agent))|\.github\/(?:actions\/setup-wsl-test-runtime\/|workflows\/windows-wsl-e2e\.yml))/.test( + /^(?:config\/scripts\/(?:verify-wsl-e2e-participation|verify-playwright-participation)\.mjs$|src\/main\/(?:wsl[/-]|pty\/.*wsl|providers\/wsl)|src\/shared\/(?:wsl-|windows-terminal-shell)|src\/renderer\/src\/.*(?:terminal-paste|pty-paste)|tests\/e2e\/(?:golden-tab-bar-agent-launch\.spec|terminal-windows-shell-paste-ownership\.spec|helpers\/(?:wsl-golden-stub-agent|golden-stub-agent))|\.github\/(?:actions\/setup-wsl-test-runtime\/|workflows\/windows-wsl-e2e\.yml))/.test( file ) }, diff --git a/config/scripts/rebuild-native-deps.mjs b/config/scripts/rebuild-native-deps.mjs index 863aac850a1..d7426d8cf1d 100644 --- a/config/scripts/rebuild-native-deps.mjs +++ b/config/scripts/rebuild-native-deps.mjs @@ -567,6 +567,15 @@ function loadNativeModule(moduleName) { } return } + if (moduleName === '@vscode/windows-process-tree') { + // The tarball prebuilt loads under Electron too -- the addon is N-API, so + // a bare require proves nothing about which source it was built from. + const { assertWindowsProcessTreeCreationTime } = projectRequire( + './config/scripts/windows-process-tree-creation-time.cjs' + ) + assertWindowsProcessTreeCreationTime({ module: projectRequire(moduleName) }) + return + } projectRequire(moduleName) } diff --git a/config/scripts/run-terminal-ibus-hangul-e2e.mjs b/config/scripts/run-terminal-ibus-hangul-e2e.mjs index 669f7744b39..572f1988629 100644 --- a/config/scripts/run-terminal-ibus-hangul-e2e.mjs +++ b/config/scripts/run-terminal-ibus-hangul-e2e.mjs @@ -9,6 +9,7 @@ import { readFileSync, writeFileSync } from 'node:fs' +import { verifyPlaywrightParticipation } from './verify-playwright-participation.mjs' import os from 'node:os' import path from 'node:path' import { @@ -20,6 +21,9 @@ import { const projectDir = path.resolve(import.meta.dirname, '../..') const scriptPath = import.meta.filename const insideSessionFlag = '--inside-session' +const nestedWaylandFlag = '--nested-wayland' +const nestedWayland = process.argv.includes(nestedWaylandFlag) +const waylandTitle = 'a digit typed right after a Hangul syllable reaches the pty' const processStopTimeoutMs = 5_000 const processKillTimeoutMs = 1_000 @@ -111,26 +115,38 @@ function configureHangulEngine() { } } -async function waitForHangulEngine(ibusProcess) { +async function waitForHangulEngine(sessionProcess) { + let lastError = '' const deadline = Date.now() + 15_000 while (Date.now() < deadline) { - if (ibusProcess.exitCode !== null) { - throw new Error(`ibus-daemon exited early with code ${ibusProcess.exitCode}`) + if (sessionProcess.exitCode !== null) { + throw new Error(`IME session process exited early with code ${sessionProcess.exitCode}`) } - const result = spawnSync('ibus', ['engine', 'hangul'], { stdio: 'pipe' }) + if ( + nestedWayland && + !existsSync(path.join(process.env.XDG_RUNTIME_DIR, process.env.WAYLAND_DISPLAY)) + ) { + await delay(100) + continue + } + const result = spawnSync('ibus', ['engine', 'hangul'], { encoding: 'utf8' }) + lastError = result.stderr?.trim() || String(result.error ?? result.status) if (result.status === 0) { return } await delay(100) } - throw new Error('Timed out while selecting the IBus Hangul engine') + throw new Error(`Timed out while selecting the IBus Hangul engine: ${lastError}`) } async function runInsideSession(evidenceDir) { const receiptPath = path.join(evidenceDir, 'ime-engagement-receipt.jsonl') const ibusLogPath = path.join(evidenceDir, 'ibus-daemon.log') const ibusLogFd = openSync(ibusLogPath, 'w') - const windowManagerLogPath = path.join(evidenceDir, 'xfwm4.log') + const windowManagerLogPath = path.join( + evidenceDir, + nestedWayland ? 'gnome-shell.log' : 'xfwm4.log' + ) const windowManagerLogFd = openSync(windowManagerLogPath, 'w') const evidence = { display: process.env.DISPLAY ?? null, @@ -147,32 +163,58 @@ async function runInsideSession(evidenceDir) { try { configureHangulEngine() - windowManagerProcess = spawn('xfwm4', ['--compositor=off'], { - detached: true, - env: process.env, - stdio: ['ignore', windowManagerLogFd, windowManagerLogFd] - }) - if (!windowManagerProcess.pid) { - throw new Error('xfwm4 did not return a PID') - } - evidence.windowManagerPid = windowManagerProcess.pid - console.error(`[terminal-ime] started xfwm4 PID ${windowManagerProcess.pid}`) - - ibusProcess = spawn( - 'ibus-daemon', - ['--xim', '--verbose', '--panel=disable', '--emoji-extension=disable'], - { + if (nestedWayland) { + for (const [schema, key, value] of [ + ['org.gnome.desktop.interface', 'enable-animations', 'false'], + ['org.gnome.desktop.input-sources', 'sources', "[('ibus', 'hangul')]"] + ]) { + const result = spawnSync('gsettings', ['set', schema, key, value], { encoding: 'utf8' }) + if (result.status !== 0) { + throw new Error(`Failed to configure GNOME: ${result.stderr}`) + } + } + windowManagerProcess = spawn( + 'gnome-shell', + ['--nested', '--wayland', `--wayland-display=${process.env.WAYLAND_DISPLAY}`], + { + detached: true, + env: process.env, + stdio: ['ignore', windowManagerLogFd, windowManagerLogFd] + } + ) + } else { + windowManagerProcess = spawn('xfwm4', ['--compositor=off'], { detached: true, env: process.env, - stdio: ['ignore', ibusLogFd, ibusLogFd] + stdio: ['ignore', windowManagerLogFd, windowManagerLogFd] + }) + } + if (!windowManagerProcess.pid) { + throw new Error('Window manager did not return a PID') + } + evidence.windowManagerPid = windowManagerProcess.pid + console.error(`[terminal-ime] started window manager PID ${windowManagerProcess.pid}`) + + if (nestedWayland) { + // GNOME starts IBus in the private session; a second daemon can compete for ownership. + await waitForHangulEngine(windowManagerProcess) + } else { + ibusProcess = spawn( + 'ibus-daemon', + ['--xim', '--verbose', '--panel=disable', '--emoji-extension=disable'], + { + detached: true, + env: process.env, + stdio: ['ignore', ibusLogFd, ibusLogFd] + } + ) + if (!ibusProcess.pid) { + throw new Error('ibus-daemon did not return a PID') } - ) - if (!ibusProcess.pid) { - throw new Error('ibus-daemon did not return a PID') + evidence.ibusDaemonPid = ibusProcess.pid + console.error(`[terminal-ime] started ibus-daemon PID ${ibusProcess.pid}`) + await waitForHangulEngine(ibusProcess) } - evidence.ibusDaemonPid = ibusProcess.pid - console.error(`[terminal-ime] started ibus-daemon PID ${ibusProcess.pid}`) - await waitForHangulEngine(ibusProcess) console.error(`[terminal-ime] IBus version: ${commandOutput('ibus', ['version'])}`) console.error(`[terminal-ime] IBus engine: ${commandOutput('ibus', ['engine'])}`) console.error( @@ -189,23 +231,49 @@ async function runInsideSession(evidenceDir) { 'hangul-keyboard' ])}` ) - evidence.ibusGroupBeforeCleanup = processGroupMembers(ibusProcess.pid) + evidence.ibusGroupBeforeCleanup = ibusProcess?.pid ? processGroupMembers(ibusProcess.pid) : [] console.error(`[terminal-ime] owned IBus group: ${evidence.ibusGroupBeforeCleanup.join('; ')}`) const testProcess = spawn( process.platform === 'win32' ? 'pnpm.cmd' : 'pnpm', - [ - 'run', - 'test:e2e:headful', - '--workers=1', - '--', - 'tests/e2e/terminal-ibus-hangul-native.spec.ts', - 'tests/e2e/terminal-hangul-terminating-digit-native.spec.ts' - ], + nestedWayland + ? [ + 'exec', + 'playwright', + 'test', + '--config', + 'tests/playwright.config.ts', + 'tests/e2e/terminal-hangul-terminating-digit-native.spec.ts', + '--project=electron-headful', + '--workers=1', + '--repeat-each=3', + '--retries=0', + '--reporter=list,json' + ] + : [ + 'run', + 'test:e2e:headful', + '--workers=1', + '--', + 'tests/e2e/terminal-ibus-hangul-native.spec.ts', + 'tests/e2e/terminal-hangul-terminating-digit-native.spec.ts' + ], { cwd: projectDir, env: { ...process.env, + ...(nestedWayland + ? { + ORCA_E2E_IME_INJECTOR: 'nested', + ORCA_E2E_NESTED_FOCUS_CMD: path.join( + projectDir, + 'config/scripts/focus-nested-wayland-terminal.sh' + ), + ORCA_E2E_EXTRA_APP_ARGS: + '--ozone-platform=wayland --enable-wayland-ime --wayland-text-input-version=3 --password-store=basic --use-mock-keychain --disable-gpu-sandbox', + PLAYWRIGHT_JSON_OUTPUT_FILE: path.join(evidenceDir, 'playwright.json') + } + : {}), ORCA_E2E_FORWARD_APP_LOGS: '1', ORCA_E2E_NATIVE_IBUS_HANGUL: '1', [IME_ENGAGEMENT_RECEIPT_ENV]: receiptPath, @@ -232,6 +300,13 @@ async function runInsideSession(evidenceDir) { windowManagerProcess.pid ) } + if (nestedWayland && existsSync(path.join(evidenceDir, 'playwright.json'))) { + mkdirSync(path.join(projectDir, 'test-results'), { recursive: true }) + copyFileSync( + path.join(evidenceDir, 'playwright.json'), + path.join(projectDir, 'test-results', 'terminal-wayland-playwright.json') + ) + } closeSync(ibusLogFd) closeSync(windowManagerLogFd) mkdirSync(path.join(projectDir, 'test-results'), { recursive: true }) @@ -241,7 +316,11 @@ async function runInsideSession(evidenceDir) { ) copyFileSync( windowManagerLogPath, - path.join(projectDir, 'test-results', 'terminal-ibus-hangul-native-xfwm4.log') + path.join( + projectDir, + 'test-results', + nestedWayland ? 'terminal-wayland-gnome-shell.log' : 'terminal-ibus-hangul-native-xfwm4.log' + ) ) writeFileSync( path.join(projectDir, 'test-results', 'terminal-ibus-hangul-native-processes.json'), @@ -269,6 +348,23 @@ async function runInsideSession(evidenceDir) { // Why unconditionally, and not only when Playwright failed: a skipped test reports as a pass, // so exit code 0 is exactly the state this check exists to distrust. const receiptText = existsSync(receiptPath) ? readFileSync(receiptPath, 'utf8') : '' + if (nestedWayland) { + verifyPlaywrightParticipation( + JSON.parse(readFileSync(path.join(evidenceDir, 'playwright.json'), 'utf8')), + { titles: [waylandTitle], label: 'Native Wayland Hangul', repetitions: 3 } + ) + const receipts = receiptText.trim().split('\n') + if (receipts.length !== 3) { + throw new Error('Expected three native Wayland engagement receipts') + } + for (const receipt of receipts) { + const problems = verifyImeEngagementReceipts(receipt, [waylandTitle]) + if (problems.length) { + throw new Error(problems.join('\n')) + } + } + return testExitCode + } const engagementProblems = verifyImeEngagementReceipts(receiptText, EXPECTED_NATIVE_IME_TESTS) if (engagementProblems.length > 0) { for (const problem of engagementProblems) { @@ -288,7 +384,7 @@ async function runInsideSession(evidenceDir) { async function runOuter() { if (process.platform !== 'linux') { - throw new Error('The native IBus Hangul E2E runner requires Linux/X11') + throw new Error('The native IBus Hangul E2E runner requires Linux') } const evidenceDir = mkdtempSync(path.join(os.tmpdir(), 'orca-terminal-ime-e2e-')) @@ -302,24 +398,36 @@ async function runOuter() { 'xvfb-run', [ '--auto-servernum', + ...(nestedWayland ? ['--server-args=-screen 0 1280x800x24'] : []), 'dbus-run-session', '--', process.execPath, scriptPath, insideSessionFlag, - evidenceDir + evidenceDir, + ...(nestedWayland ? [nestedWaylandFlag] : []) ], { cwd: projectDir, detached: true, env: { ...process.env, + ...(nestedWayland + ? { + WAYLAND_DISPLAY: 'wayland-orca-ime', + XDG_SESSION_TYPE: 'wayland', + XDG_CURRENT_DESKTOP: 'GNOME', + LIBGL_ALWAYS_SOFTWARE: '1', + NO_AT_BRIDGE: '1' + } + : {}), GTK_IM_MODULE: 'ibus', IBUS_ENABLE_SYNC_MODE: '1', LANG: process.env.LANG || 'C.UTF-8', QT_IM_MODULE: 'ibus', - XDG_CACHE_HOME: path.join(evidenceDir, 'cache'), - XDG_CONFIG_HOME: path.join(evidenceDir, 'config'), + // GNOME 42 drops XDG_CONFIG_HOME when spawning IBus; both must use its default path. + XDG_CACHE_HOME: nestedWayland ? undefined : path.join(evidenceDir, 'cache'), + XDG_CONFIG_HOME: nestedWayland ? undefined : path.join(evidenceDir, 'config'), XDG_RUNTIME_DIR: runtimeDir, XMODIFIERS: '@im=ibus' }, @@ -329,17 +437,20 @@ async function runOuter() { if (!sessionProcess.pid) { throw new Error('xvfb-run did not return a PID') } - console.error(`[terminal-ime] started isolated X11 session PID ${sessionProcess.pid}`) + console.error(`[terminal-ime] started isolated display session PID ${sessionProcess.pid}`) const exitCode = await waitForExit(sessionProcess) const remaining = await stopOwnedProcessGroup(sessionProcess.pid) if (remaining.length > 0) { - throw new Error(`Owned X11 session processes survived cleanup: ${remaining.join('; ')}`) + throw new Error(`Owned display session processes survived cleanup: ${remaining.join('; ')}`) } return exitCode } const insideSession = process.argv[2] === insideSessionFlag try { + if (nestedWayland && process.env.GITHUB_ACTIONS !== 'true') { + throw new Error('Nested Wayland native input validation runs only in GitHub Actions') + } if (insideSession && !process.argv[3]) { throw new Error(`${insideSessionFlag} requires an evidence directory argument`) } diff --git a/config/scripts/terminal-ime-e2e-workflow.test.mjs b/config/scripts/terminal-ime-e2e-workflow.test.mjs index 96062ebe706..65277b5e898 100644 --- a/config/scripts/terminal-ime-e2e-workflow.test.mjs +++ b/config/scripts/terminal-ime-e2e-workflow.test.mjs @@ -68,6 +68,21 @@ describe('terminal IME e2e workflow', () => { expect(runner).not.toContain('pkill') }) + it('runs native Wayland independently with CJK fonts and retained evidence', () => { + const job = workflow.jobs['linux-wayland'] + expect(job.needs).toBeUndefined() + const install = job.steps.find((step) => step.run?.includes('apt-get install')).run + for (const tool of ['gnome-shell', 'ibus-hangul', 'fonts-noto-cjk', 'xwininfo']) { + expect(install).toContain(tool === 'xwininfo' ? 'x11-utils' : tool) + } + expect(job.steps.find((step) => step.run?.includes('--nested-wayland')).run).toBe( + 'node config/scripts/run-terminal-ibus-hangul-e2e.mjs --nested-wayland' + ) + const upload = job.steps.find((step) => step.uses?.startsWith('actions/upload-artifact')) + expect(upload.if).toBe('always()') + expect(upload.with.name).toBe('terminal-wayland-ime-evidence') + }) + it('bounds blocking native input commands', () => { const nativeSpec = readFileSync( join(projectDir, 'tests/e2e/terminal-ibus-hangul-native.spec.ts'), diff --git a/config/scripts/verify-packaged-browser-participation.mjs b/config/scripts/verify-packaged-browser-participation.mjs new file mode 100644 index 00000000000..c165ab46f19 --- /dev/null +++ b/config/scripts/verify-packaged-browser-participation.mjs @@ -0,0 +1,20 @@ +import { readFileSync } from 'node:fs' +import { pathToFileURL } from 'node:url' +import { verifyPlaywrightParticipation } from './verify-playwright-participation.mjs' + +export const PACKAGED_BROWSER_TEST_TITLES = [ + 'keeps an old packaged client on the current server-hosted path', + 'keeps a current client on an old packaged server-hosted path' +] + +export function verifyPackagedBrowserParticipation(report) { + verifyPlaywrightParticipation(report, { + titles: PACKAGED_BROWSER_TEST_TITLES, + label: 'Packaged browser' + }) +} + +if (process.argv[1] && import.meta.url === pathToFileURL(process.argv[1]).href) { + verifyPackagedBrowserParticipation(JSON.parse(readFileSync(process.argv[2], 'utf8'))) + console.log('Both packaged browser directions passed three times without skips or retries.') +} diff --git a/config/scripts/verify-packaged-browser-participation.test.mjs b/config/scripts/verify-packaged-browser-participation.test.mjs new file mode 100644 index 00000000000..6508777e19a --- /dev/null +++ b/config/scripts/verify-packaged-browser-participation.test.mjs @@ -0,0 +1,57 @@ +import { describe, expect, it } from 'vitest' +import { + verifyPackagedBrowserParticipation, + PACKAGED_BROWSER_TEST_TITLES +} from './verify-packaged-browser-participation.mjs' + +function report() { + return { + stats: { expected: 6, skipped: 0, unexpected: 0, flaky: 0 }, + suites: [ + { + suites: [ + { + specs: PACKAGED_BROWSER_TEST_TITLES.map((title) => ({ + title, + tests: Array.from({ length: 3 }, () => ({ + expectedStatus: 'passed', + results: [{ status: 'passed' }] + })) + })) + } + ] + } + ] + } +} + +describe('Packaged browser participation', () => { + it('accepts both named scenarios executed three times', () => { + expect(() => verifyPackagedBrowserParticipation(report())).not.toThrow() + }) + it.each(['skipped', 'unexpected', 'flaky'])('rejects a nonzero %s result', (key) => { + const value = report() + value.stats[key] = 1 + expect(() => verifyPackagedBrowserParticipation(value)).toThrow('participation failed') + }) + it('rejects missing scenarios even when aggregate counts claim six passes', () => { + const value = report() + value.suites[0].suites[0].specs.pop() + expect(() => verifyPackagedBrowserParticipation(value)).toThrow('requires three executions') + }) + it('rejects an unrelated scenario substituted for an expected scenario', () => { + const value = report() + value.suites[0].suites[0].specs[0].title = 'native shell passes' + expect(() => verifyPackagedBrowserParticipation(value)).toThrow( + 'Unexpected Packaged browser scenario' + ) + }) + it('rejects a pass obtained after a failed attempt', () => { + const value = report() + value.suites[0].suites[0].specs[0].tests[0].results.unshift({ status: 'failed' }) + expect(() => verifyPackagedBrowserParticipation(value)).toThrow('without retries') + }) + it('rejects missing report content', () => { + expect(() => verifyPackagedBrowserParticipation({})).toThrow('participation failed') + }) +}) diff --git a/config/scripts/verify-playwright-participation.mjs b/config/scripts/verify-playwright-participation.mjs new file mode 100644 index 00000000000..d78f2757f1e --- /dev/null +++ b/config/scripts/verify-playwright-participation.mjs @@ -0,0 +1,42 @@ +export function verifyPlaywrightParticipation(report, { titles, label, repetitions = 3 }) { + const stats = report?.stats + if ( + !stats || + stats.expected !== titles.length * repetitions || + stats.skipped !== 0 || + stats.unexpected !== 0 || + stats.flaky !== 0 || + report.errors?.length + ) { + throw new Error(`${label} participation failed: ${JSON.stringify(stats)}`) + } + const counts = new Map(titles.map((title) => [title, 0])) + const visit = (suites) => { + for (const suite of suites ?? []) { + for (const spec of suite.specs ?? []) { + if (!counts.has(spec.title)) { + throw new Error(`Unexpected ${label} scenario: ${spec.title}`) + } + for (const test of spec.tests ?? []) { + if ( + test.expectedStatus !== 'passed' || + test.results?.length !== 1 || + test.results[0].status !== 'passed' + ) { + throw new Error(`${label} scenario did not pass without retries: ${spec.title}`) + } + counts.set(spec.title, counts.get(spec.title) + 1) + } + } + visit(suite.suites) + } + } + visit(report.suites) + for (const [title, count] of counts) { + if (count !== repetitions) { + throw new Error( + `${label} scenario requires ${repetitions === 3 ? 'three' : repetitions} executions: ${title} (${count})` + ) + } + } +} diff --git a/config/scripts/verify-wsl-e2e-participation.mjs b/config/scripts/verify-wsl-e2e-participation.mjs index 21570ef7689..9e8c9252b7f 100644 --- a/config/scripts/verify-wsl-e2e-participation.mjs +++ b/config/scripts/verify-wsl-e2e-participation.mjs @@ -1,3 +1,4 @@ +import { verifyPlaywrightParticipation } from './verify-playwright-participation.mjs' import { readFileSync } from 'node:fs' import { pathToFileURL } from 'node:url' @@ -8,44 +9,7 @@ export const WSL_TEST_TITLES = [ ] export function verifyWslParticipation(report) { - const stats = report?.stats - if ( - !stats || - stats.expected !== 9 || - stats.skipped !== 0 || - stats.unexpected !== 0 || - stats.flaky !== 0 || - report.errors?.length - ) { - throw new Error(`WSL participation failed: ${JSON.stringify(stats)}`) - } - const counts = new Map(WSL_TEST_TITLES.map((title) => [title, 0])) - const visit = (suites) => { - for (const suite of suites ?? []) { - for (const spec of suite.specs ?? []) { - if (!counts.has(spec.title)) { - throw new Error(`Unexpected WSL scenario: ${spec.title}`) - } - for (const test of spec.tests ?? []) { - if ( - test.expectedStatus !== 'passed' || - test.results?.length !== 1 || - test.results[0].status !== 'passed' - ) { - throw new Error(`WSL scenario did not pass without retries: ${spec.title}`) - } - counts.set(spec.title, counts.get(spec.title) + 1) - } - } - visit(suite.suites) - } - } - visit(report.suites) - for (const [title, count] of counts) { - if (count !== 3) { - throw new Error(`WSL scenario requires three executions: ${title} (${count})`) - } - } + verifyPlaywrightParticipation(report, { titles: WSL_TEST_TITLES, label: 'WSL' }) } if (process.argv[1] && import.meta.url === pathToFileURL(process.argv[1]).href) { diff --git a/config/scripts/windows-process-tree-creation-time.cjs b/config/scripts/windows-process-tree-creation-time.cjs new file mode 100644 index 00000000000..88f231f14d3 --- /dev/null +++ b/config/scripts/windows-process-tree-creation-time.cjs @@ -0,0 +1,42 @@ +'use strict' + +/** + * Prove the COMPILED addon understands `CREATIONTIME`, not just the patched JS. + * + * Unlike node-pty, this package ships a prebuilt `.node` at the same + * `build/Release/` path node-gyp writes to, so neither a load nor a path check + * can tell a stale prebuilt from a source build. pnpm patches the source tree + * and leaves that prebuilt in place, which is how `ProcessDataFlag.CreationTime` + * came to exist in `lib/index.js` on a binary that ignores flag 4 -- the gate + * read true and every row came back without `creationTimeMs`. + * + * `supportedProcessDataFlags` is exported by the patched `addon.cc`, so its + * presence is the binary's own answer. Shared by the Node and Electron probes + * the way `node-pty-job-ownership.cjs` is. + */ + +/** `ProcessDataFlags::CREATIONTIME` in src/process.h. */ +const CREATION_TIME_FLAG = 4 + +function assertWindowsProcessTreeCreationTime({ module, platform = process.platform }) { + if (platform !== 'win32') { + return + } + const supported = module?.supportedProcessDataFlags + if (typeof supported === 'number' && (supported & CREATION_TIME_FLAG) !== 0) { + return + } + throw new Error( + [ + '@vscode/windows-process-tree does not report CreationTime support', + `(supportedProcessDataFlags=${String(supported)}).`, + 'That is the tarball prebuilt, not a build of the patched source, so every', + 'process row comes back without creationTimeMs: Windows descendant exit', + 'verification cannot identify a PID and structured Claude/Codex chat runs', + 'with an unprovable child-tree reaper.', + 'Rebuild it from source so config/patches/@vscode__windows-process-tree@0.8.0.patch applies.' + ].join(' ') + ) +} + +module.exports = { assertWindowsProcessTreeCreationTime, CREATION_TIME_FLAG } diff --git a/config/scripts/wsl-e2e-lane-contract.test.mjs b/config/scripts/wsl-e2e-lane-contract.test.mjs index 0369eb7c0c4..6790e19e5fe 100644 --- a/config/scripts/wsl-e2e-lane-contract.test.mjs +++ b/config/scripts/wsl-e2e-lane-contract.test.mjs @@ -8,6 +8,7 @@ const read = (path) => readFileSync(new URL(`../../${path}`, import.meta.url), ' describe('real WSL terminal lane', () => { it.each([ 'config/scripts/verify-wsl-e2e-participation.mjs', + 'config/scripts/verify-playwright-participation.mjs', 'src/main/wsl-availability.ts', 'src/main/wsl/wsl-runner.ts', 'src/main/pty/wsl-orca-env.ts', diff --git a/config/vitest.performance.config.ts b/config/vitest.performance.config.ts index 7682b7b9698..9d739cbd52b 100644 --- a/config/vitest.performance.config.ts +++ b/config/vitest.performance.config.ts @@ -11,6 +11,7 @@ const contracts = [ 'src/renderer/src/components/editor/rich-markdown-lowlight-cache.test.ts', 'src/renderer/src/components/terminal-pane/agent-completion-coordinator-queued-inspection-disposal.test.ts', 'src/renderer/src/lib/pane-manager/pane-terminal-output-scheduler-queue-retention.test.ts', + 'src/renderer/src/store/store-identity-churn-probe.test.ts', 'config/scripts/app-store-performance-plugin.test.mjs', 'config/scripts/quadratic-buffer-concat-plugin.test.mjs', 'config/scripts/sort-comparator-performance-plugin.test.mjs' diff --git a/docs/assets/readme-downloads.svg b/docs/assets/readme-downloads.svg index 33ad276aa2d..ebee3673b77 100644 --- a/docs/assets/readme-downloads.svg +++ b/docs/assets/readme-downloads.svg @@ -1,5 +1,5 @@ - - downloads: 41m + + downloads: 42m @@ -15,7 +15,7 @@ downloads downloads - 41m - 41m + 42m + 42m diff --git a/docs/readme/README.es.md b/docs/readme/README.es.md index f2247e0900d..85e48c6d765 100644 --- a/docs/readme/README.es.md +++ b/docs/readme/README.es.md @@ -36,7 +36,7 @@ Supervisa y dirige a tus agentes desde el teléfono — recibe una notificación cuando un agente termine y envía instrucciones de seguimiento desde cualquier lugar. -[App Store de iOS](https://apps.apple.com/us/app/orca-ide/id6766130217) · [APK para Android](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) · [Docs →](https://www.onorca.dev/docs/mobile) +[App Store de iOS](https://apps.apple.com/us/app/orca-ide/id6766130217) · [APK para Android](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) · [Docs →](https://www.onorca.dev/docs/mobile) @@ -227,7 +227,7 @@ yay -S stably-orca-bin Vincúlala con tu app de escritorio para supervisar y dirigir a tus agentes desde el teléfono. - **iOS:** [Descargar desde App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) -- **Android:** [Descargar el APK](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) +- **Android:** [Descargar el APK](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) --- diff --git a/docs/readme/README.fr.md b/docs/readme/README.fr.md index e601abc2344..adf966b5053 100644 --- a/docs/readme/README.fr.md +++ b/docs/readme/README.fr.md @@ -40,7 +40,7 @@ Surveillez et pilotez vos agents depuis votre téléphone — soyez notifié quand un agent termine, et envoyez des instructions de suivi où que vous soyez. -[App Store iOS](https://apps.apple.com/us/app/orca-ide/id6766130217) · [TestFlight](https://testflight.apple.com/join/YjeGMQBA) · [APK Android 0.0.47](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) · [Docs →](https://www.onorca.dev/docs/mobile) +[App Store iOS](https://apps.apple.com/us/app/orca-ide/id6766130217) · [TestFlight](https://testflight.apple.com/join/YjeGMQBA) · [APK Android 0.0.48](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) · [Docs →](https://www.onorca.dev/docs/mobile) @@ -235,7 +235,7 @@ yay -S stably-orca-bin Associez-la à l'app de bureau pour surveiller et piloter vos agents depuis votre téléphone. - **iOS :** [Télécharger sur l'App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) ou [rejoindre TestFlight](https://testflight.apple.com/join/YjeGMQBA) -- **Android :** [Télécharger l'APK 0.0.47](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) +- **Android :** [Télécharger l'APK 0.0.48](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) --- diff --git a/docs/readme/README.ja.md b/docs/readme/README.ja.md index cce2032a67c..ce5a7ddf07f 100644 --- a/docs/readme/README.ja.md +++ b/docs/readme/README.ja.md @@ -36,7 +36,7 @@ スマートフォンからエージェントを監視・操作 — エージェントの完了を通知で受け取り、どこからでもフォローアップを送信できます。 -[iOS App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) · [Android APK](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) · [ドキュメント →](https://www.onorca.dev/docs/mobile) +[iOS App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) · [Android APK](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) · [ドキュメント →](https://www.onorca.dev/docs/mobile) @@ -227,7 +227,7 @@ yay -S stably-orca-bin デスクトップアプリとペアリングして、スマートフォンからエージェントを監視・操作できます。 - **iOS:** [App Store からダウンロード](https://apps.apple.com/us/app/orca-ide/id6766130217) -- **Android:** [APK をダウンロード](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) +- **Android:** [APK をダウンロード](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) --- diff --git a/docs/readme/README.ko.md b/docs/readme/README.ko.md index 837ecf2133f..81226572e9f 100644 --- a/docs/readme/README.ko.md +++ b/docs/readme/README.ko.md @@ -36,7 +36,7 @@ 휴대폰에서 에이전트를 모니터링하고 조종하세요 — 에이전트가 완료되면 알림을 받고 어디서든 후속 지시를 보낼 수 있습니다. -[iOS App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) · [TestFlight](https://testflight.apple.com/join/YjeGMQBA) · [Android APK 0.0.47](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) · [문서 →](https://www.onorca.dev/docs/mobile) +[iOS App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) · [TestFlight](https://testflight.apple.com/join/YjeGMQBA) · [Android APK 0.0.48](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) · [문서 →](https://www.onorca.dev/docs/mobile) @@ -230,7 +230,7 @@ yay -S stably-orca-bin 데스크톱 앱과 페어링해 휴대폰에서 에이전트를 모니터링하고 조종하세요. - **iOS:** [App Store에서 다운로드](https://apps.apple.com/us/app/orca-ide/id6766130217) 또는 [TestFlight 참여](https://testflight.apple.com/join/YjeGMQBA) -- **Android:** [APK 0.0.47 다운로드](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) · [설치 가이드](https://www.onorca.dev/docs/android-apk) +- **Android:** [APK 0.0.48 다운로드](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) · [설치 가이드](https://www.onorca.dev/docs/android-apk) --- diff --git a/docs/readme/README.pt.md b/docs/readme/README.pt.md index 86d998a4e5f..4f4461607d3 100644 --- a/docs/readme/README.pt.md +++ b/docs/readme/README.pt.md @@ -36,7 +36,7 @@ Monitore e conduza seus agentes pelo celular — receba uma notificação quando um agente terminar e envie instruções de acompanhamento de qualquer lugar. -[App Store para iOS](https://apps.apple.com/us/app/orca-ide/id6766130217) · [TestFlight](https://testflight.apple.com/join/YjeGMQBA) · [APK Android 0.0.47](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) · [Docs →](https://www.onorca.dev/docs/mobile) +[App Store para iOS](https://apps.apple.com/us/app/orca-ide/id6766130217) · [TestFlight](https://testflight.apple.com/join/YjeGMQBA) · [APK Android 0.0.48](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) · [Docs →](https://www.onorca.dev/docs/mobile) @@ -230,7 +230,7 @@ yay -S stably-orca-bin Conecte ao app desktop para monitorar e conduzir seus agentes pelo celular. - **iOS:** [Baixar na App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) ou [entrar no TestFlight](https://testflight.apple.com/join/YjeGMQBA) -- **Android:** [Baixar APK 0.0.47](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) +- **Android:** [Baixar APK 0.0.48](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) --- diff --git a/docs/readme/README.zh-CN.md b/docs/readme/README.zh-CN.md index 10f47e20fe6..970628edd32 100644 --- a/docs/readme/README.zh-CN.md +++ b/docs/readme/README.zh-CN.md @@ -36,7 +36,7 @@ 用手机监控并指挥你的智能体 — 智能体完成时收到通知,随时随地发送后续指令。 -[iOS App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) · [Android APK](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) · [文档 →](https://www.onorca.dev/docs/mobile) +[iOS App Store](https://apps.apple.com/us/app/orca-ide/id6766130217) · [Android APK](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) · [文档 →](https://www.onorca.dev/docs/mobile) @@ -227,7 +227,7 @@ yay -S stably-orca-bin 与桌面应用配对,用手机监控并指挥你的智能体。 - **iOS:** [从 App Store 下载](https://apps.apple.com/us/app/orca-ide/id6766130217) -- **Android:** [下载 APK](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk) +- **Android:** [下载 APK](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk) --- diff --git a/docs/reference/headless-linux-server.md b/docs/reference/headless-linux-server.md index 50a38cf446e..2b452f05fc5 100644 --- a/docs/reference/headless-linux-server.md +++ b/docs/reference/headless-linux-server.md @@ -390,6 +390,10 @@ its own `orca`. `ws://` through an HTTPS-only endpoint. - Hostnames, IPv4, bracketed IPv6, and raw IPv6 literals are supported. IPv6 still requires an IPv6-reachable listener/network path. +- Background push notifications to a paired phone do not fire from a headless + server: agent-completion detection runs in the desktop renderer, which serve + mode never starts, so nothing reaches the push gateway even though the phone + registers successfully. - `xvfb-run` and `dbus-run-session -- xvfb-run` remain valid diagnostic launch shapes, but neither should be needed when `Xvfb` is installed and no display is configured. Repeated D-Bus messages without a ready block indicate startup diff --git a/docs/reference/mobile-push-contract.md b/docs/reference/mobile-push-contract.md new file mode 100644 index 00000000000..4f6f4d5d30c --- /dev/null +++ b/docs/reference/mobile-push-contract.md @@ -0,0 +1,352 @@ +# Mobile push: contract and build spec + +Tracking issue: stablyai/orca#8129. Design page: `/tmp/orca-mobile-push/orca-mobile-push.html`. +This document is the single contract every lane builds against. Do not deviate without updating it. + +## Summary + +A small Orca-hosted push gateway (`cloud/apps/push`) holds the APNs key and FCM credentials and sends +to phones. The desktop host registers each paired phone's native push token with the gateway and asks +the gateway to push on every mobile notification it already fans out over the socket. The phone dedupes +by `notificationId#notificationSeq`. No ack gate, no generic mode, no staging gateway, one auth path for +signed-in and accountless hosts. + +## Identities + +- **Host public key**: the desktop's existing X25519 E2EE public key (`src/main/runtime/e2ee-keypair.ts`), + 32 bytes, base64. The phone already stores it per host as `publicKeyB64`. +- **hostFingerprint**: `sha256(hostPublicKey)` base64url, first 16 chars. Identical derivation to + `deriveRelayHostId` in `src/main/runtime/relay/relay-http-client.ts`. Both desktop and phone can compute it. +- **deviceId**: the desktop's `DeviceEntry.deviceId` for the paired phone. Opaque UUID. +- **registrationId**: gateway-assigned opaque id for one (hostFingerprint, deviceId) pair. + +## Gateway HTTP API + +Base URL: `https://push.onorca.dev` (dev override via env). JSON bodies, `Content-Type: application/json`. +All schemas are zod, `.strict()`, exported from `cloud/packages/push-contract`. + +### Host authentication: challenge, proof, session + +The host keypair is X25519 (box), so it cannot sign. Reuse the relay's challenge shape. + +`POST /v1/host/challenge` +```json +{ "v": 1, "hostPublicKeyB64": "<32 bytes b64>" } +``` +→ 200 +```json +{ "challengeId": "", "gatewayEphemeralPublicKeyB64": "<32 b64>", "nonceB64": "<24 b64>", + "ciphertextB64": "", "expiresAt": } +``` +- Gateway generates an ephemeral box keypair per challenge, a 24-byte nonce, and a 32-byte secret. +- `plaintext = "orca-push-host-challenge/v1\0" || u32be(len(transcript)) || transcript || secret(32)` +- `ciphertext = nacl.box(plaintext, nonce, hostPublicKey, gatewayEphemeralSecretKey)` +- Transcript is the relay's length-prefixed field encoding (`field(name, value)` = + u32be(len(name)) || name || u32be(len(value)) || value), fields in this exact order: + `protocol="orca-push-host-proof/v1"`, `version=0x01`, `gatewayOrigin`, `gatewayEphemeralPublicKey`, + `challengeNonce`, `challengeId`, `issuedAt` (u64be ms), `expiresAt` (u64be ms), `hostFingerprint`, + `hostPublicKey`. +- Challenge TTL 10 s, and 10 s is the whole window the gateway honours. The 30 s clock skew tolerance + is the host's alone: it validates a timestamp the gateway chose, so it needs the allowance and the + gateway does not. A gateway that subtracted the tolerance from its own check would run a 40 s TTL. + Store challenge (id, secret hash, host fingerprint, host public key, expiry) in DB so any Cloud Run + instance can verify. Expired rows are pruned 30 s late so a slow proof reads as expired rather than + as an unknown challenge. +- Issuing a challenge writes no `push_hosts` row. It is unauthenticated, so a `push_hosts` row would be + a free permanent write for any caller. The row is upserted in `POST /v1/host/session` once the proof + verifies, from the public key the challenge row carries. + +`POST /v1/host/session` +```json +{ "v": 1, "challengeId": "", "proofB64": "<32 b64>" } +``` +- Host opens the box with its secret key, validates every transcript field (same checks as + `validateTranscript` in `src/main/runtime/relay/relay-host-proof.ts`, adapted to the push fields), + and returns `proof = HMAC-SHA256(secret, "orca-push-host-proof/v1\0ack\0" || transcript)`. +- Gateway verifies with `timingSafeEqual`, consumes the challenge (single use), and returns +```json +{ "sessionToken": "", "expiresAt": , "hostFingerprint": "<16 chars>" } +``` +- Session TTL 24 h. Stored hashed (sha256) in DB. Bearer on every other call: + `Authorization: Bearer `. 401 with `{ "error": "session_expired" }` on expiry; host + re-runs the challenge. + +### Device registration + +`POST /v1/devices` (Bearer) +```json +{ "v": 1, "deviceId": "", "platform": "ios" | "android", "token": "", + "apnsEnvironment": "sandbox" | "production", // ios only, required for ios + "filter": { "sources": ["agent-task-complete", "terminal-bell", "plugin"], + "agentStates": ["needs-input", "finished"] } } +``` +→ 200 `{ "registrationId": "" }`. Upsert keyed by (hostFingerprint, deviceId); a new token +replaces the old. `deviceId` is caller-chosen, so a host is capped at 64 registrations: the 65th +distinct `deviceId` → 409 `{ "error": "too_many_devices" }`. Re-registering a `deviceId` the host +already owns is always accepted, and deleting a registration frees its slot. `GET /v1/devices` is +bounded at 1024 rows to match its response schema, which the per-host cap keeps well out of reach. +`filter` is stored but enforced by the host (see desktop); gateway stores it only so a +host restart can re-read it. iOS tokens are variable-length, hex-encoded byte strings; Android +tokens are FCM registration strings. + +`DELETE /v1/devices/:registrationId` (Bearer) → 204. Only the owning host may delete. + +`GET /v1/devices` (Bearer) → `{ "devices": [{ registrationId, deviceId, platform, dead: boolean }] }`. + +### Send + +`POST /v1/send` (Bearer) +```json +{ "v": 1, + "registrationIds": ["", "..."], + "notification": { + "notificationId": "", + "notificationSeq": , "notificationEpoch": "", + "source": "agent-task-complete" | "terminal-bell" | "plugin", + "agentState": "needs-input" | "finished" | null, + "title": "", "body": "", + "worktreeId": "" } } +``` +→ 200 +```json +{ "results": [{ "registrationId": "", "status": "queued" | "dead" | "rate_limited" | "error" }] } +``` +- `queued` means accepted into the coalescing window. `dead` means the provider reported the token + unregistered; the host must drop the registration. Never block the socket fan-out on this call. +- Quota: 60 sends per hostFingerprint per rolling hour, 200 per registration per rolling day. Over quota + → `rate_limited` per result, HTTP 200. Whole request over a hard cap of 20 registrationIds → 400. + The cap counts the ids as sent; the gateway then dedupes them, so a repeated id spends quota once, + yields one result, and counts once toward `coalescedCount`. `results` may therefore be shorter than + `registrationIds`, and callers must match a result by its `registrationId`, never by position. +- Notification JSON is limited to 3000 UTF-8 bytes to leave provider envelope space; identities + are preserved exactly, including long filesystem paths. Oversized payloads fail validation. +- Gateway retries are deduplicated by host, registration, notification epoch, and sequence in the + quota ledger for its 25-hour retention window. Duplicates return `queued` without reserving + quota or enqueueing another delivery. +- Both quota counters are reserved under a per-host lock held for the whole transaction. PostgreSQL + reads at READ COMMITTED, so a concurrent count-then-insert would otherwise admit a whole burst. + +### Request limits and unauthenticated abuse + +- Every POST is capped at 16 KiB by a streaming body limit, not by `Content-Length` alone: a chunked + body declares no length. Over the cap → 413 `{ "error": "request_too_large" }`. +- `POST /v1/host/challenge` and `POST /v1/host/session` are the only unauthenticated routes. They share + one token bucket per client IP, 30 requests per minute, refilling continuously. Over the bucket → 429 + `{ "error": "rate_limited" }`. The client IP is the **last** `x-forwarded-for` hop, not the first: + Cloud Run appends the connecting peer, so everything left of that value is caller-supplied and can be + a fresh forgery on every request, which would hand a flood a new bucket each time. + `ORCA_PUSH_TRUSTED_PROXY_HOPS` (default 0) says how many appenders sit between the platform and the + client, so a future load balancer sets it to 1. A header with fewer hops than that depth is not + trusted at all. Falls back to `x-real-ip` and then to a single shared bucket. The bucket is per + instance and in memory, so the effective cap scales with the instance count; it exists to blunt a + flood, not to meter. +- Every other `/v1` route is capped by a second, wider bucket per client IP, 240 requests per minute, + applied **before** the bearer is looked up. A bearer has to be read from the database before it can + be refused, and that read takes one of only two pool connections per instance, so without this cap + a flood of forged bearers would starve real hosts of the pool while every one of them got a 401. +- The gateway cannot prove that a host owns the token it registers: any host with a session may + register any well-formed token and send text to it, within its own quota. The phone drops such a push + in the foreground because the fingerprint resolves to no paired host, and never routes a tap on it, + but the OS banner shows while the app is backgrounded. Reaching it needs the victim's native token, + which the gateway never returns and which only the phone and its host ever see. + +### Coalescing (gateway) + +Per registrationId, hold sends for 3 s. If one event arrives, send it as-is. If N>1 arrive, send one +summary: title `Orca`, body ` agents need attention` (or ` updates` when no needs-input), data +carries the latest event's fields plus `coalescedCount`. Collapse id for a summary is +`host:` so a later summary replaces it. The window is held in memory per gateway +instance, so with more than one instance a burst can produce up to one summary per instance; accepted +for this release, and the collapse id keeps the phone showing one banner. Transient provider errors +retry at most three attempts within two minutes, honoring Retry-After and FCM minimum delays. Permanent failures +are not retried. Unregister/dead-token state is re-read before every attempt. Shutdown stops admission +and drains admitted requests, pending windows, and active deliveries before closing resources; +a nine-second hard deadline remains below Cloud Run's termination grace. Delivery remains in memory. + +### Provider payloads + +APNs (HTTP/2, `api.push.apple.com` or `api.sandbox.push.apple.com` by `apnsEnvironment`; JWT auth +from key id + team id + `.p8`, token cached and refreshed every 50 min): +- headers: `apns-topic: com.stably.orca.mobile`, `apns-push-type: alert`, `apns-priority: 10`, + `apns-expiration: now+4h`, `apns-collapse-id: >` +- body: `{"aps":{"alert":{"title","body"},"sound":"default","thread-id":""}, + "orca":{ hostFingerprint, worktreeId, notificationId, notificationSeq, notificationEpoch, source, + agentState, coalescedCount }}` +- Dead token: 410, or 400 with `BadDeviceToken`/`Unregistered`/`DeviceTokenNotForTopic`. + +FCM (V1 `projects/onorca-cloud/messages:send`, bearer from the runtime service account via the GCE +metadata server or `GOOGLE_APPLICATION_CREDENTIALS` locally): +- `{"message":{"token","notification":{"title","body"},"android":{"priority":"HIGH","ttl":"14400s", + "collapse_key":"","notification":{"channel_id":"orca-desktop","tag":""}}, + "data":{ all orca fields as strings }}}` +- Dead token: `UNREGISTERED`, or `INVALID_ARGUMENT` whose message names the token. + +### Gateway storage (Postgres in prod, SQLite in tests, same pattern as `cloud/apps/relay/src/database.ts`) + +- `push_hosts(host_fingerprint pk, host_public_key, created_at, last_seen_at)`, written only on a + verified proof and pruned after 1 h of no contact when no `push_devices` row still names the host. + Nothing reads it, and any keypair mints a host for free, so it is not allowed to accumulate. +- `push_sessions` holds one row per host, enforced by a unique index and transaction lock. Minting a + session deletes the host's earlier one, since a desktop holds a single session and only re-proves once it is gone. +- `push_challenges(challenge_id pk, host_fingerprint, host_public_key, secret_hash, transcript, + expires_at, consumed_at)` +- `push_sessions(token_hash pk, host_fingerprint, expires_at, created_at)` +- `push_devices(registration_id pk, host_fingerprint, device_id, platform, token, apns_environment, + filter_json, dead_at, created_at, updated_at, unique(host_fingerprint, device_id))` +- `push_send_log(host_fingerprint, registration_id, sent_at)` for quota, pruned after 25 h. + +Logging: aggregate counters only. Never log tokens, titles, bodies, or raw fingerprints (log the first +4 chars of a fingerprint at most). + +### Gateway env + +`PORT`, `ORCA_PUSH_PUBLIC_URL`, `ORCA_PUSH_DATABASE_URL` (absent → SQLite under `ORCA_PUSH_DATA_DIR`), +`ORCA_PUSH_APNS_KEY` (PEM text), `ORCA_PUSH_APNS_KEY_ID`, `ORCA_PUSH_APPLE_TEAM_ID`, +`ORCA_PUSH_APNS_TOPIC` (default `com.stably.orca.mobile`), `ORCA_PUSH_FCM_PROJECT_ID` (default +`onorca-cloud`), `ORCA_PUSH_COALESCE_MS` (default 3000), `ORCA_PUSH_TRUSTED_PROXY_HOPS` (default 0, +proxies appending to `x-forwarded-for` after the client). +Secret Manager names (already exist in `onorca-cloud`): `orca-cloud-push-apns-key`, +`orca-cloud-push-apns-key-id`, `orca-cloud-push-apple-team-id`. Runtime SA: +`orca-cloud-push@onorca-cloud.iam.gserviceaccount.com` (already has FCM admin + secret accessor). + +## Desktop (`src/main`, `src/shared`) + +- Capability `NOTIFICATIONS_REMOTE_PUSH_RUNTIME_CAPABILITY = 'notifications.remote-push.v1'` in + `src/shared/protocol-version.ts`, advertised statically. +- RPC `notifications.registerPush` params `{ platform, token, apnsEnvironment?, filter }` (same shapes + as the gateway `POST /v1/devices` minus deviceId, which comes from `ctx.pairedDeviceId`). Returns + `{ registered: true, registrationId } | { registered: false, reason: 'gateway_unreachable' | + 'gateway_rejected' | 'not_mobile' | 'registration_storage_failed' | 'throttled' }`. A device may + register at most 10 times per minute (`throttled` beyond that, its earlier registration untouched): + each call is a gateway write plus a synchronous registry write on the main thread, and a paired + phone could otherwise loop it. The unregister RPC is not throttled, since with nothing registered it + is a lookup and with something registered it can only run once per successful register. The params + schema is strict, so a caller-supplied `deviceId` is an error, not a key silently dropped. Persists `pushRegistration: + { registrationId, platform, filter, registeredAt }` on `DeviceEntry` in `device-registry.ts` (new + optional field, tolerated by old registries). When the gateway accepted the token but the host could + not store it — the device left mobile scope mid-call (`not_mobile`) or the registry write threw + (`registration_storage_failed`) — the host queues the gateway delete in the unregister outbox rather + than leaking a registration nothing will ever push to. Registration, unregister, and outbox deletes + are serialized per device; re-registration first settles earlier cleanup. Authentication failure + never drops a durable delete. Stale send responses only clear the exact local registration observed, + while provider dead-token updates match the token/platform/environment that was sent. Phones must + treat any `registered: false` as "retry later", so an unknown reason string is safe to add. +- RPC `notifications.unregisterPush` params null → `{ unregistered: boolean }`. Removes the field and + enqueues a gateway delete in a durable outbox (`src/main/runtime/push/push-unregister-outbox.ts`, + modelled on `relay-revoke-outbox.ts`). Unpair/revoke (`revokeMobileDevice`) enqueues the same. The + drain re-reads the queue as it goes, so a delete queued mid-drain lands in the same pass, and a pass + that leaves retryable items schedules an unref'd backoff retry (30 s, doubling, capped at 10 min) + instead of waiting for the next launch. +- Both RPCs added to `runtime-rpc-mobile-method-allowlist.ts`. +- Push client `src/main/runtime/push/push-gateway-client.ts`: challenge/proof/session with token cache, + register, delete, send. Node `fetch`. Gateway URL from `profile-cloud-auth-config.ts` + (`pushGatewayUrl`, default `https://push.onorca.dev`, env override `ORCA_PUSH_GATEWAY_URL`). +- Host proof answering: new `src/main/runtime/push/push-host-proof.ts`, a copy of the relay's + `answerRelayHostChallenge` with the push transcript fields. Shared code with the relay proof is + welcome if it stays a pure refactor. +- Dispatch hook: in `RuntimeMobileNotificationController.dispatch`, after the socket fan-out, call + `pushDispatcher.enqueue(eventWithSeq)`. The dispatcher applies each device's `filter`, skips `dismiss` + events, maps `agentState` to `needs-input | finished` (blocked/waiting → needs-input, else finished), + batches matching registrationIds into `POST /v1/send` requests of at most 20 registrations each (the + gateway's per-request cap; extra devices get their own request rather than being dropped), and drops + unchanged registrations the gateway reports `dead`. Failure categories are counted without payload + values and logged at most once per minute (with a final flush on shutdown). Fire-and-forget with + one retry after 2 s per request; never throws into dispatch. +- Add `agentState` to `MobileNotificationDispatchEvent` and set it in `src/main/ipc/notifications.ts` + from `args.agentState`. Fix `buildAgentTaskCompleteNotificationOptions` so `working|running|busy` + never yields "finished" (title says "working" and the dispatcher treats it as not-final, i.e. no push). +- Headless serve: no renderer means no `notifications:dispatch`. Document in + `docs/reference/headless-linux-server.md`; do not fix here. + +## Mobile (`mobile/`) + +- Commit `google-services.json` (from `/tmp/orca-mobile-push/google-services.json`) at `mobile/` and set + `"android": { "googleServicesFile": "./google-services.json" }` in `app.json`. Add `"expo-notifications"` + to `plugins` so prebuild writes the `aps-environment` entitlement. +- Token: `Notifications.getDevicePushTokenAsync()`; `data` is the APNs hex or FCM string. iOS + `apnsEnvironment`: `__DEV__ ? 'sandbox' : 'production'` (dev-client builds are debug, TestFlight and + App Store are release). Listen with `addPushTokenListener` and re-register on change. +- Settings (`mobile/app/notifications.tsx`): single "Background notifications" switch, default off, + hint text exactly: "Get alerts while Orca is closed. Alerts show the same text as on your desktop. That + text, your phone's push token, and opaque host and device ids pass through Orca's push service and Apple + or Google. Turning this off or unpairing deletes the token." Event controls live in the shared notification-preferences section and apply to both connected and background notifications. + Hide the whole section, with copy "Update your desktop app to enable background notifications", when + no paired host advertises `notifications.remote-push.v1`. +- Registration: on switch-on (after OS permission), and on every host reaching `connected` while the + switch is on, call `notifications.registerPush` on that host if it advertises the capability. On + switch-off call `notifications.unregisterPush` on every connected host and remember to retry on hosts + that were offline. On host removal, best-effort unregister before deleting credentials. +- Receive: `addNotificationReceivedListener` (foreground) checks `data.orca.notificationId` + + `notificationSeq` against the host session seen set in `notification-reconnect-catchup.ts`; if seen, + suppress via `setNotificationHandler` returning no banner; otherwise show and mark seen. Background and + killed: OS shows it. +- Tap: `data.orca.hostFingerprint` → hostId by computing the same sha256/base64url/16 derivation over each + stored host's `publicKeyB64`; then existing `getNotificationNavigationTarget` + `useOpenNotificationRoute`. +- Reopen: existing replay catch-up runs unchanged. Dismiss events also + `dismissNotificationAsync` any presented notification whose `data.orca.notificationId` matches. +- Old host without the capability: nothing changes. + +## Infra (`cloud/infra/terraform`, `.github/workflows`) + +- Cloud Run service `orca-cloud-push`, region `us-central1`, project from the environment tfvars, runtime + SA `orca-cloud-push@.iam.gserviceaccount.com` (exists in prod; declare and import), the three + secrets mounted as env (exist; declare and import), Cloud SQL connector to the shared instance with its + own database `orca_push`, min instances 1, max 4, concurrency 80, ingress all, unauthenticated invoke. +- IAM: `roles/firebasecloudmessaging.admin` and `roles/serviceusage.serviceUsageConsumer` on the runtime + SA (exist in prod; declare and import). Secret accessor per secret. +- Hostname `push.onorca.dev`. The DNS zone lives in the apps root in `stablyai/orca-cloud`; add the + Cloud Run domain mapping here and leave a TODO comment naming the record the other repo must add. +- Workflow `.github/workflows/cloud-push-deploy.yml`: gated on `vars.ORCA_CLOUD_OPERATIONS_ENABLED`, + Workload Identity like `cloud-relay-*`, builds the image, deploys with `--no-traffic`, probes the new + revision's `/ready` and a validate-only FCM send, then shifts 100% traffic. Uses + `.github/actions/cloud-sql-rollout-lease` around the schema step. +- Add the new root files to `cloud/dev/contracts` and `cloud/dev/fixtures` partitions so + `terraform-root-partition.test.mjs` and `Cloud Verify` pass. + +## Non-goals for this release + +Ack gate, generic-alert mode, staging gateway, iOS Notification Service Extension, Android data-only +messages, Live Activities, account-based quota tiers, dismissal via silent push. + +### Device delivery preferences + +The desktop advertises `notifications.delivery-preferences.v1`. Completion detection remains +active when desktop notifications are off; semantic validity checks still precede delivery. +IPC publishes `desktopAllowed: false` for terminal events disabled by the desktop master or +source switch. Desktop focus and native authorization remain desktop-only delivery gates. + +`notifications.subscribe` and `notifications.getMissedSince` accept optional +`includeDesktopSuppressed: true`. Only opted-in callers receive those events, including replay; +legacy callers keep the old filtered stream. A new phone against an older host can narrow the +available events but cannot recover events that host never published. + +The phone defaults to following each host. `filter.followDesktop` is optional: absent retains +legacy desktop gating; explicit false permits independent event choices. The desktop persists +it with the paired registration and evaluates it for every send, so desktop preference changes +work while the phone is disconnected. This flag is host-local and is not sent to the gateway. +The phone uses the same shared event predicate for socket/replay delivery as the push dispatcher. +Optional `emittedAt` carries the event time for per-device five-second burst suppression after +source filtering. Desktop eligibility, source, and agent state use separate upstream cooldown +buckets so filtered events cannot suppress the next eligible event. Legacy RPC callers retain +workspace-wide burst suppression on the host. + +`filter.sound` is also host-local. False groups that device's requests separately and adds +optional `notification.sound: false` to gateway sends. The gateway omits APNs `aps.sound` and +uses Android's `orca-desktop-silent` channel. Missing sound preserves existing audible delivery. +Deploy the updated gateway before distributing hosts that send the optional sound field: older +gateways strictly reject unknown notification fields. No token or database migration is needed. + +The phone's master switch disables background registration as well as local scheduling. Sound +and viewing preferences belong to the receiving phone. The phone suppresses a banner for its +currently viewed host/workspace only while active; it never assumes desktop focus means the +phone is viewing that workspace. Changes to an offline host's persisted filter take effect on +reconnection. No live APNs/FCM delivery is implied by simulator notification injection. + +For a phone registered for background push, socket notification delivery waits while the app is +inactive. On foreground, it checks the native push tray before scheduling a local fallback, so +a still-connected background socket cannot duplicate APNs/FCM delivery. Unsubscribing cancels +the wait without claiming delivery. Hosts without push registration keep local delivery. + +Native notification readers accept Expo's iOS `request.trigger.payload` as well as +`request.content.data`. APNs custom fields can exist only in the former; foreground deduplication, +tray replay suppression, dismissal, and tap routing all use the same reader. diff --git a/docs/reference/windows-process-enumeration.md b/docs/reference/windows-process-enumeration.md index 34afb56c8e6..0f7f17bd433 100644 --- a/docs/reference/windows-process-enumeration.md +++ b/docs/reference/windows-process-enumeration.md @@ -344,7 +344,7 @@ on any other OS keeps using the scan. ## Why the package is patched -`config/patches/@vscode__windows-process-tree@0.8.0.patch` carries four hunks. +`config/patches/@vscode__windows-process-tree@0.8.0.patch` carries five changes. 1. **Spectre mitigation.** The upstream `binding.gyp` requires Spectre-mitigated libraries, which Orca's Windows build agents do not install. `node-pty` is @@ -360,6 +360,32 @@ on any other OS keeps using the scan. `node_addon_api.gyp` resolves outside the repo and hourly Windows builds die at configure. `node-pty` is patched the same way for the same reason. 4. **No PEB reads, no `PROCESS_VM_READ`.** See below. +5. **The `CreationTime` flag (4).** Upstream exposes no process start time, and + `isWindowsProcessStartTimeAvailable()` gates structured Claude and Codex + chat on it, so without this change win32 silently fell back to the legacy + transcript path. `GetProcessCreationTime` opens + `PROCESS_QUERY_LIMITED_INFORMATION` and converts `GetProcessTimes`' FILETIME + to Unix ms; a process that denies the handle is emitted with the field + absent, never zero, because callers must be able to tell "cannot identify" + from a timestamp. +5. **`supportedProcessDataFlags`.** `addon.cc` exports the flag bits the + compiled binary understands, and `lib/index.js` re-exports it. + + Why a fifth hunk and not just the enum: unlike `node-pty`, this package + publishes a prebuilt `.node` at the same `build/Release/` path node-gyp + writes to. pnpm patches the source tree and leaves that prebuilt alone, so a + host can hold a patched `lib/index.js` — `ProcessDataFlag.CreationTime` and + all — over a binary that ignores flag 4. CI produced exactly that: the gate + read available and every row came back without `creationTimeMs`. Neither a + load check nor a path check can see the difference, so the binary has to say + so itself. + + Two readers depend on it. `isWindowsProcessStartTimeAvailable()` returns + false unless this bit is set, because claiming otherwise leaves + `captureWindowsDescendantSnapshot` returning null forever while structured + chat believes it has a reaper. And `windows-process-tree-creation-time.cjs` + asserts it during install, which is what forces a from-source rebuild — + the same role `node-pty-job-ownership.cjs` plays for node-pty's job exports. The typings claim `commandLine` is truncated at 512 characters. Measured, it is not: the longest observed on a real host was 26,059. @@ -494,10 +520,10 @@ already has, which is why the addon is checked again at load. ## What the snapshot does not provide -`CreationDate` (process start time) has no equivalent. Anything using a start -time to prove a PID has not been recycled — daemon identity, managed-hook -ownership, and CPU accounting in the memory collector — still reads it through -its own query. Those callers are not migrated. +`CreationDate` (process start time) now has an equivalent — `creationTimeMs`, +above — but only inside this module. Daemon identity, managed-hook ownership and +CPU accounting in the memory collector still read a start time through their own +queries; those callers are not migrated. Committed private bytes have no equivalent either, and the one memory value the addon can produce is unusable for the sizes Orca now sees: `process.cc` stores @@ -509,10 +535,12 @@ counters in the same pass. Migrating it to the native table would cost both, and it is why this module no longer sets the `Memory` flag at all: the field had no reader, and asking for it opened a handle per process on every snapshot. -Start time is a proxy for identity, not identity. The durable answer for the -process trees Orca itself spawns is an inherited handle: a job object names the -tree Orca created, so no start-time comparison is needed. Those readers should -be resolved that way rather than by adding a start time to this module. +Start time is a proxy for identity, not identity. For the process trees Orca +itself spawns the durable answer is still an inherited handle: a job object +names the tree Orca created, so no start-time comparison is needed. The +`creationTimeMs` this snapshot now carries is for the trees Orca did **not** +create the handle for — a recovered agent session, a descendant walked out of +the table — where a bare PID is all there is to re-identify. Do not adopt `getProcessCpuUsage()` from the package. It takes both CPU samples inside one call with a blocking `Sleep(1000)` in the middle, which would hold a diff --git a/docs/site/content/docs/mobile.mdx b/docs/site/content/docs/mobile.mdx index 13365eac3ae..d0967c1d8c5 100644 --- a/docs/site/content/docs/mobile.mdx +++ b/docs/site/content/docs/mobile.mdx @@ -11,7 +11,7 @@ The Orca mobile companion is an iOS/Android app that pairs with your desktop Orc The mobile companion is in beta. Install iOS from the [App Store](https://apps.apple.com/us/app/orca-ide/id6766130217), join the [TestFlight preview channel](https://testflight.apple.com/join/YjeGMQBA), or install Android from the [current APK - 0.0.46](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.46/app-release.apk). + 0.0.48](https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk). ## What you can do from mobile @@ -31,7 +31,7 @@ The Orca mobile companion is an iOS/Android app that pairs with your desktop Orc - Create a workspace from mobile with the same Smart source modes as desktop: Smart, GitHub, Linear, GitLab, Branch, and Name. With **multiple connected desktops**, **New Workspace** asks which host should create it first (one connected host skips the picker). - Open a host card's **⋯** menu for **Edit**, **Connect**, **Remove**, and related actions (long-press still works as a shortcut). - Edit a saved host's display name or connection address without re-pairing (for example when the desktop moves between home LAN and Tailscale). -- Get push notifications when an agent finishes, mirroring [desktop notifications](/docs/notifications). +- Get push notifications when an agent finishes or needs input, mirroring [desktop notifications](/docs/notifications). Turn on **Background notifications** in the phone's Notifications settings to keep receiving them while Orca is closed; see [Notifications](/docs/notifications#background-notifications-on-your-phone) for what that sends and where. The mobile app is intentionally not a full editor — it's a remote control for the desktop you already have running. diff --git a/docs/site/content/docs/model/quick-open.mdx b/docs/site/content/docs/model/quick-open.mdx index 7e74ceeb4ea..e47222f2e82 100644 --- a/docs/site/content/docs/model/quick-open.mdx +++ b/docs/site/content/docs/model/quick-open.mdx @@ -19,9 +19,9 @@ Type a web search instead of a path or URL to open it in the worktree browser wi ## Worktree Jump Palette (Cmd-J) -Jump across every worktree and every tab in one search. The placeholder in the empty input reads _repo/worktree_ — type either half and Orca filters accordingly. Once you start typing, search includes non-archived worktrees even if they are hidden by the sidebar's current filters. Slack-style emoji shortcodes (`:rocket:`) use the same suggestion popover as workspace naming. +Jump across worktrees and tabs in one search. The palette opens with the sidebar's current host and project scope, including individual repository selections. The placeholder in the empty input reads _repo/worktree_ — type either half and Orca filters accordingly. Typing can still find non-archived worktrees hidden by the sidebar's other visibility toggles, but it keeps that host and repository scope. Slack-style emoji shortcodes (`:rocket:`) use the same suggestion popover as workspace naming. -Press **Tab** in the palette for a host and project filter menu. Selected hosts and projects narrow the result set and show as chips you can remove one at a time; closing the palette clears the filter so the next open is unscoped. +Press **Tab** in the palette for a host and project filter menu. Project choices are repository-granular. Selected hosts and repositories narrow the result set and show as chips you can remove one at a time. Changes are temporary: closing the palette discards them, and the next open reseeds the filter from the sidebar. Results include: diff --git a/docs/site/content/docs/model/worktrees.mdx b/docs/site/content/docs/model/worktrees.mdx index 39fbbda3b59..f71d21c2ca4 100644 --- a/docs/site/content/docs/model/worktrees.mdx +++ b/docs/site/content/docs/model/worktrees.mdx @@ -102,7 +102,7 @@ The sidebar header filter menu groups host and project scope under a shared **Sh - **Other-client** workspaces — **Hide other-client workspaces** appears when a shared [Remote Orca Server](/docs/remote-servers) has workspaces created from another paired client; turn it on to keep this device's list to workspaces you created here. Empty `Cmd-J` recents and numeric shortcuts follow the same filter; typing a query still finds hidden rows. - **Detached HEAD** workspaces — checkouts sitting on a commit rather than a branch -Active filter count shows on the filter control; **Clear** resets only the filters that are on. Text search and [Worktree Jump Palette](/docs/model/quick-open) (`Cmd-J`) still reach workspaces hidden only by these filters once you type a query — the jump palette also has its own host/project filters (**Tab**). +Active filter count shows on the filter control; **Clear** resets only the filters that are on. Text search and [Worktree Jump Palette](/docs/model/quick-open) (`Cmd-J`) still reach workspaces hidden only by the hide toggles once you type a query. Cmd-J keeps the sidebar's host and project scope when it opens; press **Tab** to adjust its temporary host and individual-repository filters. When you add a parent folder that contains multiple Git repos, Orca can import the selected repos separately or group them under one project group. diff --git a/docs/site/content/docs/notifications.mdx b/docs/site/content/docs/notifications.mdx index 8d5e02866f7..aeea1c65d6d 100644 --- a/docs/site/content/docs/notifications.mdx +++ b/docs/site/content/docs/notifications.mdx @@ -29,3 +29,33 @@ Pick a custom desktop notification sound per category under [Settings → Notifi Supported formats: MP3, WAV, OGG, M4A, AAC, FLAC. One file applies to all delivered desktop notifications. When you use a custom sound, set its playback volume from the same settings pane. + +## Background notifications on your phone + +The Orca mobile app shows an agent-finished or needs-input alert while it is open and connected to your desktop. To keep receiving them while the app is in the background or closed, turn on **Background notifications** in the phone's Notifications settings. It is off by default. + +When it is on, your desktop sends each alert to Orca's push service, which delivers it through Apple or Google to your phone. The alert shows the same title and text as the desktop notification. What leaves your computer is that text, your phone's push token, and opaque host and device ids. Orca's push service keeps the text only long enough to send it and never writes it to storage. Apple and Google can read it in transit, as they can for any app's notifications. The service is open source in the Orca repository under `cloud/apps/push`. + +Turning the switch off, or unpairing the phone from the desktop, deletes the token from the push service. The **Enable notifications** switch turns off both connected alerts and background push. Removing a host from the phone while that desktop is offline may leave background alerts arriving from it until the desktop is unpaired or the switch is turned off on the phone. + +Background notifications need a paired desktop that has been updated to advertise the feature; the phone hides the switch otherwise. They do not fire from a headless `orca serve` host, because agent-completion detection runs in the desktop app. On Android they need Google Play services, so de-Googled phones keep the in-app behaviour only. + +## Notification preferences on your phone + +**Use desktop settings** is on by default. Each paired desktop's notification master switch, +**Agent Task Complete**, and **Terminal Bell** switches determine which terminal events reach +this phone. Desktop focus and desktop OS permissions do not suppress phone alerts. + +Turn off **Use desktop settings** to choose **Task finished**, **Needs input**, **Terminal bell**, +and **Plugin notifications** independently on your phone. These event filters apply to both +connected notifications (including reconnect catch-up) and background push. Older desktops +still filter events before forwarding them; update the desktop to enable independent delivery. +Previously customized background agent-state filters are preserved as independent preferences. + +A terminal bell is a program's attention signal, not proof that an agent finished. Disable +**Terminal bell** on your phone if a CLI repeatedly rings while it is working. + +**Notification sound** and **Suppress while viewing workspace** are local to the phone. +Viewing suppression applies only while the phone is open on that host's workspace. Background +notifications can still arrive while the phone is closed. Phone sound choices do not sync custom +desktop audio files. Preference changes reach disconnected desktops when they reconnect. diff --git a/mobile/app.config.js b/mobile/app.config.js new file mode 100644 index 00000000000..4927fa3c956 --- /dev/null +++ b/mobile/app.config.js @@ -0,0 +1,19 @@ +// Why this file exists: a bare "expo-notifications" plugin entry writes +// `aps-environment: development` into the iOS entitlements, while push-token.ts +// reports `production` for every non-__DEV__ build. A TestFlight or App Store build +// would then register a production APNs token against a sandbox entitlement, and the +// gateway's pushes would be accepted by Apple and delivered nowhere. Deriving the +// mode from an env var the release workflow sets makes the two agree by construction +// instead of relying on the export step to rewrite the entitlement. +// +// app.json stays the source for everything else: Expo reads it first and hands it to +// this function, so the fastlane version/buildNumber rewrite still flows through. +const APS_ENVIRONMENT = + process.env.ORCA_IOS_APS_ENVIRONMENT === 'production' ? 'production' : 'development' + +module.exports = ({ config }) => ({ + ...config, + plugins: (config.plugins ?? []).map((plugin) => + plugin === 'expo-notifications' ? ['expo-notifications', { mode: APS_ENVIRONMENT }] : plugin + ) +}) diff --git a/mobile/app.json b/mobile/app.json index fc36687d74f..6121923f775 100644 --- a/mobile/app.json +++ b/mobile/app.json @@ -75,10 +75,12 @@ "allowBackup": false, "permissions": ["RECORD_AUDIO", "MODIFY_AUDIO_SETTINGS"], "package": "com.stably.orca.mobile", - "versionCode": 16 + "versionCode": 16, + "googleServicesFile": "./google-services.json" }, "plugins": [ "expo-router", + "expo-notifications", "./plugins/android-respect-rotation-lock.js", [ "expo-splash-screen", diff --git a/mobile/app/_layout.tsx b/mobile/app/_layout.tsx index 9080cdedcf9..661a18359a5 100644 --- a/mobile/app/_layout.tsx +++ b/mobile/app/_layout.tsx @@ -1,6 +1,9 @@ +import { readNativeNotificationData } from '../src/notifications/native-notification-data' +import { loadNotificationDeliveryPreferences } from '../src/notifications/notification-delivery-preferences' +import { setNotificationViewingWorkspace } from '../src/notifications/notification-viewing-policy' import { useCallback, useEffect, useRef } from 'react' import { View, StyleSheet } from 'react-native' -import { Stack, useRouter } from 'expo-router' +import { Stack, useRouter, useGlobalSearchParams, usePathname } from 'expo-router' import { StatusBar } from 'expo-status-bar' import * as SplashScreen from 'expo-splash-screen' import * as Notifications from 'expo-notifications' @@ -10,6 +13,13 @@ import { OrcaLogo } from '../src/components/OrcaLogo' import { RpcClientProvider } from '../src/transport/client-context' import { getNotificationNavigationTarget } from '../src/notifications/notification-routing' import { useOpenNotificationRoute } from '../src/notifications/use-open-notification-route' +import { + isRemotePushTrigger, + pushNotificationRouteData, + shouldSuppressForegroundPush +} from '../src/notifications/push-receive' +import { startPushTokenSync } from '../src/notifications/push-registration' +import { ensureDesktopNotificationChannel } from '../src/notifications/desktop-notification-channel' import { loadHostCatalog } from '../src/transport/host-store' import { extractPairingCodeFromUrl } from '../src/transport/pairing' import { recoverMobileRelayPairing } from '../src/transport/mobile-relay-pairing-recovery' @@ -19,22 +29,44 @@ import { recoverMobileRelayPairing } from '../src/transport/mobile-relay-pairing // between the native splash and the first React paint. SplashScreen.preventAutoHideAsync() +// Why at boot and not only on subscribe: the gateway's FCM payload targets the +// 'orca-desktop' channel, and a background push can land before any socket has +// connected. Android drops a notification whose channel does not exist yet. +ensureDesktopNotificationChannel() + // Why: without this, expo-notifications silently drops notifications when // the app is in the foreground. Setting all three to true makes iOS/Android // display the banner, play the sound, and show the badge even while the // app is active. This runs once at module load time before any notification // is scheduled. Notifications.setNotificationHandler({ - handleNotification: async () => ({ - shouldShowBanner: true, - shouldShowList: true, - shouldPlaySound: true, - shouldSetBadge: false - }) + handleNotification: async (notification) => { + // Why the check: a gateway push can arrive for an event the socket already + // delivered, and only the handler can stop the OS drawing a second banner. + const suppressed = await shouldSuppressForegroundPush( + readNativeNotificationData(notification.request) + ).catch(() => false) + return { + shouldShowBanner: !suppressed, + shouldShowList: !suppressed, + shouldPlaySound: !suppressed && (await loadNotificationDeliveryPreferences()).sound, + shouldSetBadge: false + } + } }) export default function RootLayout() { const router = useRouter() + const pathname = usePathname() + const { hostId, worktreeId } = useGlobalSearchParams<{ hostId?: string; worktreeId?: string }>() + useEffect(() => { + setNotificationViewingWorkspace( + pathname.includes('/session/') && typeof hostId === 'string' && typeof worktreeId === 'string' + ? { hostId, worktreeId } + : null + ) + return () => setNotificationViewingWorkspace(null) + }, [pathname, hostId, worktreeId]) const openNotificationRoute = useOpenNotificationRoute() const handledNotificationIdsRef = useRef>(new Set()) @@ -44,6 +76,10 @@ export default function RootLayout() { void recoverMobileRelayPairing() }, []) + // Why: a rolled APNs/FCM token stops delivering silently, so every paired host + // has to be re-registered with the new one as soon as the provider hands it over. + useEffect(() => startPushTokenSync(), []) + // Why: route `orca://pair?...` deep links to the confirm screen so // the same pairing flow runs whether the link arrived via QR scan, // paste, AirDrop, Messages, or `xcrun simctl openurl`. getInitialURL @@ -94,9 +130,18 @@ export default function RootLayout() { } } - async function getNavigationTarget(data: unknown) { + async function getNavigationTarget(notification: Notifications.Notification) { const hosts = await loadHostCatalog().catch(() => null) - return getNotificationNavigationTarget(data, { + const data = readNativeNotificationData(notification.request) + // A gateway push names its host by key fingerprint, not by this device's hostId. + // With no catalog to resolve against, such a push stays unrouted instead of + // falling back to whatever hostId its raw data carries. + const routeData = pushNotificationRouteData( + data, + hosts ?? [], + isRemotePushTrigger(notification.request.trigger) + ) + return getNotificationNavigationTarget(routeData, { knownHostIds: hosts ? new Set(hosts.map((host) => host.id)) : undefined, credentialStatusByHostId: hosts ? new Map(hosts.map((host) => [host.id, host.credentialStatus])) @@ -124,7 +169,7 @@ export default function RootLayout() { } } - const target = await getNavigationTarget(response.notification.request.content.data) + const target = await getNavigationTarget(response.notification) clearLastNotificationResponse() if (disposed) { return diff --git a/mobile/app/notifications.tsx b/mobile/app/notifications.tsx index d9696251a94..db1b94238cc 100644 --- a/mobile/app/notifications.tsx +++ b/mobile/app/notifications.tsx @@ -1,13 +1,36 @@ +import { NotificationDeliverySection } from '../src/notifications/NotificationDeliverySection' +import { + DEFAULT_NOTIFICATION_DELIVERY, + loadNotificationDeliveryPreferences, + type NotificationDeliveryPreferences +} from '../src/notifications/notification-delivery-preferences' import { useState, useCallback, useEffect } from 'react' -import { AppState, Linking, View, Text, StyleSheet, Pressable, Switch } from 'react-native' +import { + AppState, + Linking, + View, + Text, + StyleSheet, + Pressable, + Switch, + ScrollView, + Alert +} from 'react-native' import { useSafeAreaInsets } from 'react-native-safe-area-context' import { useRouter, useFocusEffect } from 'expo-router' import { ChevronLeft } from 'lucide-react-native' import { colors, spacing, typography } from '../src/theme/mobile-theme' import { loadPushNotificationsEnabled, + loadRemotePushEnabled, savePushNotificationsEnabled } from '../src/storage/preferences' +import { BackgroundNotificationsSection } from '../src/notifications/BackgroundNotificationsSection' +import { + setNotificationDeliveryPreferences, + setRemotePushEnabled +} from '../src/notifications/push-registration' +import { useRemotePushCapableHosts } from '../src/notifications/use-remote-push-capable-hosts' import { ensureNotificationPermissions, getNotificationPermissionState, @@ -26,14 +49,22 @@ export default function NotificationsScreen() { const insets = useSafeAreaInsets() const [pushEnabled, setPushEnabled] = useState(false) const [permissionState, setPermissionState] = useState(DEFAULT_PERMISSION_STATE) + const [backgroundEnabled, setBackgroundEnabled] = useState(false) + const [delivery, setDelivery] = useState(DEFAULT_NOTIFICATION_DELIVERY) + const [saving, setSaving] = useState(false) + const remotePushSupport = useRemotePushCapableHosts() const refreshSettings = useCallback(async () => { - const [enabled, permission] = await Promise.all([ + const [enabled, permission, background, states] = await Promise.all([ loadPushNotificationsEnabled(), - getNotificationPermissionState() + getNotificationPermissionState(), + loadRemotePushEnabled(), + loadNotificationDeliveryPreferences() ]) setPushEnabled(enabled) setPermissionState(permission) + setBackgroundEnabled(background) + setDelivery(states) }, []) useFocusEffect( @@ -59,11 +90,45 @@ export default function NotificationsScreen() { if (!granted) { setPushEnabled(false) await savePushNotificationsEnabled(false) + await setRemotePushEnabled(false) + setBackgroundEnabled(false) return } } setPushEnabled(value) await savePushNotificationsEnabled(value) + if (!value) { + await setRemotePushEnabled(false) + setBackgroundEnabled(false) + } + } + + const toggleBackground = async (value: boolean) => { + if (value) { + const granted = await ensureNotificationPermissions() + setPermissionState(await getNotificationPermissionState()) + if (!granted) { + return + } + } + if (value) { + await savePushNotificationsEnabled(true) + setPushEnabled(true) + } + setBackgroundEnabled(value) + await setRemotePushEnabled(value) + } + + const changeDelivery = async (value: NotificationDeliveryPreferences) => { + setSaving(true) + try { + await setNotificationDeliveryPreferences(value) + setDelivery(value) + } catch { + Alert.alert('Could not save notification settings', 'Please try again.') + } finally { + setSaving(false) + } } const switchEnabled = pushEnabled && permissionState.granted @@ -73,7 +138,13 @@ export default function NotificationsScreen() { : 'Get notified on this device when an agent needs your input or finishes a task.' return ( - + router.back()}> @@ -83,8 +154,9 @@ export default function NotificationsScreen() { - Agent notifications + Enable notifications void togglePush(v)} @@ -105,7 +177,19 @@ export default function NotificationsScreen() { )} - + + void changeDelivery(value)} + /> + void toggleBackground(value)} + /> + ) } diff --git a/mobile/google-services.json b/mobile/google-services.json new file mode 100644 index 00000000000..4120a97dafc --- /dev/null +++ b/mobile/google-services.json @@ -0,0 +1,39 @@ +{ + "project_info": { + "project_number": "120364513935", + "project_id": "onorca-cloud", + "storage_bucket": "onorca-cloud.firebasestorage.app" + }, + "client": [ + { + "client_info": { + "mobilesdk_app_id": "1:120364513935:android:1d951dc430aeb9bc664efa", + "android_client_info": { + "package_name": "com.stably.orca.mobile" + } + }, + "oauth_client": [ + { + "client_id": "120364513935-evfa8502bp5r9hn7afhd9i03oibs8223.apps.googleusercontent.com", + "client_type": 3 + } + ], + "api_key": [ + { + "current_key": "AIzaSyBmT_w0OUQSiVfxblx-F0qlRvGkBBkTNQU" + } + ], + "services": { + "appinvite_service": { + "other_platform_oauth_client": [ + { + "client_id": "120364513935-evfa8502bp5r9hn7afhd9i03oibs8223.apps.googleusercontent.com", + "client_type": 3 + } + ] + } + } + } + ], + "configuration_version": "1" +} diff --git a/mobile/src/home/use-mobile-home-host-connections.ts b/mobile/src/home/use-mobile-home-host-connections.ts index 989583f11ab..9cf094ee240 100644 --- a/mobile/src/home/use-mobile-home-host-connections.ts +++ b/mobile/src/home/use-mobile-home-host-connections.ts @@ -1,6 +1,7 @@ import { useEffect, useMemo, useRef, useState } from 'react' import { decodeAccountsSnapshot } from '../components/AccountUsage' import { subscribeToDesktopNotifications } from '../notifications/mobile-notifications' +import { attachPushRegistration } from '../notifications/push-registration' import { usePrimeHosts } from '../transport/client-context' import { createHostConnectRefetchGate } from '../transport/host-connect-refetch-gate' import { selectHomeAutoConnectHostIds } from '../transport/home-host-auto-connect' @@ -37,11 +38,15 @@ function wireMobileHomeHostSubscriptions( ): () => void { let unsubscribeNotifications: (() => void) | null = null let unsubscribeAccounts: (() => void) | null = null + let detachPushRegistration: (() => void) | null = null const refetchGate = createHostConnectRefetchGate() const wireState = (state: ConnectionState): void => { const reconnected = refetchGate.observe(state) if (state === 'connected') { unsubscribeNotifications ??= subscribeToDesktopNotifications(entry.client, entry.hostId) + // Why here: this is the one place a host is known to be authenticated, which is + // what registerPush needs; it no-ops on hosts without the push capability. + detachPushRegistration ??= attachPushRegistration(entry.hostId, entry.client) unsubscribeAccounts ??= entry.client.subscribe('accounts.subscribe', null, (payload) => { if (!payload || typeof payload !== 'object') { return @@ -78,6 +83,8 @@ function wireMobileHomeHostSubscriptions( unsubscribeNotifications = null unsubscribeAccounts?.() unsubscribeAccounts = null + detachPushRegistration?.() + detachPushRegistration = null } wireState(entry.state) const unsubscribeState = entry.client.onStateChange(wireState) @@ -85,6 +92,7 @@ function wireMobileHomeHostSubscriptions( unsubscribeState() unsubscribeNotifications?.() unsubscribeAccounts?.() + detachPushRegistration?.() } } diff --git a/mobile/src/notifications/BackgroundNotificationsSection.test.tsx b/mobile/src/notifications/BackgroundNotificationsSection.test.tsx new file mode 100644 index 00000000000..ced4ec7f210 --- /dev/null +++ b/mobile/src/notifications/BackgroundNotificationsSection.test.tsx @@ -0,0 +1,71 @@ +import { createElement } from 'react' +import { act, create, type ReactTestRenderer } from 'react-test-renderer' +import { afterEach, describe, expect, it, vi } from 'vitest' +import { + BACKGROUND_NOTIFICATIONS_HINT, + BACKGROUND_NOTIFICATIONS_UNSUPPORTED, + BackgroundNotificationsSection, + type BackgroundNotificationsSectionProps +} from './BackgroundNotificationsSection' + +vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, + StyleSheet: { create: (styles: T) => styles }, + Switch: 'Switch', + Text: 'Text', + View: 'View' +})) + +describe('BackgroundNotificationsSection', () => { + let renderer: ReactTestRenderer | null = null + + afterEach(() => { + act(() => renderer?.unmount()) + renderer = null + }) + + function render(overrides: Partial = {}) { + act(() => { + renderer = create( + createElement(BackgroundNotificationsSection, { + supported: true, + resolved: true, + enabled: true, + onToggleEnabled: () => {}, + ...overrides + }) + ) + }) + return renderer! + } + + function textOf(tree: ReactTestRenderer): string[] { + return tree.root + .findAllByType('Text' as never) + .map((node) => node.props.children) + .filter((child): child is string => typeof child === 'string') + } + + it('shows the switch, the disclosure without a second set of event filters', () => { + const texts = textOf(render()) + + expect(texts).toEqual(['Background notifications', BACKGROUND_NOTIFICATIONS_HINT]) + }) + + it('states verbatim which parties see the alert text and the push token', () => { + expect(BACKGROUND_NOTIFICATIONS_HINT).toBe( + "Get alerts while Orca is closed. Alerts show the same text as on your desktop. That text, your phone's push token, and opaque host and device ids pass through Orca's push service and Apple or Google. Turning this off or unpairing deletes the token." + ) + }) + + it('replaces the whole section when no paired host advertises remote push', () => { + const tree = render({ supported: false }) + + expect(textOf(tree)).toEqual([BACKGROUND_NOTIFICATIONS_UNSUPPORTED]) + expect(tree.root.findAllByType('Switch' as never)).toHaveLength(0) + }) + + it('renders nothing while the paired hosts are still being probed', () => { + expect(render({ supported: false, resolved: false }).toJSON()).toBeNull() + }) +}) diff --git a/mobile/src/notifications/BackgroundNotificationsSection.tsx b/mobile/src/notifications/BackgroundNotificationsSection.tsx new file mode 100644 index 00000000000..00f6e86f3c8 --- /dev/null +++ b/mobile/src/notifications/BackgroundNotificationsSection.tsx @@ -0,0 +1,94 @@ +import { StyleSheet, Switch, Text, View } from 'react-native' +import { colors, spacing, typography } from '../theme/mobile-theme' + +// Verbatim from the push contract: it is the disclosure for handing a native push +// token to Orca's gateway and to Apple or Google, so the wording is not ours to edit. +export const BACKGROUND_NOTIFICATIONS_HINT = + "Get alerts while Orca is closed. Alerts show the same text as on your desktop. That text, your phone's push token, and opaque host and device ids pass through Orca's push service and Apple or Google. Turning this off or unpairing deletes the token." + +export const BACKGROUND_NOTIFICATIONS_UNSUPPORTED = + 'Update your desktop app to enable background notifications' + +export type BackgroundNotificationsSectionProps = { + /** True once some paired host advertised `notifications.remote-push.v1`. */ + supported: boolean + /** False while every paired host is still being probed; renders nothing rather + * than telling someone to update a desktop that may well be current. */ + resolved: boolean + enabled: boolean + onToggleEnabled: (value: boolean) => void +} + +export function BackgroundNotificationsSection({ + supported, + resolved, + enabled, + onToggleEnabled +}: BackgroundNotificationsSectionProps) { + if (!supported) { + return resolved ? ( + + {BACKGROUND_NOTIFICATIONS_UNSUPPORTED} + + ) : null + } + + return ( + + + Background notifications + + + {BACKGROUND_NOTIFICATIONS_HINT} + + ) +} + +const styles = StyleSheet.create({ + section: { + backgroundColor: colors.bgPanel, + borderRadius: 12, + overflow: 'hidden', + marginTop: spacing.md + }, + row: { + flexDirection: 'row', + alignItems: 'center', + gap: spacing.sm + 2, + paddingVertical: spacing.md, + paddingHorizontal: spacing.md + 2 + }, + subRow: { + paddingVertical: spacing.sm, + paddingLeft: spacing.lg + spacing.xs + }, + rowLabel: { + flex: 1, + fontSize: typography.bodySize, + fontWeight: '500', + color: colors.textPrimary + }, + subRowLabel: { + fontWeight: '400', + color: colors.textSecondary + }, + hint: { + fontSize: typography.metaSize, + color: colors.textMuted, + lineHeight: 18, + paddingHorizontal: spacing.md + 2, + paddingBottom: spacing.md + }, + unsupported: { + fontSize: typography.metaSize, + color: colors.textMuted, + lineHeight: 18, + padding: spacing.md + 2 + } +}) diff --git a/mobile/src/notifications/NotificationDeliverySection.test.tsx b/mobile/src/notifications/NotificationDeliverySection.test.tsx new file mode 100644 index 00000000000..f60bb7353b8 --- /dev/null +++ b/mobile/src/notifications/NotificationDeliverySection.test.tsx @@ -0,0 +1,45 @@ +import { createElement } from 'react' +import { act, create } from 'react-test-renderer' +import { expect, it, vi } from 'vitest' +import { NotificationDeliverySection } from './NotificationDeliverySection' +import { DEFAULT_NOTIFICATION_DELIVERY } from './notification-delivery-preferences' + +vi.mock('@react-native-async-storage/async-storage', () => ({ default: {} })) +vi.mock('react-native', () => ({ + StyleSheet: { create: (value: unknown) => value }, + View: 'View', + Text: 'Text', + Switch: 'Switch' +})) + +it('exposes independent event controls only after turning off desktop mirroring', () => { + const onChange = vi.fn() + let renderer: ReturnType + act(() => { + renderer = create( + createElement(NotificationDeliverySection, { value: DEFAULT_NOTIFICATION_DELIVERY, onChange }) + ) + }) + const switches = () => renderer.root.findAllByType('Switch' as never) + expect(switches().map((node) => node.props.accessibilityLabel)).toEqual([ + 'Use desktop settings', + 'Notification sound', + 'Suppress while viewing workspace' + ]) + act(() => switches()[0].props.onValueChange(false)) + const independent = onChange.mock.calls[0][0] + expect(independent.followDesktop).toBe(false) + act(() => + renderer.update(createElement(NotificationDeliverySection, { value: independent, onChange })) + ) + expect(switches().map((node) => node.props.accessibilityLabel)).toContain('Terminal bell') + act(() => + switches() + .find((node) => node.props.accessibilityLabel === 'Terminal bell')! + .props.onValueChange(false) + ) + expect(onChange).toHaveBeenLastCalledWith( + expect.objectContaining({ terminalBell: false, taskFinished: true, needsInput: true }) + ) + act(() => renderer.unmount()) +}) diff --git a/mobile/src/notifications/NotificationDeliverySection.tsx b/mobile/src/notifications/NotificationDeliverySection.tsx new file mode 100644 index 00000000000..5619eeaef84 --- /dev/null +++ b/mobile/src/notifications/NotificationDeliverySection.tsx @@ -0,0 +1,71 @@ +import { StyleSheet, Switch, Text, View } from 'react-native' +import { colors, radii, spacing, typography } from '../theme/mobile-theme' +import type { NotificationDeliveryPreferences } from './notification-delivery-preferences' + +type Props = { + value: NotificationDeliveryPreferences + disabled?: boolean + onChange: (value: NotificationDeliveryPreferences) => void +} + +export function NotificationDeliverySection({ value, disabled, onChange }: Props) { + const row = (key: keyof NotificationDeliveryPreferences, label: string) => ( + + {label} + onChange({ ...value, [key]: enabled })} + trackColor={{ false: colors.bgRaised, true: colors.textSecondary }} + thumbColor={colors.textPrimary} + /> + + ) + return ( + + {row('followDesktop', 'Use desktop settings')} + + {value.followDesktop + ? 'Follow each desktop’s notification and event switches. Desktop focus does not silence this phone.' + : 'Choose which alerts reach this phone, both while connected and in the background. Independent delivery requires an updated desktop.'} + + {!value.followDesktop && ( + <> + {row('taskFinished', 'Task finished')} + {row('needsInput', 'Needs input')} + {row('terminalBell', 'Terminal bell')} + + A program requests attention by sending a bell character. This can happen while an agent + is still working. + + {row('plugin', 'Plugin notifications')} + + )} + {row('sound', 'Notification sound')} + {row('suppressWhileViewing', 'Suppress while viewing workspace')} + + Sound and viewing preferences apply only to this phone. Changes reach disconnected desktops + when they reconnect. + + + ) +} + +const styles = StyleSheet.create({ + section: { + backgroundColor: colors.bgPanel, + borderRadius: radii.card, + overflow: 'hidden', + marginTop: spacing.md + }, + row: { flexDirection: 'row', alignItems: 'center', gap: spacing.sm, padding: spacing.md }, + label: { flex: 1, fontSize: typography.bodySize, fontWeight: '500', color: colors.textPrimary }, + hint: { + fontSize: typography.metaSize, + color: colors.textMuted, + paddingHorizontal: spacing.md, + paddingBottom: spacing.md + } +}) diff --git a/mobile/src/notifications/desktop-notification-channel.test.ts b/mobile/src/notifications/desktop-notification-channel.test.ts new file mode 100644 index 00000000000..c719157cf6b --- /dev/null +++ b/mobile/src/notifications/desktop-notification-channel.test.ts @@ -0,0 +1,62 @@ +import { readFileSync } from 'node:fs' +import { beforeEach, describe, expect, it, vi } from 'vitest' +import * as Notifications from 'expo-notifications' +import { Platform } from 'react-native' +import { + DESKTOP_NOTIFICATION_CHANNEL_ID, + ensureDesktopNotificationChannel +} from './desktop-notification-channel' + +vi.mock('expo-notifications', () => ({ + AndroidImportance: { HIGH: 'high' }, + setNotificationChannelAsync: vi.fn() +})) + +vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, + Platform: { OS: 'android' } +})) + +beforeEach(() => { + vi.clearAllMocks() + Object.assign(Platform, { OS: 'android' }) + vi.mocked(Notifications.setNotificationChannelAsync).mockResolvedValue(null as never) +}) + +describe('ensureDesktopNotificationChannel', () => { + it('creates the channel the gateway payload names', () => { + ensureDesktopNotificationChannel() + + expect(Notifications.setNotificationChannelAsync).toHaveBeenCalledWith( + 'orca-desktop', + expect.objectContaining({ importance: 'high' }) + ) + expect(DESKTOP_NOTIFICATION_CHANNEL_ID).toBe('orca-desktop') + }) + + it('does nothing on iOS, which has no notification channels', () => { + Object.assign(Platform, { OS: 'ios' }) + + ensureDesktopNotificationChannel() + + expect(Notifications.setNotificationChannelAsync).not.toHaveBeenCalled() + }) + + it('survives a shell whose channel API rejects', () => { + vi.mocked(Notifications.setNotificationChannelAsync).mockRejectedValue(new Error('no channels')) + + expect(() => ensureDesktopNotificationChannel()).not.toThrow() + }) +}) + +describe('app boot', () => { + it('creates the channel at startup, not only once a socket subscribes', () => { + // A background push can be the first thing to target 'orca-desktop', and Android + // drops a notification whose channel does not exist. Asserted against the source + // because vitest only collects src/, so app/_layout.tsx has no runtime coverage. + const layout = readFileSync(new URL('../../app/_layout.tsx', import.meta.url), 'utf8') + + expect(layout).toContain("from '../src/notifications/desktop-notification-channel'") + expect(layout).toMatch(/^ensureDesktopNotificationChannel\(\)$/m) + }) +}) diff --git a/mobile/src/notifications/desktop-notification-channel.ts b/mobile/src/notifications/desktop-notification-channel.ts new file mode 100644 index 00000000000..318c79f8bc4 --- /dev/null +++ b/mobile/src/notifications/desktop-notification-channel.ts @@ -0,0 +1,27 @@ +import * as Notifications from 'expo-notifications' +import { Platform } from 'react-native' + +// Why an id both sides share: the gateway's FCM payload names this channel, so a +// background push can be the first thing that ever targets it. Android drops a +// notification whose channel does not exist, and the channel used to be created +// only inside subscribeToDesktopNotifications — i.e. only once a socket connected. +export const DESKTOP_NOTIFICATION_CHANNEL_ID = 'orca-desktop' + +/** Idempotent on Android (the OS updates the existing channel); a no-op elsewhere. */ +export function ensureDesktopNotificationChannel(): void { + if (Platform.OS !== 'android') { + return + } + void Notifications.setNotificationChannelAsync(`${DESKTOP_NOTIFICATION_CHANNEL_ID}-silent`, { + name: 'Orca silent notifications', + importance: Notifications.AndroidImportance.HIGH, + sound: null, + enableVibrate: false + })?.catch(() => {}) + void Notifications.setNotificationChannelAsync(DESKTOP_NOTIFICATION_CHANNEL_ID, { + name: 'Desktop Notifications', + importance: Notifications.AndroidImportance.HIGH, + vibrationPattern: [0, 250], + lightColor: '#6366f1' + })?.catch(() => {}) +} diff --git a/mobile/src/notifications/local-notification-scheduling.ts b/mobile/src/notifications/local-notification-scheduling.ts index f511346250e..77a80a9a4f0 100644 --- a/mobile/src/notifications/local-notification-scheduling.ts +++ b/mobile/src/notifications/local-notification-scheduling.ts @@ -1,11 +1,19 @@ +import { reserveNotificationCooldown } from '../../../src/shared/notification-burst-cooldown' +import { loadNotificationDeliveryPreferences } from './notification-delivery-preferences' +import { allowsLocalNotification } from './notification-viewing-policy' import * as Notifications from 'expo-notifications' import { Platform } from 'react-native' import { loadPushNotificationsEnabled } from '../storage/preferences' +import { DESKTOP_NOTIFICATION_CHANNEL_ID } from './desktop-notification-channel' import { buildLocalNotificationData, type DesktopNotificationSource } from './notification-routing' import { ensureNotificationPermissions } from './notification-permissions' +import { dismissPresentedPushNotification } from './push-tray-dismissal' export type NotificationEvent = { type: 'notification' + desktopAllowed?: boolean + emittedAt?: number + agentState?: string source: DesktopNotificationSource title: string body: string @@ -30,6 +38,19 @@ type ScheduledNotificationState = { dismissAfterSchedule?: boolean } +const recentNotifications = new Map() + +function reserveLocalNotification(event: NotificationEvent, hostId: string): boolean { + return ( + event.emittedAt === undefined || + reserveNotificationCooldown( + recentNotifications, + JSON.stringify([hostId, event.worktreeId ?? 'global']), + event.emittedAt + ) + ) +} + const scheduledNotificationsByHostAndNotificationId = new Map() // Why: keys never repeat and are only freed on desktop dismiss (which remote users often miss), so bound the map to stop unbounded growth. @@ -62,21 +83,17 @@ export function setScheduledNotificationsMaxForTests(max?: number): void { maxScheduledNotifications = max ?? MAX_SCHEDULED_NOTIFICATIONS } -export function configureNotificationChannel(): void { - if (Platform.OS === 'android') { - void Notifications.setNotificationChannelAsync('orca-desktop', { - name: 'Desktop Notifications', - importance: Notifications.AndroidImportance.HIGH, - vibrationPattern: [0, 250], - lightColor: '#6366f1' - }) - } -} - export async function showLocalNotification( event: NotificationEvent, hostId: string ): Promise { + if (!(await allowsLocalNotification(event, hostId))) { + return + } + const preferences = await loadNotificationDeliveryPreferences() + const channelId = preferences.sound + ? DESKTOP_NOTIFICATION_CHANNEL_ID + : `${DESKTOP_NOTIFICATION_CHANNEL_ID}-silent` const storedKey = event.notificationId ? getStoredNotificationKey(hostId, event.notificationId) : null @@ -92,12 +109,16 @@ export async function showLocalNotification( return } + if (!reserveLocalNotification(event, hostId)) { + return + } await Notifications.scheduleNotificationAsync({ content: { title: event.title, body: event.body, + sound: preferences.sound ? 'default' : false, data: buildLocalNotificationData(event, hostId), - ...(Platform.OS === 'android' ? { channelId: 'orca-desktop' } : {}) + ...(Platform.OS === 'android' ? { channelId } : {}) }, trigger: null }) @@ -125,6 +146,9 @@ export async function showLocalNotification( return null } + if (!reserveLocalNotification(event, hostId)) { + return null + } if (notificationState.identifier) { await Notifications.dismissNotificationAsync(notificationState.identifier).catch(() => {}) notificationState.identifier = undefined @@ -134,8 +158,9 @@ export async function showLocalNotification( content: { title: event.title, body: event.body, + sound: preferences.sound ? 'default' : false, data: buildLocalNotificationData(event, hostId), - ...(Platform.OS === 'android' ? { channelId: 'orca-desktop' } : {}) + ...(Platform.OS === 'android' ? { channelId } : {}) }, trigger: null }) @@ -173,6 +198,9 @@ export async function dismissLocalNotification( if (!event.notificationId) { return } + // Why first and unconditionally: a push the OS presented while Orca was closed has + // no entry below, so the local registry alone would leave it in the tray forever. + await dismissPresentedPushNotification(event.notificationId) const storedKey = getStoredNotificationKey(hostId, event.notificationId) const state = scheduledNotificationsByHostAndNotificationId.get(storedKey) if (!state) { diff --git a/mobile/src/notifications/mobile-notifications.test.ts b/mobile/src/notifications/mobile-notifications.test.ts index d85b1363005..ad6189d1870 100644 --- a/mobile/src/notifications/mobile-notifications.test.ts +++ b/mobile/src/notifications/mobile-notifications.test.ts @@ -3,7 +3,6 @@ import * as Notifications from 'expo-notifications' import { Platform } from 'react-native' import { getNotificationPermissionState, - setScheduledNotificationsMaxForTests, subscribeToDesktopNotifications } from './mobile-notifications' import AsyncStorage from '@react-native-async-storage/async-storage' @@ -14,6 +13,7 @@ import { resetHostNotificationSessionsForTests } from './notification-reconnect- vi.mock('expo-notifications', () => ({ AndroidImportance: { HIGH: 'high' }, setNotificationChannelAsync: vi.fn(), + getPresentedNotificationsAsync: vi.fn(async () => []), getPermissionsAsync: vi.fn(), requestPermissionsAsync: vi.fn(), scheduleNotificationAsync: vi.fn(), @@ -21,9 +21,15 @@ vi.mock('expo-notifications', () => ({ })) vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, Platform: { OS: 'ios', Version: 18 } })) +// The reconnect catch-up reads the tray to learn which pushes the OS already showed, +// and mapping those to this host needs the catalog, whose real module pulls the +// native keychain. No push is presented in these tests, so an empty catalog is enough. +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn(async () => []) })) + // Why: mobile-notifications now persists the catch-up watermark to // AsyncStorage. The package isn't resolvable in the node test env (other // mobile tests mock it the same way), so we provide a no-op mock. @@ -35,6 +41,7 @@ vi.mock('@react-native-async-storage/async-storage', () => ({ })) vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(async () => false), loadPushNotificationsEnabled: vi.fn() })) @@ -68,303 +75,6 @@ describe('getNotificationPermissionState', () => { ) }) -describe('subscribeToDesktopNotifications', () => { - beforeEach(() => { - vi.clearAllMocks() - }) - - // Why the macrotask and not N microtask ticks (#8591): deliveries now run through - // the per-host serialization queue, so a delivery is several more `await` hops deep - // than it used to be and a fixed tick count silently under-drains. Yielding to the - // macrotask queue drains whatever depth the chain happens to have. - function flushAsync(): Promise { - return new Promise((resolve) => { - setTimeout(resolve, 0) - }) - } - - function makeDeferred(): { promise: Promise; resolve: (value: T) => void } { - let resolve!: (value: T) => void - const promise = new Promise((next) => { - resolve = next - }) - return { promise, resolve } - } - - it('drops the local stream when disposed before the desktop returns ready', () => { - const unsubscribeStream = vi.fn() - const client = { - subscribe: vi.fn(() => unsubscribeStream), - getState: vi.fn(() => 'connected'), - sendRequest: vi.fn() - } as unknown as RpcClient - - const unsubscribe = subscribeToDesktopNotifications(client, 'host-1') - unsubscribe() - - expect(unsubscribeStream).toHaveBeenCalledTimes(1) - expect(client.sendRequest).not.toHaveBeenCalled() - }) - - it('stores scheduled notification identifiers, replaces duplicates, and dismisses by id', async () => { - vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) - vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ - status: 'granted', - canAskAgain: true - } as never) - vi.mocked(Notifications.scheduleNotificationAsync) - .mockResolvedValueOnce('scheduled-1') - .mockResolvedValueOnce('scheduled-2') - vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) - let onEvent: ((data: unknown) => void) | null = null - const client = { - subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { - onEvent = callback - return vi.fn() - }), - getState: vi.fn(() => 'connected'), - sendRequest: vi.fn() - } as unknown as RpcClient - - subscribeToDesktopNotifications(client, 'host-1') - onEvent?.({ - type: 'notification', - source: 'agent-task-complete', - title: 'Done', - body: 'Finished.', - worktreeId: 'repo::/tmp/worktree', - notificationId: 'agent:one' - }) - await flushAsync() - onEvent?.({ - type: 'notification', - source: 'agent-task-complete', - title: 'Done again', - body: 'Finished again.', - notificationId: 'agent:one' - }) - await flushAsync() - expect(Notifications.scheduleNotificationAsync).toHaveBeenCalledTimes(2) - onEvent?.({ type: 'dismiss', notificationId: 'agent:one' }) - await flushAsync() - - expect(Notifications.scheduleNotificationAsync).toHaveBeenCalledTimes(2) - expect(Notifications.scheduleNotificationAsync).toHaveBeenNthCalledWith( - 1, - expect.objectContaining({ - content: expect.objectContaining({ - data: expect.objectContaining({ - hostId: 'host-1', - notificationId: 'agent:one', - worktreeId: 'repo::/tmp/worktree' - }) - }) - }) - ) - expect(Notifications.dismissNotificationAsync).toHaveBeenNthCalledWith(1, 'scheduled-1') - expect(Notifications.dismissNotificationAsync).toHaveBeenNthCalledWith(2, 'scheduled-2') - }) - - it('dedupes concurrent notification events with the same desktop notification id', async () => { - vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) - vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ - status: 'granted', - canAskAgain: true - } as never) - vi.mocked(Notifications.scheduleNotificationAsync).mockResolvedValue('scheduled-1') - let onEvent: ((data: unknown) => void) | null = null - const client = { - subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { - onEvent = callback - return vi.fn() - }), - getState: vi.fn(() => 'connected'), - sendRequest: vi.fn() - } as unknown as RpcClient - - subscribeToDesktopNotifications(client, 'host-concurrent') - onEvent?.({ - type: 'notification', - source: 'agent-task-complete', - title: 'Done', - body: 'Finished.', - notificationId: 'agent:concurrent' - }) - onEvent?.({ - type: 'notification', - source: 'agent-task-complete', - title: 'Done', - body: 'Finished.', - notificationId: 'agent:concurrent' - }) - await flushAsync() - - expect(Notifications.scheduleNotificationAsync).toHaveBeenCalledTimes(1) - }) - - it('dismisses a notification when dismiss arrives while scheduling is pending', async () => { - vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) - vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ - status: 'granted', - canAskAgain: true - } as never) - let resolveSchedule!: (identifier: string) => void - vi.mocked(Notifications.scheduleNotificationAsync).mockImplementation( - () => - new Promise((resolve) => { - resolveSchedule = resolve - }) - ) - vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) - let onEvent: ((data: unknown) => void) | null = null - const client = { - subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { - onEvent = callback - return vi.fn() - }), - getState: vi.fn(() => 'connected'), - sendRequest: vi.fn() - } as unknown as RpcClient - - subscribeToDesktopNotifications(client, 'host-dismiss-race') - onEvent?.({ - type: 'notification', - source: 'agent-task-complete', - title: 'Done', - body: 'Finished.', - notificationId: 'agent:pending' - }) - await flushAsync() - onEvent?.({ type: 'dismiss', notificationId: 'agent:pending' }) - resolveSchedule('scheduled-pending') - await flushAsync() - - expect(Notifications.dismissNotificationAsync).toHaveBeenCalledWith('scheduled-pending') - }) - - it('does not carry a failed pending dismiss into a future schedule', async () => { - const secondEnabled = makeDeferred() - vi.mocked(loadPushNotificationsEnabled) - .mockResolvedValueOnce(true) - .mockReturnValueOnce(secondEnabled.promise) - .mockResolvedValueOnce(true) - vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ - status: 'granted', - canAskAgain: true - } as never) - vi.mocked(Notifications.scheduleNotificationAsync) - .mockResolvedValueOnce('scheduled-1') - .mockResolvedValueOnce('scheduled-2') - vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) - let onEvent: ((data: unknown) => void) | null = null - const client = { - subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { - onEvent = callback - return vi.fn() - }), - getState: vi.fn(() => 'connected'), - sendRequest: vi.fn() - } as unknown as RpcClient - - subscribeToDesktopNotifications(client, 'host-dismiss-failed-replacement') - onEvent?.({ - type: 'notification', - source: 'agent-task-complete', - title: 'Done', - body: 'Finished.', - notificationId: 'agent:stale-dismiss' - }) - await flushAsync() - onEvent?.({ - type: 'notification', - source: 'agent-task-complete', - title: 'Done again', - body: 'Finished again.', - notificationId: 'agent:stale-dismiss' - }) - await flushAsync() - onEvent?.({ type: 'dismiss', notificationId: 'agent:stale-dismiss' }) - secondEnabled.resolve(false) - await flushAsync() - - onEvent?.({ - type: 'notification', - source: 'agent-task-complete', - title: 'Done later', - body: 'Finished later.', - notificationId: 'agent:stale-dismiss' - }) - await flushAsync() - - expect(Notifications.scheduleNotificationAsync).toHaveBeenCalledTimes(2) - expect(Notifications.dismissNotificationAsync).toHaveBeenCalledTimes(1) - expect(Notifications.dismissNotificationAsync).toHaveBeenCalledWith('scheduled-1') - }) - - it('treats unknown dismiss events as no-ops', async () => { - vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) - let onEvent: ((data: unknown) => void) | null = null - const client = { - subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { - onEvent = callback - return vi.fn() - }), - getState: vi.fn(() => 'connected'), - sendRequest: vi.fn() - } as unknown as RpcClient - - subscribeToDesktopNotifications(client, 'host-unknown') - onEvent?.({ type: 'dismiss', notificationId: 'agent:missing' }) - await flushAsync() - - expect(Notifications.dismissNotificationAsync).not.toHaveBeenCalled() - }) - - // Why: notificationId is unique per completion, so the map grew unbounded when - // the desktop never sent a dismiss (the remote-mobile case). It is now capped. - it('evicts the oldest scheduled entry once the cap is exceeded', async () => { - setScheduledNotificationsMaxForTests(1) - try { - vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) - vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ - status: 'granted', - canAskAgain: true - } as never) - vi.mocked(Notifications.scheduleNotificationAsync) - .mockResolvedValueOnce('scheduled-old') - .mockResolvedValueOnce('scheduled-new') - vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) - let onEvent: ((data: unknown) => void) | null = null - const client = { - subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { - onEvent = callback - return vi.fn() - }), - getState: vi.fn(() => 'connected'), - sendRequest: vi.fn() - } as unknown as RpcClient - - subscribeToDesktopNotifications(client, 'host-1') - onEvent?.({ type: 'notification', title: 't', body: 'b', notificationId: 'agent:old' }) - await flushAsync() - onEvent?.({ type: 'notification', title: 't', body: 'b', notificationId: 'agent:new' }) - await flushAsync() - - // The older entry was evicted by the cap: dismissing it is a no-op... - onEvent?.({ type: 'dismiss', notificationId: 'agent:old' }) - await flushAsync() - expect(Notifications.dismissNotificationAsync).not.toHaveBeenCalledWith('scheduled-old') - - // ...while the most-recent entry is retained and still dismissable. - onEvent?.({ type: 'dismiss', notificationId: 'agent:new' }) - await flushAsync() - expect(Notifications.dismissNotificationAsync).toHaveBeenCalledWith('scheduled-new') - } finally { - setScheduledNotificationsMaxForTests() - } - }) -}) - // Why: #8129 catch-up. On a reconnect the live stream re-emits `ready`; the // client must fetch missed notifications from its watermark and push exactly // the ones it had not yet delivered — never re-pushing an already-delivered id. @@ -452,6 +162,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { notifications: [ { type: 'notification', + source: 'agent-task-complete', title: 'missed', body: 'b', notificationId: 'agent:missed', @@ -471,6 +182,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { // Live stream already delivered agent:dup (seq 11) before reap. sub.onData?.({ type: 'notification', + source: 'agent-task-complete', title: 'dup', body: 'b', notificationId: 'agent:dup', @@ -486,7 +198,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { const missedCall = vi .mocked(sub.client.sendRequest) .mock.calls.find((c: unknown[]) => c[0] === 'notifications.getMissedSince') - expect(missedCall?.[1]).toEqual({ lastSeenSeq: 11 }) + expect(missedCall?.[1]).toEqual({ includeDesktopSuppressed: true, lastSeenSeq: 11 }) // Only agent:missed was pushed; agent:dup appears exactly once (live only). const scheduledIds = vi .mocked(Notifications.scheduleNotificationAsync) @@ -532,10 +244,18 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { .mock.calls.filter((c: unknown[]) => c[0] === 'notifications.getMissedSince') // The cold open catches up from its stored watermark against the SAME counter — // 57 is meaningful there, so it is the correct cut (#8591 second pass). - expect(missedCalls[0]?.[1]).toEqual({ lastSeenSeq: 57, epoch: 'epoch-before-restart' }) + expect(missedCalls[0]?.[1]).toEqual({ + includeDesktopSuppressed: true, + lastSeenSeq: 57, + epoch: 'epoch-before-restart' + }) // After the restart the watermark is reset to 0 and tagged with the live epoch — // not the stale 57, which would make `57 >= 2` true and kill catch-up silently. - expect(missedCalls.at(-1)?.[1]).toEqual({ lastSeenSeq: 0, epoch: 'epoch-after-restart' }) + expect(missedCalls.at(-1)?.[1]).toEqual({ + includeDesktopSuppressed: true, + lastSeenSeq: 0, + epoch: 'epoch-after-restart' + }) }) it('refuses to seed a stored watermark that lost the race to a newer live epoch', async () => { @@ -579,7 +299,11 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { const missedCall = vi .mocked(sub.client.sendRequest) .mock.calls.find((c: unknown[]) => c[0] === 'notifications.getMissedSince') - expect(missedCall?.[1]).toEqual({ lastSeenSeq: 0, epoch: 'epoch-after-restart' }) + expect(missedCall?.[1]).toEqual({ + includeDesktopSuppressed: true, + lastSeenSeq: 0, + epoch: 'epoch-after-restart' + }) }) it('keeps the persisted watermark when the desktop epoch is unchanged', async () => { @@ -609,7 +333,11 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { const missedCall = vi .mocked(sub.client.sendRequest) .mock.calls.find((c: unknown[]) => c[0] === 'notifications.getMissedSince') - expect(missedCall?.[1]).toEqual({ lastSeenSeq: 57, epoch: 'epoch-stable' }) + expect(missedCall?.[1]).toEqual({ + includeDesktopSuppressed: true, + lastSeenSeq: 57, + epoch: 'epoch-stable' + }) }) it('drops an already-seen id if a replay re-includes it (defense-in-depth)', async () => { @@ -631,6 +359,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { notifications: [ { type: 'notification', + source: 'agent-task-complete', title: 'dup', body: 'b', notificationId: 'agent:dup', @@ -638,6 +367,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { }, { type: 'notification', + source: 'agent-task-complete', title: 'new', body: 'b', notificationId: 'agent:new', @@ -656,6 +386,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { // Live stream delivered agent:dup (seq 11) before reap. sub.onData?.({ type: 'notification', + source: 'agent-task-complete', title: 'dup', body: 'b', notificationId: 'agent:dup', @@ -692,6 +423,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { // Live stream delivers seq 5. sub.onData?.({ type: 'notification', + source: 'agent-task-complete', title: 't', body: 'b', notificationId: 'agent:live', @@ -728,6 +460,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { notifications: [ { type: 'notification', + source: 'agent-task-complete', title: 'missed', body: 'b', notificationId: 'agent:missed', @@ -760,7 +493,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { const missedCalls = vi .mocked(sub.client.sendRequest) .mock.calls.filter((c: unknown[]) => c[0] === 'notifications.getMissedSince') - expect(missedCalls.at(-1)?.[1]).toEqual({ lastSeenSeq: 8 }) + expect(missedCalls.at(-1)?.[1]).toEqual({ includeDesktopSuppressed: true, lastSeenSeq: 8 }) }) it('replays a terminal bell at a seq the previous desktop counter already used', async () => { @@ -785,7 +518,15 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { ok: true, result: { epoch: 'epoch-B', - notifications: [{ type: 'notification', title: 'bell', body: 'B', notificationSeq: 1 }] + notifications: [ + { + type: 'notification', + source: 'agent-task-complete', + title: 'bell', + body: 'B', + notificationSeq: 1 + } + ] } } as never } @@ -796,7 +537,13 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { sub.onData?.({ type: 'ready', subscriptionId: 'sub-1', epoch: 'epoch-A' }) await flushAsync() // A live bell under epoch A — no notificationId, so its seen-key is `seq:1`. - sub.onData?.({ type: 'notification', title: 'bell', body: 'A', notificationSeq: 1 }) + sub.onData?.({ + type: 'notification', + source: 'agent-task-complete', + title: 'bell', + body: 'A', + notificationSeq: 1 + }) await flushAsync() expect(vi.mocked(Notifications.scheduleNotificationAsync).mock.calls.length).toBe(1) @@ -841,7 +588,11 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { .mocked(sub.client.sendRequest) .mock.calls.find((c: unknown[]) => c[0] === 'notifications.getMissedSince') // Must not be 57: that seq was never shown to belong to this counter. - expect(missedCall?.[1]).toEqual({ lastSeenSeq: 0, epoch: 'epoch-live' }) + expect(missedCall?.[1]).toEqual({ + includeDesktopSuppressed: true, + lastSeenSeq: 0, + epoch: 'epoch-live' + }) }) it('catches up on the FIRST connection after an upgrade, without a second ready', async () => { @@ -875,6 +626,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { notifications: [ { type: 'notification', + source: 'agent-task-complete', notificationId: 'missed-58', notificationSeq: 58, notificationEpoch: 'epoch-live', @@ -896,7 +648,11 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { .mocked(sub.client.sendRequest) .mock.calls.find((c: unknown[]) => c[0] === 'notifications.getMissedSince') // The single 'ready' must replay from the stored watermark, not skip it. - expect(missedCall?.[1]).toEqual({ lastSeenSeq: 57, epoch: 'epoch-live' }) + expect(missedCall?.[1]).toEqual({ + includeDesktopSuppressed: true, + lastSeenSeq: 57, + epoch: 'epoch-live' + }) // And the missed notification must actually reach the user. expect(vi.mocked(Notifications.scheduleNotificationAsync).mock.calls.length).toBe(1) }) @@ -944,6 +700,7 @@ describe('subscribeToDesktopNotifications — reconnect catch-up', () => { await flushAsync() sub.onData?.({ type: 'notification', + source: 'agent-task-complete', title: 't', body: 'b', notificationId: 'agent:x', diff --git a/mobile/src/notifications/mobile-notifications.ts b/mobile/src/notifications/mobile-notifications.ts index 0043762e3ec..1ab9c6fcd94 100644 --- a/mobile/src/notifications/mobile-notifications.ts +++ b/mobile/src/notifications/mobile-notifications.ts @@ -1,5 +1,5 @@ +import { waitForSocketPushHandoff } from './socket-push-delivery-handoff' import type { RpcClient } from '../transport/rpc-client' -// Re-exported so the existing importers (and their vi.mock paths) keep working. export { ensureNotificationPermissions, getNotificationPermissionState, @@ -7,12 +7,12 @@ export { } from './notification-permissions' export { setScheduledNotificationsMaxForTests } from './local-notification-scheduling' import { - configureNotificationChannel, dismissLocalNotification, showLocalNotification, type DismissNotificationEvent, type NotificationEvent } from './local-notification-scheduling' +import { ensureDesktopNotificationChannel } from './desktop-notification-channel' import { adoptNotificationEpoch, catchUpWatermarkSeq, @@ -26,6 +26,7 @@ import { seenKeyForEvent, shouldQueueShowForNotificationId } from './notification-reconnect-catchup' +import { markPresentedPushesSeen, readPresentedPushSeenKeys } from './push-tray-seen-seed' type SubscribeResult = { type: 'ready' @@ -34,14 +35,13 @@ type SubscribeResult = { epoch?: string } -// Per-connection subscription; a reconnect `ready` triggers watermarked catch-up (#8129) so already-pushed events aren't re-sent. export function subscribeToDesktopNotifications(client: RpcClient, hostId: string): () => void { - configureNotificationChannel() + ensureDesktopNotificationChannel() let subscriptionId: string | null = null let disposed = false - // Why (#8591): survives the unsubscribe/resubscribe the app performs on every - // socket drop, so a reconnect still knows its watermark and that it reconnected. + const deliveryAbort = new AbortController() + // Preserve the watermark across socket reconnects. const session = getHostNotificationSession(hostId) /** @@ -84,23 +84,26 @@ export function subscribeToDesktopNotifications(client: RpcClient, hostId: strin adoptNotificationEpoch(session, hostId, event.notificationEpoch) const epochAtDelivery = session.lastDeliveredEpoch if (type === 'notification') { - await showLocalNotification(event as NotificationEvent, hostId) + const show = await waitForSocketPushHandoff( + event as NotificationEvent, + hostId, + deliveryAbort.signal + ) + if (disposed) { + throw new Error('notification_subscription_disposed') + } + if (show) { + await showLocalNotification(event as NotificationEvent, hostId) + } } else { await dismissLocalNotification(event as DismissNotificationEvent, hostId) } - // Why after the await, exactly like the watermark below: `seen` asserts this event - // reached the user (#8129). Marked before, a rejected show leaves the key behind and - // every later replay is dropped as a duplicate — loss the quarantine cannot recover, - // since the first event to drain a batch lifts it past the one never shown. + // Claim only after local delivery or a matching presented push. const key = seenKeyForEvent(event) // A mid-flight epoch adoption already cleared the counter lifetime this key indexes. if (key && session.lastDeliveredEpoch === epochAtDelivery) { session.seen.add(key) } - // Why after the await (#8591): the watermark is a promise that everything up - // to this seq has been shown. Advancing it before the local notification lands - // means a process death in between silently drops it — the next launch asks the - // desktop for seq greater than one the user never saw. if (event.notificationSeq != null && event.notificationSeq > session.lastDeliveredSeq) { session.lastDeliveredSeq = event.notificationSeq // Why clamped: while a failed catch-up's range is still unrecovered, persisting @@ -113,12 +116,9 @@ export function subscribeToDesktopNotifications(client: RpcClient, hostId: strin } } - // Claimed inline rather than via queueDelivery: the batch is already one queue - // entry, and re-enqueueing per item is what let a live event cut in. async function deliverMissedEvent( event: NotificationEvent | DismissNotificationEvent ): Promise { - // No pre-marking here either: deliverLive marks the key once the show lands. const key = seenKeyForEvent(event) if (key && session.seen.has(key)) { return @@ -144,12 +144,14 @@ export function subscribeToDesktopNotifications(client: RpcClient, hostId: strin if (disposed) { return } - // Captured before the request: everything at or below it is known delivered, so - // it is the floor the watermark falls back to if this catch-up never completes. + // Preserve the delivered floor if catch-up fails. const askFrom = catchUpWatermarkSeq(session) + // Read concurrently; claim inside the queue after epoch adoption to avoid stale keys. + const presentedPushKeys = readPresentedPushSeenKeys(hostId) const missed = await client .sendRequest('notifications.getMissedSince', { lastSeenSeq: askFrom, + includeDesktopSuppressed: true, // Why: sending the epoch lets the desktop reject a watermark from a counter // it no longer has and return the whole retained buffer instead of nothing. ...(session.lastDeliveredEpoch != null ? { epoch: session.lastDeliveredEpoch } : {}) @@ -176,8 +178,7 @@ export function subscribeToDesktopNotifications(client: RpcClient, hostId: strin // request stays OUTSIDE the queue: sendRequest waits up to 30s, and holding the // chain for that would stall live delivery on a slow link. await enqueueHostDelivery(session, async () => { - // Advances only past events this batch settled, so a teardown or a failing show - // quarantines the true contiguous point instead of the range it never reached. + markPresentedPushesSeen(session, await presentedPushKeys) let contiguousSeq = askFrom let drained = false try { @@ -213,7 +214,8 @@ export function subscribeToDesktopNotifications(client: RpcClient, hostId: strin } } - const unsubscribeStream = client.subscribe('notifications.subscribe', {}, (data: unknown) => { + const params = { includeDesktopSuppressed: true } + const unsubscribeStream = client.subscribe('notifications.subscribe', params, (data: unknown) => { const event = data as | NotificationEvent | DismissNotificationEvent @@ -285,6 +287,7 @@ export function subscribeToDesktopNotifications(client: RpcClient, hostId: strin return () => { disposed = true + deliveryAbort.abort() // Why: drop the local stream first — readiness can race unmount; don't hold the callback while a subscription id is pending. unsubscribeStream() if (subscriptionId) { diff --git a/mobile/src/notifications/native-notification-data.test.ts b/mobile/src/notifications/native-notification-data.test.ts new file mode 100644 index 00000000000..2b157a5fda6 --- /dev/null +++ b/mobile/src/notifications/native-notification-data.test.ts @@ -0,0 +1,22 @@ +import { expect, it } from 'vitest' +import { readNativeNotificationData } from './native-notification-data' +import { readOrcaPushPayload } from './push-payload' + +it('reads actual Expo APNs payloads when content.data is null', () => { + const orca = { + hostFingerprint: 'qa-host', + notificationId: 'done', + notificationSeq: 4, + notificationEpoch: 'epoch' + } + const data = readNativeNotificationData({ + content: { data: null }, + trigger: { type: 'push', payload: { aps: {}, orca } } + }) + expect(readOrcaPushPayload(data)).toMatchObject(orca) +}) +it('keeps Android push and local notification data', () => { + const data = { hostId: 'host', notificationId: 'done' } + expect(readNativeNotificationData({ content: { data }, trigger: { type: 'push' } })).toBe(data) + expect(readNativeNotificationData({ content: { data }, trigger: null })).toBe(data) +}) diff --git a/mobile/src/notifications/native-notification-data.ts b/mobile/src/notifications/native-notification-data.ts new file mode 100644 index 00000000000..74d50397660 --- /dev/null +++ b/mobile/src/notifications/native-notification-data.ts @@ -0,0 +1,13 @@ +export function readNativeNotificationData(request: { + content: { data?: unknown } + trigger?: unknown +}): unknown { + const trigger = request.trigger + if (trigger && typeof trigger === 'object' && 'type' in trigger && trigger.type === 'push') { + // Expo iOS keeps raw APNs custom fields here when content.data is null. + if ('payload' in trigger && trigger.payload && typeof trigger.payload === 'object') { + return trigger.payload + } + } + return request.content.data +} diff --git a/mobile/src/notifications/notification-catchup-failure-quarantine.test.ts b/mobile/src/notifications/notification-catchup-failure-quarantine.test.ts index 997b9fce930..c9f6595f576 100644 --- a/mobile/src/notifications/notification-catchup-failure-quarantine.test.ts +++ b/mobile/src/notifications/notification-catchup-failure-quarantine.test.ts @@ -8,6 +8,7 @@ import { loadPushNotificationsEnabled } from '../storage/preferences' vi.mock('expo-notifications', () => ({ AndroidImportance: { HIGH: 'high' }, setNotificationChannelAsync: vi.fn(), + getPresentedNotificationsAsync: vi.fn(async () => []), getPermissionsAsync: vi.fn(), requestPermissionsAsync: vi.fn(), scheduleNotificationAsync: vi.fn(), @@ -15,9 +16,15 @@ vi.mock('expo-notifications', () => ({ })) vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, Platform: { OS: 'ios', Version: 18 } })) +// The reconnect catch-up reads the tray to learn which pushes the OS already showed, +// and mapping those to this host needs the catalog, whose real module pulls the +// native keychain. No push is presented in these tests, so an empty catalog is enough. +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn(async () => []) })) + const WATERMARK_KEY = 'orca:mobileNotificationsWatermark:host-1' const storage = new Map() @@ -31,6 +38,7 @@ vi.mock('@react-native-async-storage/async-storage', () => ({ })) vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(async () => false), loadPushNotificationsEnabled: vi.fn() })) @@ -68,7 +76,9 @@ function makeHostClient() { if (method !== 'notifications.getMissedSince') { return { ok: true, result: undefined } as never } - askedFrom.push((params as { lastSeenSeq: number }).lastSeenSeq) + askedFrom.push( + (params as { includeDesktopSuppressed: true; lastSeenSeq: number }).lastSeenSeq + ) if (outcome.kind === 'heldReject') { await new Promise((resolve) => { releaseHeld = resolve @@ -102,6 +112,7 @@ function makeHostClient() { function notification(seq: number) { return { type: 'notification', + source: 'agent-task-complete', title: `m${seq}`, body: 'b', notificationId: `agent:${seq}`, diff --git a/mobile/src/notifications/notification-delivery-ordering.test.ts b/mobile/src/notifications/notification-delivery-ordering.test.ts index 68d64d7b3de..5960c8c524d 100644 --- a/mobile/src/notifications/notification-delivery-ordering.test.ts +++ b/mobile/src/notifications/notification-delivery-ordering.test.ts @@ -8,6 +8,7 @@ import { loadPushNotificationsEnabled } from '../storage/preferences' vi.mock('expo-notifications', () => ({ AndroidImportance: { HIGH: 'high' }, setNotificationChannelAsync: vi.fn(), + getPresentedNotificationsAsync: vi.fn(async () => []), getPermissionsAsync: vi.fn(), requestPermissionsAsync: vi.fn(), scheduleNotificationAsync: vi.fn(), @@ -15,9 +16,15 @@ vi.mock('expo-notifications', () => ({ })) vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, Platform: { OS: 'ios', Version: 18 } })) +// The reconnect catch-up reads the tray to learn which pushes the OS already showed, +// and mapping those to this host needs the catalog, whose real module pulls the +// native keychain. No push is presented in these tests, so an empty catalog is enough. +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn(async () => []) })) + const WATERMARK_KEY = 'orca:mobileNotificationsWatermark:host-1' const storage = new Map() let getItemImpl: (key: string) => Promise = async (key) => storage.get(key) ?? null @@ -32,6 +39,7 @@ vi.mock('@react-native-async-storage/async-storage', () => ({ })) vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(async () => false), loadPushNotificationsEnabled: vi.fn() })) @@ -94,6 +102,7 @@ describe('#8591 per-host delivery ordering', () => { notifications: [ { type: 'notification', + source: 'agent-task-complete', title: 'm6', body: 'b', notificationId: 'a:6', @@ -101,6 +110,7 @@ describe('#8591 per-host delivery ordering', () => { }, { type: 'notification', + source: 'agent-task-complete', title: 'm7', body: 'b', notificationId: 'a:7', @@ -122,6 +132,7 @@ describe('#8591 per-host delivery ordering', () => { // Live seq 11 arrives while the replay is wedged on seq 6. onData?.({ type: 'notification', + source: 'agent-task-complete', title: 'live-11', body: 'b', notificationId: 'a:11', @@ -174,6 +185,7 @@ describe('#8591 per-host delivery ordering', () => { notifications: [ { type: 'notification', + source: 'agent-task-complete', title: 'dup', body: 'b', notificationId: 'agent:dup', @@ -196,6 +208,7 @@ describe('#8591 per-host delivery ordering', () => { // seq, so the seen-set does not catch it — only the queued-show claim does. onData?.({ type: 'notification', + source: 'agent-task-complete', title: 'dup', body: 'b', notificationId: 'agent:dup', @@ -212,7 +225,10 @@ describe('#8591 per-host delivery ordering', () => { it('still delivers when the persisted watermark read never resolves', async () => { // Every delivery awaits the seed, so a wedged AsyncStorage read would disable // this host's notifications for the whole app lifetime — silently. - getItemImpl = () => new Promise(() => {}) + getItemImpl = (key) => + key.startsWith('orca:mobileNotificationsWatermark:') + ? new Promise(() => {}) + : Promise.resolve(null) let onData: ((data: unknown) => void) | null = null const client = { @@ -230,6 +246,7 @@ describe('#8591 per-host delivery ordering', () => { onData?.({ type: 'ready', subscriptionId: 'sub-1', epoch: 'epoch-1' }) onData?.({ type: 'notification', + source: 'agent-task-complete', title: 'live-1', body: 'b', notificationId: 'a:1', diff --git a/mobile/src/notifications/notification-delivery-preferences.test.ts b/mobile/src/notifications/notification-delivery-preferences.test.ts new file mode 100644 index 00000000000..b6c38616fb9 --- /dev/null +++ b/mobile/src/notifications/notification-delivery-preferences.test.ts @@ -0,0 +1,87 @@ +import { beforeEach, expect, it, vi } from 'vitest' +import { AppState } from 'react-native' +import { + DEFAULT_NOTIFICATION_DELIVERY, + loadNotificationDeliveryPreferences, + notificationPreferencesFilter, + saveNotificationDeliveryPreferences +} from './notification-delivery-preferences' +import { + allowsLocalNotification, + setNotificationViewingWorkspace +} from './notification-viewing-policy' +import { allowsMobileNotification } from '../../../src/shared/mobile-notification-policy' + +const storage = new Map() +vi.mock('@react-native-async-storage/async-storage', () => ({ + default: { + getItem: vi.fn(async (key: string) => storage.get(key) ?? null), + setItem: vi.fn(async (key: string, value: string) => { + storage.set(key, value) + }) + } +})) +vi.mock('react-native', () => ({ AppState: { currentState: 'background' } })) +beforeEach(() => { + storage.clear() + setNotificationViewingWorkspace(null) + AppState.currentState = 'background' +}) + +it('defaults to following desktop and persists independent event preferences', async () => { + expect(await loadNotificationDeliveryPreferences()).toEqual(DEFAULT_NOTIFICATION_DELIVERY) + const value = { + ...DEFAULT_NOTIFICATION_DELIVERY, + followDesktop: false, + terminalBell: false, + sound: false + } + await saveNotificationDeliveryPreferences(value) + expect(await loadNotificationDeliveryPreferences()).toEqual(value) + expect(notificationPreferencesFilter(value)).toMatchObject({ + followDesktop: false, + sound: false, + sources: ['agent-task-complete', 'plugin'] + }) +}) + +it('preserves explicitly narrowed filters from before the new settings screen', async () => { + storage.set('orca:remotePushAgentStates', '["needs-input"]') + expect(await loadNotificationDeliveryPreferences()).toMatchObject({ + followDesktop: false, + needsInput: true, + taskFinished: false + }) +}) + +it.each(['agent-task-complete', 'terminal-bell', 'plugin'])( + 'uses identical type filtering for socket/replay and background push: %s', + async (source) => { + for (const followDesktop of [true, false]) { + const value = { + ...DEFAULT_NOTIFICATION_DELIVERY, + followDesktop, + terminalBell: false, + taskFinished: false + } + await saveNotificationDeliveryPreferences(value) + for (const desktopAllowed of [true, false]) { + const event = { source, desktopAllowed, agentState: 'done' } + expect(await allowsLocalNotification(event, 'host')).toBe( + allowsMobileNotification(notificationPreferencesFilter(value), event) + ) + } + } + } +) + +it('suppresses only the workspace being viewed on this phone, and never while backgrounded', async () => { + const event = { source: 'terminal-bell', worktreeId: 'folder-id' } + setNotificationViewingWorkspace({ hostId: 'ssh-host', worktreeId: 'folder-id' }) + AppState.currentState = 'active' + expect(await allowsLocalNotification(event, 'ssh-host')).toBe(false) + expect(await allowsLocalNotification(event, 'another-host')).toBe(true) + expect(await allowsLocalNotification({ ...event, worktreeId: 'other' }, 'ssh-host')).toBe(true) + AppState.currentState = 'background' + expect(await allowsLocalNotification(event, 'ssh-host')).toBe(true) +}) diff --git a/mobile/src/notifications/notification-delivery-preferences.ts b/mobile/src/notifications/notification-delivery-preferences.ts new file mode 100644 index 00000000000..ad56e3ff6b0 --- /dev/null +++ b/mobile/src/notifications/notification-delivery-preferences.ts @@ -0,0 +1,88 @@ +import AsyncStorage from '@react-native-async-storage/async-storage' +import { + MOBILE_PUSH_AGENT_STATES, + MOBILE_PUSH_SOURCES, + type MobilePushFilter +} from '../../../src/shared/mobile-push-contract' + +const KEY = 'orca:notificationDeliveryPreferences' +export type NotificationDeliveryPreferences = { + followDesktop: boolean + taskFinished: boolean + needsInput: boolean + terminalBell: boolean + plugin: boolean + sound: boolean + suppressWhileViewing: boolean +} + +export const DEFAULT_NOTIFICATION_DELIVERY: NotificationDeliveryPreferences = { + followDesktop: true, + taskFinished: true, + needsInput: true, + terminalBell: true, + plugin: true, + sound: true, + suppressWhileViewing: true +} + +export async function loadNotificationDeliveryPreferences(): Promise { + const raw = await AsyncStorage.getItem(KEY) + if (!raw) { + // Preserve an existing explicit background filter when upgrading. + const legacy = await AsyncStorage.getItem('orca:remotePushAgentStates') + if (legacy) { + const states: unknown = JSON.parse(legacy) + if (Array.isArray(states)) { + return { + ...DEFAULT_NOTIFICATION_DELIVERY, + followDesktop: false, + taskFinished: states.includes('finished'), + needsInput: states.includes('needs-input') + } + } + } + return { ...DEFAULT_NOTIFICATION_DELIVERY } + } + const stored = JSON.parse(raw) as Record + const result = { ...DEFAULT_NOTIFICATION_DELIVERY } + for (const key of Object.keys(result) as (keyof NotificationDeliveryPreferences)[]) { + if (typeof stored?.[key] === 'boolean') { + result[key] = stored[key] + } + } + return result +} + +export async function saveNotificationDeliveryPreferences( + value: NotificationDeliveryPreferences +): Promise { + await AsyncStorage.setItem(KEY, JSON.stringify(value)) +} + +export function notificationPreferencesFilter( + value: NotificationDeliveryPreferences +): MobilePushFilter { + if (value.followDesktop) { + return { + sound: value.sound, + followDesktop: true, + sources: MOBILE_PUSH_SOURCES, + agentStates: MOBILE_PUSH_AGENT_STATES + } + } + return { + followDesktop: false, + sound: value.sound, + sources: MOBILE_PUSH_SOURCES.filter((source) => + source === 'terminal-bell' + ? value.terminalBell + : source === 'plugin' + ? value.plugin + : value.needsInput || value.taskFinished + ), + agentStates: MOBILE_PUSH_AGENT_STATES.filter((state) => + state === 'needs-input' ? value.needsInput : value.taskFinished + ) + } +} diff --git a/mobile/src/notifications/notification-local-delivery.test.ts b/mobile/src/notifications/notification-local-delivery.test.ts new file mode 100644 index 00000000000..18c19daba7d --- /dev/null +++ b/mobile/src/notifications/notification-local-delivery.test.ts @@ -0,0 +1,211 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import * as Notifications from 'expo-notifications' +import AsyncStorage from '@react-native-async-storage/async-storage' +import { showLocalNotification } from './local-notification-scheduling' +import { Platform } from 'react-native' +import { subscribeToDesktopNotifications } from './mobile-notifications' +import type { RpcClient } from '../transport/rpc-client' +import { loadPushNotificationsEnabled } from '../storage/preferences' +import { resetHostNotificationSessionsForTests } from './notification-reconnect-catchup' + +vi.mock('expo-notifications', () => ({ + AndroidImportance: { HIGH: 'high' }, + setNotificationChannelAsync: vi.fn(), + getPresentedNotificationsAsync: vi.fn(async () => []), + getPermissionsAsync: vi.fn(), + requestPermissionsAsync: vi.fn(), + scheduleNotificationAsync: vi.fn(), + dismissNotificationAsync: vi.fn() +})) + +vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, + Platform: { OS: 'ios', Version: 18 } +})) + +// The reconnect catch-up reads the tray to learn which pushes the OS already showed, +// and mapping those to this host needs the catalog, whose real module pulls the +// native keychain. No push is presented in these tests, so an empty catalog is enough. +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn(async () => []) })) + +// Why: mobile-notifications now persists the catch-up watermark to +// AsyncStorage. The package isn't resolvable in the node test env (other +// mobile tests mock it the same way), so we provide a no-op mock. +vi.mock('@react-native-async-storage/async-storage', () => ({ + default: { + getItem: vi.fn(async () => null), + setItem: vi.fn(async () => undefined) + } +})) + +vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(async () => false), + loadPushNotificationsEnabled: vi.fn() +})) + +beforeEach(() => { + Object.assign(Platform, { OS: 'ios', Version: 18 }) + // Why (#8591): the reconnect watermark/seen-set now live per host at module + // scope so they survive the app's unsubscribe-on-disconnect. Reset between + // tests so each case starts from a genuine cold open. + resetHostNotificationSessionsForTests() +}) + +describe('subscribeToDesktopNotifications', () => { + beforeEach(() => { + vi.clearAllMocks() + }) + + // Why the macrotask and not N microtask ticks (#8591): deliveries now run through + // the per-host serialization queue, so a delivery is several more `await` hops deep + // than it used to be and a fixed tick count silently under-drains. Yielding to the + // macrotask queue drains whatever depth the chain happens to have. + function flushAsync(): Promise { + return new Promise((resolve) => { + setTimeout(resolve, 0) + }) + } + + it('drops the local stream when disposed before the desktop returns ready', () => { + const unsubscribeStream = vi.fn() + const client = { + subscribe: vi.fn(() => unsubscribeStream), + getState: vi.fn(() => 'connected'), + sendRequest: vi.fn() + } as unknown as RpcClient + + const unsubscribe = subscribeToDesktopNotifications(client, 'host-1') + unsubscribe() + + expect(unsubscribeStream).toHaveBeenCalledTimes(1) + expect(client.sendRequest).not.toHaveBeenCalled() + }) + + it('stores scheduled notification identifiers, replaces duplicates, and dismisses by id', async () => { + vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) + vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ + status: 'granted', + canAskAgain: true + } as never) + vi.mocked(Notifications.scheduleNotificationAsync) + .mockResolvedValueOnce('scheduled-1') + .mockResolvedValueOnce('scheduled-2') + vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) + let onEvent: ((data: unknown) => void) | null = null + const client = { + subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { + onEvent = callback + return vi.fn() + }), + getState: vi.fn(() => 'connected'), + sendRequest: vi.fn() + } as unknown as RpcClient + + subscribeToDesktopNotifications(client, 'host-1') + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 'Done', + body: 'Finished.', + worktreeId: 'repo::/tmp/worktree', + notificationId: 'agent:one' + }) + await flushAsync() + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 'Done again', + body: 'Finished again.', + notificationId: 'agent:one' + }) + await flushAsync() + expect(Notifications.scheduleNotificationAsync).toHaveBeenCalledTimes(2) + onEvent?.({ type: 'dismiss', notificationId: 'agent:one' }) + await flushAsync() + + expect(Notifications.scheduleNotificationAsync).toHaveBeenCalledTimes(2) + expect(Notifications.scheduleNotificationAsync).toHaveBeenNthCalledWith( + 1, + expect.objectContaining({ + content: expect.objectContaining({ + data: expect.objectContaining({ + hostId: 'host-1', + notificationId: 'agent:one', + worktreeId: 'repo::/tmp/worktree' + }) + }) + }) + ) + expect(Notifications.dismissNotificationAsync).toHaveBeenNthCalledWith(1, 'scheduled-1') + expect(Notifications.dismissNotificationAsync).toHaveBeenNthCalledWith(2, 'scheduled-2') + }) + + it('dedupes concurrent notification events with the same desktop notification id', async () => { + vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) + vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ + status: 'granted', + canAskAgain: true + } as never) + vi.mocked(Notifications.scheduleNotificationAsync).mockResolvedValue('scheduled-1') + let onEvent: ((data: unknown) => void) | null = null + const client = { + subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { + onEvent = callback + return vi.fn() + }), + getState: vi.fn(() => 'connected'), + sendRequest: vi.fn() + } as unknown as RpcClient + + subscribeToDesktopNotifications(client, 'host-concurrent') + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 'Done', + body: 'Finished.', + notificationId: 'agent:concurrent' + }) + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 'Done', + body: 'Finished.', + notificationId: 'agent:concurrent' + }) + await flushAsync() + + expect(Notifications.scheduleNotificationAsync).toHaveBeenCalledTimes(1) + }) +}) + +it('filters before cooldown and retains the existing banner when a later burst is suppressed', async () => { + vi.clearAllMocks() + vi.mocked(AsyncStorage.getItem).mockResolvedValue( + JSON.stringify({ followDesktop: false, terminalBell: false }) + ) + vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) + vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ + status: 'granted', + canAskAgain: true + } as never) + vi.mocked(Notifications.scheduleNotificationAsync).mockResolvedValue('cooldown-banner') + const event = { + type: 'notification' as const, + title: 'Done', + body: '', + worktreeId: 'folder', + notificationId: 'cooldown-event', + emittedAt: 10000 + } + await showLocalNotification({ ...event, source: 'terminal-bell' }, 'cooldown-host') + await showLocalNotification( + { ...event, source: 'agent-task-complete', agentState: 'done', emittedAt: 10250 }, + 'cooldown-host' + ) + await showLocalNotification( + { ...event, source: 'agent-task-complete', agentState: 'done', emittedAt: 10500 }, + 'cooldown-host' + ) + expect(Notifications.scheduleNotificationAsync).toHaveBeenCalledTimes(1) + expect(Notifications.dismissNotificationAsync).not.toHaveBeenCalled() +}) diff --git a/mobile/src/notifications/notification-local-dismissal.test.ts b/mobile/src/notifications/notification-local-dismissal.test.ts new file mode 100644 index 00000000000..74a700d2f0d --- /dev/null +++ b/mobile/src/notifications/notification-local-dismissal.test.ts @@ -0,0 +1,251 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import * as Notifications from 'expo-notifications' +import { Platform } from 'react-native' +import { + setScheduledNotificationsMaxForTests, + subscribeToDesktopNotifications +} from './mobile-notifications' +import type { RpcClient } from '../transport/rpc-client' +import { loadPushNotificationsEnabled } from '../storage/preferences' +import { resetHostNotificationSessionsForTests } from './notification-reconnect-catchup' + +vi.mock('expo-notifications', () => ({ + AndroidImportance: { HIGH: 'high' }, + setNotificationChannelAsync: vi.fn(), + getPresentedNotificationsAsync: vi.fn(async () => []), + getPermissionsAsync: vi.fn(), + requestPermissionsAsync: vi.fn(), + scheduleNotificationAsync: vi.fn(), + dismissNotificationAsync: vi.fn() +})) + +vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, + Platform: { OS: 'ios', Version: 18 } +})) + +// The reconnect catch-up reads the tray to learn which pushes the OS already showed, +// and mapping those to this host needs the catalog, whose real module pulls the +// native keychain. No push is presented in these tests, so an empty catalog is enough. +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn(async () => []) })) + +// Why: mobile-notifications now persists the catch-up watermark to +// AsyncStorage. The package isn't resolvable in the node test env (other +// mobile tests mock it the same way), so we provide a no-op mock. +vi.mock('@react-native-async-storage/async-storage', () => ({ + default: { + getItem: vi.fn(async () => null), + setItem: vi.fn(async () => undefined) + } +})) + +vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(async () => false), + loadPushNotificationsEnabled: vi.fn() +})) + +beforeEach(() => { + Object.assign(Platform, { OS: 'ios', Version: 18 }) + // Why (#8591): the reconnect watermark/seen-set now live per host at module + // scope so they survive the app's unsubscribe-on-disconnect. Reset between + // tests so each case starts from a genuine cold open. + resetHostNotificationSessionsForTests() +}) + +describe('subscribeToDesktopNotifications', () => { + beforeEach(() => { + vi.clearAllMocks() + }) + + // Why the macrotask and not N microtask ticks (#8591): deliveries now run through + // the per-host serialization queue, so a delivery is several more `await` hops deep + // than it used to be and a fixed tick count silently under-drains. Yielding to the + // macrotask queue drains whatever depth the chain happens to have. + function flushAsync(): Promise { + return new Promise((resolve) => { + setTimeout(resolve, 0) + }) + } + + function makeDeferred(): { promise: Promise; resolve: (value: T) => void } { + let resolve!: (value: T) => void + const promise = new Promise((next) => { + resolve = next + }) + return { promise, resolve } + } + + it('dismisses a notification when dismiss arrives while scheduling is pending', async () => { + vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) + vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ + status: 'granted', + canAskAgain: true + } as never) + let resolveSchedule!: (identifier: string) => void + vi.mocked(Notifications.scheduleNotificationAsync).mockImplementation( + () => + new Promise((resolve) => { + resolveSchedule = resolve + }) + ) + vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) + let onEvent: ((data: unknown) => void) | null = null + const client = { + subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { + onEvent = callback + return vi.fn() + }), + getState: vi.fn(() => 'connected'), + sendRequest: vi.fn() + } as unknown as RpcClient + + subscribeToDesktopNotifications(client, 'host-dismiss-race') + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 'Done', + body: 'Finished.', + notificationId: 'agent:pending' + }) + await flushAsync() + onEvent?.({ type: 'dismiss', notificationId: 'agent:pending' }) + resolveSchedule('scheduled-pending') + await flushAsync() + + expect(Notifications.dismissNotificationAsync).toHaveBeenCalledWith('scheduled-pending') + }) + + it('does not carry a failed pending dismiss into a future schedule', async () => { + const secondEnabled = makeDeferred() + vi.mocked(loadPushNotificationsEnabled) + .mockResolvedValueOnce(true) + .mockReturnValueOnce(secondEnabled.promise) + .mockResolvedValueOnce(true) + vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ + status: 'granted', + canAskAgain: true + } as never) + vi.mocked(Notifications.scheduleNotificationAsync) + .mockResolvedValueOnce('scheduled-1') + .mockResolvedValueOnce('scheduled-2') + vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) + let onEvent: ((data: unknown) => void) | null = null + const client = { + subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { + onEvent = callback + return vi.fn() + }), + getState: vi.fn(() => 'connected'), + sendRequest: vi.fn() + } as unknown as RpcClient + + subscribeToDesktopNotifications(client, 'host-dismiss-failed-replacement') + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 'Done', + body: 'Finished.', + notificationId: 'agent:stale-dismiss' + }) + await flushAsync() + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 'Done again', + body: 'Finished again.', + notificationId: 'agent:stale-dismiss' + }) + await flushAsync() + onEvent?.({ type: 'dismiss', notificationId: 'agent:stale-dismiss' }) + secondEnabled.resolve(false) + await flushAsync() + + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 'Done later', + body: 'Finished later.', + notificationId: 'agent:stale-dismiss' + }) + await flushAsync() + + expect(Notifications.scheduleNotificationAsync).toHaveBeenCalledTimes(2) + expect(Notifications.dismissNotificationAsync).toHaveBeenCalledTimes(1) + expect(Notifications.dismissNotificationAsync).toHaveBeenCalledWith('scheduled-1') + }) + + it('treats unknown dismiss events as no-ops', async () => { + vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) + let onEvent: ((data: unknown) => void) | null = null + const client = { + subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { + onEvent = callback + return vi.fn() + }), + getState: vi.fn(() => 'connected'), + sendRequest: vi.fn() + } as unknown as RpcClient + + subscribeToDesktopNotifications(client, 'host-unknown') + onEvent?.({ type: 'dismiss', notificationId: 'agent:missing' }) + await flushAsync() + + expect(Notifications.dismissNotificationAsync).not.toHaveBeenCalled() + }) + + // Why: notificationId is unique per completion, so the map grew unbounded when + // the desktop never sent a dismiss (the remote-mobile case). It is now capped. + it('evicts the oldest scheduled entry once the cap is exceeded', async () => { + setScheduledNotificationsMaxForTests(1) + try { + vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) + vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ + status: 'granted', + canAskAgain: true + } as never) + vi.mocked(Notifications.scheduleNotificationAsync) + .mockResolvedValueOnce('scheduled-old') + .mockResolvedValueOnce('scheduled-new') + vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) + let onEvent: ((data: unknown) => void) | null = null + const client = { + subscribe: vi.fn((_method, _params, callback: (data: unknown) => void) => { + onEvent = callback + return vi.fn() + }), + getState: vi.fn(() => 'connected'), + sendRequest: vi.fn() + } as unknown as RpcClient + + subscribeToDesktopNotifications(client, 'host-1') + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 't', + body: 'b', + notificationId: 'agent:old' + }) + await flushAsync() + onEvent?.({ + type: 'notification', + source: 'agent-task-complete', + title: 't', + body: 'b', + notificationId: 'agent:new' + }) + await flushAsync() + + // The older entry was evicted by the cap: dismissing it is a no-op... + onEvent?.({ type: 'dismiss', notificationId: 'agent:old' }) + await flushAsync() + expect(Notifications.dismissNotificationAsync).not.toHaveBeenCalledWith('scheduled-old') + + // ...while the most-recent entry is retained and still dismissable. + onEvent?.({ type: 'dismiss', notificationId: 'agent:new' }) + await flushAsync() + expect(Notifications.dismissNotificationAsync).toHaveBeenCalledWith('scheduled-new') + } finally { + setScheduledNotificationsMaxForTests() + } + }) +}) diff --git a/mobile/src/notifications/notification-reconnect-teardown.test.ts b/mobile/src/notifications/notification-reconnect-teardown.test.ts index a5e7433bf0f..a291982245b 100644 --- a/mobile/src/notifications/notification-reconnect-teardown.test.ts +++ b/mobile/src/notifications/notification-reconnect-teardown.test.ts @@ -9,6 +9,7 @@ import { loadPushNotificationsEnabled } from '../storage/preferences' vi.mock('expo-notifications', () => ({ AndroidImportance: { HIGH: 'high' }, setNotificationChannelAsync: vi.fn(), + getPresentedNotificationsAsync: vi.fn(async () => []), getPermissionsAsync: vi.fn(), requestPermissionsAsync: vi.fn(), scheduleNotificationAsync: vi.fn(), @@ -16,9 +17,15 @@ vi.mock('expo-notifications', () => ({ })) vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, Platform: { OS: 'ios', Version: 18 } })) +// The reconnect catch-up reads the tray to learn which pushes the OS already showed, +// and mapping those to this host needs the catalog, whose real module pulls the +// native keychain. No push is presented in these tests, so an empty catalog is enough. +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn(async () => []) })) + // In-memory AsyncStorage so the persisted watermark survives across the // subscribe/unsubscribe cycles this test exercises (the real device behaviour). const storage = new Map() @@ -32,6 +39,7 @@ vi.mock('@react-native-async-storage/async-storage', () => ({ })) vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(async () => false), loadPushNotificationsEnabled: vi.fn() })) @@ -46,7 +54,7 @@ function flushAsync(): Promise { // scratch on the next 'connected'. function makeHostClient() { let onData: ((data: unknown) => void) | null = null - const getMissedCalls: { lastSeenSeq: number }[] = [] + const getMissedCalls: { includeDesktopSuppressed: true; lastSeenSeq: number }[] = [] const client = { subscribe: vi.fn((_m: string, _p: unknown, cb: (data: unknown) => void) => { onData = cb @@ -57,7 +65,7 @@ function makeHostClient() { getState: vi.fn(() => 'connected'), sendRequest: vi.fn(async (method: string, params: unknown = {}) => { if (method === 'notifications.getMissedSince') { - getMissedCalls.push(params as { lastSeenSeq: number }) + getMissedCalls.push(params as { includeDesktopSuppressed: true; lastSeenSeq: number }) return { ok: true, result: { notifications: missedQueue } } as never } return { ok: true, result: undefined } as never @@ -100,6 +108,7 @@ describe('#8591 reconnect catch-up under the real app teardown lifecycle', () => await flushAsync() host.onData?.({ type: 'notification', + source: 'agent-task-complete', title: 'live', body: 'b', notificationId: 'agent:live', @@ -117,6 +126,7 @@ describe('#8591 reconnect catch-up under the real app teardown lifecycle', () => host.setMissed([ { type: 'notification', + source: 'agent-task-complete', title: 'missed-8', body: 'b', notificationId: 'agent:m8', @@ -124,6 +134,7 @@ describe('#8591 reconnect catch-up under the real app teardown lifecycle', () => }, { type: 'notification', + source: 'agent-task-complete', title: 'missed-9', body: 'b', notificationId: 'agent:m9', @@ -139,7 +150,7 @@ describe('#8591 reconnect catch-up under the real app teardown lifecycle', () => // The user must be told about seq 8 and 9. Nothing else can deliver them: // the desktop only fans out live, so this catch-up is the only path. expect(host.getMissedCalls).toHaveLength(1) - expect(host.getMissedCalls[0]).toEqual({ lastSeenSeq: 7 }) + expect(host.getMissedCalls[0]).toEqual({ includeDesktopSuppressed: true, lastSeenSeq: 7 }) const titles = vi .mocked(Notifications.scheduleNotificationAsync) .mock.calls.map((c) => (c[0] as { content: { title: string } }).content.title) @@ -160,6 +171,7 @@ describe('#8591 reconnect catch-up under the real app teardown lifecycle', () => await flushAsync() host.onData?.({ type: 'notification', + source: 'agent-task-complete', title: 'live-7', body: 'b', notificationId: 'agent:seven', @@ -174,6 +186,7 @@ describe('#8591 reconnect catch-up under the real app teardown lifecycle', () => host.setMissed([ { type: 'notification', + source: 'agent-task-complete', title: 'live-7', body: 'b', notificationId: 'agent:seven', @@ -181,6 +194,7 @@ describe('#8591 reconnect catch-up under the real app teardown lifecycle', () => }, { type: 'notification', + source: 'agent-task-complete', title: 'missed-8', body: 'b', notificationId: 'agent:m8', diff --git a/mobile/src/notifications/notification-reopen-push-duplicate.test.ts b/mobile/src/notifications/notification-reopen-push-duplicate.test.ts new file mode 100644 index 00000000000..e6a0bd9287f --- /dev/null +++ b/mobile/src/notifications/notification-reopen-push-duplicate.test.ts @@ -0,0 +1,204 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import * as Notifications from 'expo-notifications' +import { sha256 } from '@noble/hashes/sha256' +import { loadHostCatalog } from '../transport/host-store' +import type { HostCatalogEntry } from '../transport/types' +import type { RpcClient } from '../transport/rpc-client' +import { loadPushNotificationsEnabled } from '../storage/preferences' +import { subscribeToDesktopNotifications } from './mobile-notifications' +import { resetHostNotificationSessionsForTests } from './notification-reconnect-catchup' + +// Why this file exists: a push the OS drew while Orca was closed never runs through +// the foreground handler, so nothing marks it seen. The reconnect catch-up then +// replays the same event and the user gets a second banner for it. + +vi.mock('expo-notifications', () => ({ + AndroidImportance: { HIGH: 'high' }, + setNotificationChannelAsync: vi.fn(), + getPresentedNotificationsAsync: vi.fn(async () => []), + getPermissionsAsync: vi.fn(), + requestPermissionsAsync: vi.fn(), + scheduleNotificationAsync: vi.fn(), + dismissNotificationAsync: vi.fn() +})) + +vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, + Platform: { OS: 'ios', Version: 18 } +})) + +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn() })) + +const WATERMARK_KEY = 'orca:mobileNotificationsWatermark:host-1' +const storage = new Map() + +vi.mock('@react-native-async-storage/async-storage', () => ({ + default: { + getItem: vi.fn(async (key: string) => storage.get(key) ?? null), + setItem: vi.fn(async (key: string, value: string) => { + storage.set(key, value) + }) + } +})) + +vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(async () => false), + loadPushNotificationsEnabled: vi.fn() +})) + +const publicKey = Uint8Array.from({ length: 32 }, (_, index) => index) +const publicKeyB64 = Buffer.from(publicKey).toString('base64') +const hostFingerprint = Buffer.from(sha256(publicKey)).toString('base64url').slice(0, 16) + +function flushAsync(): Promise { + return new Promise((resolve) => { + setTimeout(resolve, 10) + }) +} + +function presentTray(entries: readonly Record[]): void { + vi.mocked(Notifications.getPresentedNotificationsAsync).mockResolvedValue( + entries.map((orca, index) => ({ + request: { + identifier: `tray-${index}`, + content: { data: null }, + trigger: { type: 'push', payload: { orca } } + } + })) as never + ) +} + +function shownTitles(): string[] { + return vi + .mocked(Notifications.scheduleNotificationAsync) + .mock.calls.map((call) => (call[0] as { content: { title: string } }).content.title) +} + +function persistedSeq(): number { + return (JSON.parse(storage.get(WATERMARK_KEY) ?? '{}') as { seq?: number }).seq ?? 0 +} + +/** A catch-up that replays seq 6 and 7 for host-1. */ +function catchUpClient(): { client: RpcClient; ready: () => void } { + let onData: ((data: unknown) => void) | null = null + const client = { + subscribe: vi.fn((_method: string, _params: unknown, callback: (data: unknown) => void) => { + onData = callback + return vi.fn() + }), + getState: vi.fn(() => 'connected'), + sendRequest: vi.fn(async (method: string) => { + if (method === 'notifications.getMissedSince') { + return { + ok: true, + result: { + notifications: [ + { + type: 'notification', + source: 'agent-task-complete', + title: 'm6', + body: 'b', + notificationId: 'a:6', + notificationSeq: 6 + }, + { + type: 'notification', + source: 'agent-task-complete', + title: 'm7', + body: 'b', + notificationId: 'a:7', + notificationSeq: 7 + } + ] + } + } as never + } + return { ok: true, result: undefined } as never + }) + } as unknown as RpcClient + return { + client, + ready: () => onData?.({ type: 'ready', subscriptionId: 'sub-1', epoch: 'epoch-1' }) + } +} + +async function reopenWithTray(): Promise { + storage.set(WATERMARK_KEY, JSON.stringify({ seq: 5, epoch: 'epoch-1' })) + const { client, ready } = catchUpClient() + subscribeToDesktopNotifications(client, 'host-1') + ready() + await flushAsync() +} + +beforeEach(() => { + vi.clearAllMocks() + storage.clear() + resetHostNotificationSessionsForTests() + vi.mocked(loadHostCatalog).mockResolvedValue([ + { id: 'host-1', publicKeyB64 } + ] as unknown as HostCatalogEntry[]) + vi.mocked(loadPushNotificationsEnabled).mockResolvedValue(true) + vi.mocked(Notifications.getPermissionsAsync).mockResolvedValue({ + status: 'granted', + canAskAgain: true + } as never) + vi.mocked(Notifications.scheduleNotificationAsync).mockResolvedValue('sched-1') + vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) + vi.mocked(Notifications.getPresentedNotificationsAsync).mockResolvedValue([]) +}) + +describe('reopen after a push the OS showed while Orca was closed', () => { + it('replays only the events still missing from the tray', async () => { + presentTray([ + { hostFingerprint, notificationId: 'a:6', notificationSeq: 6, notificationEpoch: 'epoch-1' } + ]) + + await reopenWithTray() + + expect(shownTitles()).toEqual(['m7']) + }) + + it('leaves the watermark to the replay rather than jumping it to the push seq', async () => { + presentTray([ + { hostFingerprint, notificationId: 'a:9', notificationSeq: 9, notificationEpoch: 'epoch-1' } + ]) + + await reopenWithTray() + + // Seq 9 in the tray says one event was shown, not that 6..8 were; advancing past + // them would make the desktop cut them out of every later catch-up. + expect(shownTitles()).toEqual(['m6', 'm7']) + expect(persistedSeq()).toBe(7) + }) + + it('still replays an event a coalesced summary only counted', async () => { + presentTray([ + { + hostFingerprint, + notificationId: 'a:6', + notificationSeq: 6, + notificationEpoch: 'epoch-1', + coalescedCount: 3 + } + ]) + + await reopenWithTray() + + expect(shownTitles()).toEqual(['m6', 'm7']) + }) + + it('ignores a tray entry pushed for a different paired host', async () => { + presentTray([ + { + hostFingerprint: '0123456789abcdef', + notificationId: 'a:6', + notificationSeq: 6, + notificationEpoch: 'epoch-1' + } + ]) + + await reopenWithTray() + + expect(shownTitles()).toEqual(['m6', 'm7']) + }) +}) diff --git a/mobile/src/notifications/notification-viewing-policy.ts b/mobile/src/notifications/notification-viewing-policy.ts new file mode 100644 index 00000000000..c54a04fd695 --- /dev/null +++ b/mobile/src/notifications/notification-viewing-policy.ts @@ -0,0 +1,30 @@ +import { AppState } from 'react-native' +import { + allowsMobileNotification, + type MobileNotificationPolicyEvent +} from '../../../src/shared/mobile-notification-policy' +import { + loadNotificationDeliveryPreferences, + notificationPreferencesFilter +} from './notification-delivery-preferences' + +let viewing: { hostId: string; worktreeId: string } | null = null +export function setNotificationViewingWorkspace(value: typeof viewing): void { + viewing = value +} + +export async function allowsLocalNotification( + event: MobileNotificationPolicyEvent & { worktreeId?: string }, + hostId: string +): Promise { + const preferences = await loadNotificationDeliveryPreferences() + if (!allowsMobileNotification(notificationPreferencesFilter(preferences), event)) { + return false + } + return !( + preferences.suppressWhileViewing && + AppState.currentState === 'active' && + viewing?.hostId === hostId && + viewing.worktreeId === event.worktreeId + ) +} diff --git a/mobile/src/notifications/notification-watermark-seed-race.test.ts b/mobile/src/notifications/notification-watermark-seed-race.test.ts index 742f0711982..12efb88e5d0 100644 --- a/mobile/src/notifications/notification-watermark-seed-race.test.ts +++ b/mobile/src/notifications/notification-watermark-seed-race.test.ts @@ -15,6 +15,7 @@ import { loadPushNotificationsEnabled } from '../storage/preferences' vi.mock('expo-notifications', () => ({ AndroidImportance: { HIGH: 'high' }, setNotificationChannelAsync: vi.fn(), + getPresentedNotificationsAsync: vi.fn(async () => []), getPermissionsAsync: vi.fn(), requestPermissionsAsync: vi.fn(), scheduleNotificationAsync: vi.fn(), @@ -22,9 +23,15 @@ vi.mock('expo-notifications', () => ({ })) vi.mock('react-native', () => ({ + AppState: { currentState: 'background' }, Platform: { OS: 'ios', Version: 18 } })) +// The reconnect catch-up reads the tray to learn which pushes the OS already showed, +// and mapping those to this host needs the catalog, whose real module pulls the +// native keychain. No push is presented in these tests, so an empty catalog is enough. +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn(async () => []) })) + // A storage whose reads can be held open, so a live event can be injected into the // exact window a real cold open has: subscription up, persisted watermark not yet read. const storage = new Map() @@ -51,6 +58,7 @@ vi.mock('@react-native-async-storage/async-storage', () => ({ })) vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(async () => false), loadPushNotificationsEnabled: vi.fn() })) @@ -70,7 +78,8 @@ function releaseReads(): void { function makeHostClient() { let onData: ((data: unknown) => void) | null = null - const getMissedCalls: { lastSeenSeq: number; epoch?: string }[] = [] + const getMissedCalls: { includeDesktopSuppressed: true; lastSeenSeq: number; epoch?: string }[] = + [] const client = { subscribe: vi.fn((_m: string, _p: unknown, cb: (data: unknown) => void) => { onData = cb @@ -81,7 +90,9 @@ function makeHostClient() { getState: vi.fn(() => 'connected'), sendRequest: vi.fn(async (method: string, params: unknown = {}) => { if (method === 'notifications.getMissedSince') { - getMissedCalls.push(params as { lastSeenSeq: number; epoch?: string }) + getMissedCalls.push( + params as { includeDesktopSuppressed: true; lastSeenSeq: number; epoch?: string } + ) return { ok: true, result: { notifications: [] } } as never } return { ok: true, result: undefined } as never @@ -128,6 +139,7 @@ describe('#8591 watermark seeding races a cold open', () => { host.onData?.({ type: 'ready', subscriptionId: 'sub-1', epoch: 'epoch-a' }) host.onData?.({ type: 'notification', + source: 'agent-task-complete', title: 'live-12', body: 'b', notificationId: 'agent:live', @@ -142,7 +154,9 @@ describe('#8591 watermark seeding races a cold open', () => { releaseReads() await flushAsync() - expect(host.getMissedCalls).toEqual([{ lastSeenSeq: 5, epoch: 'epoch-a' }]) + expect(host.getMissedCalls).toEqual([ + { includeDesktopSuppressed: true, lastSeenSeq: 5, epoch: 'epoch-a' } + ]) }) it('treats a zeroed-but-present watermark as a returning device, not a first pairing', async () => { @@ -156,7 +170,9 @@ describe('#8591 watermark seeding races a cold open', () => { host.onData?.({ type: 'ready', subscriptionId: 'sub-1', epoch: 'epoch-a' }) await flushAsync() - expect(host.getMissedCalls).toEqual([{ lastSeenSeq: 0, epoch: 'epoch-a' }]) + expect(host.getMissedCalls).toEqual([ + { includeDesktopSuppressed: true, lastSeenSeq: 0, epoch: 'epoch-a' } + ]) }) it('does not catch up on a first-ever pairing', async () => { diff --git a/mobile/src/notifications/push-host-fingerprint.test.ts b/mobile/src/notifications/push-host-fingerprint.test.ts new file mode 100644 index 00000000000..2fc5b44dba1 --- /dev/null +++ b/mobile/src/notifications/push-host-fingerprint.test.ts @@ -0,0 +1,62 @@ +import { describe, expect, it } from 'vitest' +import { sha256 } from '@noble/hashes/sha256' +import { deriveHostFingerprint, resolveHostIdForFingerprint } from './push-host-fingerprint' + +// Why Buffer here: it computes the same value through a completely different +// base64 path than the module's btoa/replace, so the vector is a real cross-check +// of the derivation the desktop and gateway independently perform. +function expectedFingerprint(publicKey: Uint8Array): string { + return Buffer.from(sha256(publicKey)).toString('base64url').slice(0, 16) +} + +const publicKey = Uint8Array.from({ length: 32 }, (_, index) => index) +const publicKeyB64 = Buffer.from(publicKey).toString('base64') + +describe('deriveHostFingerprint', () => { + it('matches base64url(sha256(publicKey)) truncated to 16 chars', () => { + const fingerprint = deriveHostFingerprint(publicKeyB64) + + expect(fingerprint).toBe(expectedFingerprint(publicKey)) + expect(fingerprint).toHaveLength(16) + }) + + it('produces url-safe characters only, so a fingerprint survives a JSON payload', () => { + // 0xff bytes are what push '+' and '/' into a standard base64 digest. + const dense = new Uint8Array(32).fill(0xff) + const fingerprint = deriveHostFingerprint(Buffer.from(dense).toString('base64')) + + expect(fingerprint).toBe(expectedFingerprint(dense)) + expect(fingerprint).toMatch(/^[A-Za-z0-9_-]{16}$/) + }) + + it.each([ + ['a key of the wrong length', Buffer.from(new Uint8Array(16)).toString('base64')], + ['text that is not base64 at all', '!!!not base64!!!'], + ['an empty key', ''] + ])('returns null for %s', (_label, value) => { + expect(deriveHostFingerprint(value)).toBeNull() + }) +}) + +describe('resolveHostIdForFingerprint', () => { + const other = Uint8Array.from({ length: 32 }, (_, index) => index + 1) + const hosts = [ + { id: 'host-corrupt', publicKeyB64: 'not-a-key' }, + { id: 'host-other', publicKeyB64: Buffer.from(other).toString('base64') }, + { id: 'host-1', publicKeyB64 } + ] + + it('maps a push fingerprint back to the paired host id', () => { + expect(resolveHostIdForFingerprint(expectedFingerprint(publicKey), hosts)).toBe('host-1') + }) + + it('returns null for a fingerprint no paired host derives', () => { + expect(resolveHostIdForFingerprint('0123456789abcdef', hosts)).toBeNull() + }) + + it('rejects a fingerprint of the wrong length before hashing anything', () => { + expect( + resolveHostIdForFingerprint(expectedFingerprint(publicKey).slice(0, 8), hosts) + ).toBeNull() + }) +}) diff --git a/mobile/src/notifications/push-host-fingerprint.ts b/mobile/src/notifications/push-host-fingerprint.ts new file mode 100644 index 00000000000..3aa8b739fba --- /dev/null +++ b/mobile/src/notifications/push-host-fingerprint.ts @@ -0,0 +1,58 @@ +import { sha256 } from '@noble/hashes/sha256' + +// Why: a push arrives from the gateway, so it can only name the host by something +// both sides derive independently — base64url(sha256(hostPublicKey)) truncated to +// 16 chars, identical to deriveRelayHostId in +// src/main/runtime/relay/relay-http-client.ts. The phone maps it back to its own +// hostId by re-deriving over each stored host's publicKeyB64. +// +// Base64 is inlined rather than imported (same call as mobile-relay-credential-hash.ts): +// the only shared encoders live in modules that drag in tweetnacl, expo-crypto, or +// the host store, none of which a pure derivation should need. + +const HOST_FINGERPRINT_LENGTH = 16 + +function decodeBase64(value: string): Uint8Array | null { + try { + const binary = atob(value) + const bytes = new Uint8Array(binary.length) + for (let index = 0; index < binary.length; index++) { + bytes[index] = binary.charCodeAt(index) + } + return bytes + } catch { + return null + } +} + +function encodeBase64Url(bytes: Uint8Array): string { + let binary = '' + for (const byte of bytes) { + binary += String.fromCharCode(byte) + } + return btoa(binary).replace(/\+/g, '-').replace(/\//g, '_').replace(/=+$/, '') +} + +/** Null when the stored key is unreadable, so a corrupt host entry can't shadow a real match. */ +export function deriveHostFingerprint(publicKeyB64: string): string | null { + const publicKey = decodeBase64(publicKeyB64) + if (!publicKey || publicKey.length !== 32) { + return null + } + return encodeBase64Url(sha256(publicKey)).slice(0, HOST_FINGERPRINT_LENGTH) +} + +export function resolveHostIdForFingerprint( + fingerprint: string, + hosts: readonly { readonly id: string; readonly publicKeyB64: string }[] +): string | null { + if (fingerprint.length !== HOST_FINGERPRINT_LENGTH) { + return null + } + for (const host of hosts) { + if (deriveHostFingerprint(host.publicKeyB64) === fingerprint) { + return host.id + } + } + return null +} diff --git a/mobile/src/notifications/push-payload.ts b/mobile/src/notifications/push-payload.ts new file mode 100644 index 00000000000..8de0243a63f --- /dev/null +++ b/mobile/src/notifications/push-payload.ts @@ -0,0 +1,47 @@ +// Why two shapes: APNs nests Orca's fields under `orca` beside `aps`, while FCM +// carries them flat in `data` as strings. Both reach JS as the notification's +// `content.data`, so the reader accepts either and coerces the numeric fields. +export type OrcaPushPayload = { + readonly hostFingerprint: string + readonly notificationId?: string + readonly notificationSeq?: number + readonly notificationEpoch?: string + readonly worktreeId?: string + readonly source?: string + readonly agentState?: string + // Present only on a gateway summary standing in for N events; see the coalescing + // window in docs/reference/mobile-push-contract.md. + readonly coalescedCount?: number +} + +function readString(value: unknown): string | undefined { + return typeof value === 'string' && value.length > 0 ? value : undefined +} + +function readSeq(value: unknown): number | undefined { + const raw = typeof value === 'number' ? value : Number(readString(value)) + return Number.isFinite(raw) ? raw : undefined +} + +export function readOrcaPushPayload(data: unknown): OrcaPushPayload | null { + if (!data || typeof data !== 'object') { + return null + } + const nested = (data as { orca?: unknown }).orca + const record = (nested && typeof nested === 'object' ? nested : data) as Record + // The fingerprint is what makes this a gateway push; locally scheduled data never has one. + const hostFingerprint = readString(record.hostFingerprint) + if (!hostFingerprint) { + return null + } + return { + hostFingerprint, + notificationId: readString(record.notificationId), + notificationSeq: readSeq(record.notificationSeq), + notificationEpoch: readString(record.notificationEpoch), + worktreeId: readString(record.worktreeId), + source: readString(record.source), + agentState: readString(record.agentState), + coalescedCount: readSeq(record.coalescedCount) + } +} diff --git a/mobile/src/notifications/push-preference-update.test.ts b/mobile/src/notifications/push-preference-update.test.ts new file mode 100644 index 00000000000..1e1426fef93 --- /dev/null +++ b/mobile/src/notifications/push-preference-update.test.ts @@ -0,0 +1,75 @@ +import { beforeEach, expect, it, vi } from 'vitest' +import { + attachPushRegistration, + resetPushRegistrationForTests, + setNotificationDeliveryPreferences, + NOTIFICATIONS_REMOTE_PUSH_CAPABILITY +} from './push-registration' +import { DEFAULT_NOTIFICATION_DELIVERY } from './notification-delivery-preferences' + +const storage = new Map() +vi.mock('@react-native-async-storage/async-storage', () => ({ + default: { + getItem: vi.fn(async (key: string) => storage.get(key) ?? null), + setItem: vi.fn(async (key: string, value: string) => { + storage.set(key, value) + }) + } +})) +vi.mock('./push-token', () => ({ + getDevicePushToken: vi.fn(async () => ({ + platform: 'ios', + token: 'a'.repeat(64), + apnsEnvironment: 'sandbox' + })), + addPushTokenListener: vi.fn() +})) + +beforeEach(() => { + resetPushRegistrationForTests() + storage.clear() + storage.set('orca:remotePushEnabled', 'true') +}) + +it('replaces an in-flight old registration with the latest event and sound preferences', async () => { + const calls: { method: string; params: unknown }[] = [] + let finishFirst: ((value: unknown) => void) | undefined + const client = { + sendRequest: vi.fn(async (method: string, params?: unknown) => { + calls.push({ method, params }) + if (method === 'status.get') { + return { ok: true, result: { capabilities: [NOTIFICATIONS_REMOTE_PUSH_CAPABILITY] } } + } + if (method === 'notifications.registerPush') { + if (!finishFirst) { + return new Promise((resolve) => { + finishFirst = resolve + }) + } + return { ok: true, result: { registered: true, registrationId: 'new' } } + } + return { ok: true, result: { unregistered: true } } + }) + } + const detach = attachPushRegistration('host', client as never) + await vi.waitFor(() => expect(finishFirst).toBeDefined()) + const update = setNotificationDeliveryPreferences({ + ...DEFAULT_NOTIFICATION_DELIVERY, + followDesktop: false, + terminalBell: false, + sound: false + }) + finishFirst!({ ok: true, result: { registered: true, registrationId: 'old' } }) + await update + await vi.waitFor(() => + expect( + calls.filter((call) => call.method === 'notifications.registerPush').length + ).toBeGreaterThan(1) + ) + const latest = calls.findLast((call) => call.method === 'notifications.registerPush') + expect(latest?.params).toMatchObject({ + filter: { followDesktop: false, sound: false, sources: ['agent-task-complete', 'plugin'] } + }) + expect(calls.some((call) => call.method === 'notifications.unregisterPush')).toBe(true) + detach() +}) diff --git a/mobile/src/notifications/push-receive.test.ts b/mobile/src/notifications/push-receive.test.ts new file mode 100644 index 00000000000..ddfc2708e21 --- /dev/null +++ b/mobile/src/notifications/push-receive.test.ts @@ -0,0 +1,281 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import AsyncStorage from '@react-native-async-storage/async-storage' +import { sha256 } from '@noble/hashes/sha256' +import { loadHostCatalog } from '../transport/host-store' +import type { HostCatalogEntry } from '../transport/types' +import { getNotificationNavigationTarget } from './notification-routing' +import { + getHostNotificationSession, + resetHostNotificationSessionsForTests +} from './notification-reconnect-catchup' +import { + isRemotePushTrigger, + pushNotificationRouteData, + shouldSuppressForegroundPush +} from './push-receive' + +vi.mock('react-native', () => ({ AppState: { currentState: 'background' } })) + +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn() })) + +const storage = new Map() + +vi.mock('@react-native-async-storage/async-storage', () => ({ + default: { + getItem: vi.fn(async (key: string) => storage.get(key) ?? null), + setItem: vi.fn(async (key: string, value: string) => { + storage.set(key, value) + }), + removeItem: vi.fn(async () => undefined) + } +})) + +const publicKey = Uint8Array.from({ length: 32 }, (_, index) => index) +const publicKeyB64 = Buffer.from(publicKey).toString('base64') +const hostFingerprint = Buffer.from(sha256(publicKey)).toString('base64url').slice(0, 16) + +const hosts = [{ id: 'host-1', publicKeyB64 }] as unknown as HostCatalogEntry[] + +// APNs nests Orca's fields beside `aps`; FCM sends them flat and stringified. +function apnsData(orca: Record): unknown { + return { aps: { alert: { title: 'Orca', body: 'Agent needs input' } }, orca } +} + +function fcmData(orca: Record): unknown { + return Object.fromEntries(Object.entries(orca).map(([key, value]) => [key, String(value)])) +} + +beforeEach(() => { + vi.clearAllMocks() + storage.clear() + storage.set('orca:pushNotificationsEnabled', 'true') + storage.set('orca:remotePushEnabled', 'true') + resetHostNotificationSessionsForTests() + vi.mocked(loadHostCatalog).mockResolvedValue(hosts) +}) + +describe('shouldSuppressForegroundPush', () => { + it('suppresses a push whose id and seq the socket already delivered', async () => { + const session = getHostNotificationSession('host-1') + session.lastDeliveredEpoch = 'epoch-1' + session.seen.add('id:agent:one#7') + + await expect( + shouldSuppressForegroundPush( + apnsData({ + hostFingerprint, + notificationId: 'agent:one', + notificationSeq: 7, + notificationEpoch: 'epoch-1' + }) + ) + ).resolves.toBe(true) + }) + + it('shows an unseen push and marks it so the socket replay is dropped', async () => { + const data = apnsData({ + hostFingerprint, + notificationId: 'agent:one', + notificationSeq: 7, + notificationEpoch: 'epoch-1' + }) + + await expect(shouldSuppressForegroundPush(data)).resolves.toBe(false) + + expect(getHostNotificationSession('host-1').seen.has('id:agent:one#7')).toBe(true) + await expect(shouldSuppressForegroundPush(data)).resolves.toBe(true) + }) + + it('reads the flat stringified fields an FCM data message carries', async () => { + const session = getHostNotificationSession('host-1') + session.lastDeliveredEpoch = 'epoch-1' + session.seen.add('id:agent:one#7') + + await expect( + shouldSuppressForegroundPush( + fcmData({ + hostFingerprint, + notificationId: 'agent:one', + notificationSeq: 7, + notificationEpoch: 'epoch-1' + }) + ) + ).resolves.toBe(true) + }) + + it('keys a terminal bell on its seq alone, since it carries no notification id', async () => { + const session = getHostNotificationSession('host-1') + session.lastDeliveredEpoch = 'epoch-1' + session.seen.add('seq:4') + + await expect( + shouldSuppressForegroundPush( + apnsData({ + hostFingerprint, + source: 'terminal-bell', + notificationSeq: 4, + notificationEpoch: 'epoch-1' + }) + ) + ).resolves.toBe(true) + }) + + it('shows a push that names no counter lifetime without letting it claim a key', async () => { + const session = getHostNotificationSession('host-1') + session.lastDeliveredEpoch = 'epoch-1' + session.seen.add('seq:4') + + // Without an epoch the seq cannot be tied to this counter, so a forged seq:4 + // must neither be swallowed against it nor stop the real bell at seq 4. + await expect( + shouldSuppressForegroundPush(apnsData({ hostFingerprint, notificationSeq: 4 })) + ).resolves.toBe(false) + await expect( + shouldSuppressForegroundPush(apnsData({ hostFingerprint, notificationSeq: 5 })) + ).resolves.toBe(false) + expect(session.seen.has('seq:5')).toBe(false) + }) + + it('voids seen keys from a previous desktop lifetime before testing its own', async () => { + const session = getHostNotificationSession('host-1') + session.lastDeliveredEpoch = 'epoch-old' + session.seen.add('seq:4') + + await expect( + shouldSuppressForegroundPush( + apnsData({ hostFingerprint, notificationSeq: 4, notificationEpoch: 'epoch-new' }) + ) + ).resolves.toBe(false) + }) + + it('leaves a locally scheduled notification to the existing path', async () => { + await expect( + shouldSuppressForegroundPush({ hostId: 'host-1', source: 'agent-task-complete' }) + ).resolves.toBe(false) + expect(loadHostCatalog).not.toHaveBeenCalled() + }) + + it('suppresses a push for a host this phone no longer has, since its tap routes nowhere', async () => { + vi.mocked(loadHostCatalog).mockResolvedValue([]) + + await expect( + shouldSuppressForegroundPush(apnsData({ hostFingerprint, notificationSeq: 1 })) + ).resolves.toBe(true) + }) + + it('seeds the persisted watermark before adopting, so a push cannot void it', async () => { + storage.set( + 'orca:mobileNotificationsWatermark:host-1', + JSON.stringify({ seq: 42, epoch: 'epoch-1' }) + ) + + await shouldSuppressForegroundPush( + apnsData({ hostFingerprint, notificationSeq: 43, notificationEpoch: 'epoch-1' }) + ) + + // Unseeded, the null epoch reads as a new counter lifetime: the seq resets to 0 + // and {seq: 0} is persisted over a watermark the next reconnect still needs. + expect(getHostNotificationSession('host-1').lastDeliveredSeq).toBe(42) + expect(AsyncStorage.setItem).not.toHaveBeenCalled() + }) + + it('shows a coalesced summary without claiming the key of the one event it names', async () => { + await expect( + shouldSuppressForegroundPush( + apnsData({ + hostFingerprint, + notificationId: 'agent:one', + notificationSeq: 7, + coalescedCount: 3 + }) + ) + ).resolves.toBe(false) + + // Claiming it would make the socket swallow the banner for agent:one itself, + // which the summary only ever counted. + expect(getHostNotificationSession('host-1').seen.has('id:agent:one#7')).toBe(false) + }) +}) + +describe('pushNotificationRouteData', () => { + it('routes a tap by mapping the fingerprint to the paired host id', () => { + const data = pushNotificationRouteData( + apnsData({ + hostFingerprint, + notificationId: 'agent:one', + worktreeId: 'repo::/Users/me/orca/workspaces/feature', + source: 'agent-task-complete' + }), + hosts + ) + + expect(getNotificationNavigationTarget(data, { knownHostIds: new Set(['host-1']) })).toEqual({ + hostId: 'host-1', + sessionTarget: { + name: '[hostId]/session/[worktreeId]', + params: { hostId: 'host-1', worktreeId: 'repo::/Users/me/orca/workspaces/feature' } + } + }) + }) + + it('falls back to the host screen for a push with no worktree', () => { + const data = pushNotificationRouteData( + fcmData({ hostFingerprint, source: 'terminal-bell' }), + hosts + ) + + expect(getNotificationNavigationTarget(data)).toEqual({ + hostId: 'host-1', + sessionTarget: null + }) + }) + + it('passes locally scheduled data through untouched', () => { + const data = { hostId: 'host-9', source: 'agent-task-complete' } + + expect(pushNotificationRouteData(data, hosts)).toBe(data) + }) + + it('leaves an unresolvable fingerprint unrouted rather than guessing a host', () => { + const data = pushNotificationRouteData(apnsData({ hostFingerprint: '0123456789abcdef' }), hosts) + + expect(getNotificationNavigationTarget(data)).toBeNull() + }) + + it('leaves a remote push unrouted when no host catalog could be read', () => { + const data = { hostId: 'host-1', orca: { hostFingerprint, notificationId: 'agent:one' } } + + expect(pushNotificationRouteData(data, [], true)).toBeNull() + }) + + it('leaves a remote push with no fingerprint unrouted instead of treating it as local', () => { + const data = { hostId: 'host-1', worktreeId: 'wt-1', source: 'agent-task-complete' } + + expect(pushNotificationRouteData(data, hosts, true)).toBeNull() + // The same shape from this app's own scheduler still routes. + expect(pushNotificationRouteData(data, hosts, false)).toBe(data) + }) + + it('recognises only a provider-delivered trigger as remote', () => { + expect(isRemotePushTrigger({ type: 'push' })).toBe(true) + expect(isRemotePushTrigger({ type: 'timeInterval', seconds: 1 })).toBe(false) + expect(isRemotePushTrigger({ channelId: 'orca-desktop' })).toBe(false) + expect(isRemotePushTrigger(null)).toBe(false) + expect(isRemotePushTrigger(undefined)).toBe(false) + }) + + it('drops a gateway payload that pairs an unresolvable fingerprint with a stray hostId', () => { + const data = { + hostId: 'host-1', + orca: { hostFingerprint: '0123456789abcdef', notificationId: 'agent:one' } + } + + // Returning the raw data would let the stray hostId route a tap the push never named. + expect(pushNotificationRouteData(data, hosts)).toBeNull() + expect( + getNotificationNavigationTarget(pushNotificationRouteData(data, hosts), { + knownHostIds: new Set(['host-1']) + }) + ).toBeNull() + }) +}) diff --git a/mobile/src/notifications/push-receive.ts b/mobile/src/notifications/push-receive.ts new file mode 100644 index 00000000000..c920b6bda29 --- /dev/null +++ b/mobile/src/notifications/push-receive.ts @@ -0,0 +1,121 @@ +import { allowsLocalNotification } from './notification-viewing-policy' +import { loadPushNotificationsEnabled, loadRemotePushEnabled } from '../storage/preferences' +import { loadHostCatalog } from '../transport/host-store' +import { + adoptNotificationEpoch, + getHostNotificationSession, + seedWatermarkFromStorage, + seenKeyForEvent +} from './notification-reconnect-catchup' +import { resolveHostIdForFingerprint } from './push-host-fingerprint' +import { readOrcaPushPayload, type OrcaPushPayload } from './push-payload' + +async function resolvePushHostId(payload: OrcaPushPayload): Promise { + const hosts = await loadHostCatalog().catch(() => []) + return resolveHostIdForFingerprint(payload.hostFingerprint, hosts) +} + +/** + * Whether a foreground notification is a push for an event the socket already + * delivered, and must therefore be swallowed instead of banner'd a second time. + * + * Marking happens here rather than in a received listener because the handler is + * the only hook that can actually suppress, and the key must be claimed exactly + * once — a listener running afterwards would mark an event the handler dropped. + */ +export async function shouldSuppressForegroundPush(data: unknown): Promise { + const payload = readOrcaPushPayload(data) + if (!payload) { + return false + } + const hostId = await resolvePushHostId(payload) + // Why suppressed rather than shown: the only pushes that outlive their host are + // ones a gateway registration still holds after a removal whose unregister never + // reached the desktop. A banner naming a host this phone no longer has cannot be + // tapped anywhere, so it is noise the user cannot act on or turn off per-host. + if (!hostId) { + return true + } + if (!(await loadPushNotificationsEnabled()) || !(await loadRemotePushEnabled())) { + return true + } + if ( + !(await allowsLocalNotification( + { ...payload, source: payload.source ?? 'agent-task-complete' }, + hostId + )) + ) { + return true + } + const session = getHostNotificationSession(hostId) + // Why seeded first: the socket may never have connected this launch (phone on + // cellular), leaving lastDeliveredEpoch null. Adopting against an unseeded session + // resets the seq to 0 and persists that over a valid watermark, so the next + // reconnect replays the desktop's whole retained buffer. + seedWatermarkFromStorage(session, hostId) + await session.watermarkSeeded + // A push that names no counter lifetime cannot claim a seq-derived key: the + // desktop always sends the epoch, so this is shown as-is and never marked. + if (payload.notificationEpoch == null) { + return false + } + // The seen keys are seq-derived, so a push from a new desktop lifetime must void + // them before its own key is tested against a counter that no longer exists. + adoptNotificationEpoch(session, hostId, payload.notificationEpoch) + // Why a coalesced summary is neither suppressed nor marked: it carries only the + // latest event's fields, so claiming that key would make the socket swallow the + // specific banner for an event the summary only ever counted. + if ((payload.coalescedCount ?? 0) > 1) { + return false + } + const key = seenKeyForEvent(payload) + if (!key) { + return false + } + if (session.seen.has(key)) { + return true + } + session.seen.add(key) + return false +} + +/** Whether the OS says a notification came from a provider rather than this app. */ +export function isRemotePushTrigger(trigger: unknown): boolean { + return ( + typeof trigger === 'object' && + trigger !== null && + (trigger as { readonly type?: unknown }).type === 'push' + ) +} + +/** + * Notification data a tap can route with: the gateway names the host by fingerprint, + * so it is mapped back to this device's hostId. Locally scheduled data passes + * through untouched, which is what keeps its taps on their existing path. + * + * Why null and not the raw data when the fingerprint does not resolve: a gateway + * payload is attacker-adjacent input, and passing it on would let a stray `hostId` + * beside the `orca` block route a tap at a host the push never named. A remote + * push with no fingerprint at all is the same input minus the block, so it is + * unrouted too rather than handed to the local path as if this app scheduled it. + */ +export function pushNotificationRouteData( + data: unknown, + hosts: readonly { readonly id: string; readonly publicKeyB64: string }[], + remote = false +): unknown { + const payload = readOrcaPushPayload(data) + if (!payload) { + return remote ? null : data + } + const hostId = resolveHostIdForFingerprint(payload.hostFingerprint, hosts) + if (!hostId) { + return null + } + return { + hostId, + ...(payload.source ? { source: payload.source } : {}), + ...(payload.worktreeId ? { worktreeId: payload.worktreeId } : {}), + ...(payload.notificationId ? { notificationId: payload.notificationId } : {}) + } +} diff --git a/mobile/src/notifications/push-registration.test.ts b/mobile/src/notifications/push-registration.test.ts new file mode 100644 index 00000000000..22070bfdd79 --- /dev/null +++ b/mobile/src/notifications/push-registration.test.ts @@ -0,0 +1,412 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { RpcClient, SendRequestOptions } from '../transport/rpc-client' +import type { RpcResponse } from '../transport/types' +import { + loadRemotePushAgentStates, + loadRemotePushEnabled, + loadRemotePushFilter, + loadRemotePushHostRegistrations, + saveRemotePushAgentStates, + saveRemotePushEnabled, + saveRemotePushHostRegistrations, + type RemotePushAgentState, + type RemotePushHostRegistrations +} from '../storage/preferences' +import { addPushTokenListener, getDevicePushToken, type MobilePushToken } from './push-token' +import { + NOTIFICATIONS_REMOTE_PUSH_CAPABILITY, + attachPushRegistration, + resetPushRegistrationForTests, + setRemotePushAgentStates, + setRemotePushEnabled, + startPushTokenSync, + unregisterPushForRemovedHost +} from './push-registration' + +vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(), + saveRemotePushEnabled: vi.fn(), + loadRemotePushAgentStates: vi.fn(), + saveRemotePushAgentStates: vi.fn(), + loadRemotePushFilter: vi.fn(), + loadRemotePushHostRegistrations: vi.fn(), + saveRemotePushHostRegistrations: vi.fn() +})) + +vi.mock('./push-token', () => ({ + getDevicePushToken: vi.fn(), + addPushTokenListener: vi.fn() +})) + +const IOS_TOKEN: MobilePushToken = { + platform: 'ios', + token: 'a'.repeat(64), + apnsEnvironment: 'production' +} + +// Every await in the module resolves immediately, so one macrotask drains the whole +// per-host reconcile chain no matter how many hops deep it happens to be. +function flush(): Promise { + return new Promise((resolve) => setTimeout(resolve, 0)) +} + +function ok(result: unknown): RpcResponse { + return { id: 'req', ok: true, result, _meta: { runtimeId: 'runtime-1' } } +} + +type SentRequest = { method: string; params?: unknown; options?: SendRequestOptions } + +function makeClient(capabilities: readonly string[]): { + client: Pick + sent: SentRequest[] +} { + const sent: SentRequest[] = [] + const client = { + sendRequest: vi.fn(async (method: string, params?: unknown, options?: SendRequestOptions) => { + sent.push({ method, params, options }) + if (method === 'status.get') { + return ok({ capabilities: [...capabilities] }) + } + if (method === 'notifications.registerPush') { + return ok({ registered: true, registrationId: 'registration-1' }) + } + if (method === 'notifications.unregisterPush') { + return ok({ unregistered: true }) + } + return ok(null) + }) + } + return { client, sent } +} + +function methodsIn(sent: SentRequest[]): string[] { + return sent.map((request) => request.method) +} + +let enabled = false +let agentStates: readonly RemotePushAgentState[] = ['needs-input', 'finished'] +let stored: RemotePushHostRegistrations + +beforeEach(() => { + vi.clearAllMocks() + resetPushRegistrationForTests() + enabled = false + agentStates = ['needs-input', 'finished'] + stored = { registeredHostIds: [], pendingUnregisterHostIds: [] } + + vi.mocked(loadRemotePushEnabled).mockImplementation(async () => enabled) + vi.mocked(saveRemotePushEnabled).mockImplementation(async (value) => { + enabled = value + }) + vi.mocked(loadRemotePushAgentStates).mockImplementation(async () => agentStates) + vi.mocked(saveRemotePushAgentStates).mockImplementation(async (value) => { + agentStates = value + }) + vi.mocked(loadRemotePushFilter).mockImplementation(async () => ({ + sources: ['agent-task-complete', 'terminal-bell', 'plugin'], + agentStates + })) + vi.mocked(loadRemotePushHostRegistrations).mockImplementation(async () => stored) + vi.mocked(saveRemotePushHostRegistrations).mockImplementation(async (value) => { + stored = value + }) + vi.mocked(getDevicePushToken).mockResolvedValue(IOS_TOKEN) +}) + +describe('push registration capability gating', () => { + it('registers a connected host that advertises remote push', async () => { + const { client, sent } = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + await setRemotePushEnabled(true) + + attachPushRegistration('host-1', client) + await flush() + + const register = sent.find((request) => request.method === 'notifications.registerPush') + expect(register?.params).toEqual({ + platform: 'ios', + token: IOS_TOKEN.token, + apnsEnvironment: 'production', + filter: { + sources: ['agent-task-complete', 'terminal-bell', 'plugin'], + agentStates: ['needs-input', 'finished'] + } + }) + expect(stored.registeredHostIds).toEqual(['host-1']) + }) + + it('never calls registerPush on a host without the capability', async () => { + const { client, sent } = makeClient(['some-other.v1']) + await setRemotePushEnabled(true) + + attachPushRegistration('host-legacy', client) + await flush() + + expect(methodsIn(sent)).toEqual(['status.get']) + expect(stored.registeredHostIds).toEqual([]) + }) + + it('leaves a capable host alone while the switch is off', async () => { + const { client, sent } = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + + attachPushRegistration('host-1', client) + await flush() + + expect(methodsIn(sent)).toEqual(['status.get']) + }) + + it('omits apnsEnvironment for an Android token', async () => { + vi.mocked(getDevicePushToken).mockResolvedValue({ platform: 'android', token: 'fcm-token' }) + const { client, sent } = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + await setRemotePushEnabled(true) + + attachPushRegistration('host-1', client) + await flush() + + const register = sent.find((request) => request.method === 'notifications.registerPush') + expect(register?.params).toMatchObject({ platform: 'android', token: 'fcm-token' }) + expect(register?.params).not.toHaveProperty('apnsEnvironment') + }) + + it('registers nothing when the device has no push token at all', async () => { + vi.mocked(getDevicePushToken).mockResolvedValue(null) + const { client, sent } = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + await setRemotePushEnabled(true) + + attachPushRegistration('host-simulator', client) + await flush() + + expect(methodsIn(sent)).toEqual(['status.get']) + }) + + it('asks only once when the host answers that it has no push capability', async () => { + const { client, sent } = makeClient(['some-other.v1']) + await setRemotePushEnabled(true) + attachPushRegistration('host-legacy', client) + await flush() + + await setRemotePushAgentStates(['needs-input']) + await flush() + + expect(methodsIn(sent)).toEqual(['status.get']) + }) + + it('re-probes a host whose first status.get never answered', async () => { + const sent: string[] = [] + let probeFails = true + const client = { + sendRequest: vi.fn(async (method: string) => { + sent.push(method) + if (method === 'status.get') { + if (probeFails) { + throw new Error('request timed out') + } + return ok({ capabilities: [NOTIFICATIONS_REMOTE_PUSH_CAPABILITY] }) + } + return ok({ registered: true, registrationId: 'registration-1' }) + }) + } + await setRemotePushEnabled(true) + attachPushRegistration('host-1', client) + await flush() + expect(sent).toEqual(['status.get']) + + // A latched `false` would keep this host unregistered for the connection's life. + probeFails = false + await setRemotePushAgentStates(['needs-input']) + await flush() + + expect(sent).toEqual(['status.get', 'status.get', 'notifications.registerPush']) + }) + + it('retries the device token on the next reconcile after the device had none', async () => { + vi.mocked(getDevicePushToken).mockResolvedValueOnce(null).mockResolvedValue(IOS_TOKEN) + const { client, sent } = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + await setRemotePushEnabled(true) + attachPushRegistration('host-1', client) + await flush() + expect(methodsIn(sent)).toEqual(['status.get']) + + // A token can be missing only for now — APNs registration still in flight. + await setRemotePushAgentStates(['needs-input']) + await flush() + + expect(methodsIn(sent)).toContain('notifications.registerPush') + }) +}) + +describe('push registration token and filter changes', () => { + it('re-registers every connected host when the provider rolls the token', async () => { + let onTokenChange: ((token: MobilePushToken) => void) | null = null + vi.mocked(addPushTokenListener).mockImplementation((listener) => { + onTokenChange = listener + return () => {} + }) + const { client, sent } = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + await setRemotePushEnabled(true) + attachPushRegistration('host-1', client) + await flush() + startPushTokenSync() + + onTokenChange?.({ platform: 'ios', token: 'b'.repeat(64), apnsEnvironment: 'sandbox' }) + await flush() + + const registers = sent.filter((request) => request.method === 'notifications.registerPush') + expect(registers).toHaveLength(2) + expect(registers[1]?.params).toMatchObject({ + token: 'b'.repeat(64), + apnsEnvironment: 'sandbox' + }) + }) + + it('re-registers with the narrowed filter when a sub-switch is turned off', async () => { + const { client, sent } = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + await setRemotePushEnabled(true) + attachPushRegistration('host-1', client) + await flush() + + await setRemotePushAgentStates(['needs-input']) + await flush() + + const registers = sent.filter((request) => request.method === 'notifications.registerPush') + expect(registers).toHaveLength(2) + expect(registers[1]?.params).toMatchObject({ + filter: { + sources: ['agent-task-complete', 'terminal-bell', 'plugin'], + agentStates: ['needs-input'] + } + }) + }) +}) + +describe('push unregistration', () => { + it('unregisters a connected host as soon as the switch goes off', async () => { + const { client, sent } = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + await setRemotePushEnabled(true) + attachPushRegistration('host-1', client) + await flush() + + await setRemotePushEnabled(false) + await flush() + + expect(methodsIn(sent)).toContain('notifications.unregisterPush') + expect(stored.registeredHostIds).toEqual([]) + expect(stored.pendingUnregisterHostIds).toEqual([]) + }) + + it('retries the unregister on a host that was offline when the switch went off', async () => { + const first = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + await setRemotePushEnabled(true) + const detach = attachPushRegistration('host-1', first.client) + await flush() + detach() + + await setRemotePushEnabled(false) + await flush() + expect(methodsIn(first.sent)).not.toContain('notifications.unregisterPush') + expect(stored.pendingUnregisterHostIds).toEqual(['host-1']) + + // A fresh process: only the persisted intent survives the restart. + resetPushRegistrationForTests() + const reconnected = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + attachPushRegistration('host-1', reconnected.client) + await flush() + + // No probe first: a pending entry is a switch-off the user already performed, so + // it must not wait on a status.get that may never answer. + expect(methodsIn(reconnected.sent)).toEqual(['notifications.unregisterPush']) + expect(stored.pendingUnregisterHostIds).toEqual([]) + }) + + it('keeps the pending intent when the retry itself fails', async () => { + stored = { registeredHostIds: ['host-1'], pendingUnregisterHostIds: ['host-1'] } + const client = { + sendRequest: vi.fn(async (method: string) => + method === 'status.get' + ? ok({ capabilities: [NOTIFICATIONS_REMOTE_PUSH_CAPABILITY] }) + : Promise.reject(new Error('socket closed')) + ) + } + + attachPushRegistration('host-1', client) + await flush() + + expect(stored.pendingUnregisterHostIds).toEqual(['host-1']) + }) + + it('unregisters best-effort before a removed host loses its credentials', async () => { + const { client, sent } = makeClient([NOTIFICATIONS_REMOTE_PUSH_CAPABILITY]) + await setRemotePushEnabled(true) + attachPushRegistration('host-1', client) + await flush() + + await unregisterPushForRemovedHost('host-1') + + expect(methodsIn(sent)).toContain('notifications.unregisterPush') + expect(stored.registeredHostIds).toEqual([]) + expect(stored.pendingUnregisterHostIds).toEqual([]) + }) + + it('drops a removed host that was never connected without any request', async () => { + stored = { registeredHostIds: ['host-gone'], pendingUnregisterHostIds: ['host-gone'] } + + await unregisterPushForRemovedHost('host-gone') + + expect(stored).toEqual({ registeredHostIds: [], pendingUnregisterHostIds: [] }) + }) + + it('unregisters a pending host even when its capability probe never answers', async () => { + stored = { registeredHostIds: ['host-1'], pendingUnregisterHostIds: ['host-1'] } + const sent: string[] = [] + const client = { + sendRequest: vi.fn(async (method: string) => { + sent.push(method) + if (method === 'status.get') { + throw new Error('request timed out') + } + return ok({ unregistered: true }) + }) + } + + attachPushRegistration('host-1', client) + await flush() + + // Gating this on the probe leaves the gateway pushing while the switch reads off. + expect(sent).toEqual(['notifications.unregisterPush']) + expect(stored.pendingUnregisterHostIds).toEqual([]) + }) + + it('re-arms the unregister when the switch goes off while a register is in flight', async () => { + const sent: string[] = [] + let releaseRegister: (() => void) | null = null + const client = { + sendRequest: vi.fn(async (method: string) => { + sent.push(method) + if (method === 'status.get') { + return ok({ capabilities: [NOTIFICATIONS_REMOTE_PUSH_CAPABILITY] }) + } + if (method === 'notifications.registerPush') { + await new Promise((resolve) => { + releaseRegister = resolve + }) + return ok({ registered: true, registrationId: 'registration-1' }) + } + return ok({ unregistered: true }) + }) + } + await setRemotePushEnabled(true) + attachPushRegistration('host-1', client) + await flush() + + // The sweep snapshots `registered` while this host is still only in flight. + const switchedOff = setRemotePushEnabled(false) + await flush() + releaseRegister?.() + await switchedOff + await flush() + + // Recording the late success would leave a live gateway registration behind a + // switch that reads off, with nothing pending to ever retract it. + expect(sent).toContain('notifications.unregisterPush') + expect(stored).toEqual({ registeredHostIds: [], pendingUnregisterHostIds: [] }) + }) +}) diff --git a/mobile/src/notifications/push-registration.ts b/mobile/src/notifications/push-registration.ts new file mode 100644 index 00000000000..c98e50d41e0 --- /dev/null +++ b/mobile/src/notifications/push-registration.ts @@ -0,0 +1,289 @@ +import { + saveNotificationDeliveryPreferences, + type NotificationDeliveryPreferences +} from './notification-delivery-preferences' +import type { + MobilePushRegisterInput, + MobilePushRegisterResult +} from '../../../src/shared/mobile-push-contract' +import { NOTIFICATIONS_REMOTE_PUSH_RUNTIME_CAPABILITY } from '../../../src/shared/protocol-version' +import type { RpcClient } from '../transport/rpc-client' +import { + loadRemotePushEnabled, + loadRemotePushFilter, + loadRemotePushHostRegistrations, + saveRemotePushAgentStates, + saveRemotePushEnabled, + saveRemotePushHostRegistrations, + type RemotePushAgentState, + type RemotePushFilter +} from '../storage/preferences' +import { addPushTokenListener, getDevicePushToken, type MobilePushToken } from './push-token' + +export const NOTIFICATIONS_REMOTE_PUSH_CAPABILITY = NOTIFICATIONS_REMOTE_PUSH_RUNTIME_CAPABILITY + +type PushClient = Pick + +const REQUEST_TIMEOUT_MS = 5_000 +const REMOVAL_TIMEOUT_MS = 2_000 + +type HostPushState = { + client: PushClient | null + // An unanswered probe is unknown, not unsupported. + supported: boolean | null + chain: Promise +} + +type RegistrationRecords = { registered: Set; pending: Set } + +const hostsById = new Map() +let registrationRecords: RegistrationRecords | null = null +let tokenPromise: Promise | null = null +// A late registration must not overwrite a newer preference or consent choice. +let consentGeneration = 0 + +function hostState(hostId: string): HostPushState { + let state = hostsById.get(hostId) + if (!state) { + state = { client: null, supported: null, chain: Promise.resolve() } + hostsById.set(hostId, state) + } + return state +} + +async function readRecords(): Promise { + if (!registrationRecords) { + const stored = await loadRemotePushHostRegistrations() + registrationRecords ??= { + registered: new Set(stored.registeredHostIds), + pending: new Set(stored.pendingUnregisterHostIds) + } + } + return registrationRecords +} + +async function mutateRecords(mutate: (value: RegistrationRecords) => void): Promise { + const value = await readRecords() + mutate(value) + await saveRemotePushHostRegistrations({ + registeredHostIds: [...value.registered], + pendingUnregisterHostIds: [...value.pending] + }).catch(() => {}) +} + +// A missing token is retried: APNs registration may still be in flight. +async function currentToken(): Promise { + if (!tokenPromise) { + const pending: Promise = getDevicePushToken().then((token) => { + if (!token && tokenPromise === pending) { + tokenPromise = null + } + return token + }) + tokenPromise = pending + } + return tokenPromise +} + +async function readRemotePushCapability(client: PushClient): Promise { + try { + const response = await client.sendRequest('status.get') + if (!response.ok) { + return null + } + const result = response.result + if (!result || typeof result !== 'object') { + return false + } + const capabilities = (result as { capabilities?: unknown }).capabilities + return ( + Array.isArray(capabilities) && capabilities.includes(NOTIFICATIONS_REMOTE_PUSH_CAPABILITY) + ) + } catch { + return null + } +} + +async function sendRegister( + client: PushClient, + token: MobilePushToken, + filter: RemotePushFilter +): Promise { + const params: Omit = { + platform: token.platform, + token: token.token, + ...(token.apnsEnvironment ? { apnsEnvironment: token.apnsEnvironment } : {}), + filter: { ...filter, sources: [...filter.sources], agentStates: [...filter.agentStates] } + } + const response = await client + .sendRequest('notifications.registerPush', params, { + timeoutMs: REQUEST_TIMEOUT_MS, + failWhenDisconnected: true + }) + .catch(() => null) + if (!response?.ok) { + return false + } + return (response.result as MobilePushRegisterResult | null)?.registered === true +} + +async function sendUnregister(client: PushClient, timeoutMs: number): Promise { + const response = await client + .sendRequest('notifications.unregisterPush', null, { + timeoutMs, + failWhenDisconnected: true + }) + .catch(() => null) + return response?.ok === true +} + +async function reconcileHost(hostId: string): Promise { + const state = hostsById.get(hostId) + const client = state?.client + if (!state || !client) { + return + } + const generation = consentGeneration + const value = await readRecords() + // Unregister intent takes priority even before the capability probe answers. + if (value.pending.has(hostId)) { + if (state.supported === false || !(await sendUnregister(client, REQUEST_TIMEOUT_MS))) { + return + } + await mutateRecords((current) => { + current.pending.delete(hostId) + current.registered.delete(hostId) + }) + // A preference change can invalidate a register without disabling push. + if (!(await loadRemotePushEnabled())) { + return + } + } + if (state.supported == null) { + const probed = await readRemotePushCapability(client) + if (state.client !== client) { + return + } + if (probed == null) { + return + } + state.supported = probed + } + if (!state.supported || state.client !== client) { + return + } + if (!(await loadRemotePushEnabled())) { + return + } + const token = await currentToken() + if (!token) { + return + } + if (!(await sendRegister(client, token, await loadRemotePushFilter()))) { + return + } + if (generation !== consentGeneration) { + await mutateRecords((current) => current.pending.add(hostId)) + void enqueueReconcile(hostId) + return + } + await mutateRecords((current) => current.registered.add(hostId)) +} + +function enqueueReconcile(hostId: string): Promise { + const state = hostState(hostId) + const run = state.chain.then(() => reconcileHost(hostId)).catch(() => {}) + state.chain = run + return run +} + +async function reconcileAllHosts(): Promise { + await Promise.all([...hostsById.keys()].map((hostId) => enqueueReconcile(hostId))) +} + +/** + * Track a host whose client has reached `connected`, registering (or retrying a + * pending unregister) as the current preference requires. The returned function + * detaches the client on disconnect; the host's tracked state survives it. + */ +export function attachPushRegistration(hostId: string, client: PushClient): () => void { + const state = hostState(hostId) + if (state.client !== client) { + state.client = client + state.supported = null + } + void enqueueReconcile(hostId) + return () => { + if (state.client === client) { + state.client = null + } + } +} + +export async function setRemotePushEnabled(enabled: boolean): Promise { + consentGeneration++ + await saveRemotePushEnabled(enabled) + await mutateRecords((current) => { + if (!enabled) { + for (const hostId of current.registered) { + current.pending.add(hostId) + } + return + } + current.pending.clear() + }) + await reconcileAllHosts() +} + +export async function setNotificationDeliveryPreferences( + value: NotificationDeliveryPreferences +): Promise { + consentGeneration++ + await saveNotificationDeliveryPreferences(value) + await reconcileAllHosts() +} + +/** Re-registers every connected host so the gateway stores the narrowed filter. */ +export async function setRemotePushAgentStates( + states: readonly RemotePushAgentState[] +): Promise { + consentGeneration++ + await saveRemotePushAgentStates(states) + await reconcileAllHosts() +} + +/** + * Best-effort unregister before the host's credentials are deleted. + * + * Why best-effort is all there is: the credentials are the only way back to that + * host, so a desktop that was offline here keeps its gateway registration and keeps + * pushing to this phone. shouldSuppressForegroundPush drops those in the foreground; + * background alerts stop only when that desktop unpairs the phone, or the switch is + * turned off here. Documented in docs/site/content/docs/notifications.mdx. + */ +export async function unregisterPushForRemovedHost(hostId: string): Promise { + const state = hostsById.get(hostId) + if (state?.client && state.supported !== false) { + await sendUnregister(state.client, REMOVAL_TIMEOUT_MS) + } + hostsById.delete(hostId) + await mutateRecords((current) => { + current.registered.delete(hostId) + current.pending.delete(hostId) + }) +} + +/** A rolled token stops delivering, so re-register every connected host at once. */ +export function startPushTokenSync(): () => void { + return addPushTokenListener((token) => { + tokenPromise = Promise.resolve(token) + void reconcileAllHosts() + }) +} + +export function resetPushRegistrationForTests(): void { + hostsById.clear() + registrationRecords = null + tokenPromise = null + consentGeneration = 0 +} diff --git a/mobile/src/notifications/push-token.test.ts b/mobile/src/notifications/push-token.test.ts new file mode 100644 index 00000000000..2a193430ac6 --- /dev/null +++ b/mobile/src/notifications/push-token.test.ts @@ -0,0 +1,92 @@ +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import * as Notifications from 'expo-notifications' +import { addPushTokenListener, getDevicePushToken } from './push-token' + +vi.mock('expo-notifications', () => ({ + getDevicePushTokenAsync: vi.fn(), + addPushTokenListener: vi.fn() +})) + +const dev = globalThis as { __DEV__?: boolean } + +beforeEach(() => { + vi.clearAllMocks() +}) + +afterEach(() => { + delete dev.__DEV__ +}) + +describe('getDevicePushToken', () => { + it.each([ + [true, 'sandbox'], + [false, 'production'] + ])('reports apnsEnvironment for a __DEV__=%s iOS build as %s', async (isDev, environment) => { + dev.__DEV__ = isDev + vi.mocked(Notifications.getDevicePushTokenAsync).mockResolvedValue({ + type: 'ios', + data: 'a'.repeat(64) + } as never) + + await expect(getDevicePushToken()).resolves.toEqual({ + platform: 'ios', + token: 'a'.repeat(64), + apnsEnvironment: environment + }) + }) + + it('omits apnsEnvironment for Android, where FCM has no environment split', async () => { + vi.mocked(Notifications.getDevicePushTokenAsync).mockResolvedValue({ + type: 'android', + data: 'fcm-registration-token' + } as never) + + await expect(getDevicePushToken()).resolves.toEqual({ + platform: 'android', + token: 'fcm-registration-token' + }) + }) + + it.each([ + ['a web push subscription', { type: 'web', data: { endpoint: 'https://example.test' } }], + ['an empty token', { type: 'ios', data: '' }] + ])('returns null for %s', async (_label, raw) => { + vi.mocked(Notifications.getDevicePushTokenAsync).mockResolvedValue(raw as never) + + await expect(getDevicePushToken()).resolves.toBeNull() + }) + + it('returns null when the shell cannot mint a token at all', async () => { + vi.mocked(Notifications.getDevicePushTokenAsync).mockRejectedValue(new Error('no entitlement')) + + await expect(getDevicePushToken()).resolves.toBeNull() + }) +}) + +describe('addPushTokenListener', () => { + it('forwards a rolled native token and removes the subscription on teardown', () => { + const remove = vi.fn() + let emit: ((raw: unknown) => void) | null = null + vi.mocked(Notifications.addPushTokenListener).mockImplementation((listener) => { + emit = listener as (raw: unknown) => void + return { remove } as never + }) + const seen: unknown[] = [] + + const stop = addPushTokenListener((token) => seen.push(token)) + emit?.({ type: 'android', data: 'rolled' }) + emit?.({ type: 'web', data: {} }) + stop() + + expect(seen).toEqual([{ platform: 'android', token: 'rolled' }]) + expect(remove).toHaveBeenCalledTimes(1) + }) + + it('degrades to a no-op on a shell that cannot subscribe to token changes', () => { + vi.mocked(Notifications.addPushTokenListener).mockImplementation(() => { + throw new Error('no push support') + }) + + expect(() => addPushTokenListener(() => {})()).not.toThrow() + }) +}) diff --git a/mobile/src/notifications/push-token.ts b/mobile/src/notifications/push-token.ts new file mode 100644 index 00000000000..5f29c0ec1dd --- /dev/null +++ b/mobile/src/notifications/push-token.ts @@ -0,0 +1,59 @@ +import * as Notifications from 'expo-notifications' +import type { + MobilePushApnsEnvironment, + MobilePushPlatform +} from '../../../src/shared/mobile-push-contract' + +// Why: the native APNs/FCM token, not an Expo push token — Orca's own gateway +// talks to Apple and Google directly, so it needs the raw device token. + +export type MobilePushToken = { + readonly platform: MobilePushPlatform + readonly token: string + readonly apnsEnvironment?: MobilePushApnsEnvironment +} + +// Dev-client builds are debug and get sandbox APNs; TestFlight and App Store are release. +function apnsEnvironment(): MobilePushApnsEnvironment { + return typeof __DEV__ !== 'undefined' && __DEV__ ? 'sandbox' : 'production' +} + +function toMobilePushToken(raw: { type: string; data: unknown }): MobilePushToken | null { + if (typeof raw.data !== 'string' || raw.data.length === 0) { + return null + } + if (raw.type === 'ios') { + return { platform: 'ios', token: raw.data, apnsEnvironment: apnsEnvironment() } + } + // Web tokens carry an object payload and no Orca gateway path; only native counts. + return raw.type === 'android' ? { platform: 'android', token: raw.data } : null +} + +/** + * The device's native push token, or null when this build cannot have one — + * a simulator, a de-Googled Android device, or a shell without the entitlement. + */ +export async function getDevicePushToken(): Promise { + try { + return toMobilePushToken(await Notifications.getDevicePushTokenAsync()) + } catch { + return null + } +} + +/** Providers can roll a token while the app runs; the old one stops delivering. */ +export function addPushTokenListener(listener: (token: MobilePushToken) => void): () => void { + try { + const subscription = Notifications.addPushTokenListener((raw) => { + const token = toMobilePushToken(raw) + if (token) { + listener(token) + } + }) + return () => subscription.remove() + } catch { + // A shell with no push capability cannot subscribe; the caller is a root-level + // effect, so throwing here would take the whole app down over an optional feature. + return () => {} + } +} diff --git a/mobile/src/notifications/push-tray-dismissal.test.ts b/mobile/src/notifications/push-tray-dismissal.test.ts new file mode 100644 index 00000000000..64ccbf7ebd9 --- /dev/null +++ b/mobile/src/notifications/push-tray-dismissal.test.ts @@ -0,0 +1,57 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import * as Notifications from 'expo-notifications' +import { dismissPresentedPushNotification } from './push-tray-dismissal' + +vi.mock('expo-notifications', () => ({ + getPresentedNotificationsAsync: vi.fn(), + dismissNotificationAsync: vi.fn() +})) + +function presented(identifier: string, data: unknown): unknown { + return { request: { identifier, content: { data } } } +} + +beforeEach(() => { + vi.clearAllMocks() + vi.mocked(Notifications.dismissNotificationAsync).mockResolvedValue(undefined) +}) + +describe('dismissPresentedPushNotification', () => { + it('dismisses only the tray entries whose push payload carries the same notification id', async () => { + vi.mocked(Notifications.getPresentedNotificationsAsync).mockResolvedValue([ + presented('tray-1', { + orca: { hostFingerprint: 'fp0123456789abcd', notificationId: 'agent:one' } + }), + presented('tray-2', { + orca: { hostFingerprint: 'fp0123456789abcd', notificationId: 'agent:two' } + }), + // Flat FCM shape for the same notification, presented on Android. + presented('tray-3', { hostFingerprint: 'fp0123456789abcd', notificationId: 'agent:one' }) + ] as never) + + await dismissPresentedPushNotification('agent:one') + + expect(vi.mocked(Notifications.dismissNotificationAsync).mock.calls.map(([id]) => id)).toEqual([ + 'tray-1', + 'tray-3' + ]) + }) + + it('ignores locally scheduled notifications, which the local registry already owns', async () => { + vi.mocked(Notifications.getPresentedNotificationsAsync).mockResolvedValue([ + presented('tray-1', { hostId: 'host-1', notificationId: 'agent:one' }) + ] as never) + + await dismissPresentedPushNotification('agent:one') + + expect(Notifications.dismissNotificationAsync).not.toHaveBeenCalled() + }) + + it('stays silent on a native shell that cannot query the tray', async () => { + vi.mocked(Notifications.getPresentedNotificationsAsync).mockRejectedValue( + new Error('unavailable') + ) + + await expect(dismissPresentedPushNotification('agent:one')).resolves.toBeUndefined() + }) +}) diff --git a/mobile/src/notifications/push-tray-dismissal.ts b/mobile/src/notifications/push-tray-dismissal.ts new file mode 100644 index 00000000000..850c6488e3c --- /dev/null +++ b/mobile/src/notifications/push-tray-dismissal.ts @@ -0,0 +1,30 @@ +import { readNativeNotificationData } from './native-notification-data' +import * as Notifications from 'expo-notifications' +import { readOrcaPushPayload } from './push-payload' + +/** + * Retire a push the OS presented for a notification the desktop has now dismissed. + * The local scheduling registry knows nothing about it — the OS drew it while Orca + * was closed — so the notification tray is the only place it can be found. + * + * Kept out of push-receive.ts deliberately: this runs on the socket dismiss path, + * which must not pull the host store (and its native keychain deps) behind it. + */ +export async function dismissPresentedPushNotification(notificationId: string): Promise { + try { + const presented = await Notifications.getPresentedNotificationsAsync() + await Promise.all( + presented.map(async (notification) => { + const payload = readOrcaPushPayload(readNativeNotificationData(notification.request)) + if (payload?.notificationId !== notificationId) { + return + } + await Notifications.dismissNotificationAsync(notification.request.identifier).catch( + () => {} + ) + }) + ) + } catch { + // Older native shells lack the tray query; local dismissal still runs. + } +} diff --git a/mobile/src/notifications/push-tray-seen-seed.test.ts b/mobile/src/notifications/push-tray-seen-seed.test.ts new file mode 100644 index 00000000000..377dc9dab18 --- /dev/null +++ b/mobile/src/notifications/push-tray-seen-seed.test.ts @@ -0,0 +1,124 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import * as Notifications from 'expo-notifications' +import { sha256 } from '@noble/hashes/sha256' +import { loadHostCatalog } from '../transport/host-store' +import type { HostCatalogEntry } from '../transport/types' +import { + getHostNotificationSession, + resetHostNotificationSessionsForTests +} from './notification-reconnect-catchup' +import { markPresentedPushesSeen, readPresentedPushSeenKeys } from './push-tray-seen-seed' + +vi.mock('expo-notifications', () => ({ getPresentedNotificationsAsync: vi.fn() })) + +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn() })) + +vi.mock('@react-native-async-storage/async-storage', () => ({ + default: { + getItem: vi.fn(async () => null), + setItem: vi.fn(async () => undefined) + } +})) + +const publicKey = Uint8Array.from({ length: 32 }, (_, index) => index) +const publicKeyB64 = Buffer.from(publicKey).toString('base64') +const hostFingerprint = Buffer.from(sha256(publicKey)).toString('base64url').slice(0, 16) + +const hosts = [{ id: 'host-1', publicKeyB64 }] as unknown as HostCatalogEntry[] + +function presented(orca: Record): unknown { + const identifier = `tray-${String(orca.notificationId ?? 'bell')}` + return { request: { identifier, content: { data: { orca } } } } +} + +beforeEach(() => { + vi.clearAllMocks() + resetHostNotificationSessionsForTests() + vi.mocked(loadHostCatalog).mockResolvedValue(hosts) +}) + +describe('readPresentedPushSeenKeys', () => { + it('keys the tray entries the gateway pushed for this host', async () => { + vi.mocked(Notifications.getPresentedNotificationsAsync).mockResolvedValue([ + presented({ hostFingerprint, notificationId: 'agent:one', notificationSeq: 6 }), + presented({ hostFingerprint, notificationSeq: 7 }) + ] as never) + + await expect(readPresentedPushSeenKeys('host-1')).resolves.toEqual([ + { key: 'id:agent:one#6', epoch: undefined }, + { key: 'seq:7', epoch: undefined } + ]) + }) + + it('ignores a tray entry belonging to another paired host', async () => { + vi.mocked(Notifications.getPresentedNotificationsAsync).mockResolvedValue([ + presented({ hostFingerprint: '0123456789abcdef', notificationId: 'agent:one' }) + ] as never) + + await expect(readPresentedPushSeenKeys('host-1')).resolves.toEqual([]) + }) + + it('ignores a coalesced summary, whose key names a banner nobody has seen', async () => { + vi.mocked(Notifications.getPresentedNotificationsAsync).mockResolvedValue([ + presented({ + hostFingerprint, + notificationId: 'agent:one', + notificationSeq: 6, + coalescedCount: 3 + }) + ] as never) + + await expect(readPresentedPushSeenKeys('host-1')).resolves.toEqual([]) + }) + + it('ignores a locally scheduled notification, which the socket path already owns', async () => { + vi.mocked(Notifications.getPresentedNotificationsAsync).mockResolvedValue([ + { request: { identifier: 'tray-1', content: { data: { hostId: 'host-1' } } } } + ] as never) + + await expect(readPresentedPushSeenKeys('host-1')).resolves.toEqual([]) + expect(loadHostCatalog).toHaveBeenCalled() + }) + + it('stays silent on a native shell that cannot query the tray', async () => { + vi.mocked(Notifications.getPresentedNotificationsAsync).mockRejectedValue( + new Error('unavailable') + ) + + await expect(readPresentedPushSeenKeys('host-1')).resolves.toEqual([]) + }) +}) + +describe('markPresentedPushesSeen', () => { + it('claims the keys without touching the watermark', () => { + const session = getHostNotificationSession('host-1') + session.lastDeliveredEpoch = 'epoch-1' + + markPresentedPushesSeen(session, [{ key: 'id:agent:one#9', epoch: 'epoch-1' }]) + + expect(session.seen.has('id:agent:one#9')).toBe(true) + // A push seq proves one event was shown, not that everything below it was. + expect(session.lastDeliveredSeq).toBe(0) + }) + + it('drops a key that names no counter lifetime at all', () => { + const session = getHostNotificationSession('host-1') + session.lastDeliveredEpoch = 'epoch-1' + + markPresentedPushesSeen(session, [{ key: 'seq:4', epoch: undefined }]) + + // The desktop always sends an epoch; a key without one cannot be shown to belong + // to this counter, and claiming it would drop the real bell at seq 4. + expect(session.seen.has('seq:4')).toBe(false) + }) + + it('drops a key from a desktop lifetime that has already been retired', () => { + const session = getHostNotificationSession('host-1') + session.lastDeliveredEpoch = 'epoch-2' + + markPresentedPushesSeen(session, [{ key: 'seq:4', epoch: 'epoch-1' }]) + + // The new counter re-issues seq 4, so the stale key would drop a real bell. + expect(session.seen.has('seq:4')).toBe(false) + }) +}) diff --git a/mobile/src/notifications/push-tray-seen-seed.ts b/mobile/src/notifications/push-tray-seen-seed.ts new file mode 100644 index 00000000000..a7b83dd7d38 --- /dev/null +++ b/mobile/src/notifications/push-tray-seen-seed.ts @@ -0,0 +1,72 @@ +import { readNativeNotificationData } from './native-notification-data' +import * as Notifications from 'expo-notifications' +import { loadHostCatalog } from '../transport/host-store' +import { seenKeyForEvent, type HostNotificationSession } from './notification-reconnect-catchup' +import { resolveHostIdForFingerprint } from './push-host-fingerprint' +import { readOrcaPushPayload } from './push-payload' + +/** + * Dedup keys for the pushes the OS has already drawn for one host. + * + * Why this exists: a push shown while Orca was closed never ran through the + * foreground handler, so nothing in this process claimed its key. The reconnect + * catch-up then replays that same event and shows a second banner for it. + * + * Kept separate from push-tray-dismissal.ts, which must stay free of the host + * store (and its native keychain deps) because it runs on the socket dismiss path. + */ +export type PresentedPushSeenKey = { readonly key: string; readonly epoch: string | undefined } + +export async function readPresentedPushSeenKeys( + hostId: string +): Promise { + try { + const presented = await Notifications.getPresentedNotificationsAsync() + if (presented.length === 0) { + return [] + } + const hosts = await loadHostCatalog().catch(() => []) + const keys: PresentedPushSeenKey[] = [] + for (const notification of presented) { + const payload = readOrcaPushPayload(readNativeNotificationData(notification.request)) + // A coalesced summary stands in for N events while carrying only the latest + // one's fields, so its key belongs to a banner the user has NOT seen. + if (!payload || (payload.coalescedCount ?? 0) > 1) { + continue + } + if (resolveHostIdForFingerprint(payload.hostFingerprint, hosts) !== hostId) { + continue + } + const key = seenKeyForEvent(payload) + if (key) { + keys.push({ key, epoch: payload.notificationEpoch }) + } + } + return keys + } catch { + // Older native shells lack the tray query; the catch-up replays as it did before. + return [] + } +} + +/** + * Claim the tray's keys on the session, skipping any that do not name the live + * counter lifetime. A push without an epoch cannot be tied to this counter, and + * the desktop always sends one, so it is left unclaimed rather than allowed to + * swallow a real event at the same seq. + * + * The watermark is deliberately untouched: a push seq proves one event was shown, + * not that everything below it was, and advancing past a gap would make the desktop + * cut the notifications in it forever. + */ +export function markPresentedPushesSeen( + session: HostNotificationSession, + keys: readonly PresentedPushSeenKey[] +): void { + for (const { key, epoch } of keys) { + if (epoch == null || epoch !== session.lastDeliveredEpoch) { + continue + } + session.seen.add(key) + } +} diff --git a/mobile/src/notifications/socket-push-delivery-handoff.test.ts b/mobile/src/notifications/socket-push-delivery-handoff.test.ts new file mode 100644 index 00000000000..43dbfa1df73 --- /dev/null +++ b/mobile/src/notifications/socket-push-delivery-handoff.test.ts @@ -0,0 +1,81 @@ +import { beforeEach, expect, it, vi } from 'vitest' +import { AppState } from 'react-native' +import { waitForSocketPushHandoff } from './socket-push-delivery-handoff' +import { readPresentedPushSeenKeys } from './push-tray-seen-seed' +import { loadRemotePushEnabled } from '../storage/preferences' +import { seenKeyForEvent } from './notification-reconnect-catchup' + +let active: ((state: string) => void) | undefined +const remove = vi.fn() +vi.mock('react-native', () => ({ + AppState: { + currentState: 'background', + addEventListener: vi.fn((_event, callback) => { + active = callback + return { remove } + }) + } +})) +vi.mock('../storage/preferences', () => ({ + loadRemotePushEnabled: vi.fn(async () => true), + loadRemotePushHostRegistrations: vi.fn(async () => ({ registeredHostIds: ['host'] })) +})) +vi.mock('./push-tray-seen-seed', () => ({ readPresentedPushSeenKeys: vi.fn(async () => []) })) +const event = { + type: 'notification' as const, + source: 'agent-task-complete' as const, + title: 'Done', + body: '', + notificationId: 'done', + notificationSeq: 1, + notificationEpoch: 'epoch' +} +beforeEach(() => { + vi.clearAllMocks() + active = undefined + AppState.currentState = 'background' +}) + +it('waits for foreground and suppresses a live socket event already delivered by APNs', async () => { + vi.mocked(readPresentedPushSeenKeys).mockResolvedValue([ + { key: seenKeyForEvent(event)!, epoch: 'epoch' } + ]) + const delivery = waitForSocketPushHandoff(event, 'host', new AbortController().signal) + await vi.waitFor(() => expect(active).toBeDefined()) + expect(readPresentedPushSeenKeys).not.toHaveBeenCalled() + AppState.currentState = 'active' + active?.('active') + expect(await delivery).toBe(false) + expect(remove).toHaveBeenCalledOnce() +}) + +it('falls back to local delivery on foreground when no provider notification arrived', async () => { + vi.mocked(readPresentedPushSeenKeys).mockResolvedValue([]) + const delivery = waitForSocketPushHandoff(event, 'host', new AbortController().signal) + await vi.waitFor(() => expect(active).toBeDefined()) + AppState.currentState = 'active' + active?.('active') + expect(await delivery).toBe(true) +}) + +it('releases the background wait when the subscription is disposed', async () => { + const controller = new AbortController() + const delivery = waitForSocketPushHandoff(event, 'host', controller.signal) + await vi.waitFor(() => expect(active).toBeDefined()) + controller.abort() + expect(await delivery).toBe(false) + expect(remove).toHaveBeenCalledOnce() +}) + +it('keeps local background delivery when remote push is disabled', async () => { + vi.mocked(loadRemotePushEnabled).mockResolvedValueOnce(false) + expect(await waitForSocketPushHandoff(event, 'host', new AbortController().signal)).toBe(true) + expect(active).toBeUndefined() +}) + +it('leaves hosts without a registered push token on local delivery', async () => { + expect( + await waitForSocketPushHandoff(event, 'unregistered-host', new AbortController().signal) + ).toBe(true) + expect(active).toBeUndefined() +}) diff --git a/mobile/src/notifications/socket-push-delivery-handoff.ts b/mobile/src/notifications/socket-push-delivery-handoff.ts new file mode 100644 index 00000000000..25dc27b00a3 --- /dev/null +++ b/mobile/src/notifications/socket-push-delivery-handoff.ts @@ -0,0 +1,49 @@ +import { AppState } from 'react-native' +import { loadRemotePushEnabled, loadRemotePushHostRegistrations } from '../storage/preferences' +import { readPresentedPushSeenKeys } from './push-tray-seen-seed' +import { seenKeyForEvent } from './notification-reconnect-catchup' +import type { NotificationEvent } from './local-notification-scheduling' + +function waitUntilActive(signal: AbortSignal): Promise { + if (AppState.currentState === 'active' || signal.aborted) { + return Promise.resolve() + } + return new Promise((resolve) => { + const finish = () => { + subscription.remove() + signal.removeEventListener('abort', finish) + resolve() + } + const subscription = AppState.addEventListener('change', (state) => { + if (state === 'active') { + finish() + } + }) + signal.addEventListener('abort', finish, { once: true }) + if (signal.aborted || AppState.currentState === 'active') { + finish() + } + }) +} + +export async function waitForSocketPushHandoff( + event: NotificationEvent, + hostId: string, + signal: AbortSignal +): Promise { + if (!(await loadRemotePushEnabled())) { + return true + } + const registrations = await loadRemotePushHostRegistrations() + if (!registrations.registeredHostIds.includes(hostId)) { + return true + } + // iOS can keep the socket alive while backgrounded; let APNs own that interval. + await waitUntilActive(signal) + if (signal.aborted) { + return false + } + const key = seenKeyForEvent(event) + const presented = await readPresentedPushSeenKeys(hostId) + return !presented.some((push) => push.key === key && push.epoch === event.notificationEpoch) +} diff --git a/mobile/src/notifications/use-remote-push-capable-hosts.test.tsx b/mobile/src/notifications/use-remote-push-capable-hosts.test.tsx new file mode 100644 index 00000000000..511406b5c06 --- /dev/null +++ b/mobile/src/notifications/use-remote-push-capable-hosts.test.tsx @@ -0,0 +1,176 @@ +import { createElement } from 'react' +import { act, create, type ReactTestRenderer } from 'react-test-renderer' +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { loadHostCatalog } from '../transport/host-store' +import type { HostCatalogEntry } from '../transport/types' +import type { RpcClient } from '../transport/rpc-client' +import { startRuntimeCapabilityProbe } from '../transport/runtime-capability-probe' +import { useAllHostClients } from '../transport/use-all-host-clients' +import { + useRemotePushCapableHosts, + type RemotePushHostSupport +} from './use-remote-push-capable-hosts' + +vi.mock('../transport/host-store', () => ({ loadHostCatalog: vi.fn() })) +vi.mock('../transport/use-all-host-clients', () => ({ useAllHostClients: vi.fn() })) +vi.mock('../transport/runtime-capability-probe', () => ({ + startRuntimeCapabilityProbe: vi.fn() +})) + +// The real module reaches expo-notifications and the preference store for the token +// path; only the capability string matters here. +vi.mock('./push-registration', () => ({ + NOTIFICATIONS_REMOTE_PUSH_CAPABILITY: 'notifications.remote-push.v1' +})) + +const CAPABILITY = 'notifications.remote-push.v1' + +type ClientEntry = { hostId: string; client: RpcClient; state: string } + +/** Distinct object per host, so identity changes are the thing under test. */ +function clientFor(hostId: string): RpcClient { + return { hostId } as unknown as RpcClient +} + +let renderer: ReactTestRenderer | null = null +let latest: RemotePushHostSupport = { supported: false, resolved: false } +const answerByHostId = new Map void>() +const stopProbe = vi.fn() + +function Harness(): null { + latest = useRemotePushCapableHosts() + return null +} + +async function mount(): Promise { + await act(async () => { + renderer = create(createElement(Harness)) + await Promise.resolve() + }) +} + +async function setClients(entries: readonly ClientEntry[]): Promise { + vi.mocked(useAllHostClients).mockReturnValue(entries as never) + await act(async () => { + renderer?.update(createElement(Harness)) + await Promise.resolve() + }) +} + +async function answer(hostId: string, capabilities: readonly string[]): Promise { + await act(async () => { + answerByHostId.get(hostId)?.(capabilities) + await Promise.resolve() + }) +} + +beforeEach(() => { + vi.clearAllMocks() + answerByHostId.clear() + latest = { supported: false, resolved: false } + vi.mocked(useAllHostClients).mockReturnValue([] as never) + vi.mocked(startRuntimeCapabilityProbe).mockImplementation((client, onCapabilities) => { + answerByHostId.set((client as unknown as { hostId: string }).hostId, onCapabilities) + return stopProbe + }) + vi.mocked(loadHostCatalog).mockResolvedValue([ + { id: 'host-1', publicKeyB64: 'k1' }, + { id: 'host-2', publicKeyB64: 'k2' } + ] as unknown as HostCatalogEntry[]) +}) + +afterEach(() => { + act(() => renderer?.unmount()) + renderer = null +}) + +describe('useRemotePushCapableHosts', () => { + it('stays unresolved when the host catalog cannot be read', async () => { + vi.mocked(loadHostCatalog).mockRejectedValue(new Error('keychain locked')) + + await mount() + + // Resolving here would render "Update your desktop app" at someone whose desktop + // is already current, on the strength of a catalog read that simply failed. + expect(latest).toEqual({ supported: false, resolved: false }) + }) + + it('waits for every connected host before answering', async () => { + await mount() + await setClients([ + { hostId: 'host-1', client: clientFor('host-1'), state: 'connected' }, + { hostId: 'host-2', client: clientFor('host-2'), state: 'connected' } + ]) + + await answer('host-1', [CAPABILITY]) + expect(latest.resolved).toBe(false) + + await answer('host-2', ['some-other.v1']) + expect(latest).toEqual({ supported: true, resolved: true }) + }) + + it('keeps the answer of a host that has since disconnected', async () => { + await mount() + const client = clientFor('host-1') + await setClients([{ hostId: 'host-1', client, state: 'connected' }]) + await answer('host-1', [CAPABILITY]) + + await setClients([{ hostId: 'host-1', client, state: 'connecting' }]) + + expect(latest).toEqual({ supported: true, resolved: true }) + }) + + it('resolves immediately when nothing is paired', async () => { + vi.mocked(loadHostCatalog).mockResolvedValue([]) + + await mount() + + expect(latest).toEqual({ supported: false, resolved: true }) + }) + + it('leaves a running probe alone when another host changes state', async () => { + await mount() + const first = clientFor('host-1') + await setClients([{ hostId: 'host-1', client: first, state: 'connected' }]) + expect(startRuntimeCapabilityProbe).toHaveBeenCalledTimes(1) + + // useAllHostClients rebuilds its array on every connection tick, so a plain + // dependency on it would tear down and restart host-1's probe here. + await setClients([ + { hostId: 'host-1', client: first, state: 'connected' }, + { hostId: 'host-2', client: clientFor('host-2'), state: 'connecting' } + ]) + await setClients([ + { hostId: 'host-1', client: first, state: 'connected' }, + { hostId: 'host-2', client: clientFor('host-2'), state: 'connected' } + ]) + + expect(stopProbe).not.toHaveBeenCalled() + expect( + vi.mocked(startRuntimeCapabilityProbe).mock.calls.map(([client]) => client) + ).toHaveLength(2) + }) + + it('restarts the probe when a reconnect replaces the host client', async () => { + await mount() + await setClients([{ hostId: 'host-1', client: clientFor('host-1'), state: 'connected' }]) + + await setClients([{ hostId: 'host-1', client: clientFor('host-1'), state: 'connected' }]) + + expect(stopProbe).toHaveBeenCalledTimes(1) + expect(startRuntimeCapabilityProbe).toHaveBeenCalledTimes(2) + }) + + it('ignores an answer from a host the catalog no longer lists', async () => { + await mount() + await setClients([ + { hostId: 'host-ghost', client: clientFor('host-ghost'), state: 'connected' } + ]) + + await answer('host-ghost', [CAPABILITY]) + + // An unpaired desktop cannot push to this phone, so its vote must not offer + // the switch — nor count as the answer that resolves the section. + expect(latest).toEqual({ supported: false, resolved: false }) + }) +}) diff --git a/mobile/src/notifications/use-remote-push-capable-hosts.ts b/mobile/src/notifications/use-remote-push-capable-hosts.ts new file mode 100644 index 00000000000..a89ed79ff6b --- /dev/null +++ b/mobile/src/notifications/use-remote-push-capable-hosts.ts @@ -0,0 +1,105 @@ +import { useEffect, useRef, useState } from 'react' +import { loadHostCatalog } from '../transport/host-store' +import type { RpcClient } from '../transport/rpc-client' +import { startRuntimeCapabilityProbe } from '../transport/runtime-capability-probe' +import { useAllHostClients } from '../transport/use-all-host-clients' +import { NOTIFICATIONS_REMOTE_PUSH_CAPABILITY } from './push-registration' + +export type RemotePushHostSupport = { + /** At least one paired host advertises `notifications.remote-push.v1`. */ + supported: boolean + /** Whether the answer above is final rather than "nobody has replied yet". */ + resolved: boolean +} + +/** + * Whether background push can be offered at all. The desktop advertises the + * capability in `status.get`, so the answer needs a connected host — until one + * replies the screen must stay silent rather than tell someone to update a + * desktop that is already current. + */ +export function useRemotePushCapableHosts(): RemotePushHostSupport { + const [hostIds, setHostIds] = useState([]) + const [hostsLoaded, setHostsLoaded] = useState(false) + const [supportedByHostId, setSupportedByHostId] = useState>({}) + const probesRef = useRef(new Map void }>()) + + useEffect(() => { + let cancelled = false + void loadHostCatalog() + .then((hosts) => { + if (!cancelled) { + setHostIds(hosts.map((host) => host.id)) + setHostsLoaded(true) + } + }) + // Why nothing on failure: an unread catalog marked loaded resolves the answer as + // "no paired host supports push", which tells the user to update a current desktop. + .catch(() => {}) + return () => { + cancelled = true + } + }, []) + + const clients = useAllHostClients(hostIds) + + // Why pruned rather than left: an answer for a host that is no longer paired is a + // vote from a desktop this phone cannot receive a push from. + useEffect(() => { + setSupportedByHostId((previous) => { + const kept = Object.entries(previous).filter(([hostId]) => hostIds.includes(hostId)) + return kept.length === Object.keys(previous).length ? previous : Object.fromEntries(kept) + }) + }, [hostIds]) + + // Why diffed by client identity rather than restarted on every `clients` value: + // useAllHostClients rebuilds the array on each connection tick, so a plain + // dependency tears down and re-runs every host's probe whenever any host moves. + useEffect(() => { + const connected = new Map( + clients + .filter((entry) => entry.state === 'connected') + .map((entry) => [entry.hostId, entry.client]) + ) + const probes = probesRef.current + for (const [hostId, probe] of probes) { + if (connected.get(hostId) !== probe.client) { + probe.stop() + probes.delete(hostId) + } + } + for (const [hostId, client] of connected) { + if (!probes.has(hostId)) { + const stop = startRuntimeCapabilityProbe(client, (capabilities) => { + setSupportedByHostId((previous) => ({ + ...previous, + [hostId]: capabilities.includes(NOTIFICATIONS_REMOTE_PUSH_CAPABILITY) + })) + }) + probes.set(hostId, { client, stop }) + } + } + }, [clients]) + + useEffect(() => { + const probes = probesRef.current + return () => { + for (const probe of probes.values()) { + probe.stop() + } + probes.clear() + } + }, []) + + const answeredHostIds = hostIds.filter((hostId) => hostId in supportedByHostId) + return { + supported: answeredHostIds.some((hostId) => supportedByHostId[hostId] === true), + // A connected host that has not answered yet is exactly the case the silence is + // for, so one outstanding probe holds the whole section back. Disconnected hosts + // do not: their earlier answer stands, and one that never answered never will. + resolved: + (hostsLoaded && hostIds.length === 0) || + (answeredHostIds.length > 0 && + clients.every((entry) => entry.state !== 'connected' || entry.hostId in supportedByHostId)) + } +} diff --git a/mobile/src/session/mobile-native-chat-session-option-persistence.test.ts b/mobile/src/session/mobile-native-chat-session-option-persistence.test.ts new file mode 100644 index 00000000000..28be1af40e2 --- /dev/null +++ b/mobile/src/session/mobile-native-chat-session-option-persistence.test.ts @@ -0,0 +1,99 @@ +import { describe, expect, it, vi } from 'vitest' +import { + applyNativeChatSessionOptionSettingsMutation, + resolveStructuredLaunchSeedOptions +} from '../../../src/shared/native-chat-session-option-defaults' +import type { PersistedNativeChatSessionOptions } from '../../../src/shared/native-chat-session-options' +import type { RpcClient } from '../transport/rpc-client' +import { persistMobileStructuredOptionPicks } from './mobile-native-chat-session-option-persistence' + +function hostClient(initial?: PersistedNativeChatSessionOptions) { + let stored = initial + const sendRequest = vi.fn(async (method: string, params?: unknown) => { + expect(method).toBe('settings.mutateNativeChatSessionOptions') + const next = applyNativeChatSessionOptionSettingsMutation( + stored, + params as Parameters[1] + ) + stored = next ?? stored + return { id: '2', ok: true as const, result: null, _meta: { runtimeId: 'host' } } + }) + return { client: { sendRequest } as unknown as RpcClient, sendRequest, read: () => stored } +} + +describe('persistMobileStructuredOptionPicks', () => { + it('writes the pick to the host record a later launch seeds from', async () => { + const host = hostClient() + await persistMobileStructuredOptionPicks({ + client: host.client, + agent: 'codex', + picks: [{ modelId: 'gpt-fast', optionId: 'effort', value: 'low' }] + }) + expect(resolveStructuredLaunchSeedOptions(host.read(), 'codex')).toEqual({ + model: 'gpt-fast', + effort: 'low' + }) + }) + + it('merges onto the host record instead of replacing another agent', async () => { + const host = hostClient({ + claude: { model: 'opus', valuesByModel: { opus: { effort: 'high' } } } + }) + await persistMobileStructuredOptionPicks({ + client: host.client, + agent: 'codex', + picks: [{ modelId: 'gpt-fast', optionId: 'model', value: 'gpt-fast' }] + }) + expect(resolveStructuredLaunchSeedOptions(host.read(), 'claude')).toEqual({ + model: 'opus', + effort: 'high' + }) + }) + + it('sends concurrent deltas that preserve both picks on the host', async () => { + const host = hostClient() + const first = persistMobileStructuredOptionPicks({ + client: host.client, + agent: 'codex', + picks: [{ modelId: 'gpt-fast', optionId: 'effort', value: 'low' }] + }) + const second = persistMobileStructuredOptionPicks({ + client: host.client, + agent: 'claude', + picks: [{ modelId: 'opus', optionId: 'effort', value: 'high' }] + }) + await Promise.all([first, second]) + expect(resolveStructuredLaunchSeedOptions(host.read(), 'codex')).toEqual({ + model: 'gpt-fast', + effort: 'low' + }) + expect(resolveStructuredLaunchSeedOptions(host.read(), 'claude')).toEqual({ + model: 'opus', + effort: 'high' + }) + }) + + it('stays silent without a host or without picks', async () => { + const host = hostClient() + await persistMobileStructuredOptionPicks({ client: null, agent: 'codex', picks: [] }) + await persistMobileStructuredOptionPicks({ client: host.client, agent: 'codex', picks: [] }) + expect(host.sendRequest).not.toHaveBeenCalled() + }) + + it('uses one targeted host mutation instead of a settings read-modify-write', async () => { + const host = hostClient() + await persistMobileStructuredOptionPicks({ + client: host.client, + agent: 'codex', + picks: [{ modelId: 'gpt-fast', optionId: 'model', value: 'gpt-fast' }] + }) + expect(host.sendRequest).toHaveBeenCalledExactlyOnceWith( + 'settings.mutateNativeChatSessionOptions', + { + type: 'apply-picks', + agent: 'codex', + picks: [{ modelId: 'gpt-fast', optionId: 'model', value: 'gpt-fast' }] + } + ) + }) +}) diff --git a/mobile/src/session/mobile-native-chat-session-option-persistence.ts b/mobile/src/session/mobile-native-chat-session-option-persistence.ts new file mode 100644 index 00000000000..aaad5a92f51 --- /dev/null +++ b/mobile/src/session/mobile-native-chat-session-option-persistence.ts @@ -0,0 +1,25 @@ +import type { AgentType } from '../../../src/shared/agent-status-types' +import type { StructuredSessionOptionPick } from '../../../src/shared/structured-agent-session-options' +import type { RpcClient } from '../transport/rpc-client' + +/** The host owns the record a later launch seeds from, so a phone-side pick writes there + * rather than to any client-local store. Best-effort: a failed write only costs the + * next session its remembered start. */ +export function persistMobileStructuredOptionPicks(args: { + client: RpcClient | null + agent: AgentType + picks: readonly StructuredSessionOptionPick[] +}): Promise { + const { agent, client, picks } = args + if (!client || picks.length === 0) { + return Promise.resolve() + } + return client + .sendRequest('settings.mutateNativeChatSessionOptions', { + type: 'apply-picks', + agent, + picks + }) + .then(() => undefined) + .catch(() => undefined) +} diff --git a/mobile/src/session/use-mobile-structured-agent-options.ts b/mobile/src/session/use-mobile-structured-agent-options.ts index 108275223be..7c8651ffda6 100644 --- a/mobile/src/session/use-mobile-structured-agent-options.ts +++ b/mobile/src/session/use-mobile-structured-agent-options.ts @@ -15,6 +15,7 @@ import { commitStructuredAgentSessionOption, commitStructuredAgentSessionOptionValues, createStructuredAgentSessionOptionState, + structuredAgentSessionOptionPicks, structuredAgentSessionOptionSnapshot } from '../../../src/shared/structured-agent-session-options' import type { RpcClient } from '../transport/rpc-client' @@ -22,6 +23,7 @@ import { callAgentSession, type StructuredAgentSessionMutate } from './mobile-structured-agent-session-rpc' +import { persistMobileStructuredOptionPicks } from './mobile-native-chat-session-option-persistence' type StructuredOptionsController = { optionSnapshot: SessionOptionDescriptor[] @@ -101,14 +103,22 @@ export function useMobileStructuredAgentOptions(args: { return result.status !== 'rejected' } if (result.status === 'accepted') { + const committed = result.value.options ?? { [id]: value } setOptionState((current) => current.record === targetRecord && result.sameFence - ? commitStructuredAgentSessionOptionValues( - current, - result.value.options ?? { [id]: value } - ) + ? commitStructuredAgentSessionOptionValues(current, committed) : current ) + // Only an accepted pick: an `unknown` outcome commits optimistically to the + // visible record, and remembering one the provider refused would seed a + // launch the user never chose. + if (agent === 'claude' || agent === 'codex') { + void persistMobileStructuredOptionPicks({ + client, + agent, + picks: structuredAgentSessionOptionPicks(optionState, committed) + }) + } return true } if (result.status === 'unknown') { @@ -128,7 +138,7 @@ export function useMobileStructuredAgentOptions(args: { ) } }, - [mutate, optionState] + [agent, client, mutate, optionState] ) const invokeStructuredOption = useCallback(async () => false, []) diff --git a/mobile/src/session/use-mobile-structured-agent-session.test.tsx b/mobile/src/session/use-mobile-structured-agent-session.test.tsx index 83562839363..d888e74c921 100644 --- a/mobile/src/session/use-mobile-structured-agent-session.test.tsx +++ b/mobile/src/session/use-mobile-structured-agent-session.test.tsx @@ -138,7 +138,7 @@ function runningStatusItem(): AgentJournalRenderItem { } } -function defaultSendRequest(method: string, params?: Record) { +async function defaultSendRequest(method: string, params?: Record) { if (method === 'agentSession.send') { return ok({ ok: true, @@ -420,6 +420,11 @@ describe('useMobileStructuredAgentSession', () => { }), expect.any(Object) ) + expect(sendRequest).toHaveBeenCalledWith('settings.mutateNativeChatSessionOptions', { + type: 'apply-picks', + agent: 'codex', + picks: [{ modelId: 'gpt-fast', optionId: 'model', value: 'gpt-fast' }] + }) await act(async () => { expect(await hook.respondPermission(hook.permission!.options[0]!.send)).toBe(true) diff --git a/mobile/src/storage/preferences.ts b/mobile/src/storage/preferences.ts index 5173ac5bc8a..37d237f7bd7 100644 --- a/mobile/src/storage/preferences.ts +++ b/mobile/src/storage/preferences.ts @@ -1,4 +1,14 @@ +import { + loadNotificationDeliveryPreferences, + notificationPreferencesFilter, + saveNotificationDeliveryPreferences +} from '../notifications/notification-delivery-preferences' import AsyncStorage from '@react-native-async-storage/async-storage' +import { + MOBILE_PUSH_AGENT_STATES, + type MobilePushAgentState, + type MobilePushFilter +} from '../../../src/shared/mobile-push-contract' const PINS_PREFIX = 'orca:pins:' const NOTIF_KEY = 'orca:pushNotificationsEnabled' @@ -30,6 +40,98 @@ export async function savePushNotificationsEnabled(enabled: boolean): Promise { + try { + return (await AsyncStorage.getItem(REMOTE_PUSH_KEY)) === 'true' + } catch { + return false + } +} + +export async function saveRemotePushEnabled(enabled: boolean): Promise { + await AsyncStorage.setItem(REMOTE_PUSH_KEY, String(enabled)) +} + +function remotePushAgentStates(value: unknown): RemotePushAgentState[] { + return stringArray(value).filter((state): state is RemotePushAgentState => + (MOBILE_PUSH_AGENT_STATES as readonly string[]).includes(state) + ) +} + +// Both states default on; an absent key is a device that never opened the section. +export async function loadRemotePushAgentStates(): Promise { + try { + const raw = await AsyncStorage.getItem(REMOTE_PUSH_AGENT_STATES_KEY) + return raw === null ? MOBILE_PUSH_AGENT_STATES : remotePushAgentStates(JSON.parse(raw)) + } catch { + return MOBILE_PUSH_AGENT_STATES + } +} + +export async function saveRemotePushAgentStates( + states: readonly RemotePushAgentState[] +): Promise { + const current = await loadNotificationDeliveryPreferences() + await saveNotificationDeliveryPreferences({ + ...current, + followDesktop: false, + taskFinished: states.includes('finished'), + needsInput: states.includes('needs-input') + }) + await AsyncStorage.setItem(REMOTE_PUSH_AGENT_STATES_KEY, JSON.stringify([...states])) +} + +export async function loadRemotePushFilter(): Promise { + return notificationPreferencesFilter(await loadNotificationDeliveryPreferences()) +} + +// Why persisted: switching off while a host is offline leaves a token the gateway +// would still push to. The pending list is the phone's side of the desktop's +// unregister outbox — it survives a restart so the retry actually happens. +export type RemotePushHostRegistrations = { + readonly registeredHostIds: readonly string[] + readonly pendingUnregisterHostIds: readonly string[] +} + +const EMPTY_REMOTE_PUSH_HOST_REGISTRATIONS: RemotePushHostRegistrations = { + registeredHostIds: [], + pendingUnregisterHostIds: [] +} + +export async function loadRemotePushHostRegistrations(): Promise { + try { + const raw = await AsyncStorage.getItem(REMOTE_PUSH_HOST_REGISTRATIONS_KEY) + if (!raw) { + return EMPTY_REMOTE_PUSH_HOST_REGISTRATIONS + } + const parsed = JSON.parse(raw) as Record + return { + registeredHostIds: stringArray(parsed.registeredHostIds), + pendingUnregisterHostIds: stringArray(parsed.pendingUnregisterHostIds) + } + } catch { + return EMPTY_REMOTE_PUSH_HOST_REGISTRATIONS + } +} + +export async function saveRemotePushHostRegistrations( + value: RemotePushHostRegistrations +): Promise { + await AsyncStorage.setItem(REMOTE_PUSH_HOST_REGISTRATIONS_KEY, JSON.stringify(value)) +} + const TEXT_SCALE_KEY = 'orca:terminalTextScale' // Why: the mobile terminal fits the desktop's full column count to the phone diff --git a/mobile/src/transport/host-removal-lifecycle.test.ts b/mobile/src/transport/host-removal-lifecycle.test.ts index 6c96ef1c446..26d6569afb9 100644 --- a/mobile/src/transport/host-removal-lifecycle.test.ts +++ b/mobile/src/transport/host-removal-lifecycle.test.ts @@ -1,6 +1,7 @@ import { beforeEach, describe, expect, it, vi } from 'vitest' const removeHostMock = vi.hoisted(() => vi.fn()) +const unregisterPushMock = vi.hoisted(() => vi.fn(async () => {})) const asyncStorage = vi.hoisted(() => ({ getItem: vi.fn(async () => null), setItem: vi.fn(async () => undefined), @@ -16,6 +17,12 @@ vi.mock('./host-store', () => ({ removeHost: (hostId: string) => removeHostMock(hostId) })) +// Why mocked: the real module reaches expo-notifications for the device token, which +// no node test environment can load. +vi.mock('../notifications/push-registration', () => ({ + unregisterPushForRemovedHost: (hostId: string) => unregisterPushMock(hostId) +})) + import { removeHostAndCloseClient } from './host-removal-lifecycle' import { getHostNotificationSession, @@ -25,6 +32,7 @@ import { describe('host removal lifecycle', () => { beforeEach(() => { removeHostMock.mockReset() + unregisterPushMock.mockClear() asyncStorage.removeItem.mockClear() resetHostNotificationSessionsForTests() }) @@ -75,6 +83,27 @@ describe('host removal lifecycle', () => { expect(afterRemoval.lastDeliveredEpoch).toBeNull() }) + it('drops the gateway push registration before the credentials it needs are gone', async () => { + removeHostMock.mockResolvedValue(undefined) + + await removeHostAndCloseClient('host-1', vi.fn()) + + expect(unregisterPushMock).toHaveBeenCalledWith('host-1') + expect(unregisterPushMock.mock.invocationCallOrder[0]).toBeLessThan( + removeHostMock.mock.invocationCallOrder[0] + ) + }) + + it('still removes the host when the push unregister cannot land', async () => { + removeHostMock.mockResolvedValue(undefined) + unregisterPushMock.mockRejectedValueOnce(new Error('socket closed')) + const closeHostClient = vi.fn() + + await removeHostAndCloseClient('host-1', closeHostClient) + + expect(closeHostClient).toHaveBeenCalledWith('host-1') + }) + it('erases the persisted watermark, not just the in-memory session', async () => { // Why separately from the test above: the session is process-local, the // watermark is not. Retiring only the session lets a re-pair of the same host diff --git a/mobile/src/transport/host-removal-lifecycle.ts b/mobile/src/transport/host-removal-lifecycle.ts index cd0a09cb67e..488e4e3f9fe 100644 --- a/mobile/src/transport/host-removal-lifecycle.ts +++ b/mobile/src/transport/host-removal-lifecycle.ts @@ -2,12 +2,16 @@ import { clearWatermark, forgetHostNotificationSession } from '../notifications/notification-reconnect-catchup' +import { unregisterPushForRemovedHost } from '../notifications/push-registration' import { removeHost } from './host-store' export async function removeHostAndCloseClient( hostId: string, forgetHostClient: (hostId: string) => void ): Promise { + // Why before removeHost: the unregister needs the still-authenticated client, and + // the desktop's own revoke path covers the case where this call cannot land. + await unregisterPushForRemovedHost(hostId).catch(() => {}) // Why: closing before the metadata commit can strand a still-paired host on // storage failure; closing immediately after success prevents socket leaks. await removeHost(hostId) diff --git a/package.json b/package.json index 7b27dffc8c7..25152e84849 100644 --- a/package.json +++ b/package.json @@ -105,7 +105,7 @@ "test:e2e:workspace-session-golden": "pnpm run ensure:electron-runtime && npx playwright test tests/e2e/golden-quit-relaunch-session.spec.ts tests/e2e/golden-terminal-file-link.spec.ts tests/e2e/golden-worktree-create-switch.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1", "test:e2e:multi-client-navigation": "node config/scripts/run-multi-client-navigation-e2e.mjs", "test:e2e:floating-mobile-emulator": "pnpm run ensure:electron-runtime && npx playwright test tests/e2e/floating-mobile-emulator-tab.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1", - "test:e2e:terminal-rendering-golden": "pnpm run ensure:electron-runtime && npx playwright test tests/e2e/terminal-raw-emoji-table-scroll-restore.spec.ts tests/e2e/terminal-webgl-atlas-budget.spec.ts --grep @terminal-rendering-golden --config tests/playwright.config.ts --project electron-headless --workers=1", + "test:e2e:terminal-rendering-golden": "pnpm run ensure:electron-runtime && npx playwright test tests/e2e/terminal-raw-emoji-table-scroll-restore.spec.ts tests/e2e/terminal-webgl-atlas-budget.spec.ts --grep @terminal-rendering-golden --config tests/playwright.config.ts --project electron-headless --project electron-headful --workers=1", "test:e2e:source-control-golden": "pnpm run ensure:electron-runtime && npx playwright test tests/e2e/golden-file-open-edit-save.spec.ts tests/e2e/golden-source-control-commit.spec.ts tests/e2e/golden-source-control-open-diff.spec.ts --grep @golden --config tests/playwright.config.ts --project electron-headless --workers=1", "test:e2e:posix-profile-index-golden": "pnpm run ensure:electron-runtime && npx playwright test tests/e2e/golden-posix-fresh-startup.spec.ts tests/e2e/golden-posix-profile-index-fsync.spec.ts --grep @posix-profile-index-golden --config tests/playwright.config.ts --project electron-headless --workers=1", "test:e2e:windows-fresh-startup-golden": "pnpm run ensure:electron-runtime && npx playwright test tests/e2e/golden-windows-fresh-startup.spec.ts --grep @windows-fresh-startup-golden --config tests/playwright.config.ts --project electron-headless --workers=1", diff --git a/src/cli/handlers/orchestration-caller-identity-cli.test.ts b/src/cli/handlers/orchestration-caller-identity-cli.test.ts new file mode 100644 index 00000000000..dc272e4b94b --- /dev/null +++ b/src/cli/handlers/orchestration-caller-identity-cli.test.ts @@ -0,0 +1,377 @@ +/** + * How the orchestration CLI decides WHO is speaking. + * + * Split out of `orchestration.test.ts`, which sat exactly on the test-file line ceiling: these two + * suites are one subject — the coordinator and task-creator identity a command carries — and both + * exercise the env-handle validation and pane-remint chain rather than flag-to-param mapping. + */ + +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' + +const callMock = vi.fn() +const getTerminalHandleMock = vi.hoisted(() => vi.fn()) +const originalTerminalHandle = process.env.ORCA_TERMINAL_HANDLE +const originalPaneKey = process.env.ORCA_PANE_KEY +// Why: isolate the handler's flag-to-param mapping; printResult only writes output. +vi.mock('../format', () => ({ printResult: vi.fn() })) +vi.mock('../selectors', () => ({ getTerminalHandle: getTerminalHandleMock })) + +import { ORCHESTRATION_HANDLERS } from './orchestration' +import { RuntimeClientError } from '../runtime-client' + +function staleHandleError(): RuntimeClientError { + return new RuntimeClientError('terminal_handle_stale', 'terminal_handle_stale') +} + +// Queues the stale-handle remint chain shared by coordinator commands: +// `terminal.resolveIdentity` answers not-live → resolvePane returns liveHandle → downstream RPC. +function stubStaleHandleRemint(liveHandle: string, downstream: unknown): void { + callMock + .mockResolvedValueOnce(notLiveIdentity()) + .mockResolvedValueOnce({ result: { terminal: { handle: liveHandle } } }) + .mockResolvedValueOnce(downstream) +} + +// Queues a not-live identity followed by a resolvePane remint that fails with `error`. +function stubStaleHandleRemintFailure(error: RuntimeClientError): void { + callMock.mockResolvedValueOnce(notLiveIdentity()).mockRejectedValueOnce(error) +} + +/** What the runtime answers for a handle whose leaf check reports `terminal_handle_stale`. */ +function notLiveIdentity(): { result: { identity: { live: false } } } { + return { result: { identity: { live: false } } } +} + +function liveIdentity(handle: string): { result: { identity: { handle: string; live: true } } } { + return { result: { identity: { handle, live: true } } } +} + +afterEach(() => { + getTerminalHandleMock.mockReset() + if (originalTerminalHandle === undefined) { + delete process.env.ORCA_TERMINAL_HANDLE + } else { + process.env.ORCA_TERMINAL_HANDLE = originalTerminalHandle + } + if (originalPaneKey === undefined) { + delete process.env.ORCA_PANE_KEY + } else { + process.env.ORCA_PANE_KEY = originalPaneKey + } +}) + +describe('orchestration dispatch coordinator handle', () => { + beforeEach(() => { + callMock.mockReset() + getTerminalHandleMock.mockReset() + delete process.env.ORCA_TERMINAL_HANDLE + delete process.env.ORCA_PANE_KEY + }) + + const invokeDispatch = (flags: Map) => + ORCHESTRATION_HANDLERS['orchestration dispatch']({ + flags, + client: { call: callMock }, + cwd: '/tmp/repo', + json: true + } as never) + + const invokeDispatchShow = (flags: Map) => + ORCHESTRATION_HANDLERS['orchestration dispatch-show']({ + flags, + client: { call: callMock }, + cwd: '/tmp/repo', + json: true + } as never) + + const invokeRun = (flags: Map) => + ORCHESTRATION_HANDLERS['orchestration coordinator-start']({ + flags, + client: { call: callMock }, + cwd: '/tmp/repo', + json: true + } as never) + + it('remints a stale coordinator env handle from the caller pane key', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_stale_coord' + process.env.ORCA_PANE_KEY = 'tab_coord:leaf_coord' + stubStaleHandleRemint('term_live_coord', { + result: { dispatch: { id: 'ctx_1', task_id: 'task_1', status: 'dispatched' } } + }) + getTerminalHandleMock.mockRejectedValue(new Error('active terminal fallback is unsafe')) + + await invokeDispatch( + new Map([ + ['task', 'task_1'], + ['to', 'term_worker'], + ['inject', true] + ]) + ) + + expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.resolveIdentity', { + terminal: 'term_stale_coord' + }) + expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { + paneKey: 'tab_coord:leaf_coord' + }) + expect(getTerminalHandleMock).not.toHaveBeenCalled() + expect(callMock).toHaveBeenNthCalledWith(3, 'orchestration.dispatch', { + task: 'task_1', + to: 'term_worker', + from: 'term_live_coord', + inject: true, + dryRun: undefined, + returnPreamble: undefined, + devMode: false + }) + }) + + it('rejects stale coordinator env handles when the caller pane cannot be proven', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_stale_coord' + callMock.mockRejectedValueOnce(staleHandleError()) + getTerminalHandleMock.mockResolvedValue('term_wrong_active') + + await expect( + invokeDispatch( + new Map([ + ['task', 'task_1'], + ['to', 'term_worker'] + ]) + ) + ).rejects.toMatchObject({ + code: 'no_active_sender_terminal' + }) + + expect(callMock).toHaveBeenCalledTimes(1) + expect(getTerminalHandleMock).not.toHaveBeenCalled() + }) + + it('propagates unexpected caller pane remint failures for coordinator commands', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_stale_coord' + process.env.ORCA_PANE_KEY = 'tab_coord:leaf_coord' + stubStaleHandleRemintFailure( + new RuntimeClientError('runtime_unavailable', 'runtime_unavailable') + ) + getTerminalHandleMock.mockResolvedValue('term_wrong_active') + + await expect( + invokeDispatch( + new Map([ + ['task', 'task_1'], + ['to', 'term_worker'] + ]) + ) + ).rejects.toMatchObject({ + code: 'runtime_unavailable' + }) + + expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.resolveIdentity', { + terminal: 'term_stale_coord' + }) + expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { + paneKey: 'tab_coord:leaf_coord' + }) + expect(callMock).toHaveBeenCalledTimes(2) + expect(getTerminalHandleMock).not.toHaveBeenCalled() + }) + + it('uses a live coordinator handle for dispatch-show preamble previews', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_stale_coord' + process.env.ORCA_PANE_KEY = 'tab_coord:leaf_coord' + stubStaleHandleRemint('term_live_coord', { + result: { dispatch: null, preamble: 'preamble' } + }) + getTerminalHandleMock.mockRejectedValue(new Error('active terminal fallback is unsafe')) + + await invokeDispatchShow( + new Map([ + ['task', 'task_1'], + ['preamble', true] + ]) + ) + + expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.resolveIdentity', { + terminal: 'term_stale_coord' + }) + expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { + paneKey: 'tab_coord:leaf_coord' + }) + expect(callMock).toHaveBeenNthCalledWith(3, 'orchestration.dispatchShow', { + task: 'task_1', + preamble: true, + from: 'term_live_coord', + devMode: false + }) + }) + + it('retires the legacy coordinator command without runtime effects', async () => { + await expect( + invokeRun(new Map([['spec', 'run the plan']])) + ).rejects.toMatchObject({ + code: 'orchestration_migration_required', + data: { + reason: 'command_retired', + effectsApplied: false, + nextCommandArgs: ['skills', 'get', 'orchestration', '--full'] + } + }) + expect(callMock).not.toHaveBeenCalled() + }) +}) + +describe('orchestration task-create caller handle', () => { + beforeEach(() => { + callMock.mockReset() + getTerminalHandleMock.mockReset() + delete process.env.ORCA_TERMINAL_HANDLE + delete process.env.ORCA_PANE_KEY + }) + + const invokeTaskCreate = (flags: Map) => + ORCHESTRATION_HANDLERS['orchestration task-create']({ + flags, + client: { call: callMock }, + cwd: '/tmp/repo', + json: true + } as never) + + it('records a live env terminal handle as task creator', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_creator' + callMock + .mockResolvedValueOnce(liveIdentity('term_creator')) + .mockResolvedValueOnce({ result: { task: { id: 'task_1', status: 'ready' } } }) + + await invokeTaskCreate(new Map([['spec', 'do work']])) + + expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.resolveIdentity', { + terminal: 'term_creator' + }) + expect(callMock).toHaveBeenNthCalledWith(2, 'orchestration.taskCreate', { + spec: 'do work', + taskTitle: undefined, + displayName: undefined, + deps: undefined, + parent: undefined, + run: undefined, + callerTerminalHandle: 'term_creator' + }) + }) + + it('fails closed when a stale task creator handle cannot be reminted', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_stale' + callMock.mockRejectedValueOnce(staleHandleError()) + getTerminalHandleMock.mockResolvedValue('term_wrong_active') + + await expect( + invokeTaskCreate(new Map([['spec', 'do work']])) + ).rejects.toMatchObject({ code: 'no_active_sender_terminal' }) + + expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.resolveIdentity', { + terminal: 'term_stale' + }) + expect(getTerminalHandleMock).not.toHaveBeenCalled() + expect(callMock).toHaveBeenCalledTimes(1) + }) + + it('propagates runtime unavailability while proving the bound coordinator', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_creator' + callMock.mockRejectedValueOnce( + new RuntimeClientError('runtime_unavailable', 'runtime_unavailable') + ) + + await expect( + invokeTaskCreate(new Map([['spec', 'do work']])) + ).rejects.toMatchObject({ code: 'runtime_unavailable' }) + + expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.resolveIdentity', { + terminal: 'term_creator' + }) + expect(getTerminalHandleMock).not.toHaveBeenCalled() + expect(callMock).toHaveBeenCalledTimes(1) + }) + + it('propagates runtime unavailability while reminting the bound coordinator', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_stale' + process.env.ORCA_PANE_KEY = 'tab_creator:leaf_creator' + stubStaleHandleRemintFailure( + new RuntimeClientError('runtime_unavailable', 'runtime_unavailable') + ) + getTerminalHandleMock.mockResolvedValue('term_wrong_active') + + await expect( + invokeTaskCreate(new Map([['spec', 'do work']])) + ).rejects.toMatchObject({ code: 'runtime_unavailable' }) + + expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.resolveIdentity', { + terminal: 'term_stale' + }) + expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { + paneKey: 'tab_creator:leaf_creator' + }) + expect(getTerminalHandleMock).not.toHaveBeenCalled() + expect(callMock).toHaveBeenCalledTimes(2) + }) + + it('propagates unexpected caller pane remint failures for task creation', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_stale' + process.env.ORCA_PANE_KEY = 'tab_creator:leaf_creator' + stubStaleHandleRemintFailure(new RuntimeClientError('permission_denied', 'denied')) + getTerminalHandleMock.mockResolvedValue('term_wrong_active') + + await expect( + invokeTaskCreate(new Map([['spec', 'do work']])) + ).rejects.toMatchObject({ + code: 'permission_denied' + }) + + expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.resolveIdentity', { + terminal: 'term_stale' + }) + expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { + paneKey: 'tab_creator:leaf_creator' + }) + expect(callMock).toHaveBeenCalledTimes(2) + expect(getTerminalHandleMock).not.toHaveBeenCalled() + }) + + it('propagates unexpected env handle validation failures', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_creator' + callMock.mockRejectedValueOnce(new RuntimeClientError('permission_denied', 'denied')) + + await expect( + invokeTaskCreate(new Map([['spec', 'do work']])) + ).rejects.toMatchObject({ + code: 'permission_denied' + }) + + expect(callMock).toHaveBeenCalledTimes(1) + }) + + it('remints a stale task creator env handle from the caller pane key', async () => { + process.env.ORCA_TERMINAL_HANDLE = 'term_stale' + process.env.ORCA_PANE_KEY = 'tab_creator:leaf_creator' + stubStaleHandleRemint('term_live', { + result: { task: { id: 'task_1', status: 'ready' } } + }) + getTerminalHandleMock.mockRejectedValue(new Error('active terminal fallback is unsafe')) + + await invokeTaskCreate(new Map([['spec', 'do work']])) + + expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.resolveIdentity', { + terminal: 'term_stale' + }) + expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { + paneKey: 'tab_creator:leaf_creator' + }) + expect(getTerminalHandleMock).not.toHaveBeenCalled() + expect(callMock).toHaveBeenNthCalledWith(3, 'orchestration.taskCreate', { + spec: 'do work', + taskTitle: undefined, + displayName: undefined, + deps: undefined, + parent: undefined, + run: undefined, + callerTerminalHandle: 'term_live' + }) + }) +}) diff --git a/src/cli/handlers/orchestration-gate-cli.test.ts b/src/cli/handlers/orchestration-gate-cli.test.ts index b4be315c155..a2793ac9323 100644 --- a/src/cli/handlers/orchestration-gate-cli.test.ts +++ b/src/cli/handlers/orchestration-gate-cli.test.ts @@ -72,7 +72,7 @@ describe('orchestration gate commands carry caller identity', () => { process.env.ORCA_TERMINAL_HANDLE = 'term_coord' queueFixtures( callMock, - okFixture('req_show', { terminal: { handle: 'term_coord' } }), + okFixture('req_identity', { identity: { handle: 'term_coord', live: true } }), okFixture('req_gate', { gate: { id: 'gate_1', task_id: 'task_1', status: 'pending' } }) ) @@ -91,8 +91,8 @@ describe('orchestration gate commands carry caller identity', () => { process.env.ORCA_TERMINAL_HANDLE = 'term_stale' process.env.ORCA_PANE_KEY = 'tab_coord:leaf_coord' callMock.mockImplementation(async (method: string) => { - if (method === 'terminal.show') { - throw new RuntimeClientError('terminal_handle_stale', 'stale') + if (method === 'terminal.resolveIdentity') { + return okFixture('req_identity', { identity: { handle: 'term_stale', live: false } }) } if (method === 'terminal.resolvePane') { return okFixture('req_pane', { terminal: { handle: 'term_live' } }) @@ -146,7 +146,7 @@ describe('orchestration gate commands carry caller identity', () => { process.env.ORCA_TERMINAL_HANDLE = 'term_coord' queueFixtures( callMock, - okFixture('req_show', { terminal: { handle: 'term_coord' } }), + okFixture('req_identity', { identity: { handle: 'term_coord', live: true } }), okFixture('req_list', { gates: [], count: 0 }) ) @@ -199,7 +199,9 @@ describe('orchestration gate commands carry caller identity', () => { it('reports idempotent recovery when a mutation connection drops', async () => { process.env.ORCA_TERMINAL_HANDLE = 'term_coord' callMock - .mockResolvedValueOnce(okFixture('req_show', { terminal: { handle: 'term_coord' } })) + .mockResolvedValueOnce( + okFixture('req_identity', { identity: { handle: 'term_coord', live: true } }) + ) .mockRejectedValueOnce( new RuntimeClientError( 'runtime_unavailable', diff --git a/src/cli/handlers/orchestration-task-create-cli.test.ts b/src/cli/handlers/orchestration-task-create-cli.test.ts index 839ccabdcd6..8b5957782d2 100644 --- a/src/cli/handlers/orchestration-task-create-cli.test.ts +++ b/src/cli/handlers/orchestration-task-create-cli.test.ts @@ -27,7 +27,7 @@ describe('orchestration task-create CLI mapping', () => { it('passes PowerShell-stripped deps through to the runtime', async () => { callMock - .mockResolvedValueOnce({ result: { terminal: { handle: 'term_creator' } } }) + .mockResolvedValueOnce({ result: { identity: { handle: 'term_creator', live: true } } }) .mockResolvedValueOnce({ result: { task: { id: 'task_2', status: 'pending' } } }) await ORCHESTRATION_HANDLERS['orchestration task-create']({ diff --git a/src/cli/handlers/orchestration-worker-cli.test.ts b/src/cli/handlers/orchestration-worker-cli.test.ts index f17420ab261..cddd36a4cf7 100644 --- a/src/cli/handlers/orchestration-worker-cli.test.ts +++ b/src/cli/handlers/orchestration-worker-cli.test.ts @@ -334,6 +334,59 @@ describe('orchestration worker-start CLI contract', () => { ).toContain('Warning: Terminal term_worker is running but could not be revealed.') }) + it('states the worker mode that actually ran, so a fallback is never silent', async () => { + callMock.mockResolvedValue({ + result: { + taskId: 'task_1', + dispatchId: 'ctx_1', + state: 'ready', + mode: { + mode: 'terminal', + preferred: 'structured', + reason: 'reused_terminal', + detail: + 'Your default is a structured chat session, but --terminal reuses a running terminal agent; started a terminal agent worker instead.' + }, + effects: [], + residualResources: [] + } + }) + + await ORCHESTRATION_HANDLERS['orchestration worker-start']({ + flags: new Map([ + ['task', 'task_1'], + ['terminal', 'term_worker'], + ['from', 'term_coord'] + ]), + client: { call: callMock }, + cwd: '/tmp/repo', + json: false + } as never) + + const formatter = vi.mocked(printResult).mock.calls[0]?.[2] as + | ((result: { + taskId: string + dispatchId: string + state: string + mode?: { mode: string; preferred: string; reason: string; detail: string } + }) => string) + | undefined + expect( + formatter?.({ + taskId: 'task_1', + dispatchId: 'ctx_1', + state: 'ready', + mode: { + mode: 'terminal', + preferred: 'structured', + reason: 'reused_terminal', + detail: + 'Your default is a structured chat session, but --terminal reuses a running terminal agent; started a terminal agent worker instead.' + } + }) + ).toContain('but --terminal reuses a running terminal agent') + }) + it('prints the retained-process warning for a manual worker-stop', async () => { callMock.mockResolvedValue({ result: { diff --git a/src/cli/handlers/orchestration.test.ts b/src/cli/handlers/orchestration.test.ts index 393bb31d141..d8cf528671d 100644 --- a/src/cli/handlers/orchestration.test.ts +++ b/src/cli/handlers/orchestration.test.ts @@ -16,24 +16,6 @@ import { ORCHESTRATION_HANDLERS } from './orchestration' import { RuntimeClientError } from '../runtime-client' import { printResult } from '../format' -function staleHandleError(): RuntimeClientError { - return new RuntimeClientError('terminal_handle_stale', 'terminal_handle_stale') -} - -// Queues the stale-handle remint chain shared by coordinator commands: -// stale terminal.show → resolvePane returns liveHandle → downstream RPC result. -function stubStaleHandleRemint(liveHandle: string, downstream: unknown): void { - callMock - .mockRejectedValueOnce(staleHandleError()) - .mockResolvedValueOnce({ result: { terminal: { handle: liveHandle } } }) - .mockResolvedValueOnce(downstream) -} - -// Queues a stale terminal.show followed by a resolvePane remint that fails with `error`. -function stubStaleHandleRemintFailure(error: RuntimeClientError): void { - callMock.mockRejectedValueOnce(staleHandleError()).mockRejectedValueOnce(error) -} - afterEach(() => { getTerminalHandleMock.mockReset() if (originalTerminalHandle === undefined) { @@ -332,309 +314,6 @@ describe('orchestration send structured payload flags', () => { ) }) -describe('orchestration dispatch coordinator handle', () => { - beforeEach(() => { - callMock.mockReset() - getTerminalHandleMock.mockReset() - delete process.env.ORCA_TERMINAL_HANDLE - delete process.env.ORCA_PANE_KEY - }) - - const invokeDispatch = (flags: Map) => - ORCHESTRATION_HANDLERS['orchestration dispatch']({ - flags, - client: { call: callMock }, - cwd: '/tmp/repo', - json: true - } as never) - - const invokeDispatchShow = (flags: Map) => - ORCHESTRATION_HANDLERS['orchestration dispatch-show']({ - flags, - client: { call: callMock }, - cwd: '/tmp/repo', - json: true - } as never) - - const invokeRun = (flags: Map) => - ORCHESTRATION_HANDLERS['orchestration coordinator-start']({ - flags, - client: { call: callMock }, - cwd: '/tmp/repo', - json: true - } as never) - - it('remints a stale coordinator env handle from the caller pane key', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_stale_coord' - process.env.ORCA_PANE_KEY = 'tab_coord:leaf_coord' - stubStaleHandleRemint('term_live_coord', { - result: { dispatch: { id: 'ctx_1', task_id: 'task_1', status: 'dispatched' } } - }) - getTerminalHandleMock.mockRejectedValue(new Error('active terminal fallback is unsafe')) - - await invokeDispatch( - new Map([ - ['task', 'task_1'], - ['to', 'term_worker'], - ['inject', true] - ]) - ) - - expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.show', { - terminal: 'term_stale_coord' - }) - expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { - paneKey: 'tab_coord:leaf_coord' - }) - expect(getTerminalHandleMock).not.toHaveBeenCalled() - expect(callMock).toHaveBeenNthCalledWith(3, 'orchestration.dispatch', { - task: 'task_1', - to: 'term_worker', - from: 'term_live_coord', - inject: true, - dryRun: undefined, - returnPreamble: undefined, - devMode: false - }) - }) - - it('rejects stale coordinator env handles when the caller pane cannot be proven', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_stale_coord' - callMock.mockRejectedValueOnce(staleHandleError()) - getTerminalHandleMock.mockResolvedValue('term_wrong_active') - - await expect( - invokeDispatch( - new Map([ - ['task', 'task_1'], - ['to', 'term_worker'] - ]) - ) - ).rejects.toMatchObject({ - code: 'no_active_sender_terminal' - }) - - expect(callMock).toHaveBeenCalledTimes(1) - expect(getTerminalHandleMock).not.toHaveBeenCalled() - }) - - it('propagates unexpected caller pane remint failures for coordinator commands', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_stale_coord' - process.env.ORCA_PANE_KEY = 'tab_coord:leaf_coord' - stubStaleHandleRemintFailure( - new RuntimeClientError('runtime_unavailable', 'runtime_unavailable') - ) - getTerminalHandleMock.mockResolvedValue('term_wrong_active') - - await expect( - invokeDispatch( - new Map([ - ['task', 'task_1'], - ['to', 'term_worker'] - ]) - ) - ).rejects.toMatchObject({ - code: 'runtime_unavailable' - }) - - expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.show', { - terminal: 'term_stale_coord' - }) - expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { - paneKey: 'tab_coord:leaf_coord' - }) - expect(callMock).toHaveBeenCalledTimes(2) - expect(getTerminalHandleMock).not.toHaveBeenCalled() - }) - - it('uses a live coordinator handle for dispatch-show preamble previews', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_stale_coord' - process.env.ORCA_PANE_KEY = 'tab_coord:leaf_coord' - stubStaleHandleRemint('term_live_coord', { - result: { dispatch: null, preamble: 'preamble' } - }) - getTerminalHandleMock.mockRejectedValue(new Error('active terminal fallback is unsafe')) - - await invokeDispatchShow( - new Map([ - ['task', 'task_1'], - ['preamble', true] - ]) - ) - - expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.show', { - terminal: 'term_stale_coord' - }) - expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { - paneKey: 'tab_coord:leaf_coord' - }) - expect(callMock).toHaveBeenNthCalledWith(3, 'orchestration.dispatchShow', { - task: 'task_1', - preamble: true, - from: 'term_live_coord', - devMode: false - }) - }) - - it('retires the legacy coordinator command without runtime effects', async () => { - await expect( - invokeRun(new Map([['spec', 'run the plan']])) - ).rejects.toMatchObject({ - code: 'orchestration_migration_required', - data: { - reason: 'command_retired', - effectsApplied: false, - nextCommandArgs: ['skills', 'get', 'orchestration', '--full'] - } - }) - expect(callMock).not.toHaveBeenCalled() - }) -}) - -describe('orchestration task-create caller handle', () => { - beforeEach(() => { - callMock.mockReset() - getTerminalHandleMock.mockReset() - delete process.env.ORCA_TERMINAL_HANDLE - delete process.env.ORCA_PANE_KEY - }) - - const invokeTaskCreate = (flags: Map) => - ORCHESTRATION_HANDLERS['orchestration task-create']({ - flags, - client: { call: callMock }, - cwd: '/tmp/repo', - json: true - } as never) - - it('records a live env terminal handle as task creator', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_creator' - callMock - .mockResolvedValueOnce({ result: { terminal: { handle: 'term_creator' } } }) - .mockResolvedValueOnce({ result: { task: { id: 'task_1', status: 'ready' } } }) - - await invokeTaskCreate(new Map([['spec', 'do work']])) - - expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.show', { terminal: 'term_creator' }) - expect(callMock).toHaveBeenNthCalledWith(2, 'orchestration.taskCreate', { - spec: 'do work', - taskTitle: undefined, - displayName: undefined, - deps: undefined, - parent: undefined, - run: undefined, - callerTerminalHandle: 'term_creator' - }) - }) - - it('fails closed when a stale task creator handle cannot be reminted', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_stale' - callMock.mockRejectedValueOnce(staleHandleError()) - getTerminalHandleMock.mockResolvedValue('term_wrong_active') - - await expect( - invokeTaskCreate(new Map([['spec', 'do work']])) - ).rejects.toMatchObject({ code: 'no_active_sender_terminal' }) - - expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.show', { terminal: 'term_stale' }) - expect(getTerminalHandleMock).not.toHaveBeenCalled() - expect(callMock).toHaveBeenCalledTimes(1) - }) - - it('propagates runtime unavailability while proving the bound coordinator', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_creator' - callMock.mockRejectedValueOnce( - new RuntimeClientError('runtime_unavailable', 'runtime_unavailable') - ) - - await expect( - invokeTaskCreate(new Map([['spec', 'do work']])) - ).rejects.toMatchObject({ code: 'runtime_unavailable' }) - - expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.show', { terminal: 'term_creator' }) - expect(getTerminalHandleMock).not.toHaveBeenCalled() - expect(callMock).toHaveBeenCalledTimes(1) - }) - - it('propagates runtime unavailability while reminting the bound coordinator', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_stale' - process.env.ORCA_PANE_KEY = 'tab_creator:leaf_creator' - stubStaleHandleRemintFailure( - new RuntimeClientError('runtime_unavailable', 'runtime_unavailable') - ) - getTerminalHandleMock.mockResolvedValue('term_wrong_active') - - await expect( - invokeTaskCreate(new Map([['spec', 'do work']])) - ).rejects.toMatchObject({ code: 'runtime_unavailable' }) - - expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.show', { terminal: 'term_stale' }) - expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { - paneKey: 'tab_creator:leaf_creator' - }) - expect(getTerminalHandleMock).not.toHaveBeenCalled() - expect(callMock).toHaveBeenCalledTimes(2) - }) - - it('propagates unexpected caller pane remint failures for task creation', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_stale' - process.env.ORCA_PANE_KEY = 'tab_creator:leaf_creator' - stubStaleHandleRemintFailure(new RuntimeClientError('permission_denied', 'denied')) - getTerminalHandleMock.mockResolvedValue('term_wrong_active') - - await expect( - invokeTaskCreate(new Map([['spec', 'do work']])) - ).rejects.toMatchObject({ - code: 'permission_denied' - }) - - expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.show', { terminal: 'term_stale' }) - expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { - paneKey: 'tab_creator:leaf_creator' - }) - expect(callMock).toHaveBeenCalledTimes(2) - expect(getTerminalHandleMock).not.toHaveBeenCalled() - }) - - it('propagates unexpected env handle validation failures', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_creator' - callMock.mockRejectedValueOnce(new RuntimeClientError('permission_denied', 'denied')) - - await expect( - invokeTaskCreate(new Map([['spec', 'do work']])) - ).rejects.toMatchObject({ - code: 'permission_denied' - }) - - expect(callMock).toHaveBeenCalledTimes(1) - }) - - it('remints a stale task creator env handle from the caller pane key', async () => { - process.env.ORCA_TERMINAL_HANDLE = 'term_stale' - process.env.ORCA_PANE_KEY = 'tab_creator:leaf_creator' - stubStaleHandleRemint('term_live', { - result: { task: { id: 'task_1', status: 'ready' } } - }) - getTerminalHandleMock.mockRejectedValue(new Error('active terminal fallback is unsafe')) - - await invokeTaskCreate(new Map([['spec', 'do work']])) - - expect(callMock).toHaveBeenNthCalledWith(1, 'terminal.show', { terminal: 'term_stale' }) - expect(callMock).toHaveBeenNthCalledWith(2, 'terminal.resolvePane', { - paneKey: 'tab_creator:leaf_creator' - }) - expect(getTerminalHandleMock).not.toHaveBeenCalled() - expect(callMock).toHaveBeenNthCalledWith(3, 'orchestration.taskCreate', { - spec: 'do work', - taskTitle: undefined, - displayName: undefined, - deps: undefined, - parent: undefined, - run: undefined, - callerTerminalHandle: 'term_live' - }) - }) -}) describe('orchestration timeout flag validation', () => { const invalidTimeoutValues: [string, string | boolean][] = [ ['missing', true], diff --git a/src/cli/handlers/orchestration/terminal-identity.ts b/src/cli/handlers/orchestration/terminal-identity.ts index 5da11afac1d..505bfaac262 100644 --- a/src/cli/handlers/orchestration/terminal-identity.ts +++ b/src/cli/handlers/orchestration/terminal-identity.ts @@ -2,6 +2,7 @@ import type { RuntimeClient } from '../../runtime-client' import { getOptionalStringFlag } from '../../flags' import { RuntimeClientError } from '../../runtime-client' import { getTerminalHandle } from '../../selectors' +import { isStructuredSessionWithoutIdentity } from '../../../shared/structured-session-marker' export async function resolveOrchestrationTerminalHandle( flags: Map, @@ -29,13 +30,56 @@ export async function resolveOrchestrationTerminalHandle( } return envHandle } + // Past this point every remaining route GUESSES an implicit terminal, and a structured session + // has no pane for the guess to land on — so it lands on a sibling. `check` is destructive by + // default, so that guess consumed another pane's oldest unread batch and marked it read, and the + // rightful worker never saw its mail. Refusing is the only honest answer: this child genuinely + // cannot infer its own identity. + if (isStructuredSessionWithoutIdentity()) { + throw new RuntimeClientError( + 'no_active_sender_terminal', + `This chat session has no orchestration identity of its own, so --${flagName} cannot be inferred. ` + + `Pass --${flagName} explicitly; guessing would act on another pane's mailbox.` + ) + } if (flagName === 'from') { return await resolveImplicitOrchestrationSender(flags, cwd, client) } return await getTerminalHandle(flags, cwd, client) } +/** + * Whether the handle this process was born with still names a live identity. + * + * `terminal.resolveIdentity`, never `terminal.show`: `show` is a PTY verb, so it missed for a + * structured worker and reported `terminal_handle_stale` for a handle that was perfectly live — + * which then failed every coordinator verb, because the pane remint below needs an `ORCA_PANE_KEY` + * a structured child deliberately does not carry. + */ async function isLiveTerminalHandle(handle: string, client: RuntimeClient): Promise { + try { + const response = await client.call<{ identity?: { live?: boolean } }>( + 'terminal.resolveIdentity', + { terminal: handle } + ) + const live = response.result?.identity?.live + // An unrecognised shape is an older host answering something else, not a dead handle. + return typeof live === 'boolean' ? live : await showResolvesTerminalHandle(handle, client) + } catch (err) { + if (isStaleTerminalIdentityError(err)) { + return false + } + if (getClientErrorCode(err) === 'method_not_found') { + // Clients and remote hosts update independently, so a host that predates the identity probe + // is the normal mixed-version state. Fall back to what it does have — which is correct for + // that host, because a host without the probe also has no structured workers to miss. + return await showResolvesTerminalHandle(handle, client) + } + throw err + } +} + +async function showResolvesTerminalHandle(handle: string, client: RuntimeClient): Promise { try { await client.call('terminal.show', { terminal: handle }) return true @@ -133,7 +177,9 @@ async function resolveImplicitOrchestrationSender( client: RuntimeClient ): Promise { try { - return await getTerminalHandle(flags, cwd, client) + // Unambiguous: naming the sender is an identity claim, so an arbitrary pick would let this + // command speak as a sibling worker. + return await getTerminalHandle(flags, cwd, client, { requireUnambiguous: true }) } catch (err) { if (!isNoActiveTerminalError(err)) { throw err diff --git a/src/cli/handlers/orchestration/worker-launch-handler.ts b/src/cli/handlers/orchestration/worker-launch-handler.ts index 97517373a1c..b6e4d92799c 100644 --- a/src/cli/handlers/orchestration/worker-launch-handler.ts +++ b/src/cli/handlers/orchestration/worker-launch-handler.ts @@ -41,6 +41,7 @@ export const ORCHESTRATION_WORKER_LAUNCH_HANDLER: Record failedStage?: string lastError?: string warning?: string + mode?: { mode: string; preferred: string; reason: string; detail: string } effects: unknown[] residualResources: unknown[] nextCommands?: string[] diff --git a/src/cli/handlers/orchestration/worker-output.ts b/src/cli/handlers/orchestration/worker-output.ts index fb9d7018e46..9d6d1949911 100644 --- a/src/cli/handlers/orchestration/worker-output.ts +++ b/src/cli/handlers/orchestration/worker-output.ts @@ -1,10 +1,6 @@ -import { - isSubagentGroupFallbackText, - subagentGroupFallbackText -} from '../../../shared/native-chat-subagent-summary' -import type { NativeChatMessage } from '../../../shared/native-chat-types' import type { RuntimeTerminalRead } from '../../../shared/runtime-types' import type { OrchestrationWorkerReadResult } from '../../../shared/orchestration-worker-output' +import { formatWorkerTranscriptMessage } from '../../../shared/worker-transcript-text' export type LegacyWorkerReadResult = { dispatchId: string @@ -12,6 +8,7 @@ export type LegacyWorkerReadResult = { } export type WorkerStartReceipt = { + mode?: { detail: string } taskId: string dispatchId: string state: string @@ -25,6 +22,11 @@ export type WorkerStartReceipt = { export function formatWorkerStart(value: WorkerStartReceipt): string { const lines = [`Worker ${value.dispatchId} [${value.state}] for ${value.taskId}`] + // Settings-driven rather than requested, so the human line always names the mode that ran: a + // fallback from the user's structured default is never silent. + if (value.mode) { + lines.push(value.mode.detail) + } if (value.lastError) { lines.push(`${value.failedStage ?? 'start'}: ${value.lastError}`) } else if (value.warning) { @@ -105,44 +107,6 @@ function formatWorkerReadDetails(value: OrchestrationWorkerReadResult): string { return lines.join('\n') } -function formatWorkerTranscriptMessage(message: NativeChatMessage): string { - // Every roster block is written beside a plain-text twin carrying the same - // sentence, for clients that cannot draw the block. This CLI is one, so it - // prints the twin and drops the block — the mirror of the renderer, which - // draws the block and drops the twin. Either way the sentence prints once. - const hasTwin = message.blocks.some( - (block) => block.type === 'text' && isSubagentGroupFallbackText(block.text) - ) - const blocks = message.blocks.flatMap((block): string[] => { - if (block.type === 'text') { - return [block.text] - } - if (block.type === 'tool-call') { - return [`[tool ${block.name}] ${safeJson(block.input)}`] - } - if (block.type === 'tool-result') { - return [`[tool result${block.isError ? ' error' : ''}] ${block.output}`] - } - if (block.type === 'image-ref') { - return [block.url ? `[image] ${block.url}` : `[image omitted]`] - } - if (block.type === 'subagent-group') { - // Stand in for the block only when no twin is being printed: the wire - // admits a roster that arrived without one, and dropping that - // unconditionally would lose the sentence altogether. Presence, not a - // byte compare against a recomputed sentence — a roster from a newer - // build holds a state this build reads as `unverifiable`, so recomputing - // yields a different sentence and both would print. - return hasTwin ? [] : [`[subagents] ${subagentGroupFallbackText(block.agents)}`] - } - // The journal deliberately admits block types this build does not know, and - // a newer remote host can send one over the wire. Degrade to a marker rather - // than reading fields off a shape that has none. - return ['[unsupported block]'] - }) - return `[${message.role}] ${blocks.join('\n')}`.trimEnd() -} - export type WorkerReleaseReceipt = { dispatchId: string state: string @@ -169,11 +133,3 @@ export function formatWorkerRelease(value: WorkerReleaseReceipt): string { } return lines.join('\n') } - -function safeJson(value: unknown): string { - try { - return JSON.stringify(value) - } catch { - return '[unserializable input]' - } -} diff --git a/src/cli/orchestration-structured-sender-identity.test.ts b/src/cli/orchestration-structured-sender-identity.test.ts new file mode 100644 index 00000000000..9b85125d99e --- /dev/null +++ b/src/cli/orchestration-structured-sender-identity.test.ts @@ -0,0 +1,151 @@ +/** + * The env-handle path, with NO `--from`. + * + * Every other orchestration CLI test passes `--from term_coord` explicitly, so the resolver a real + * worker actually goes through — `ORCA_TERMINAL_HANDLE` plus `validateEnvHandle` — was never + * exercised. That is why twelve coordinator verbs could fail for a structured worker while the + * whole suite stayed green, and why the worker's own preamble (which tells it to run these with no + * `--from`) failed on its first line. + */ + +import { describe, expect, it, vi } from 'vitest' + +const { + callMock, + runtimeClientConstructorMock, + serveOrcaAppMock, + getDefaultUserDataPathMock, + addEnvironmentFromPairingCodeMock, + listEnvironmentsMock, + spawnMock +} = vi.hoisted(() => ({ + callMock: vi.fn(), + runtimeClientConstructorMock: vi.fn(), + serveOrcaAppMock: vi.fn(), + getDefaultUserDataPathMock: vi.fn(() => '/tmp/orca-user-data'), + addEnvironmentFromPairingCodeMock: vi.fn(), + listEnvironmentsMock: vi.fn(), + spawnMock: vi.fn() +})) + +vi.mock('./runtime-client', async () => { + const { createRuntimeClientModuleMock } = await import('./index-test-harness.js') + return createRuntimeClientModuleMock({ + callMock, + runtimeClientConstructorMock, + serveOrcaAppMock, + getDefaultUserDataPathMock + }) +}) + +vi.mock('./runtime/environments', () => ({ + addEnvironmentFromPairingCode: addEnvironmentFromPairingCodeMock, + listEnvironments: listEnvironmentsMock, + removeEnvironment: vi.fn(), + resolveEnvironment: vi.fn() +})) + +vi.mock('child_process', async () => { + const { createChildProcessModuleMock } = await import('./index-test-harness.js') + return createChildProcessModuleMock(spawnMock) +}) + +import { main } from './index' +import { useWorktreeAwarenessEnvironment } from './index-test-harness' + +const STRUCTURED_HANDLE = 'structworker_a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + +/** + * Every command below is one of the twelve that resolve their sender through + * `resolveCoordinatorTerminalHandle`. All twelve share ONE resolver and one liveness probe, so the + * set is sufficient if it covers the distinct shapes that reach it: a read, a mutation, a + * run-scoped verb, a dispatch verb and a worker-start. A per-verb sweep would pin the argv parsing + * of twelve handlers and still tell us nothing more about the seam that actually broke. + */ +const SENDER_VERBS: { argv: string[]; method: string }[] = [ + { argv: ['orchestration', 'run-current'], method: 'orchestration.runCurrent' }, + { argv: ['orchestration', 'run-create', '--objective', 'x'], method: 'orchestration.runCreate' }, + { argv: ['orchestration', 'task-list'], method: 'orchestration.taskList' }, + { argv: ['orchestration', 'gate-list'], method: 'orchestration.gateList' }, + { argv: ['orchestration', 'dispatch-show', '--task', 't1'], method: 'orchestration.dispatchShow' } +] + +describe('a structured worker running orchestration commands as itself', () => { + useWorktreeAwarenessEnvironment({ + callMock, + serveOrcaAppMock, + getDefaultUserDataPathMock, + addEnvironmentFromPairingCodeMock, + listEnvironmentsMock, + spawnMock + }) + + function answerCalls(): void { + callMock.mockImplementation(async (method: string) => { + if (method === 'terminal.resolveIdentity') { + return { + id: 'req', + ok: true, + result: { identity: { handle: STRUCTURED_HANDLE, live: true } }, + _meta: { runtimeId: 'runtime-1' } + } + } + return { id: 'req', ok: true, result: {}, _meta: { runtimeId: 'runtime-1' } } + }) + } + + it.each(SENDER_VERBS)( + 'resolves its own identity for $method with no --from', + async ({ argv, method }) => { + // The defect this pins: the sender resolver validated the env handle with `terminal.show`, a + // PTY verb that misses for a structured worker and answers `terminal_handle_stale`. The pane + // remint that would have recovered it needs `ORCA_PANE_KEY`, which a structured child + // deliberately does not carry, so the command died on `no_active_sender_terminal`. + process.env.ORCA_TERMINAL_HANDLE = STRUCTURED_HANDLE + answerCalls() + vi.spyOn(console, 'log').mockImplementation(() => {}) + await expect(main(argv)).resolves.not.toThrow() + const called = callMock.mock.calls.map((call) => call[0] as string) + expect(called).toContain(method) + // Never through `terminal.show`: teaching that verb structured handles would hand every + // public terminal verb something that looks writable and is not. + expect(called).not.toContain('terminal.show') + } + ) + + it('sends the structured handle as the sender, not a guessed sibling', async () => { + process.env.ORCA_TERMINAL_HANDLE = STRUCTURED_HANDLE + answerCalls() + vi.spyOn(console, 'log').mockImplementation(() => {}) + await main(['orchestration', 'run-create', '--objective', 'x']) + const create = callMock.mock.calls.find((call) => call[0] === 'orchestration.runCreate') + expect((create?.[1] as { from?: string } | undefined)?.from).toBe(STRUCTURED_HANDLE) + }) + + it('still refuses a handle the runtime reports dead, with no pane key to remint from', async () => { + // The invariant the fix must not break: a stale `ORCA_TERMINAL_HANDLE` in a long-lived shell + // must keep failing rather than being baked into a coordinator preamble. + process.env.ORCA_TERMINAL_HANDLE = 'term_stale' + callMock.mockImplementation(async (method: string) => { + if (method === 'terminal.resolveIdentity') { + return { + id: 'req', + ok: true, + result: { identity: { handle: 'term_stale', live: false } }, + _meta: { runtimeId: 'runtime-1' } + } + } + return { id: 'req', ok: true, result: {}, _meta: { runtimeId: 'runtime-1' } } + }) + const errorSpy = vi.spyOn(console, 'error').mockImplementation(() => {}) + const priorExitCode = process.exitCode + await main(['orchestration', 'run-create', '--objective', 'x']) + expect(process.exitCode).toBe(1) + expect(errorSpy.mock.calls.flat().join(' ')).toMatch( + /no_active_sender_terminal|sender terminal/i + ) + expect(callMock.mock.calls.map((call) => call[0])).not.toContain('orchestration.runCreate') + process.exitCode = priorExitCode + errorSpy.mockRestore() + }) +}) diff --git a/src/cli/orchestration-structured-session-no-identity.test.ts b/src/cli/orchestration-structured-session-no-identity.test.ts new file mode 100644 index 00000000000..d7c76f2b5e0 --- /dev/null +++ b/src/cli/orchestration-structured-session-no-identity.test.ts @@ -0,0 +1,98 @@ +/** + * A structured chat session with NO orchestration identity must refuse, not guess. + * + * Non-worker structured sessions get no `ORCA_TERMINAL_HANDLE`, so `orchestration check` fell + * through to the active-terminal guess — and `check` is destructive by default, so it consumed + * another pane's oldest unread batch and marked it read. The rightful worker never saw that mail. + * + * The case pinned here is ONE terminal pane in the worktree, because that is the case + * `requireUnambiguous` misses: with a single candidate the guess still resolves, to a sibling. + */ + +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { ORCA_STRUCTURED_SESSION_ENV } from '../shared/structured-session-marker' + +const callMock = vi.hoisted(() => vi.fn()) +const getTerminalHandleMock = vi.hoisted(() => vi.fn()) + +vi.mock('./format', () => ({ printResult: vi.fn() })) +vi.mock('./selectors', () => ({ getTerminalHandle: getTerminalHandleMock })) + +import { ORCHESTRATION_HANDLERS } from './handlers/orchestration' + +const originalMarker = process.env[ORCA_STRUCTURED_SESSION_ENV] +const originalHandle = process.env.ORCA_TERMINAL_HANDLE + +function invoke(command: string, flags = new Map()) { + return ORCHESTRATION_HANDLERS[command]!({ + flags, + client: { call: callMock }, + cwd: '/tmp/repo', + json: true + } as never) +} + +describe('a structured chat session with no orchestration identity', () => { + beforeEach(() => { + callMock.mockReset() + getTerminalHandleMock.mockReset() + delete process.env.ORCA_TERMINAL_HANDLE + process.env[ORCA_STRUCTURED_SESSION_ENV] = '1' + // Exactly ONE terminal pane in the worktree: the single-candidate case, where + // `requireUnambiguous` still resolves and would hand this session a sibling's handle. + getTerminalHandleMock.mockResolvedValue('term_sibling') + }) + + afterEach(() => { + if (originalMarker === undefined) { + delete process.env[ORCA_STRUCTURED_SESSION_ENV] + } else { + process.env[ORCA_STRUCTURED_SESSION_ENV] = originalMarker + } + if (originalHandle === undefined) { + delete process.env.ORCA_TERMINAL_HANDLE + } else { + process.env.ORCA_TERMINAL_HANDLE = originalHandle + } + }) + + it('refuses a bare check instead of consuming the mailbox of a sibling pane', async () => { + await expect(invoke('orchestration check')).rejects.toMatchObject({ + code: 'no_active_sender_terminal', + message: expect.stringContaining('--terminal') + }) + // Neither guessed nor sent: a destructive read must not reach the runtime at all. + expect(getTerminalHandleMock).not.toHaveBeenCalled() + expect(callMock).not.toHaveBeenCalled() + }) + + it('refuses a bare send for the same reason, naming --from', async () => { + await expect( + invoke( + 'orchestration send', + new Map([ + ['to', 'term_coord'], + ['subject', 'hi'], + ['body', 'hello'] + ]) + ) + ).rejects.toMatchObject({ message: expect.stringContaining('--from') }) + expect(callMock).not.toHaveBeenCalled() + }) + + it('still accepts an explicit --terminal, which is the actionable escape', async () => { + callMock.mockResolvedValue({ result: { messages: [], count: 0 } }) + await invoke('orchestration check', new Map([['terminal', 'structworker_self']])) + expect(callMock).toHaveBeenCalledWith( + 'orchestration.check', + expect.objectContaining({ terminal: 'structworker_self' }) + ) + }) + + it('leaves an ordinary shell alone, which has no marker and may still guess', async () => { + delete process.env[ORCA_STRUCTURED_SESSION_ENV] + callMock.mockResolvedValue({ result: { messages: [], count: 0 } }) + await invoke('orchestration check') + expect(getTerminalHandleMock).toHaveBeenCalled() + }) +}) diff --git a/src/cli/selectors.ts b/src/cli/selectors.ts index 0f90dc94508..0db6ef82ff4 100644 --- a/src/cli/selectors.ts +++ b/src/cli/selectors.ts @@ -206,14 +206,18 @@ export async function getBrowserWorktreeSelector( export async function getTerminalHandle( flags: Map, cwd: string, - client: RuntimeClient + client: RuntimeClient, + options: { requireUnambiguous?: boolean } = {} ): Promise { const explicit = getOptionalStringFlag(flags, 'terminal') if (explicit) { return explicit } const worktree = await getBrowserWorktreeSelector(flags, cwd, client) - const response = await client.call<{ handle: string }>('terminal.resolveActive', { worktree }) + const response = await client.call<{ handle: string }>('terminal.resolveActive', { + worktree, + ...(options.requireUnambiguous ? { requireUnambiguous: true } : {}) + }) return response.result.handle } diff --git a/src/cli/specs/orchestration-worker-specs.ts b/src/cli/specs/orchestration-worker-specs.ts index e11ec3b1a91..c6a54ff1e1a 100644 --- a/src/cli/specs/orchestration-worker-specs.ts +++ b/src/cli/specs/orchestration-worker-specs.ts @@ -37,6 +37,9 @@ export const ORCHESTRATION_WORKER_COMMAND_SPECS: CommandSpec[] = [ '--model supports Claude, Codex, and Cursor opaque provider model ids; --effort requires --model. Neither can combine with --terminal.', 'New worktrees use agent-first creation and default --setup to run. Repository start-immediately runs setup beside the agent; wait-for-setup gates agent readiness and task input.', 'Creation flags (--name, --repo, --base-branch, --display-name, --comment, --setup) are rejected for current/existing worktrees. Use exact --repo on the selected server; project/host convenience routing remains on worktree create.', + "How the worker runs follows the user's own setting for new agent tabs; there is no flag for it and no caller needs to ask. A dispatch the setting cannot apply to still starts, so the placement, agent, and launch options passed here are always the ones honoured.", + 'Drive every worker the same way whichever way it was started: the same orchestration verbs, the same handle. Mail, dispatch, worker-show, worker-read and the whole lifecycle behave identically. The start receipt records which one ran, for operators and telemetry.', + 'Not every worker has a terminal. Read output with worker-read --source auto or --source transcript, which always work; --source terminal is refused when there is none, and orca terminal verbs do not accept every worker handle. Nothing above needs you to know which kind you have — the orchestration verbs cover all of them.', '--on selects only the worker server; the Run and this command remain on the current Orca server.', 'Remote current and new-child are invalid; discover an exact remote selector or use new-top-level.', '--retry-of needs --task naming the failed Task (--spec creates a new one) and does not inherit placement; repeat the intended --on/worktree and --agent/terminal choices.', diff --git a/src/cli/specs/orchestration.test.ts b/src/cli/specs/orchestration.test.ts index 54df971ba52..cd69aa50708 100644 --- a/src/cli/specs/orchestration.test.ts +++ b/src/cli/specs/orchestration.test.ts @@ -16,6 +16,46 @@ describe('orchestration send command spec', () => { }) }) +describe('orchestration worker-start command spec', () => { + const startSpec = ORCHESTRATION_COMMAND_SPECS.find( + (spec) => spec.path.join(' ') === 'orchestration worker-start' + ) + + it('offers no flag for the worker mode, because settings decide it', () => { + expect(startSpec?.allowedFlags).not.toContain('structured') + expect(startSpec?.usage).not.toContain('--structured') + expect(startSpec?.notes?.join('\n')).not.toContain('--structured') + }) + + it('documents the settings default and the fallback that keeps every dispatch working', () => { + const notes = startSpec?.notes?.join('\n') ?? '' + expect(notes).toContain("follows the user's own setting for new agent tabs") + expect(notes).toContain('A dispatch the setting cannot apply to still starts') + }) + + it('never points a caller at the worker kind, which nothing it runs depends on', () => { + const notes = startSpec?.notes?.join('\n') ?? '' + // The mode is in the receipt for operators and telemetry. Naming the field here would teach a + // coordinator agent to branch on something no verb it runs behaves differently for. + expect(notes).not.toMatch(/mode field/) + expect(notes).not.toMatch(/structured chat session/) + expect(notes).toContain('Drive every worker the same way') + }) + + it('does not promise uniformity it cannot deliver', () => { + // The note used to promise "the same verbs, the same handle, and the same worker-read + // sources". All three clauses were false for a worker with no terminal: `orca terminal` verbs + // refuse its handle and `--source terminal` has nothing to serve. A spec agents read must not + // carry a false promise — but it also must not name the worker kind, or a coordinator starts + // branching on something no verb it runs behaves differently for. So it states the limitation + // and the always-working alternative, without naming a mode. + const notes = startSpec?.notes?.join('\n') ?? '' + expect(notes).not.toContain('the same worker-read sources') + expect(notes).toContain('Not every worker has a terminal') + expect(notes).toContain('--source transcript') + }) +}) + describe('orchestration check command spec', () => { it('documents --types as a wake condition rather than a batch filter', () => { const checkSpec = ORCHESTRATION_COMMAND_SPECS.find( diff --git a/src/main/agent-hooks/first-work-branch-rename.test.ts b/src/main/agent-hooks/first-work-branch-rename.test.ts index cb1107a8622..93d15e5a2e6 100644 --- a/src/main/agent-hooks/first-work-branch-rename.test.ts +++ b/src/main/agent-hooks/first-work-branch-rename.test.ts @@ -1,6 +1,10 @@ import { beforeEach, describe, expect, it, vi } from 'vitest' import type { GlobalSettings } from '../../shared/global-settings-types' import type { Repo } from '../../shared/repo-types' +import type { AgentJournalRenderItem } from '../../shared/agent-session-journal-types' +import type { AgentSessionJournal } from '../native-chat/agent-session-journal/journal-store' +import { StructuredAgentSessionStatusFeed } from '../native-chat/agent-session-wire/structured-agent-session-status-feed' +import { maybeAutoRenameWorkspaceOnFirstStructuredTurn } from './first-work-structured-session-rename' import { WORKTREE_ID_SEPARATOR } from '../../shared/worktree/id' const { @@ -83,6 +87,134 @@ describe('maybeAutoRenameBranchOnFirstWork', () => { ) }) + it.each([ + ['claude', WORKTREE_ID], + ['codex', WORKTREE_ID], + ['claude', FOLDER_WORKTREE_ID], + ['codex', FOLDER_WORKTREE_ID] + ] as const)( + 'renames %s workspace %s on live work without a subscriber, preserving replay, dedupe and retries', + async (agent, workspaceId) => { + const { deps, setDisplayName } = makeDeps({ + getFolderWorkspacePath: () => '/workspace/platform', + isPendingFirstAgentMessageRename: () => true + }) + const items: AgentJournalRenderItem[] = [] + const journal = { + lastActivityAt: () => 0, + snapshot: () => ({ items }), + lastActivityAt: () => 1, + isReadOnly: false + } as unknown as AgentSessionJournal + const pending: Promise[] = [] + const observe = vi.fn((summary, options) => { + const work = maybeAutoRenameWorkspaceOnFirstStructuredTurn(summary, options, deps) + if (work) { + pending.push(work) + } + }) + const feed = new StructuredAgentSessionStatusFeed({ + sessions: new Map([ + ['session', { journal, params: { location: { workspaceId }, provider: agent } }] + ]), + getRecord: () => null, + now: () => 1, + onStatusChanged: observe + }) + const user = { + body: { kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'Fix auth' }] } + } as AgentJournalRenderItem + const turn = { + body: { + kind: 'status', + text: 'Working', + turnLifecycle: { turnId: 'turn-1', state: 'running' } + } + } as AgentJournalRenderItem + items.push(user, turn) + feed.publish('session', journal, { replay: true }) + await Promise.all(pending) + expect(gitExecFileAsyncMock).not.toHaveBeenCalled() + + items.pop() + feed.publish('session', journal) + items.push(turn) + generateBranchNameMock.mockResolvedValueOnce({ success: false, error: 'temporary failure' }) + feed.publish('session', journal) + await Promise.all(pending) + expect(generateBranchNameMock).toHaveBeenCalledOnce() + expect(setDisplayName).not.toHaveBeenCalled() + const callsBeforeOutput = observe.mock.calls.length + for (let index = 0; index < 100; index++) { + feed.publish('session', journal) + } + expect(observe).toHaveBeenCalledTimes(callsBeforeOutput) + + items.pop() + feed.publish('session', journal) + items.push(turn) + feed.publish('session', journal) + await Promise.all(pending) + expect(generateBranchNameMock).toHaveBeenCalledTimes(2) + expect(setDisplayName).toHaveBeenCalledWith(workspaceId, 'Fix auth') + if (workspaceId === FOLDER_WORKTREE_ID) { + expect(gitExecFileAsyncMock).not.toHaveBeenCalled() + } else { + expect(gitExecFileAsyncMock).toHaveBeenCalledWith( + ['branch', '-m', 'you/fix-auth'], + expect.anything() + ) + } + } + ) + + it('does not probe git for a folder-project structured session with a synthetic worktree id', async () => { + const workspaceId = `${REPO_ID}::/workspace/platform::workspace:123e4567-e89b-12d3-a456-426614174000` + const { deps, setDisplayName, setRenameError } = makeDeps({ + getRepo: () => ({ id: REPO_ID, kind: 'folder', path: '/workspace/platform' }) as Repo + }) + const journal = { + lastActivityAt: () => 0, + isReadOnly: false, + lastActivityAt: () => 1, + snapshot: () => ({ + items: [ + { body: { kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'Fix auth' }] } }, + { + body: { + kind: 'status', + text: 'Working', + turnLifecycle: { turnId: 'turn-1', state: 'running' } + } + } + ] + }) + } as unknown as AgentSessionJournal + const location = { workspaceId, workspaceKind: 'git-worktree' as const } + const pending: Promise[] = [] + const feed = new StructuredAgentSessionStatusFeed({ + sessions: new Map([['session', { journal, params: { location, provider: 'codex' } }]]), + getRecord: () => null, + now: () => 1, + onStatusChanged: (summary, options) => { + expect(summary.workspaceId).toBe(workspaceId) + const work = maybeAutoRenameWorkspaceOnFirstStructuredTurn(summary, options, deps) + if (work) { + pending.push(work) + } + } + }) + + feed.publish('session', journal) + await Promise.all(pending) + + expect(gitExecFileAsyncMock).not.toHaveBeenCalled() + expect(getSshGitProviderMock).not.toHaveBeenCalled() + expect(generateBranchNameMock).not.toHaveBeenCalled() + expect(setDisplayName).not.toHaveBeenCalled() + expect(setRenameError).toHaveBeenCalledWith(workspaceId, null) + }) + it('keeps incidental work-item markers from overriding the generated display name', async () => { const { deps, onRenamed, setDisplayName } = makeDeps() await maybeAutoRenameBranchOnFirstWork(workingEvent({ prompt: 'Fix auth from note #1' }), deps) @@ -233,6 +365,30 @@ describe('maybeAutoRenameBranchOnFirstWork', () => { expect(onRenamed).toHaveBeenCalledWith(FOLDER_WORKTREE_ID) }) + it.each([true, false])( + 'preserves a manual folder name during generation (pending=%s)', + async (pendingAfterRename) => { + let name = 'Platform workspace' + let pending = true + const { deps, setDisplayName } = makeDeps({ + resolveWorktreeIdForTab: () => FOLDER_WORKTREE_ID, + getFolderWorkspacePath: () => '/workspace/platform', + isPendingFirstAgentMessageRename: () => pending, + getCurrentDisplayName: () => name + }) + generateBranchNameMock.mockImplementationOnce(async () => { + name = 'My manual title' + pending = pendingAfterRename + return { success: true, slug: 'fix-auth' } + }) + + await maybeAutoRenameBranchOnFirstWork(workingEvent(), deps) + + expect(generateBranchNameMock).toHaveBeenCalledOnce() + expect(setDisplayName).not.toHaveBeenCalled() + } + ) + it('does not rename folder workspace titles without the pending marker', async () => { const { deps, setDisplayName } = makeDeps({ resolveWorktreeIdForTab: () => FOLDER_WORKTREE_ID, diff --git a/src/main/agent-hooks/first-work-branch-rename.ts b/src/main/agent-hooks/first-work-branch-rename.ts index 45f78e75e8a..355a0a35dc9 100644 --- a/src/main/agent-hooks/first-work-branch-rename.ts +++ b/src/main/agent-hooks/first-work-branch-rename.ts @@ -1,6 +1,7 @@ // On first agent work in a fresh workspace, replace the auto-generated creature branch (e.g. `you/Nautilus`) with a short work-derived name. import type { GlobalSettings } from '../../shared/global-settings-types' import type { Repo } from '../../shared/repo-types' +import { isFolderRepo } from '../../shared/repo-kind' import { getRepoIdFromWorktreeId, splitWorktreeIdForFilesystem } from '../../shared/worktree/id' import { parseWorkspaceKey } from '../../shared/workspace-scope' import { parsePaneKey } from '../../shared/stable-pane-id' @@ -170,6 +171,9 @@ async function runAutoRename( if (!repo || !parsed) { return stop('unresolved repo or worktree id') } + if (isFolderRepo(repo)) { + return stop('folder project has no branch to rename', true) + } const worktreePath = parsed.worktreePath const provider = repo.connectionId ? (getSshGitProvider(repo.connectionId) ?? null) : null diff --git a/src/main/agent-hooks/first-work-rename-runtime.ts b/src/main/agent-hooks/first-work-rename-runtime.ts new file mode 100644 index 00000000000..ebc9484071b --- /dev/null +++ b/src/main/agent-hooks/first-work-rename-runtime.ts @@ -0,0 +1,119 @@ +import { existsSync } from 'node:fs' +import { parseWorkspaceKey } from '../../shared/workspace-scope' +import { getRepoIdFromWorktreeId } from '../../shared/worktree/id' +import type { FirstWorkBranchRenameDeps } from './first-work-branch-rename' +import { rememberBranchRenameFailureOutput } from './branch-rename-failure-output' +import { renameWorktreeFolderOnFirstWork } from './first-work-folder-rename' +import { moveWorktree } from '../git/worktree' +import type { Store } from '../persistence' +import type { OrcaRuntimeService } from '../runtime/orca-runtime' + +const ENABLE_FIRST_WORK_FOLDER_RENAME = false + +export function firstWorkRenameDeps( + store: Store, + runtime: Pick< + OrcaRuntimeService, + | 'getCommitMessageAgentEnvironmentResolvers' + | 'notifyFolderWorkspaceChanged' + | 'notifyBranchRenamed' + | 'notifyWorktreeFolderRenamed' + > +): FirstWorkBranchRenameDeps { + return { + getSettings: () => store.getSettings(), + getRepo: (repoId) => store.getRepo(repoId), + getAgentEnvResolvers: () => runtime.getCommitMessageAgentEnvironmentResolvers(), + getCurrentDisplayName: (worktreeId) => { + const scope = parseWorkspaceKey(worktreeId) + return scope?.type === 'folder' + ? store.getFolderWorkspace(scope.folderWorkspaceId)?.name + : store.getWorktreeMeta(worktreeId)?.displayName + }, + getFolderWorkspacePath: (worktreeId) => { + const scope = parseWorkspaceKey(worktreeId) + return scope?.type === 'folder' + ? store.getFolderWorkspace(scope.folderWorkspaceId)?.folderPath + : undefined + }, + isPendingFirstAgentMessageRename: (worktreeId) => { + const scope = parseWorkspaceKey(worktreeId) + return scope?.type === 'folder' + ? store.getFolderWorkspace(scope.folderWorkspaceId)?.pendingFirstAgentMessageRename === true + : store.getWorktreeMeta(worktreeId)?.pendingFirstAgentMessageRename === true + }, + canRenameOrcaCreatedBranch: (worktreeId) => { + const meta = store.getWorktreeMeta(worktreeId) + // Why: a user branch could coincidentally match a creature name; only Orca-stamped worktrees are safe to auto-rename. + return !!meta?.orcaCreationSource && meta.preserveBranchOnDelete !== true + }, + setDisplayName: (worktreeId, displayName) => { + rememberBranchRenameFailureOutput(worktreeId, null) + const scope = parseWorkspaceKey(worktreeId) + if (scope?.type === 'folder') { + store.updateFolderWorkspace(scope.folderWorkspaceId, { + name: displayName, + pendingFirstAgentMessageRename: false, + firstAgentMessageRenameError: null + }) + runtime.notifyFolderWorkspaceChanged() + return + } + store.setWorktreeMeta(worktreeId, { + displayName, + // The first-agent title is an intentional user-facing label; keep it stable after the + // generated branch is renamed and across subsequent catalog refreshes. + displayNameIsPinned: true, + pendingFirstAgentMessageRename: false, + // Success clears the failure badge (redundant with the explicit setRenameError(null)). + firstAgentMessageRenameError: null + }) + }, + renameWorktreeFolder: ENABLE_FIRST_WORK_FOLDER_RENAME + ? (worktreeId, newLeaf) => + renameWorktreeFolderOnFirstWork(worktreeId, newLeaf, { + getRepo: (repoId) => store.getRepo(repoId), + getSettings: () => store.getSettings(), + migrateWorktreeIdentity: (oldId, newId) => store.migrateWorktreeIdentity(oldId, newId), + notifyWorktreeRenamed: (repoId, oldId, newId) => + runtime.notifyWorktreeFolderRenamed(repoId, oldId, newId), + pathExists: async (candidate) => existsSync(candidate), + moveWorktree + }) + : undefined, + setRenameError: (worktreeId, error, failureOutput) => { + // Refresh the full-output capture before the dedupe below — a repeat error string is still a fresh run. + rememberBranchRenameFailureOutput(worktreeId, error === null ? null : failureOutput) + // Skip the write + push when unchanged — most settled worktrees never had an error to clear. + const scope = parseWorkspaceKey(worktreeId) + if (scope?.type === 'folder') { + const current = store.getFolderWorkspace( + scope.folderWorkspaceId + )?.firstAgentMessageRenameError + if ((current ?? null) === (error ?? null)) { + return + } + store.updateFolderWorkspace(scope.folderWorkspaceId, { + firstAgentMessageRenameError: error + }) + runtime.notifyFolderWorkspaceChanged() + return + } + const current = store.getWorktreeMeta(worktreeId)?.firstAgentMessageRenameError + if ((current ?? null) === (error ?? null)) { + return + } + store.setWorktreeMeta(worktreeId, { firstAgentMessageRenameError: error }) + // Why: the hook only knows the worktreeId, so derive the repoId notifyBranchRenamed expects. + runtime.notifyBranchRenamed(getRepoIdFromWorktreeId(worktreeId)) + }, + resolveWorktreeIdForTab: (tabId) => store.getWorktreeIdForTab(tabId), + onRenamed: (repoIdOrWorktreeId) => { + if (parseWorkspaceKey(repoIdOrWorktreeId)?.type === 'folder') { + runtime.notifyFolderWorkspaceChanged() + return + } + runtime.notifyBranchRenamed(repoIdOrWorktreeId) + } + } +} diff --git a/src/main/agent-hooks/first-work-structured-session-rename.ts b/src/main/agent-hooks/first-work-structured-session-rename.ts new file mode 100644 index 00000000000..637dc1e9c3c --- /dev/null +++ b/src/main/agent-hooks/first-work-structured-session-rename.ts @@ -0,0 +1,28 @@ +import type { AgentSessionStatusSummary } from '../../shared/agent-session-wire' +import { + maybeAutoRenameBranchOnFirstWork, + type FirstWorkBranchRenameDeps +} from './first-work-branch-rename' + +export function maybeAutoRenameWorkspaceOnFirstStructuredTurn( + summary: AgentSessionStatusSummary, + options: { replay: boolean }, + deps: FirstWorkBranchRenameDeps +): Promise | undefined { + if (summary.status !== 'working') { + return + } + return maybeAutoRenameBranchOnFirstWork( + { + // No pane: a structured session is resolved by its workspace id, not by a terminal tab. + paneKey: '', + tabId: undefined, + worktreeId: summary.workspaceId, + state: 'working', + prompt: summary.latestPrompt, + assistantMessage: undefined, + isReplay: options.replay + }, + deps + ) +} diff --git a/src/main/agent-hooks/first-work-workspace-title-rename.ts b/src/main/agent-hooks/first-work-workspace-title-rename.ts index fb682c86cf3..063c6f15aec 100644 --- a/src/main/agent-hooks/first-work-workspace-title-rename.ts +++ b/src/main/agent-hooks/first-work-workspace-title-rename.ts @@ -27,6 +27,7 @@ export async function runFolderWorkspaceTitleAutoRename( return stop('folder workspace path unavailable') } + const originalDisplayName = deps.getCurrentDisplayName(worktreeId) const settings = deps.getSettings() const resolvedParams = resolveTextGenerationParams(settings, 'local', 'branchName', null) if (!resolvedParams.ok) { @@ -49,6 +50,13 @@ export async function runFolderWorkspaceTitleAutoRename( resolvedParams.params, target ) + // Generation may outlive a manual rename or workspace removal. + if ( + deps.isPendingFirstAgentMessageRename?.(worktreeId) !== true || + deps.getCurrentDisplayName(worktreeId) !== originalDisplayName + ) { + return stop('folder workspace changed during generation', true) + } if (!generated.success) { if (!generated.canceled) { deps.setRenameError(worktreeId, generated.error, generated.failureOutput ?? null) diff --git a/src/main/browser/agent-browser-bridge-execution.ts b/src/main/browser/agent-browser-bridge-execution.ts index 65f64b6fb08..31f5a191ede 100644 --- a/src/main/browser/agent-browser-bridge-execution.ts +++ b/src/main/browser/agent-browser-bridge-execution.ts @@ -12,6 +12,7 @@ import { import { translateResult } from './agent-browser-bridge-result' import { AgentBrowserBridgeTabs } from './agent-browser-bridge-tabs' import { ORCA_TAB_SESSION_PREFIX } from './agent-browser-orphan-sweep' +import { canSkipAgentBrowserSessionReset } from './agent-browser-session-reset' import { STALE_SESSION_CLOSE_TIMEOUT_MS, type AgentBrowserExecOptions, @@ -173,6 +174,15 @@ export abstract class AgentBrowserBridgeExecution extends AgentBrowserBridgeTabs } protected closeStaleAgentBrowserSession(sessionName: string): Promise { + if ( + canSkipAgentBrowserSessionReset({ + ownsSocketDirectory: this.ownsAgentBrowserSocketDirectory, + socketDirectory: this.agentBrowserEnv.AGENT_BROWSER_SOCKET_DIR, + sessionName + }) + ) { + return Promise.resolve() + } return new Promise((resolve, reject) => { let child: ReturnType | null = null let settled = false diff --git a/src/main/browser/agent-browser-bridge-session-lifecycle.test.ts b/src/main/browser/agent-browser-bridge-session-lifecycle.test.ts index 5eab202541c..2955cb66263 100644 --- a/src/main/browser/agent-browser-bridge-session-lifecycle.test.ts +++ b/src/main/browser/agent-browser-bridge-session-lifecycle.test.ts @@ -1,18 +1,26 @@ import { describe, it, expect, vi, beforeEach } from 'vitest' -const { execFileMock, webContentsFromIdMock, existsSyncMock, readFileSyncMock, stdinWrites } = - vi.hoisted(() => ({ - execFileMock: vi.fn(), - webContentsFromIdMock: vi.fn(), - existsSyncMock: vi.fn(() => false), - readFileSyncMock: vi.fn(() => Buffer.from('')), - stdinWrites: [] as string[] - })) +const { + execFileMock, + webContentsFromIdMock, + existsSyncMock, + readFileSyncMock, + lstatSyncMock, + stdinWrites +} = vi.hoisted(() => ({ + execFileMock: vi.fn(), + webContentsFromIdMock: vi.fn(), + existsSyncMock: vi.fn(() => false), + readFileSyncMock: vi.fn(() => Buffer.from('')), + lstatSyncMock: vi.fn(), + stdinWrites: [] as string[] +})) vi.mock('child_process', () => ({ execFile: execFileMock })) vi.mock('fs', () => ({ existsSync: existsSyncMock, readFileSync: readFileSyncMock, + lstatSync: lstatSyncMock, accessSync: vi.fn(), chmodSync: vi.fn(), constants: { X_OK: 1 } @@ -73,6 +81,14 @@ function closeCallCount(): number { describe('AgentBrowserBridge', () => { let bridge: AgentBrowserBridge + // The mocked fs has no mkdirSync, so the constructor never claims a socket directory itself. + function ownSocketDirectory(): void { + Object.assign(bridge, { + ownsAgentBrowserSocketDirectory: true, + agentBrowserEnv: { AGENT_BROWSER_SOCKET_DIR: '/tmp/orca-ab-test' } + }) + } + beforeEach(() => { resetAgentBrowserBridgeMocks({ webContentsFromIdMock, @@ -81,11 +97,29 @@ describe('AgentBrowserBridge', () => { stdinWrites, cdpWsProxyInstances: CdpWsProxyMock.instances }) + // Default to a socket that exists so an unprepared test still takes the reset path. + lstatSyncMock.mockReset() + lstatSyncMock.mockReturnValue({}) bridge = new AgentBrowserBridge(mockBrowserManager()) bridge.setActiveTab(100) }) + it('snapshots a fresh owned session without launching a helper just to close it', async () => { + ownSocketDirectory() + lstatSyncMock.mockImplementation(() => { + throw Object.assign(new Error('No socket'), { code: 'ENOENT' }) + }) + webContentsFromIdMock.mockReturnValue(mockWebContents(100)) + succeedWith({ snapshot: 'ready' }) + + expect(await bridge.snapshot()).toMatchObject({ snapshot: 'ready' }) + expect(closeCallCount()).toBe(0) + expect(lstatSyncMock).toHaveBeenCalledWith('/tmp/orca-ab-test/orca-tab-tab-1.sock') + }) + it('fails closed when stale agent-browser session ownership cannot be reset', async () => { + ownSocketDirectory() + lstatSyncMock.mockReturnValue({}) vi.useFakeTimers() try { const closeKill = vi.fn() diff --git a/src/main/browser/agent-browser-session-reset.test.ts b/src/main/browser/agent-browser-session-reset.test.ts new file mode 100644 index 00000000000..b38822d5175 --- /dev/null +++ b/src/main/browser/agent-browser-session-reset.test.ts @@ -0,0 +1,51 @@ +import { beforeEach, expect, it, vi } from 'vitest' +import { join } from 'node:path' + +const { lstatSync } = vi.hoisted(() => ({ lstatSync: vi.fn() })) +vi.mock('node:fs', () => ({ lstatSync })) +import { canSkipAgentBrowserSessionReset } from './agent-browser-session-reset' + +const owned = { + ownsSocketDirectory: true, + socketDirectory: '/tmp/orca-ab-profile', + sessionName: 'orca-tab-page' +} +const socketPath = join(owned.socketDirectory, 'orca-tab-page.sock') + +beforeEach(() => { + lstatSync.mockReset() +}) + +it('skips an absent owned socket', () => { + lstatSync.mockImplementation(() => { + throw Object.assign(new Error('No socket'), { code: 'ENOENT' }) + }) + expect(canSkipAgentBrowserSessionReset(owned)).toBe(true) + expect(lstatSync).toHaveBeenCalledWith(socketPath) +}) + +it('requires reset when a socket or symlink exists', () => { + lstatSync.mockReturnValue({}) + expect(canSkipAgentBrowserSessionReset(owned)).toBe(false) + expect(lstatSync).toHaveBeenCalledWith(socketPath) +}) + +it.each(['EACCES', 'EIO', 'ENOTDIR'])('requires reset for %s', (code) => { + lstatSync.mockImplementation(() => { + throw Object.assign(new Error('Socket inspection failed'), { code }) + }) + expect(canSkipAgentBrowserSessionReset(owned)).toBe(false) + expect(lstatSync).toHaveBeenCalledWith(socketPath) +}) + +// Windows and inherited socket directories both arrive as ownsSocketDirectory: false. +it.each([ + { ownsSocketDirectory: false }, + { socketDirectory: undefined }, + { sessionName: '../other' }, + { sessionName: 'has space' }, + { sessionName: '' } +])('requires reset without an owned Unix socket address: %j', (override) => { + expect(canSkipAgentBrowserSessionReset({ ...owned, ...override })).toBe(false) + expect(lstatSync).not.toHaveBeenCalled() +}) diff --git a/src/main/browser/agent-browser-session-reset.ts b/src/main/browser/agent-browser-session-reset.ts new file mode 100644 index 00000000000..8c6b7545f02 --- /dev/null +++ b/src/main/browser/agent-browser-session-reset.ts @@ -0,0 +1,30 @@ +import { lstatSync } from 'node:fs' +import { join } from 'node:path' + +// agent-browser's own session-name rule; doubles as a traversal fence for the `join` below. +const SAFE_SESSION_NAME = /^[A-Za-z0-9_-]+$/ + +/** + * True when no daemon can be holding `sessionName`, so closing it would only start one. + * + * Only an Orca-derived socket directory proves that (`ownsSocketDirectory`): it is a + * private per-profile `/tmp` directory, never an inherited one shared with a second + * profile, and never Windows, which uses named pipes and leaves no socket to inspect. + */ +export function canSkipAgentBrowserSessionReset(options: { + ownsSocketDirectory: boolean + socketDirectory: string | undefined + sessionName: string +}): boolean { + const { socketDirectory, sessionName } = options + if (!options.ownsSocketDirectory || !socketDirectory || !SAFE_SESSION_NAME.test(sessionName)) { + return false + } + try { + lstatSync(join(socketDirectory, `${sessionName}.sock`)) + return false + } catch (error) { + // Only a proven-absent socket is safe to skip; permission and other failures prove nothing. + return (error as NodeJS.ErrnoException).code === 'ENOENT' + } +} diff --git a/src/main/claude/claude-slash-command-catalog.test.ts b/src/main/claude/claude-slash-command-catalog.test.ts new file mode 100644 index 00000000000..20d79f9f53d --- /dev/null +++ b/src/main/claude/claude-slash-command-catalog.test.ts @@ -0,0 +1,132 @@ +import { describe, expect, it } from 'vitest' +import { ClaudeSlashCommandCatalog, readClaudeSlashCommands } from './claude-slash-command-catalog' + +function init(overrides: Record = {}): Record { + return { + type: 'system', + subtype: 'init', + session_id: 'provider-1', + slash_commands: ['clear', 'ref-oss', 'doctor', 'opsx:apply'], + terminal_slash_commands: ['doctor'], + skills: ['ref-oss', 'doctor'], + ...overrides + } +} + +describe('claude slash command catalog', () => { + it('tags reported skills and drops the commands reserved for a terminal UI', () => { + expect(readClaudeSlashCommands(init())).toEqual([ + { name: 'clear', kind: 'command' }, + { name: 'ref-oss', kind: 'skill' }, + { name: 'opsx:apply', kind: 'command' } + ]) + }) + + it('rejects blank, whitespace-carrying and duplicate names', () => { + expect( + readClaudeSlashCommands( + init({ slash_commands: ['clear', ' ', 'two words', 'clear'], skills: [] }) + ) + ).toEqual([{ name: 'clear', kind: 'command' }]) + }) + + it('seeds from the init frame that proved the session', () => { + expect(new ClaudeSlashCommandCatalog(init()).commands).toHaveLength(3) + expect(new ClaudeSlashCommandCatalog().commands).toBeUndefined() + // A frame of the right subtype but without the array is not a catalog. + expect( + new ClaudeSlashCommandCatalog({ type: 'system', subtype: 'init' }).commands + ).toBeUndefined() + }) + + it('replaces the catalog on commands_changed and reports only real changes', () => { + const catalog = new ClaudeSlashCommandCatalog(init()) + expect(catalog.observe(init())).toBe(false) + expect(catalog.observe({ type: 'assistant', slash_commands: ['other'] })).toBe(false) + expect( + catalog.observe({ + type: 'system', + subtype: 'commands_changed', + slash_commands: ['clear', 'brand-new'], + skills: ['brand-new'] + }) + ).toBe(true) + expect(catalog.commands).toEqual([ + { name: 'clear', kind: 'command' }, + { name: 'brand-new', kind: 'skill' } + ]) + }) + + it('notices a name that only changed kind', () => { + const catalog = new ClaudeSlashCommandCatalog( + init({ slash_commands: ['review'], skills: [], terminal_slash_commands: [] }) + ) + expect( + catalog.observe({ + type: 'system', + subtype: 'commands_changed', + slash_commands: ['review'], + skills: ['review'] + }) + ).toBe(true) + expect(catalog.commands).toEqual([{ name: 'review', kind: 'skill' }]) + }) +}) + +it('accepts descriptor reloads, removing old skills while retaining terminal filtering', () => { + const catalog = new ClaudeSlashCommandCatalog(init()) + const reload = { + type: 'system', + subtype: 'commands_changed', + commands: [ + { name: 'clear', description: 'Clear', argumentHint: '' }, + { name: 'new-skill', description: 'New', argumentHint: '' }, + { name: 'doctor', description: 'Terminal', argumentHint: '' } + ] + } + expect(catalog.observe(reload)).toBe(true) + expect(catalog.commands).toEqual([ + { name: 'clear', kind: 'command' }, + { name: 'new-skill', kind: 'skill' } + ]) + expect(catalog.observe(reload)).toBe(false) + expect(catalog.observe({ ...reload, commands: [] })).toBe(true) + expect(catalog.commands).toEqual([]) +}) + +it('lets stream init refine a control seed and preserves kinds across descriptor reloads', () => { + const seed = { commands: [{ name: 'clear' }, { name: 'project-check' }] } + const catalog = new ClaudeSlashCommandCatalog(undefined, seed) + expect(catalog.commands).toEqual([ + { name: 'clear', kind: 'command', kindUnspecified: true }, + { name: 'project-check', kind: 'command', kindUnspecified: true } + ]) + expect(catalog.observe({ type: 'system', subtype: 'commands_changed', ...seed })).toBe(false) + const fullInit = init({ slash_commands: ['clear', 'project-check'], skills: ['project-check'] }) + expect(catalog.observe(fullInit)).toBe(true) + expect(catalog.commands).toEqual([ + { name: 'clear', kind: 'command' }, + { name: 'project-check', kind: 'skill' } + ]) + expect(catalog.observe({ type: 'system', subtype: 'commands_changed', ...seed })).toBe(false) + expect(new ClaudeSlashCommandCatalog(fullInit, seed).commands).toEqual(catalog.commands) + expect(new ClaudeSlashCommandCatalog(init({ slash_commands: [] }), seed).commands).toEqual([]) +}) + +it('distinguishes missing or malformed control catalogs from authoritative empty ones', () => { + for (const initialization of [undefined, null, {}, { commands: null }, { commands: 'bad' }]) { + expect(new ClaudeSlashCommandCatalog(undefined, initialization).commands).toBeUndefined() + } + expect(new ClaudeSlashCommandCatalog(undefined, { commands: [] }).commands).toEqual([]) + expect( + new ClaudeSlashCommandCatalog(undefined, { + commands: [null, {}, { name: ' ' }, { name: 'two words' }, { name: 'ok' }, { name: 'ok' }] + }).commands + ).toEqual([{ name: 'ok', kind: 'command', kindUnspecified: true }]) +}) + +it('publishes classification becoming authoritative even when the name and kind stay unchanged', () => { + const catalog = new ClaudeSlashCommandCatalog(undefined, { commands: [{ name: 'clear' }] }) + expect(catalog.observe(init({ slash_commands: ['clear'], skills: [] }))).toBe(true) + expect(catalog.commands).toEqual([{ name: 'clear', kind: 'command' }]) +}) diff --git a/src/main/claude/claude-slash-command-catalog.ts b/src/main/claude/claude-slash-command-catalog.ts new file mode 100644 index 00000000000..b1f65d93d50 --- /dev/null +++ b/src/main/claude/claude-slash-command-catalog.ts @@ -0,0 +1,123 @@ +import type { AgentSessionSlashCommand } from '../../shared/agent-session-wire' + +// Stream init carries name arrays; control initialization and reloads carry descriptors. +const MAX_COMMANDS = 512 +const MAX_NAME_LENGTH = 200 + +function names(value: unknown): string[] { + if (!Array.isArray(value)) { + return [] + } + const seen = new Set() + for (const entry of value) { + if (seen.size >= MAX_COMMANDS) { + break + } + const name = typeof entry === 'string' ? entry.trim() : '' + if (name.length > 0 && name.length <= MAX_NAME_LENGTH && !/\s/u.test(name)) { + seen.add(name) + } + } + return [...seen] +} + +function descriptorNames(value: unknown): string[] { + return names( + Array.isArray(value) + ? value.map((entry) => (entry !== null && typeof entry === 'object' ? entry.name : undefined)) + : [] + ) +} + +function carriesCommandCatalog(message: Record): boolean { + return ( + message.type === 'system' && + (message.subtype === 'init' || message.subtype === 'commands_changed') && + Array.isArray(message.slash_commands) + ) +} + +/** What the session reports it can run, minus what it reserves for a terminal UI. */ +export function readClaudeSlashCommands( + message: Record +): AgentSessionSlashCommand[] { + // Why: the hide-list exists so a non-terminal UI like chat does not offer a + // command that only means something inside the CLI's own TUI. + const hidden = new Set(names(message.terminal_slash_commands)) + const skills = new Set(names(message.skills)) + return names(message.slash_commands) + .filter((name) => !hidden.has(name)) + .map((name) => ({ name, kind: skills.has(name) ? ('skill' as const) : ('command' as const) })) +} + +/** Per-session catalog seeded during acquisition and refreshed by provider frames. */ +export class ClaudeSlashCommandCatalog { + private entries: AgentSessionSlashCommand[] | undefined + private hasSkillClassification = false + private hidden = new Set() + private commandNames = new Set() + + constructor(initMessage?: Record, initialization?: unknown) { + // SessionStart can prove acquisition before the first stream init exists. + if ( + initialization !== null && + typeof initialization === 'object' && + 'commands' in initialization && + Array.isArray(initialization.commands) + ) { + this.entries = descriptorNames(initialization.commands).map((name) => ({ + name, + kind: 'command', + kindUnspecified: true + })) + } + if (initMessage) { + this.observe(initMessage) + } + } + + get commands(): AgentSessionSlashCommand[] | undefined { + return this.entries + } + + /** True when this frame replaced the catalog with a different one. */ + observe(message: Record): boolean { + let next: AgentSessionSlashCommand[] + if (carriesCommandCatalog(message)) { + this.hasSkillClassification = true + this.hidden = new Set(names(message.terminal_slash_commands)) + next = readClaudeSlashCommands(message) + this.commandNames = new Set( + next.filter((entry) => entry.kind === 'command').map((entry) => entry.name) + ) + } else if ( + message.type === 'system' && + message.subtype === 'commands_changed' && + Array.isArray(message.commands) + ) { + next = descriptorNames(message.commands) + .filter((name) => !this.hidden.has(name)) + .map((name) => + this.hasSkillClassification + ? { name, kind: this.commandNames.has(name) ? 'command' : 'skill' } + : { name, kind: 'command', kindUnspecified: true } + ) + } else { + return false + } + if ( + this.entries !== undefined && + next.length === this.entries.length && + next.every( + (entry, index) => + entry.name === this.entries?.[index]?.name && + entry.kind === this.entries?.[index]?.kind && + entry.kindUnspecified === this.entries?.[index]?.kindUnspecified + ) + ) { + return false + } + this.entries = next + return true + } +} diff --git a/src/main/claude/claude-structured-content-parts.test.ts b/src/main/claude/claude-structured-content-parts.test.ts index d2142150937..1d9ed800e0f 100644 --- a/src/main/claude/claude-structured-content-parts.test.ts +++ b/src/main/claude/claude-structured-content-parts.test.ts @@ -47,6 +47,92 @@ const BASE64_IMAGE = { } describe('Claude message content parts', () => { + it.each(['isMeta', 'isSynthetic', 'isCompactSummary'])( + 'consumes %s skill context without a user bubble, fallback, or new turn', + (flag) => { + const state = sinkState() + const translator = createClaudeJournalTranslator({ sink: state.sink }) + const event = userMessageWith({ type: 'text', text: '# Skill instructions' }) + translator.handle({ ...event, message: { ...event.message, [flag]: true } }) + expect(state.items).toEqual([]) + expect(state.sink.publish).not.toHaveBeenCalled() + } + ) + + it('keeps tool results in an injected skill message', () => { + const state = sinkState() + const translator = createClaudeJournalTranslator({ sink: state.sink }) + const event = userMessageWith({ + type: 'tool_result', + tool_use_id: 'skill-call', + content: 'Skill loaded' + }) + translator.handle({ ...event, message: { ...event.message, isMeta: true } }) + expect(state.items.map((item) => item.body)).toEqual([ + expect.objectContaining({ + kind: 'tool-call', + state: 'completed', + output: expect.objectContaining({ head: 'Skill loaded', truncated: false }) + }) + ]) + }) + + it.each([ + { content: '# Skill instructions' }, + { content: [{ type: 'future_context', text: '# Skill instructions' }] }, + { + content: [ + { type: 'text', text: '[Image: source: /tmp/pasted.png]' }, + { type: 'text', text: '# Skill instructions' } + ] + } + ])('does not surface injected content as text or a provider fallback: %j', ({ content }) => { + const state = sinkState() + const translator = createClaudeJournalTranslator({ sink: state.sink }) + const event = userMessageWith(null) + translator.handle({ + ...event, + message: { ...event.message, isMeta: true, message: { role: 'user', content } } + }) + expect(state.items).toEqual([]) + }) + + it('does not render user echoes even without metadata flags', () => { + const state = sinkState() + const translator = createClaudeJournalTranslator({ sink: state.sink }) + for (const content of ['/example-skill', '# Skill instructions']) { + const event = userMessageWith(null) + translator.handle({ + ...event, + message: { ...event.message, message: { role: 'user', content } } + }) + } + expect(state.items.flatMap(({ body }) => (body.kind === 'message' ? body.blocks : []))).toEqual( + [] + ) + }) + + it('silently consumes unmarked user context with unknown content parts', () => { + const state = sinkState() + const translator = createClaudeJournalTranslator({ sink: state.sink }) + const event = userMessageWith({ type: 'future_context', text: 'Expanded instructions' }) + translator.handle({ ...event, startsTurn: undefined }) + expect(state.items).toEqual([]) + expect(state.sink.publish).not.toHaveBeenCalled() + }) + + it('does not render injected image companions or start a turn', () => { + const state = sinkState() + const translator = createClaudeJournalTranslator({ sink: state.sink }) + const event = userMessageWith(null) + const content = [{ type: 'text', text: '[Image: source: /tmp/pasted.png]' }] + translator.handle({ + ...event, + message: { ...event.message, isMeta: true, message: { role: 'user', content } } + }) + expect(state.items).toEqual([]) + }) + it('does not leak a wire kind for a locally attached image', () => { const state = sinkState() const translator = createClaudeJournalTranslator({ sink: state.sink }) @@ -56,7 +142,7 @@ describe('Claude message content parts', () => { expect(providerRows(state.items)).toEqual([]) }) - it('still renders an image the CLI sends by url', () => { + it('does not render echoed image URLs', () => { const state = sinkState() const translator = createClaudeJournalTranslator({ sink: state.sink }) @@ -67,21 +153,29 @@ describe('Claude message content parts', () => { expect(providerRows(state.items)).toEqual([]) expect( state.items.flatMap((item) => (item.body.kind === 'message' ? item.body.blocks : [])) - ).toContainEqual({ type: 'image-ref', url: 'https://x.test/a.png' }) + ).toEqual([]) }) it('says what is true for a content part it cannot render, not the wire kind', () => { const state = sinkState() const translator = createClaudeJournalTranslator({ sink: state.sink }) - translator.handle(userMessageWith({ type: 'some_future_part', payload: { a: 1 } })) + const event = userMessageWith(null) + translator.handle({ + ...event, + message: { + ...event.message, + type: 'assistant', + message: { role: 'assistant', content: [{ type: 'some_future_part', payload: { a: 1 } }] } + } + }) const rows = providerRows(state.items) expect(rows).toHaveLength(1) // The kind stays on the row for debugging, behind the disclosure. - expect(rows[0].kind).toBe('message:user:content:some_future_part') + expect(rows[0].kind).toBe('message:assistant:content:some_future_part') // ...but the visible text is a sentence, not the opcode. - expect(rows[0].text).not.toContain('message:user:content') + expect(rows[0].text).not.toContain('message:assistant:content') expect(rows[0].text.toLowerCase()).toContain('claude') }) @@ -89,9 +183,18 @@ describe('Claude message content parts', () => { const state = sinkState() const translator = createClaudeJournalTranslator({ sink: state.sink }) - translator.handle( - userMessageWith({ type: 'some_future_part', message: 'the server refused the upload' }) - ) + const event = userMessageWith(null) + translator.handle({ + ...event, + message: { + ...event.message, + type: 'assistant', + message: { + role: 'assistant', + content: [{ type: 'some_future_part', message: 'the server refused the upload' }] + } + } + }) expect(providerRows(state.items)[0].text).toBe('the server refused the upload') }) diff --git a/src/main/claude/claude-structured-dispatch.test.ts b/src/main/claude/claude-structured-dispatch.test.ts index d66a64f82eb..ad09357df58 100644 --- a/src/main/claude/claude-structured-dispatch.test.ts +++ b/src/main/claude/claude-structured-dispatch.test.ts @@ -7,6 +7,7 @@ import { dispatchClaudeTurn, resolveClaudeReplayWaiter } from './claude-structur import { readClaudeImage } from './claude-structured-dispatch-content' import type { ClaudeSession } from './claude-structured-session-state' import { ClaudeBackgroundTaskTracker } from './claude-background-task-tracker' +import { ClaudeSlashCommandCatalog } from './claude-slash-command-catalog' function sessionFor(send = vi.fn().mockResolvedValue(undefined)): ClaudeSession { return { @@ -21,6 +22,7 @@ function sessionFor(send = vi.fn().mockResolvedValue(undefined)): ClaudeSession retiredDispatchWaiters: [], replayContentFallbackBlocked: false, backgroundTasks: new ClaudeBackgroundTaskTracker(), + commands: new ClaudeSlashCommandCatalog(), dispatchSequence: 0, optionMutationSequence: 0, options: new Map(), @@ -49,6 +51,28 @@ function userReplayFrame(uuid: string, text: string): Record { } describe('Claude structured dispatch image limits', () => { + it.each(['isMeta', 'isSynthetic', 'isCompactSummary'])( + 'does not acknowledge a dispatch with %s context even when the client uuid matches', + async (flag) => { + const session = sessionFor() + const dispatched = dispatchClaudeTurn( + session, + { clientMessageId: 'client-1', body: userMessage([{ type: 'text', text: '/example' }]) }, + 1000 + ) + await vi.waitFor(() => expect(session.dispatchWaiters).toHaveLength(1)) + const sentUuid = session.dispatchWaiters[0]!.sentUuid + const replay = userReplayFrame(sentUuid, '/example') + expect(resolveClaudeReplayWaiter(session, { ...replay, [flag]: true })).toBe(false) + expect(session.dispatchWaiters).toHaveLength(1) + expect(resolveClaudeReplayWaiter(session, replay)).toBe(true) + await expect(dispatched).resolves.toMatchObject({ + state: 'accepted', + providerIdentity: { uuid: sentUuid } + }) + } + ) + it('recovers the active identity when a timed-out replay arrives late', async () => { const session = sessionFor() const dispatched = dispatchClaudeTurn( @@ -65,6 +89,60 @@ describe('Claude structured dispatch image limits', () => { expect(session.activeTurnSequence).toBe(session.dispatchSequence) }) + it('settles the send a timed-out replay proves was delivered', async () => { + const session = sessionFor() + const settled = vi.fn() + const dispatched = dispatchClaudeTurn( + session, + { clientMessageId: 'client-1', body: userMessage([{ type: 'text', text: 'one' }]) }, + 500 + ) + await vi.waitFor(() => expect(session.dispatchWaiters).toHaveLength(1)) + const sentUuid = (session.dispatchWaiters[0] as { sentUuid?: string }).sentUuid + await expect(dispatched).resolves.toMatchObject({ state: 'unknown' }) + + resolveClaudeReplayWaiter(session, userReplayFrame(sentUuid!, 'one'), settled) + expect(settled).toHaveBeenCalledWith({ + clientMessageId: 'client-1', + providerIdentity: { provider: 'claude', sessionId: 'provider-session', uuid: sentUuid } + }) + }) + + it('settles a superseded dispatch even though it no longer owns the turn identity', async () => { + const session = sessionFor() + const settled = vi.fn() + const first = dispatchClaudeTurn( + session, + { clientMessageId: 'client-1', body: userMessage([{ type: 'text', text: 'one' }]) }, + 500 + ) + await vi.waitFor(() => expect(session.dispatchWaiters).toHaveLength(1)) + const firstUuid = (session.dispatchWaiters[0] as { sentUuid?: string }).sentUuid + await expect(first).resolves.toMatchObject({ state: 'unknown' }) + + const second = dispatchClaudeTurn( + session, + { clientMessageId: 'client-2', body: userMessage([{ type: 'text', text: 'two' }]) }, + 100 + ) + await vi.waitFor(() => expect(session.dispatchWaiters).toHaveLength(1)) + const secondUuid = (session.dispatchWaiters[0] as { sentUuid?: string }).sentUuid + + // The stale replay must not claim the active turn, but the message it names + // did land, so the send it came from is delivered and must stop reading as + // unconfirmed — that banner is what makes a user resend a duplicate. + expect(resolveClaudeReplayWaiter(session, userReplayFrame(firstUuid!, 'one'), settled)).toBe( + false + ) + expect(settled).toHaveBeenCalledWith({ + clientMessageId: 'client-1', + providerIdentity: { provider: 'claude', sessionId: 'provider-session', uuid: firstUuid } + }) + resolveClaudeReplayWaiter(session, userReplayFrame(secondUuid!, 'two'), settled) + await expect(second).resolves.toMatchObject({ state: 'accepted' }) + expect(settled).toHaveBeenCalledTimes(1) + }) + it('never lets a late replay for dispatch A resolve dispatch B', async () => { const session = sessionFor() const first = dispatchClaudeTurn( diff --git a/src/main/claude/claude-structured-dispatch.ts b/src/main/claude/claude-structured-dispatch.ts index 96271e41d71..b7619a1e94e 100644 --- a/src/main/claude/claude-structured-dispatch.ts +++ b/src/main/claude/claude-structured-dispatch.ts @@ -1,5 +1,8 @@ import { randomUUID } from 'node:crypto' -import type { AgentJournalMessageItem } from '../../shared/agent-session-journal-types' +import type { + AgentJournalItemIdentity, + AgentJournalMessageItem +} from '../../shared/agent-session-journal-types' import type { AgentSessionDispatchOutcome } from '../native-chat/agent-session-wire/structured-agent-session-adapter' import { claudeHasReplayContent, @@ -14,9 +17,16 @@ import { const MAX_RETIRED_DISPATCH_WAITERS = 64 +/** A dispatch whose ack window expired, proven delivered by this replay. */ +export type ClaudeLateDispatchSettlement = (input: { + clientMessageId: string + providerIdentity: AgentJournalItemIdentity +}) => void + export function resolveClaudeReplayWaiter( session: ClaudeSession, - message: Record + message: Record, + onSettledLate?: ClaudeLateDispatchSettlement ): boolean { const envelope = readClaudeMessageEnvelope(message) const isUserReplay = @@ -52,7 +62,7 @@ export function resolveClaudeReplayWaiter( ) if (retired) { forgetRetiredWaiter(session, retired) - return recoverLateIdentity(session, retired, uuid, isUserReplay) + return recoverLateIdentity(session, retired, uuid, isUserReplay, onSettledLate) } return false } @@ -65,7 +75,7 @@ export function resolveClaudeReplayWaiter( const retired = session.retiredDispatchWaiters.find((candidate) => candidate.sentUuid === uuid) if (retired) { forgetRetiredWaiter(session, retired) - return recoverLateIdentity(session, retired, uuid, isUserReplay) + return recoverLateIdentity(session, retired, uuid, isUserReplay, onSettledLate) } if (isUserReplay) { @@ -89,7 +99,7 @@ export function resolveClaudeReplayWaiter( if (lateCompatible.length === 1) { const [candidate] = lateCompatible forgetRetiredWaiter(session, candidate!) - return recoverLateIdentity(session, candidate!, uuid, true) + return recoverLateIdentity(session, candidate!, uuid, true, onSettledLate) } } return false @@ -138,11 +148,19 @@ function recoverLateIdentity( session: ClaudeSession, waiter: ClaudeDispatchWaiter, uuid: string, - isUserReplay: boolean + isUserReplay: boolean, + onSettledLate?: ClaudeLateDispatchSettlement ): boolean { if (!isUserReplay && !waiter.acceptsResult) { return false } + // The provider acted on this dispatch, so the send it came from is delivered. + // Unfenced on purpose: the dispatch-sequence check below only decides which + // turn owns the identity, while delivery is settled for good either way. + onSettledLate?.({ + clientMessageId: waiter.clientMessageId, + providerIdentity: { provider: 'claude', sessionId: session.providerSessionId, uuid } + }) if (waiter.dispatchSequence === session.dispatchSequence) { session.activeTurnId = uuid session.activeTurnSequence = waiter.dispatchSequence @@ -155,12 +173,14 @@ function waitForReplay( timeoutMs: number, acceptsResult: boolean, sentUuid: string, - replayContentKey: string + replayContentKey: string, + clientMessageId: string ): { waiter: ClaudeDispatchWaiter; promise: Promise } { let waiter!: ClaudeDispatchWaiter const promise = new Promise((resolve) => { waiter = { acceptsResult, + clientMessageId, sentUuid, dispatchSequence: session.dispatchSequence, replayContentKey, @@ -220,7 +240,8 @@ export async function dispatchClaudeTurn( timeoutMs, acceptsResult, sentUuid, - claudeDispatchContentKey(content) + claudeDispatchContentKey(content), + input.clientMessageId ) const replayed = replay.promise try { diff --git a/src/main/claude/claude-structured-item-translation.ts b/src/main/claude/claude-structured-item-translation.ts index d86093ee0a5..c0ce20908d8 100644 --- a/src/main/claude/claude-structured-item-translation.ts +++ b/src/main/claude/claude-structured-item-translation.ts @@ -17,6 +17,7 @@ export type ClaudeMessageEnvelope = { /** Messages API id shared by every frame of one streamed assistant message. */ messageId: string | null parentToolUseId: string | null + isInjectedUserTurn?: boolean } export type ClaudeToolUse = { id: string; name: string; input: unknown } @@ -42,12 +43,16 @@ export function readClaudeMessageEnvelope( const sessionId = claudeText(frame.session_id) const uuid = claudeText(frame.uuid) const role = message?.role + const isInjectedUserTurn = + frame.type === 'user' && + (frame.isMeta === true || frame.isSynthetic === true || frame.isCompactSummary === true) return sessionId && uuid && (role === 'assistant' || role === 'user') ? { sessionId, uuid, role, content: messageContent(message?.content), + isInjectedUserTurn, messageId: claudeText(message?.id), parentToolUseId: claudeText(frame.parent_tool_use_id) } @@ -93,6 +98,9 @@ export function claudeMessageBody(envelope: ClaudeMessageEnvelope): AgentJournal } export function claudeHasReplayContent(envelope: ClaudeMessageEnvelope): boolean { + if (envelope.isInjectedUserTurn) { + return false + } return envelope.content.some((value) => { const part = claudeRecord(value) return part !== null && part.type !== 'tool_result' diff --git a/src/main/claude/claude-structured-journal-translation.test.ts b/src/main/claude/claude-structured-journal-translation.test.ts index f403313dae8..951872f0c69 100644 --- a/src/main/claude/claude-structured-journal-translation.test.ts +++ b/src/main/claude/claude-structured-journal-translation.test.ts @@ -342,10 +342,7 @@ describe('Claude structured journal translation', () => { state.items.flatMap((item) => item.body.kind === 'message' && item.body.role === 'user' ? [item.body.blocks] : [] ) - ).toEqual([ - [{ type: 'text', text: 'Reply with exactly PROBE_OK_1 and nothing else.' }], - [{ type: 'text', text: '[Request interrupted by user]' }] - ]) + ).toEqual([]) expect( state.items.some((item) => item.body.kind === 'status' && !item.body.turnLifecycle) ).toBe(false) @@ -521,10 +518,7 @@ describe('Claude structured journal translation', () => { const keyed = new Map( state.items.map((item) => [agentJournalItemKey(item.identity), item.body]) ) - expect(keyed.get('claude:claude-session:user-1')).toMatchObject({ - kind: 'message', - role: 'user' - }) + expect(keyed.has('claude:claude-session:user-1')).toBe(false) expect(keyed.get('orca:claude-tool%3Aclaude-session%3Atool-1')).toMatchObject({ kind: 'tool-call', name: 'Bash', @@ -683,7 +677,6 @@ describe('Claude structured journal translation', () => { 'message:system:local_command_output', 'message:system:command_started', 'message:result', - 'message:user:content:document', 'control_request:future_control' ]) ) diff --git a/src/main/claude/claude-structured-journal-translation.ts b/src/main/claude/claude-structured-journal-translation.ts index ffaad4da570..df8e8e67f53 100644 --- a/src/main/claude/claude-structured-journal-translation.ts +++ b/src/main/claude/claude-structured-journal-translation.ts @@ -30,6 +30,7 @@ import { } from './claude-structured-prompt-items' import type { ClaudePromptRegistry } from './claude-structured-prompt-replies' import { readableProviderFrameText } from '../native-chat/agent-session-wire/unhandled-provider-frame' +import { claudeProviderFrameActivity } from '../native-chat/agent-session-wire/provider-frame-activity' import { CLAUDE_UNRENDERABLE_CONTENT_TEXT, claudeProviderFrameKind, @@ -112,7 +113,20 @@ export function createClaudeJournalTranslator( } else { deps.sink.appendTombstone(identity) } - deps.sink.publish() + // Preserve first-work evidence when completion arrives before the journal drains. + deps.sink.publish({ + coalescingKey: running ? `turn-start:${sessionId}:${turnId}` : 'publish' + }) + } + + const publishActivity = (kind: string, payload: unknown): void => { + if (!currentTurn) { + return + } + const text = claudeProviderFrameActivity(kind, payload) + if (text !== undefined) { + deps.sink.setActivity?.(text ? { turnId: currentTurn.turnId, text } : null) + } } const handleStream = (message: Record): boolean => { @@ -130,7 +144,15 @@ export function createClaudeJournalTranslator( return false } let changed = false - const body = claudeMessageBody(envelope) + // User bubbles belong to the submitted message; SDK user frames carry echoes and tool results. + const outputEnvelope = + envelope.role === 'user' + ? { + ...envelope, + content: envelope.content.filter((part) => claudeRecord(part)?.type === 'tool_result') + } + : envelope + const body = claudeMessageBody(outputEnvelope) // The final frame of a streamed block lands on the block's identity, not its own uuid. const identity = (body && envelope.role === 'assistant' ? streamedBlocks.reconcile(envelope) : null) ?? @@ -140,7 +162,7 @@ export function createClaudeJournalTranslator( deps.sink.appendItem(identity, body) changed = true } - for (const tool of claudeToolUses(envelope)) { + for (const tool of claudeToolUses(outputEnvelope)) { tools.set(tool.id, tool) deps.sink.appendItem( claudeToolIdentity(envelope.sessionId, tool.id), @@ -162,7 +184,7 @@ export function createClaudeJournalTranslator( tools.delete(result.toolUseId) changed = true } - const thinking = claudeThinkingText(envelope) + const thinking = claudeThinkingText(outputEnvelope) if (thinking) { deps.sink.appendItem(claudeThinkingIdentity(envelope.sessionId, envelope.uuid), { kind: 'status', @@ -170,7 +192,7 @@ export function createClaudeJournalTranslator( }) changed = true } - const unhandledContent = envelope.content.filter((part) => !isModeledClaudeContent(part)) + const unhandledContent = outputEnvelope.content.filter((part) => !isModeledClaudeContent(part)) for (const part of unhandledContent) { const partType = claudeText(claudeRecord(part)?.type) ?? 'unknown' providerFallback.append( @@ -196,6 +218,7 @@ export function createClaudeJournalTranslator( } currentTurn = { sessionId: envelope.sessionId, turnId: envelope.uuid } publishLifecycle(envelope.sessionId, envelope.uuid, true) + deps.sink.setActivity?.(null) } if (changed) { deps.sink.publish() @@ -235,6 +258,7 @@ export function createClaudeJournalTranslator( publishLifecycle(currentTurn.sessionId, currentTurn.turnId, false) currentTurn = null } + deps.sink.setActivity?.(null) return } if (event.type === 'message' && handleStream(event.message)) { @@ -254,6 +278,7 @@ export function createClaudeJournalTranslator( publishLifecycle(currentTurn.sessionId, currentTurn.turnId, false) currentTurn = null } + deps.sink.setActivity?.(null) // The turn is over. A block still awaiting its final keeps the text the // flush above journaled, but its live state goes: an interrupted turn // would otherwise retain that text for the life of the session. @@ -266,11 +291,14 @@ export function createClaudeJournalTranslator( providerFallback.append(kind, event.message, failure?.text) } } else if (event.type === 'message') { + const kind = claudeProviderFrameKind(event.message) if (!handleMessage(event.message, event.startsTurn === true)) { - providerFallback.append(claudeProviderFrameKind(event.message), event.message) + providerFallback.append(kind, event.message) } + publishActivity(kind, event.message) } else if (event.type === 'provider-frame') { providerFallback.append(event.kind, event.payload) + publishActivity(event.kind, event.payload) } }, flush: streamedText.flush, diff --git a/src/main/claude/claude-structured-launch-resolution.ts b/src/main/claude/claude-structured-launch-resolution.ts index 4f28f14ad65..f9160cdb005 100644 --- a/src/main/claude/claude-structured-launch-resolution.ts +++ b/src/main/claude/claude-structured-launch-resolution.ts @@ -4,6 +4,7 @@ import type { AgentSessionJournalIdentity } from '../../shared/agent-session-jou import { agentSessionProviderHandleChainHead } from '../../shared/agent-session-provider-handle' import { LOCAL_EXECUTION_HOST_ID } from '../../shared/execution-host' import { withCliRuntimeOnPath } from '../../shared/node-cli-command-resolution' +import { structuredWorkerChildIdentityEnv } from '../runtime/structured-worker-child-identity-env' import { CLAUDE_AUTH_ENV_CONFLICT_MESSAGE, CLAUDE_AUTH_SWITCH_IN_PROGRESS_MESSAGE, @@ -236,7 +237,9 @@ export function createClaudeStructuredLaunchResolver( // user's own key is their sign-in and must reach the child. const env = withCliRuntimeOnPath( command, - { + // Only a dispatched structured worker gets the orchestration identity and the Orca CLI on + // PATH; an ordinary chat session's env passes through untouched. + structuredWorkerChildIdentityEnv(record.sessionId, { ...applyClaudeEnvPatch( cloneDefinedEnv(process.env), {}, @@ -246,7 +249,7 @@ export function createClaudeStructuredLaunchResolver( } ), ...(overlay ? cloneDefinedEnv(overlay) : {}) - }, + }), { platform: process.platform } ) return { diff --git a/src/main/claude/claude-structured-location-support.test.ts b/src/main/claude/claude-structured-location-support.test.ts index 1106667d544..fe3820964e3 100644 --- a/src/main/claude/claude-structured-location-support.test.ts +++ b/src/main/claude/claude-structured-location-support.test.ts @@ -56,8 +56,11 @@ describe('supportsClaudeStructuredLocation', () => { it('accepts Windows local locations once creation-time proof is available', () => { previousPlatform = setPlatform('win32') + // supportedProcessDataFlags is the addon's own report; the enum alone is + // not proof, because pnpm patches the source over the tarball's prebuilt. __setWindowsProcessTreeLoaderForTests(() => ({ ProcessDataFlag: { None: 0, Memory: 1, CommandLine: 2, CreationTime: 4 }, + supportedProcessDataFlags: 7, getAllProcesses: () => undefined })) expect( diff --git a/src/main/claude/claude-structured-options.test.ts b/src/main/claude/claude-structured-options.test.ts index 0738095f45d..2375df12d93 100644 --- a/src/main/claude/claude-structured-options.test.ts +++ b/src/main/claude/claude-structured-options.test.ts @@ -2,6 +2,7 @@ import { describe, expect, it, vi } from 'vitest' import { setClaudeStructuredOption } from './claude-structured-options' import type { ClaudeSession } from './claude-structured-session-state' import { ClaudeBackgroundTaskTracker } from './claude-background-task-tracker' +import { ClaudeSlashCommandCatalog } from './claude-slash-command-catalog' function sessionFor(setModel: ClaudeSession['connection']['setModel']): ClaudeSession { return { @@ -16,6 +17,7 @@ function sessionFor(setModel: ClaudeSession['connection']['setModel']): ClaudeSe retiredDispatchWaiters: [], replayContentFallbackBlocked: false, backgroundTasks: new ClaudeBackgroundTaskTracker(), + commands: new ClaudeSlashCommandCatalog(), dispatchSequence: 0, optionMutationSequence: 0, options: new Map(), diff --git a/src/main/claude/claude-structured-real-cli.test.ts b/src/main/claude/claude-structured-real-cli.test.ts index 0f22c175cc6..cb3bb2b72ea 100644 --- a/src/main/claude/claude-structured-real-cli.test.ts +++ b/src/main/claude/claude-structured-real-cli.test.ts @@ -1,6 +1,6 @@ import { spawnSync } from 'node:child_process' import { randomUUID } from 'node:crypto' -import { mkdtemp, rm } from 'node:fs/promises' +import { mkdir, mkdtemp, rm, writeFile } from 'node:fs/promises' import { homedir, tmpdir } from 'node:os' import { basename, join, relative } from 'node:path' import { describe, expect, it } from 'vitest' @@ -48,13 +48,14 @@ const realClaudeAuthenticated = realClaudeAuthStatus?.loggedIn === true function realAdapter( providerSessionId: string, claudeConfigDir: string, - events: ClaudeStructuredSessionEvent[] = [] + events: ClaudeStructuredSessionEvent[] = [], + cwd = process.cwd() ): ClaudeStructuredSessionAdapter { return new ClaudeStructuredSessionAdapter({ resolveLaunch: async () => ({ pathToClaudeCodeExecutable: command, options: { ...CLAUDE_STRUCTURED_BASE_OPTIONS, sessionId: providerSessionId }, - cwd: process.cwd(), + cwd, claudeConfigDir, providerSessionId, resumeLeafUuid: null, @@ -100,7 +101,13 @@ describe.skipIf(!realClaudeAvailable)('Claude structured real CLI handshake', () const providerSessionId = randomUUID() const claudeConfigDir = process.env.CLAUDE_CONFIG_DIR?.trim() || join(homedir(), '.claude') const events: ClaudeStructuredSessionEvent[] = [] - const adapter = realAdapter(providerSessionId, claudeConfigDir, events) + const cwd = await mkdtemp(join(tmpdir(), 'orca-command-init-')) + await mkdir(join(cwd, '.claude', 'commands'), { recursive: true }) + await writeFile( + join(cwd, '.claude', 'commands', 'orca-init-catalog-proof.md'), + '---\ndescription: Initialization catalog proof\n---\nReply with OK.\n' + ) + const adapter = realAdapter(providerSessionId, claudeConfigDir, events, cwd) try { const acquisition = await adapter.acquire({ @@ -120,8 +127,17 @@ describe.skipIf(!realClaudeAvailable)('Claude structured real CLI handshake', () leafUuid: null }) expect(observedSubtypes).toContain('hook_started') + expect(adapter.readCommands('real-cli-handshake')).toContainEqual({ + name: 'orca-init-catalog-proof', + kind: 'command', + kindUnspecified: true + }) + expect( + adapter.readCommands('real-cli-handshake')?.some(({ name }) => name === 'help') + ).toBe(false) } finally { await adapter.closeAll() + await rm(cwd, { recursive: true, force: true }) } }, 10_000 diff --git a/src/main/claude/claude-structured-session-acquisition.ts b/src/main/claude/claude-structured-session-acquisition.ts index e8d09bd78d9..56b40b27177 100644 --- a/src/main/claude/claude-structured-session-acquisition.ts +++ b/src/main/claude/claude-structured-session-acquisition.ts @@ -109,7 +109,11 @@ export async function acquireClaudeSession({ if (liveSession) { liveSession.leafUuid = observedLeafUuid } - const startsTurn = liveSession ? resolveClaudeReplayWaiter(liveSession, message) : false + const startsTurn = liveSession + ? resolveClaudeReplayWaiter(liveSession, message, (settlement) => + deps.onDispatchSettledLate?.({ sessionId, ...settlement }) + ) + : false callbacks.deliver(attempt, sessionId, () => callbacks.emit(liveSession, input.events, { type: 'message', @@ -248,6 +252,7 @@ export async function acquireClaudeSession({ const publication = createClaudeSessionPublication({ connection, init, + initialization, claudeConfigDir: launch.claudeConfigDir, leafUuid: observedLeafUuid, fence: input.fence, diff --git a/src/main/claude/claude-structured-session-adapter.ts b/src/main/claude/claude-structured-session-adapter.ts index c00a588e891..a2f7643dbd5 100644 --- a/src/main/claude/claude-structured-session-adapter.ts +++ b/src/main/claude/claude-structured-session-adapter.ts @@ -195,6 +195,9 @@ export class ClaudeStructuredSessionAdapter implements StructuredAgentSessionAda : event.type === 'message' ? (session?.backgroundTasks.observe(event.message, event.startsTurn === true) ?? false) : false + if (event.type === 'message' && session?.commands.observe(event.message)) { + session.events?.publish() + } session?.translator?.handle(event) this.deps.onEvent?.(event) if (backgroundTasksChanged) { @@ -261,6 +264,8 @@ export class ClaudeStructuredSessionAdapter implements StructuredAgentSessionAda const session = this.sessions.get(sessionId) return session ? backgroundTaskState(session) : undefined } + readCommands: NonNullable = (sessionId) => + this.sessions.get(sessionId)?.commands.commands answerPrompt: StructuredAgentSessionAdapter['answerPrompt'] = (input) => answerClaudePrompt(this.session(input.sessionId), input) setOption: StructuredAgentSessionAdapter['setOption'] = (input) => diff --git a/src/main/claude/claude-structured-session-commands.test.ts b/src/main/claude/claude-structured-session-commands.test.ts new file mode 100644 index 00000000000..1b2fb6bde31 --- /dev/null +++ b/src/main/claude/claude-structured-session-commands.test.ts @@ -0,0 +1,103 @@ +import { describe, expect, it, vi } from 'vitest' +import { + adapterFor, + fakeClaude, + identityFor, + tick, + PROVIDER_SESSION_ID +} from './claude-structured-session-test-support' + +describe('session command updates', () => { + it('publishes changed catalogs exactly once while idle', async () => { + const claude = fakeClaude() + const changed = vi.fn() + const adapter = adapterFor(claude) + await adapter.acquire({ + identity: identityFor(), + fence: 7, + spawnToken: 'spawn-9', + events: { + appendItem: vi.fn(), + appendTombstone: vi.fn(), + publish: changed + } + }) + changed.mockClear() + expect(adapter.readCommands('session-1')).toBeUndefined() + const frame = { + type: 'system', + subtype: 'commands_changed', + session_id: PROVIDER_SESSION_ID, + slash_commands: ['plugin:check', 'doctor'], + skills: ['plugin:check'], + terminal_slash_commands: ['doctor'] + } + claude.connections[0].handlers.onMessage?.(frame) + await tick() + expect(adapter.readCommands('session-1')).toEqual([{ name: 'plugin:check', kind: 'skill' }]) + expect(changed).toHaveBeenCalledTimes(1) + claude.connections[0].handlers.onMessage?.(frame) + await tick() + expect(changed).toHaveBeenCalledTimes(1) + claude.connections[0].handlers.onMessage?.({ ...frame, slash_commands: [] }) + await tick() + expect(adapter.readCommands('session-1')).toEqual([]) + expect(changed).toHaveBeenCalledTimes(2) + await adapter.closeSession('session-1') + }) +}) + +it.each([ + { commands: [] }, + { commands: [{ name: 'project:check', description: 'Project command', argumentHint: '' }] } +])('seeds the pre-prompt catalog from control initialization: %j', async ({ commands }) => { + const claude = fakeClaude({ initProof: 'session-start', initCommands: commands }) + const adapter = adapterFor(claude) + try { + await adapter.acquire({ identity: identityFor(), fence: 7, spawnToken: 'spawn-9' }) + expect(adapter.readCommands('session-1')).toEqual( + commands.map(({ name }) => ({ name, kind: 'command', kindUnspecified: true })) + ) + expect(claude.connections[0].sent).toEqual([]) + expect(claude.connections[0].calls.map(({ subtype }) => subtype)).toEqual([ + 'initialize', + 'get_settings' + ]) + claude.connections[0].handlers.onMessage?.({ + type: 'system', + subtype: 'init', + session_id: PROVIDER_SESSION_ID, + slash_commands: ['project:check'], + skills: ['project:check'] + }) + expect(adapter.readCommands('session-1')).toEqual([{ name: 'project:check', kind: 'skill' }]) + } finally { + await adapter.closeSession('session-1') + } +}) + +it('keeps a buffered stream catalog newer than the initialization response', async () => { + const claude = fakeClaude({ initProof: 'session-start', initCommands: [{ name: 'old' }] }) + const open = claude.openConnection + claude.openConnection = async (...args) => { + const connection = await open(...args) + const getSettings = connection.getSettings + connection.getSettings = async (...settingsArgs) => { + args[1]?.onMessage?.({ + type: 'system', + subtype: 'commands_changed', + session_id: PROVIDER_SESSION_ID, + commands: [{ name: 'fresh' }] + }) + return getSettings(...settingsArgs) + } + return connection + } + const adapter = adapterFor(claude) + try { + await adapter.acquire({ identity: identityFor(), fence: 7, spawnToken: 'spawn-9' }) + expect(adapter.readCommands('session-1')?.map(({ name }) => name)).toEqual(['fresh']) + } finally { + await adapter.closeSession('session-1') + } +}) diff --git a/src/main/claude/claude-structured-session-publication.ts b/src/main/claude/claude-structured-session-publication.ts index 7f8cc4b5692..e55608ae679 100644 --- a/src/main/claude/claude-structured-session-publication.ts +++ b/src/main/claude/claude-structured-session-publication.ts @@ -5,10 +5,12 @@ import type { ClaudePromptRegistry } from './claude-structured-prompt-replies' import type { ClaudeJournalTranslator } from './claude-structured-journal-translation' import type { ClaudeSession } from './claude-structured-session-state' import { ClaudeBackgroundTaskTracker } from './claude-background-task-tracker' +import { ClaudeSlashCommandCatalog } from './claude-slash-command-catalog' export function createClaudeSessionPublication(input: { connection: ClaudeSession['connection'] init: ClaudeInitObservation + initialization?: unknown claudeConfigDir: string leafUuid: string | null fence: number @@ -52,6 +54,7 @@ export function createClaudeSessionPublication(input: { retiredDispatchWaiters: [], replayContentFallbackBlocked: false, backgroundTasks: new ClaudeBackgroundTaskTracker(), + commands: new ClaudeSlashCommandCatalog(input.init.message, input.initialization), dispatchSequence: 0, optionMutationSequence: 0, options: new Map(input.options), diff --git a/src/main/claude/claude-structured-session-state.ts b/src/main/claude/claude-structured-session-state.ts index 1c0b1862913..441d8f68af0 100644 --- a/src/main/claude/claude-structured-session-state.ts +++ b/src/main/claude/claude-structured-session-state.ts @@ -1,4 +1,7 @@ -import type { AgentSessionJournalIdentity } from '../../shared/agent-session-journal-types' +import type { + AgentJournalItemIdentity, + AgentSessionJournalIdentity +} from '../../shared/agent-session-journal-types' import type { StructuredAgentSessionEventSink } from '../native-chat/agent-session-wire/structured-agent-session-event-sink' import type { ClaudeStreamJsonConnection, @@ -11,6 +14,7 @@ import { cancelProcessAcquisition } from '../../shared/child-process/cancel-proc import { randomUUID } from 'node:crypto' import type { AgentSessionBackgroundTaskState } from '../../shared/agent-session-wire' import type { ClaudeBackgroundTaskTracker } from './claude-background-task-tracker' +import type { ClaudeSlashCommandCatalog } from './claude-slash-command-catalog' export type ClaudeAuthDiagnostic = { apiKeySourceConfigured: boolean @@ -56,6 +60,12 @@ export type ClaudeStructuredSessionAdapterDeps = { identity: AgentSessionJournalIdentity }) => Promise onEvent?: (event: ClaudeStructuredSessionEvent) => void + /** A dispatch whose ack timed out, proven delivered by a later provider replay. */ + onDispatchSettledLate?: (input: { + sessionId: string + clientMessageId: string + providerIdentity: AgentJournalItemIdentity + }) => void onBackgroundTasksChanged?: ( sessionId: string, state: AgentSessionBackgroundTaskState | null @@ -87,6 +97,9 @@ export type ClaudeDispatchWaiter = { resolve: (uuid: string | null) => void timer: ReturnType acceptsResult: boolean + /** Carried so a replay that lands after the ack window can settle the journal + * submission this dispatch came from, not just the in-memory turn identity. */ + clientMessageId: string /** Client uuid echoed by Claude so a replay is tied to its own dispatch. */ sentUuid: string /** Sequence used to fence a late identity from a newer dispatch. */ @@ -126,6 +139,9 @@ export type ClaudeSession = { /** Provider uuid of the most recently admitted turn, if one is active. */ activeTurnId?: string backgroundTasks: ClaudeBackgroundTaskTracker + /** The `/` surface the CLI reports for itself; seeded from init, kept current + * by later init and `commands_changed` frames. */ + commands: ClaudeSlashCommandCatalog /** Monotonic fence advanced when a dispatch starts, including unresolved dispatches. */ dispatchSequence: number /** Dispatch sequence that admitted activeTurnId. */ diff --git a/src/main/claude/claude-structured-session-test-support.ts b/src/main/claude/claude-structured-session-test-support.ts index 903cafae416..15a9fcbbb8f 100644 --- a/src/main/claude/claude-structured-session-test-support.ts +++ b/src/main/claude/claude-structured-session-test-support.ts @@ -52,6 +52,7 @@ export function fakeClaude( initModel?: string initProof?: 'init' | 'session-start' | 'none' initAccount?: unknown + initCommands?: unknown exitBeforeInit?: string settings?: unknown replayUuid?: string | null @@ -111,6 +112,7 @@ export function fakeClaude( } return { models: [{ value: 'claude-sonnet', displayName: 'Sonnet' }], + ...(options.initCommands === undefined ? {} : { commands: options.initCommands }), ...(options.initAccount === undefined ? {} : { account: options.initAccount }) } }, diff --git a/src/main/cli/orca-cli-child-path.test.ts b/src/main/cli/orca-cli-child-path.test.ts new file mode 100644 index 00000000000..727732dd64c --- /dev/null +++ b/src/main/cli/orca-cli-child-path.test.ts @@ -0,0 +1,125 @@ +import { join } from 'node:path' +import { beforeEach, describe, expect, it, vi } from 'vitest' + +const shim = vi.hoisted(() => ({ ensureLinuxTerminalOrcaCliShimDir: vi.fn() })) +vi.mock('./linux-terminal-orca-cli-shim', () => shim) + +import { prependOrcaCliDirToChildPath } from './orca-cli-child-path' + +const USER_DATA = '/data/orca' +const RESOURCES = '/app/Resources' +const SHIM_DIR = join(USER_DATA, 'linux-orca-cli-shim') + +beforeEach(() => { + shim.ensureLinuxTerminalOrcaCliShimDir.mockReset() + shim.ensureLinuxTerminalOrcaCliShimDir.mockReturnValue(SHIM_DIR) +}) + +describe('prependOrcaCliDirToChildPath', () => { + it('leads packaged Linux PATH with the bare-orca shim dir', () => { + // Why this matters at all: the Linux CLI installs as `orca-ide` so it never claims GNOME + // Orca's /usr/bin/orca screen reader, so bare `orca` only works through this shim. + const env: Record = { PATH: '/usr/local/bin:/usr/bin' } + prependOrcaCliDirToChildPath(env, { + isPackaged: true, + userDataPath: USER_DATA, + resourcesPath: RESOURCES, + platform: 'linux' + }) + expect(env.PATH).toBe(`${SHIM_DIR}:/usr/local/bin:/usr/bin`) + expect(shim.ensureLinuxTerminalOrcaCliShimDir).toHaveBeenCalledWith({ + userDataPath: USER_DATA + }) + }) + + it('promotes an already-present shim dir instead of duplicating it', () => { + const env: Record = { PATH: `/usr/bin:${SHIM_DIR}::/bin` } + prependOrcaCliDirToChildPath(env, { + isPackaged: true, + userDataPath: USER_DATA, + platform: 'linux' + }) + expect(env.PATH).toBe(`${SHIM_DIR}:/usr/bin:/bin`) + }) + + it('leaves packaged Linux PATH untouched when no shim could be written', () => { + shim.ensureLinuxTerminalOrcaCliShimDir.mockReturnValue(null) + const env: Record = { PATH: '/usr/bin' } + prependOrcaCliDirToChildPath(env, { + isPackaged: true, + userDataPath: USER_DATA, + platform: 'linux' + }) + expect(env.PATH).toBe('/usr/bin') + }) + + it('leads packaged macOS PATH with the bundled CLI dir', () => { + const env: Record = { PATH: '/usr/bin' } + prependOrcaCliDirToChildPath(env, { + isPackaged: true, + userDataPath: USER_DATA, + resourcesPath: RESOURCES, + platform: 'darwin' + }) + expect(env.PATH).toBe(`${join(RESOURCES, 'bin')}:/usr/bin`) + expect(shim.ensureLinuxTerminalOrcaCliShimDir).not.toHaveBeenCalled() + }) + + it('leads packaged Windows PATH with the bundled CLI dir under the env block spelling', () => { + const env: Record = { Path: 'C:\\Windows\\System32' } + prependOrcaCliDirToChildPath(env, { + isPackaged: true, + userDataPath: USER_DATA, + resourcesPath: RESOURCES, + platform: 'win32' + }) + expect(env.Path).toBe(`${join(RESOURCES, 'bin')};C:\\Windows\\System32`) + expect(env.PATH).toBeUndefined() + }) + + it('leaves a packaged darwin/win32 PATH alone with no resources root', () => { + const env: Record = { PATH: '/usr/bin' } + prependOrcaCliDirToChildPath(env, { + isPackaged: true, + userDataPath: USER_DATA, + resourcesPath: null, + platform: 'darwin' + }) + expect(env.PATH).toBe('/usr/bin') + }) + + it.each<[NodeJS.Platform, string]>([ + ['linux', ':'], + ['darwin', ':'], + ['win32', ';'] + ])('leads an unpackaged %s PATH with the dev launcher dir', (platform, pathDelimiter) => { + const env: Record = { PATH: '/usr/bin' } + prependOrcaCliDirToChildPath(env, { + isPackaged: false, + userDataPath: USER_DATA, + resourcesPath: RESOURCES, + platform + }) + expect(env.PATH).toBe(`${join(USER_DATA, 'cli', 'bin')}${pathDelimiter}/usr/bin`) + expect(shim.ensureLinuxTerminalOrcaCliShimDir).not.toHaveBeenCalled() + }) + + it('writes no trailing delimiter when nothing was inherited', () => { + const env: Record = { PATH: '' } + const inheritedPath = process.env.PATH + delete process.env.PATH + try { + prependOrcaCliDirToChildPath(env, { + isPackaged: false, + userDataPath: USER_DATA, + platform: 'linux' + }) + } finally { + if (inheritedPath !== undefined) { + process.env.PATH = inheritedPath + } + } + // Why: an empty trailing segment resolves as `.` in some shells. + expect(env.PATH).toBe(join(USER_DATA, 'cli', 'bin')) + }) +}) diff --git a/src/main/cli/orca-cli-child-path.ts b/src/main/cli/orca-cli-child-path.ts new file mode 100644 index 00000000000..f053e54c3d5 --- /dev/null +++ b/src/main/cli/orca-cli-child-path.ts @@ -0,0 +1,63 @@ +/** + * The PATH entry through which an Orca-launched child reaches THIS app's own CLI. + * + * Extracted from `buildPtyHostEnv` so the structured-session lane can apply the identical + * treatment. A structured worker has no PTY, but its provider child runs `orca orchestration ...` + * exactly like a PTY worker's agent does, and it was inheriting the ambient PATH instead. On + * packaged Linux that made bare `orca` resolve to GNOME's /usr/bin/orca screen reader, because + * Orca's Linux CLI installs as `orca-ide` to avoid claiming that name (stablyai/orca#7904); on + * packaged macOS/Windows it reached this app's bundled CLI only if the user had separately + * registered the CLI globally. + * + * `platform` is a test seam only: production leaves it unset and reads `process.platform`, so + * every branch behaves exactly as it did inside `buildPtyHostEnv`. + */ + +import { delimiter, join } from 'node:path' +import { readInheritedPath } from '../ipc/pty/host-env/path' +import { resolvePathEnvKey } from '../pty/windows-environment-path' +import { ensureLinuxTerminalOrcaCliShimDir } from './linux-terminal-orca-cli-shim' + +export type OrcaCliChildPathOptions = { + isPackaged: boolean + userDataPath: string + resourcesPath?: string | null + /** Test seam — production reads the real platform, which is what every branch below assumes. */ + platform?: NodeJS.Platform +} + +/** Mutates `env` in place, prepending the directory that makes bare `orca` this app's CLI. */ +export function prependOrcaCliDirToChildPath( + env: Record, + opts: OrcaCliChildPathOptions +): void { + const platform = opts.platform ?? process.platform + // Why: matches node:path's `delimiter` for the running platform, but stays correct when a test + // drives a foreign platform through the seam. + const pathDelimiter = platform === 'win32' ? ';' : delimiter + // Why: dev mode needs the launcher PATH override so `orca` resolves to the dev build instead of the production binary at /usr/local/bin/orca. + if (!opts.isPackaged) { + const devCliBin = join(opts.userDataPath, 'cli', 'bin') + const inheritedPath = readInheritedPath(env, platform) + // Why: an empty PATH segment resolves as `.` in some shells (commands run from cwd); avoid a trailing delimiter. + env[resolvePathEnvKey(env, platform)] = inheritedPath + ? `${devCliBin}${pathDelimiter}${inheritedPath}` + : devCliBin + } else if (platform === 'linux') { + // Why: bare-`orca` shim scoped to Orca PTYs — Linux CLI installs as `orca-ide` to avoid shadowing GNOME's /usr/bin/orca screen reader (stablyai/orca#7904). + const shimDir = ensureLinuxTerminalOrcaCliShimDir({ userDataPath: opts.userDataPath }) + if (shimDir) { + const inheritedEntries = readInheritedPath(env, platform) + .split(pathDelimiter) + .filter((entry) => entry.length > 0 && entry !== shimDir) + env.PATH = [shimDir, ...inheritedEntries].join(pathDelimiter) + } + } else if (opts.resourcesPath && (platform === 'darwin' || platform === 'win32')) { + // Why: global CLI registration is optional, but agents in Orca-managed PTYs must always reach this app's bundled CLI. + const bundledCliBin = join(opts.resourcesPath, 'bin') + const inheritedPath = readInheritedPath(env, platform) + env[resolvePathEnvKey(env, platform)] = inheritedPath + ? `${bundledCliBin}${pathDelimiter}${inheritedPath}` + : bundledCliBin + } +} diff --git a/src/main/codex-accounts/runtime-home-settings-test-fixtures.ts b/src/main/codex-accounts/runtime-home-settings-test-fixtures.ts index 2872ecf15c3..c7be08b3509 100644 --- a/src/main/codex-accounts/runtime-home-settings-test-fixtures.ts +++ b/src/main/codex-accounts/runtime-home-settings-test-fixtures.ts @@ -113,6 +113,7 @@ export function createSettings(overrides: TestSettingsOverrides = {}): GlobalSet opencodeWorkspaceId: '', minimaxGroupId: '', minimaxUsageModels: 'general', + minimaxEndpoint: 'overseas', geminiCliOAuthEnabled: false, agentCmdOverrides: {}, keepComputerAwakeWhileAgentsRun: false, diff --git a/src/main/codex-accounts/service-reset-credit-test-fixtures.ts b/src/main/codex-accounts/service-reset-credit-test-fixtures.ts index 4578831985a..d47c967e34c 100644 --- a/src/main/codex-accounts/service-reset-credit-test-fixtures.ts +++ b/src/main/codex-accounts/service-reset-credit-test-fixtures.ts @@ -36,6 +36,7 @@ export function createResetRateLimitState( minimax: null, grok: null, minimaxCookieConfigured: false, + minimaxApiKeyConfigured: false, grokAuthConfigured: false, claudeTarget: { runtime: 'host', wslDistro: null }, codexTarget: target, diff --git a/src/main/codex-accounts/service-test-harness.ts b/src/main/codex-accounts/service-test-harness.ts index ed454c7a149..6c0a33135ab 100644 --- a/src/main/codex-accounts/service-test-harness.ts +++ b/src/main/codex-accounts/service-test-harness.ts @@ -134,6 +134,7 @@ export function createSettings(overrides: Partial = {}): GlobalS opencodeWorkspaceId: '', minimaxGroupId: '', minimaxUsageModels: 'general', + minimaxEndpoint: 'overseas', geminiCliOAuthEnabled: false, agentCmdOverrides: {}, keepComputerAwakeWhileAgentsRun: false, diff --git a/src/main/codex/codex-structured-child-environment.test.ts b/src/main/codex/codex-structured-child-environment.test.ts index 98e988ec7d5..201efc0b0c5 100644 --- a/src/main/codex/codex-structured-child-environment.test.ts +++ b/src/main/codex/codex-structured-child-environment.test.ts @@ -1,6 +1,13 @@ import { describe, expect, it } from 'vitest' import { CODEX_SPAWN_TOKEN_ENV } from './codex-structured-owner-identity' import { buildCodexStructuredChildEnvironment } from './codex-structured-child-environment' +import { ORCA_STRUCTURED_SESSION_ENV } from '../../shared/structured-session-marker' +import { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} from '../runtime/structured-worker-identity' describe('buildCodexStructuredChildEnvironment', () => { it('keeps shell exports while pinned launch values win', () => { @@ -14,12 +21,51 @@ describe('buildCodexStructuredChildEnvironment', () => { resumeThreadId: null, env: { EXAMPLE_GATEWAY_TOKEN: 'shell-exported', CODEX_HOME: '/shell/home' } }, - 'spawn-token' + 'spawn-token', + 'session-not-a-worker' ) ).toEqual({ EXAMPLE_GATEWAY_TOKEN: 'shell-exported', CODEX_HOME: '/pinned/home', - [CODEX_SPAWN_TOKEN_ENV]: 'spawn-token' + [CODEX_SPAWN_TOKEN_ENV]: 'spawn-token', + [ORCA_STRUCTURED_SESSION_ENV]: '1' }) }) + + it('adds the orchestration handle only for a registered structured worker', () => { + const launch = { + command: 'codex', + args: ['app-server'], + cwd: '/worktree', + codexHome: null, + resumeThreadId: null, + env: {} + } + const sessionId = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + expect(buildCodexStructuredChildEnvironment(launch, 'spawn-token', sessionId)).toEqual({ + [CODEX_SPAWN_TOKEN_ENV]: 'spawn-token', + // No identity yet, so the child carries only the refuse-rather-than-guess marker. + [ORCA_STRUCTURED_SESSION_ENV]: '1' + }) + + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId, + agent: 'codex', + paneKey: mintStructuredWorkerPaneKey(sessionId), + processIncarnation: structuredWorkerProcessIncarnation(sessionId), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + try { + const env = buildCodexStructuredChildEnvironment(launch, 'spawn-token', sessionId) + expect(env.ORCA_TERMINAL_HANDLE).toBe(handle) + expect(env.ORCA_CLI_COMMAND).toBe('orca') + // A pane key here would leak into hook-emitted agent statuses, which assume a PTY leaf. + expect(env.ORCA_PANE_KEY).toBeUndefined() + } finally { + structuredWorkerIdentities.forget(handle) + } + }) }) diff --git a/src/main/codex/codex-structured-child-environment.ts b/src/main/codex/codex-structured-child-environment.ts index 88326eb3fc0..72bf17a1bce 100644 --- a/src/main/codex/codex-structured-child-environment.ts +++ b/src/main/codex/codex-structured-child-environment.ts @@ -1,13 +1,19 @@ import type { CodexStructuredLaunch } from './codex-structured-session-state' import { CODEX_SPAWN_TOKEN_ENV } from './codex-structured-owner-identity' +import { structuredWorkerChildIdentityEnv } from '../runtime/structured-worker-child-identity-env' export function buildCodexStructuredChildEnvironment( launch: CodexStructuredLaunch, - spawnToken: string + spawnToken: string, + sessionId: string ): Record { return { - ...launch.env, - ...(launch.codexHome ? { CODEX_HOME: launch.codexHome } : {}), + // Only a dispatched structured worker gets the orchestration identity and the Orca CLI on + // PATH; an ordinary chat session's env passes through untouched. + ...structuredWorkerChildIdentityEnv(sessionId, { + ...launch.env, + ...(launch.codexHome ? { CODEX_HOME: launch.codexHome } : {}) + }), [CODEX_SPAWN_TOKEN_ENV]: spawnToken } } diff --git a/src/main/codex/codex-structured-journal-items.ts b/src/main/codex/codex-structured-journal-items.ts index 0e4e3f5a900..fd8fbf558a8 100644 --- a/src/main/codex/codex-structured-journal-items.ts +++ b/src/main/codex/codex-structured-journal-items.ts @@ -60,7 +60,10 @@ export class CodexJournalItems { return this.details.get(codexStructuredItemKey(threadId, itemId)) ?? null } - handle(event: { threadId: string; method: string; params: unknown }): CodexItemTranslation { + handle( + event: { threadId: string; method: string; params: unknown }, + source: 'live' | 'history' = 'live' + ): CodexItemTranslation { const params = typeof event.params === 'object' && event.params !== null ? (event.params as Record) @@ -71,6 +74,10 @@ export class CodexJournalItems { } const turnId = readCodexTurnId(event.params) ?? this.activeTurn(event.threadId) const identity = this.identityFor(event.threadId, turnId, item) + // Count echoes for stable resume ordinals, but user bubbles come from submissions. + if (source === 'live' && item.type === 'userMessage') { + return { handled: true, admission: CODEX_JOURNAL_ADMITTED } + } const translated = codexJournalItem(item) const command = readCodexJournalString(item, 'command') if (command) { diff --git a/src/main/codex/codex-structured-journal-translation-frames.ts b/src/main/codex/codex-structured-journal-translation-frames.ts new file mode 100644 index 00000000000..22dc516b210 --- /dev/null +++ b/src/main/codex/codex-structured-journal-translation-frames.ts @@ -0,0 +1,43 @@ +/** + * The translator's provider-frame arms. + * + * Each returns null for a frame it does not own, which is the translator's + * signal to keep looking. Split out so the translator reads as routing rather + * than as the shape checks each arm performs. + */ + +import type { CodexJournalTranslationAdmission } from './codex-structured-journal-contracts' +import { settleCodexOversizedNotification } from './codex-structured-journal-settlement' +import { + readCodexJournalRecord, + readCodexJournalString +} from './codex-structured-journal-translation-values' + +type OversizedInput = Parameters[0] + +/** A notification the transport refused to carry whole: settle whatever it + * opened rather than leaving the item mid-flight. */ +export function settleCodexOversizedNotificationFrame(input: { + sessionId: string + threadId: string + kind: string + payload: unknown + sink: OversizedInput['sink'] + streams: OversizedInput['streams'] + activeItems: OversizedInput['activeItems'] +}): CodexJournalTranslationAdmission | null { + if (input.kind !== 'frame:oversized-notification') { + return null + } + const method = readCodexJournalString(readCodexJournalRecord(input.payload), 'method') + return method + ? settleCodexOversizedNotification({ + sessionId: input.sessionId, + threadId: input.threadId, + method, + sink: input.sink, + streams: input.streams, + activeItems: input.activeItems + }) + : null +} diff --git a/src/main/codex/codex-structured-journal-translation-settlement.test.ts b/src/main/codex/codex-structured-journal-translation-settlement.test.ts index 5773cf8d6fa..1f2bc480a22 100644 --- a/src/main/codex/codex-structured-journal-translation-settlement.test.ts +++ b/src/main/codex/codex-structured-journal-translation-settlement.test.ts @@ -212,7 +212,7 @@ describe('codex journal translation', () => { expect(translator.handle(notification('turn/completed', { turn: { id: TURN_ID } }))).toEqual({ accepted: true }) - expect(deferred.state()).toMatchObject({ queuedOperations: 4, backpressured: true }) + expect(deferred.state()).toMatchObject({ queuedOperations: 5, backpressured: true }) deferred.bind(deferredTarget(bodies, publishes)) await expect(deferred.lifecycleBarrier()).resolves.toEqual({ ok: true }) @@ -225,7 +225,7 @@ describe('codex journal translation', () => { expect.objectContaining({ kind: 'tool-call', state: 'running' }), expect.objectContaining({ kind: 'tool-call', state: 'failed' }) ]) - expect(publishes).toHaveLength(1) + expect(publishes).toHaveLength(2) }) it('admits terminal session settlement publication across the hard watermark', async () => { @@ -257,7 +257,7 @@ describe('codex journal translation', () => { acquisitionGeneration: 'generation-1' }) ).toEqual({ accepted: true }) - expect(deferred.state()).toMatchObject({ queuedOperations: 4, backpressured: true }) + expect(deferred.state()).toMatchObject({ queuedOperations: 5, backpressured: true }) deferred.bind(deferredTarget(bodies, publishes)) await expect(deferred.lifecycleBarrier()).resolves.toEqual({ ok: true }) @@ -277,7 +277,7 @@ describe('codex journal translation', () => { }), { kind: 'status', text: 'Provider exited: lost child' } ]) - expect(publishes).toHaveLength(1) + expect(publishes).toHaveLength(2) }) it('retries a rejected terminal admission without losing tool, prompt, turn, or session truth', () => { @@ -652,7 +652,7 @@ describe('codex journal translation', () => { translator.handle(TURN_STARTED) translator.handle( - notification('item/completed', { item: { type: 'userMessage', id: 'item-0', text: 'one' } }) + notification('item/completed', { item: { type: 'agentMessage', id: 'item-0', text: 'one' } }) ) translator.handle(notification('turn/completed', { turn: { id: TURN_ID } })) translator.handle( @@ -662,7 +662,7 @@ describe('codex journal translation', () => { ) translator.handle(notification('turn/started', { turn: { id: 'turn-2' } })) translator.handle( - notification('item/completed', { item: { type: 'userMessage', id: 'item-2', text: 'two' } }) + notification('item/completed', { item: { type: 'agentMessage', id: 'item-2', text: 'two' } }) ) expect(tap.rows.map((row) => row.key)).toEqual([ @@ -679,7 +679,7 @@ describe('codex journal translation', () => { translator.handle( notification('item/completed', { turnId: 'turn-9', - item: { type: 'userMessage', id: 'item-0', text: 'late' } + item: { type: 'agentMessage', id: 'item-0', text: 'late' } }) ) diff --git a/src/main/codex/codex-structured-journal-translation-streams.test.ts b/src/main/codex/codex-structured-journal-translation-streams.test.ts index a9bc79b71c4..86eb94fef07 100644 --- a/src/main/codex/codex-structured-journal-translation-streams.test.ts +++ b/src/main/codex/codex-structured-journal-translation-streams.test.ts @@ -157,7 +157,7 @@ describe('codex journal translation', () => { translator.handle(TURN_STARTED) translator.handle( - notification('item/completed', { item: { type: 'userMessage', id: 'item-0', text: 'hi' } }) + notification('item/completed', { item: { type: 'agentMessage', id: 'item-0', text: 'hi' } }) ) expect(tap.publishes()).toBe(1) @@ -520,7 +520,7 @@ describe('codex journal translation', () => { expect(timeline).toEqual([]) }) - it('projects only user and assistant content for a complete turn with hooks', () => { + it('projects assistant content without provider user echoes for a complete turn with hooks', () => { const { translator, tap } = translatorWith() translator.handle(notification('thread/started', { thread: { id: THREAD_ID } })) @@ -552,7 +552,6 @@ describe('codex journal translation', () => { })) ) expect(timeline.map(({ role, blocks }) => ({ role, blocks }))).toEqual([ - { role: 'user', blocks: [{ type: 'text', text: 'hi' }] }, { role: 'assistant', blocks: [{ type: 'text', text: 'hello' }] } ]) }) diff --git a/src/main/codex/codex-structured-journal-translation-turns.ts b/src/main/codex/codex-structured-journal-translation-turns.ts index 06bb28f85f9..7313946a53e 100644 --- a/src/main/codex/codex-structured-journal-translation-turns.ts +++ b/src/main/codex/codex-structured-journal-translation-turns.ts @@ -56,9 +56,16 @@ export function publishCodexTurnLifecycle(input: { return admission } } - if (input.sink.tryPublish) { - return input.sink.tryPublish({ lifecycle: true }) + // Preserve first-work evidence when completion arrives before the journal drains. + const publishOptions = { + lifecycle: true, + ...(input.state === 'running' + ? { coalescingKey: `turn-start:${input.sessionId}:${input.turnId}` } + : {}) } - input.sink.publish({ lifecycle: true }) + if (input.sink.tryPublish) { + return input.sink.tryPublish(publishOptions) + } + input.sink.publish(publishOptions) return ADMITTED } diff --git a/src/main/codex/codex-structured-journal-translation.test.ts b/src/main/codex/codex-structured-journal-translation.test.ts index e88bd1d5a65..0a791afa77e 100644 --- a/src/main/codex/codex-structured-journal-translation.test.ts +++ b/src/main/codex/codex-structured-journal-translation.test.ts @@ -302,7 +302,7 @@ describe('codex journal translation', () => { ).toBe('idle') }) - it('journals a user turn and the assistant answer under durable codex keys', () => { + it('counts a user echo without rendering it and preserves the assistant ordinal', () => { const { translator, tap } = translatorWith() translator.handle(TURN_STARTED) @@ -317,17 +317,37 @@ describe('codex journal translation', () => { }) ) - expect(tap.rows.map((row) => row.key)).toEqual([ - 'codex:thread-abc:turn-1:0', - 'codex:thread-abc:turn-1:1' - ]) - expect(tap.rows[1]?.body).toEqual({ + expect(tap.rows.map((row) => row.key)).toEqual(['codex:thread-abc:turn-1:1']) + expect(tap.rows[0]?.body).toEqual({ kind: 'message', role: 'assistant', blocks: [{ type: 'text', text: 'hello' }] }) }) + it('suppresses both echo lifecycle frames, including skill and unknown parts', () => { + const { translator, tap } = translatorWith() + translator.handle(TURN_STARTED) + const item = { + type: 'userMessage', + id: 'echo', + content: [ + { type: 'text', text: 'Expanded instructions' }, + { type: 'skill', name: 'example', path: '/tmp/SKILL.md' }, + { type: 'future_context', text: 'More context' } + ] + } + translator.handle(notification('item/started', { item })) + translator.handle(notification('item/completed', { item })) + expect(tap.rows).toEqual([]) + translator.handle( + notification('item/completed', { + item: { type: 'agentMessage', id: 'answer', text: 'Done' } + }) + ) + expect(tap.rows.map((row) => row.key)).toEqual(['codex:thread-abc:turn-1:1']) + }) + it('folds streamed deltas into one snapshot row on the same key the item started under', () => { const { translator, tap, window } = translatorWith() diff --git a/src/main/codex/codex-structured-journal-translation.ts b/src/main/codex/codex-structured-journal-translation.ts index c8ecc996742..85f3bf76e46 100644 --- a/src/main/codex/codex-structured-journal-translation.ts +++ b/src/main/codex/codex-structured-journal-translation.ts @@ -1,4 +1,8 @@ -import { CODEX_TOKEN_USAGE_METHOD } from './codex-subagent-activity' +import { createCodexProviderActivityReader } from '../native-chat/agent-session-wire/provider-frame-activity' +import { + CODEX_TOKEN_USAGE_METHOD, + readCodexNotificationThreadItem +} from './codex-subagent-activity' import { CodexSubagentRoster } from './codex-subagent-roster' import { readCodexThreadItem } from './codex-structured-item-translation' import { CodexJournalGenericFrames } from './codex-structured-journal-generic-frames' @@ -12,17 +16,14 @@ import { } from './codex-structured-journal-contracts' import { settleCodexJournalSession, - settleCodexJournalTurn, - settleCodexOversizedNotification + settleCodexJournalTurn } from './codex-structured-journal-settlement' +import { settleCodexOversizedNotificationFrame } from './codex-structured-journal-translation-frames' import { restoreCodexJournalThread } from './codex-structured-journal-translation-restore' import { CodexJournalActiveTurns } from './codex-structured-journal-translation-turn-state' import { publishCodexTurnLifecycle } from './codex-structured-journal-translation-turns' -import { - readCodexJournalRecord, - readCodexJournalString -} from './codex-structured-journal-translation-values' import { readCodexTurnId } from './codex-structured-thread-facts' +import type { CodexStructuredSessionEvent } from './codex-structured-session-adapter' export type { CodexJournalTranslationAdmission, @@ -63,22 +64,44 @@ export function createCodexJournalTranslator( }) const flushStreams = (): CodexJournalTranslationAdmission => items.streams.flush() ? CODEX_JOURNAL_ADMITTED : { accepted: false, reason: 'backpressure' } + let readActivity = createCodexProviderActivityReader() + const publishActivity = ( + event: Extract, + admission: CodexJournalTranslationAdmission + ): CodexJournalTranslationAdmission => { + if (!admission.accepted || event.threadId !== (deps.primaryThreadId?.() ?? null)) { + return admission + } + const turnId = readCodexTurnId(event.params) ?? activeTurns.current(event.threadId) + if (!turnId) { + return admission + } + const text = readActivity(event.method, event.params) + if (text !== undefined) { + deps.sink.setActivity?.(text ? { turnId, text } : null) + } + return admission + } return { - restoreThread: (threadId, thread) => - restoreCodexJournalThread({ + restoreThread: (threadId, thread) => { + if (threadId === (deps.primaryThreadId?.() ?? null)) { + readActivity = createCodexProviderActivityReader() + } + return restoreCodexJournalThread({ threadId, thread, currentTurnIds: activeTurns.byThread, ordinals: items.ordinals, handleItem: (event) => { - const translated = items.handle(event) + const translated = items.handle(event, 'history') return translated.handled ? translated.admission : { accepted: false, reason: 'untranslated' } }, flush: items.streams.flush - }), + }) + }, handle: (event) => { if (event.type === 'ended') { const streamAdmission = flushStreams() @@ -107,6 +130,8 @@ export function createCodexJournalTranslator( if (!sweep.accepted) { return sweep } + readActivity = createCodexProviderActivityReader() + deps.sink.setActivity?.(null) items.activeItems.clear() prompts.pending.clear() activeTurns.clear() @@ -115,7 +140,7 @@ export function createCodexJournalTranslator( if (event.type === 'notification') { const streamResult = items.streams.handle(event.threadId, event.method, event.params) if (streamResult.handled) { - return streamResult.admission + return publishActivity(event, streamResult.admission) } } const streamAdmission = flushStreams() @@ -155,23 +180,29 @@ export function createCodexJournalTranslator( } } if (event.method === 'item/started' || event.method === 'item/completed') { - const subagentAdmission = handleSubagentItem(event) - if (subagentAdmission) { - return subagentAdmission + const subagentItem = readCodexNotificationThreadItem(event.params, readCodexThreadItem) + if (subagentItem) { + return subagents.handleItem({ + threadId: event.threadId, + turnId: readCodexTurnId(event.params) ?? activeTurns.current(event.threadId), + item: subagentItem + }) } const translated = items.handle(event) - return translated.handled - ? translated.admission - : genericFrames.appendUnhandled( - `notification:${event.method}`, - event.params, - event.threadId - ) + return publishActivity( + event, + translated.handled + ? translated.admission + : genericFrames.appendUnhandled( + `notification:${event.method}`, + event.params, + event.threadId + ) + ) } - return genericFrames.appendUnhandled( - `notification:${event.method}`, - event.params, - event.threadId + return publishActivity( + event, + genericFrames.appendUnhandled(`notification:${event.method}`, event.params, event.threadId) ) }, resolvePrompt: (journalItemId) => prompts.resolve(journalItemId), @@ -189,45 +220,18 @@ export function createCodexJournalTranslator( } /** Routes a `subAgentActivity` item to the roster; null when it is not one. */ - function handleSubagentItem(event: { - threadId: string - params: unknown - }): CodexJournalTranslationAdmission | null { - const params = - typeof event.params === 'object' && event.params !== null - ? (event.params as Record) - : {} - const item = readCodexThreadItem(params.item) - if (!item) { - return null - } - return subagents.handleItem({ - threadId: event.threadId, - turnId: readCodexTurnId(event.params) ?? activeTurns.current(event.threadId), - item - }) - } - function settleOversizedNotification(event: { sessionId: string threadId: string kind: string payload: unknown }): CodexJournalTranslationAdmission | null { - if (event.kind !== 'frame:oversized-notification') { - return null - } - const method = readCodexJournalString(readCodexJournalRecord(event.payload), 'method') - return method - ? settleCodexOversizedNotification({ - sessionId: event.sessionId, - threadId: event.threadId, - method, - sink: deps.sink, - streams: items.streams, - activeItems: items.activeItems - }) - : null + return settleCodexOversizedNotificationFrame({ + ...event, + sink: deps.sink, + streams: items.streams, + activeItems: items.activeItems + }) } function startTurn(event: { @@ -252,6 +256,10 @@ export function createCodexJournalTranslator( }) if (admission.accepted) { activeTurns.remember(event.threadId, turnId) + if (event.threadId === (deps.primaryThreadId?.() ?? null)) { + readActivity = createCodexProviderActivityReader() + deps.sink.setActivity?.(null) + } } return admission } @@ -286,6 +294,10 @@ export function createCodexJournalTranslator( if (admission.accepted) { items.ordinals.forgetTurn(event.threadId, turnId) activeTurns.forget(event.threadId, turnId) + if (event.threadId === (deps.primaryThreadId?.() ?? null)) { + readActivity = createCodexProviderActivityReader() + deps.sink.setActivity?.(null) + } } return admission } diff --git a/src/main/codex/codex-structured-session-acquire.ts b/src/main/codex/codex-structured-session-acquire.ts index 04a48a6250e..78855842332 100644 --- a/src/main/codex/codex-structured-session-acquire.ts +++ b/src/main/codex/codex-structured-session-acquire.ts @@ -106,7 +106,7 @@ export async function acquireCodexStructuredSession(input: { command: launch.command, args: launch.args, cwd: launch.cwd, - env: buildCodexStructuredChildEnvironment(launch, acquireInput.spawnToken) + env: buildCodexStructuredChildEnvironment(launch, acquireInput.spawnToken, sessionId) }, { onNotification: (method, params) => diff --git a/src/main/codex/codex-structured-session-adapter.test.ts b/src/main/codex/codex-structured-session-adapter.test.ts index 5e218c1f31f..32492762121 100644 --- a/src/main/codex/codex-structured-session-adapter.test.ts +++ b/src/main/codex/codex-structured-session-adapter.test.ts @@ -12,6 +12,7 @@ import type { } from './codex-app-server-connection' import type { StructuredAgentSessionEventSink } from '../native-chat/agent-session-wire/structured-agent-session-event-sink' import { CODEX_SPAWN_TOKEN_ENV } from './codex-structured-owner-identity' +import { ORCA_STRUCTURED_SESSION_ENV } from '../../shared/structured-session-marker' import { encodeCodexQuestionOptionId } from './codex-structured-prompt-replies' import { CodexStructuredSessionAdapter, @@ -150,7 +151,8 @@ describe('CodexStructuredSessionAdapter.acquire', () => { expect(codex.connections[0].launch.env).toEqual({ [CODEX_SPAWN_TOKEN_ENV]: 'spawn-9', - CODEX_HOME: '/codex/home' + CODEX_HOME: '/codex/home', + [ORCA_STRUCTURED_SESSION_ENV]: '1' }) expect(codex.connections[0].launch.cwd).toBe('/work/repo') expect(codex.connections[0].calls[0]).toEqual({ @@ -285,7 +287,7 @@ describe('CodexStructuredSessionAdapter.acquire', () => { }) codex.connections[0].handlers.onNotification?.('item/completed', { - item: { type: 'userMessage', id: 'message-1', text: 'hello' } + item: { type: 'agentMessage', id: 'message-1', text: 'hello' } }) await vi.waitFor(() => { diff --git a/src/main/codex/codex-subagent-activity.ts b/src/main/codex/codex-subagent-activity.ts index 29e9211db01..55f39f41627 100644 --- a/src/main/codex/codex-subagent-activity.ts +++ b/src/main/codex/codex-subagent-activity.ts @@ -123,3 +123,18 @@ export function readCodexThreadTokenTotal(params: unknown): CodexThreadTokenTota ? { threadId, totalTokens: total } : null } + +/** Pull the `subAgentActivity` item out of a raw notification payload. + * + * Lives beside the readers rather than in the translator: the translator's job + * is routing, and this is the shape check that decides whether a frame is one + * of ours at all. Returns null for anything that is not a thread item, which is + * the translator's signal to keep looking. */ +export function readCodexNotificationThreadItem( + params: unknown, + read: (value: unknown) => CodexThreadItem | null +): CodexThreadItem | null { + const record = + typeof params === 'object' && params !== null ? (params as Record) : {} + return read(record.item) +} diff --git a/src/main/daemon/terminal-host-non-agent-foreground.test.ts b/src/main/daemon/terminal-host-non-agent-foreground.test.ts new file mode 100644 index 00000000000..4c50699ff2a --- /dev/null +++ b/src/main/daemon/terminal-host-non-agent-foreground.test.ts @@ -0,0 +1,92 @@ +import { afterEach, describe, expect, it, vi } from 'vitest' +import type { ProcessTableRow } from '../../shared/process-table-snapshot' +import type * as SnapshotReader from '../../shared/process-table-snapshot-reader' +import { inspectTerminalHostProcess } from './terminal-host-process-inspection' +import type { Session } from './session' + +const { readSnapshot } = vi.hoisted(() => ({ readSnapshot: vi.fn() })) +vi.mock('../../shared/process-table-snapshot-reader', async (importOriginal) => ({ + ...(await importOriginal()), + getStrictProcessTableSnapshotWithAge: readSnapshot +})) + +function table(command: string | null): ProcessTableRow[] { + const foregroundPgid = command === null ? 100 : 101 + const shell: ProcessTableRow = { + pid: 100, + ppid: 1, + pgid: 100, + tpgid: foregroundPgid, + tty: 'pts/1', + startTime: 'shell-start', + stat: command === null ? 'Ss+' : 'Ss', + command: '/bin/bash' + } + return command === null + ? [shell] + : [ + shell, + { + ...shell, + pid: 101, + ppid: 100, + pgid: 101, + stat: 'S+', + startTime: 'command-start', + command + } + ] +} + +async function inspect(rawName: string, command: string | null) { + readSnapshot.mockResolvedValue({ rows: table(command), capturedAgeMs: 0 }) + return inspectTerminalHostProcess({ + sessionId: 'busy-tab', + session: { + pid: 100, + incarnationId: 'incarnation-1', + isAlive: true, + getForegroundProcess: () => rawName + } as unknown as Session, + authorityGeneration: 'generation-1', + nextObservationEpoch: () => 1 + }) +} + +afterEach(() => { + vi.restoreAllMocks() + readSnapshot.mockClear() +}) + +describe.each(['linux', 'darwin'] as const)('daemon ordinary foreground on %s', (platform) => { + it.each(['sleep', 'vim', 'node'])( + 'retains the running %s name alongside agent-only evidence', + async (name) => { + vi.spyOn(process, 'platform', 'get').mockReturnValue(platform) + const result = await inspect(name, `${name} 300`) + expect(result).toMatchObject({ + foregroundProcess: name, + hasChildProcesses: true, + foregroundProcessEvidence: { verdict: 'live', processName: null } + }) + expect(readSnapshot).toHaveBeenCalledTimes(1) + } + ) + + it('still clears a stale recognized agent after its process exits', async () => { + vi.spyOn(process, 'platform', 'get').mockReturnValue(platform) + expect(await inspect('claude', null)).toMatchObject({ + foregroundProcess: null, + foregroundProcessEvidence: { verdict: 'live', processName: null } + }) + }) + + it('still reports no foreground command for an idle shell', async () => { + vi.spyOn(process, 'platform', 'get').mockReturnValue(platform) + expect(await inspect('bash', null)).toMatchObject({ + foregroundProcess: null, + hasChildProcesses: false, + foregroundProcessEvidence: { verdict: 'live', processName: null } + }) + }) +}) diff --git a/src/main/daemon/terminal-host-process-inspection.ts b/src/main/daemon/terminal-host-process-inspection.ts index 6810687631e..2f9fb1491d9 100644 --- a/src/main/daemon/terminal-host-process-inspection.ts +++ b/src/main/daemon/terminal-host-process-inspection.ts @@ -1,4 +1,5 @@ import { isShellProcess } from '../../shared/agent-detection' +import { recognizeAgentProcess } from '../../shared/agent-process-recognition' import type { RemoteForegroundEvidence } from '../../shared/foreground-process-evidence' import { getCheapProcessTableSnapshot } from '../../shared/cheap-process-table-snapshot-reader' import { getStrictProcessTableSnapshotWithAge } from '../../shared/process-table-snapshot-reader' @@ -100,9 +101,16 @@ export async function inspectTerminalHostProcess(args: { clearSteadyStateAnchor(session) } } + const nonShellForeground = foregroundProcess !== null && !isShellProcess(foregroundProcess) + // Evidence names recognized agents only, so its null must not erase an ordinary command (#18078). + const ordinaryForeground = + nonShellForeground && !recognizeAgentProcess(foregroundProcess) ? foregroundProcess : null return { - foregroundProcess: evidence.verdict === 'live' ? evidence.processName : foregroundProcess, - hasChildProcesses: foregroundProcess !== null && !isShellProcess(foregroundProcess), + foregroundProcess: + evidence.verdict === 'live' + ? (evidence.processName ?? ordinaryForeground) + : foregroundProcess, + hasChildProcesses: nonShellForeground, foregroundProcessEvidence: evidence } } diff --git a/src/main/git/command-runner/exec-file-capture.ts b/src/main/git/command-runner/exec-file-capture.ts index e9ae815ff34..f343246571d 100644 --- a/src/main/git/command-runner/exec-file-capture.ts +++ b/src/main/git/command-runner/exec-file-capture.ts @@ -15,6 +15,8 @@ type ExecFileCaptureOptions = Omit & { onChildTerminated?: () => void admissionTier?: GitAdmissionTier createTimeoutError?: () => Error + /** Called once when the deadline — not an abort — is what ended the process. */ + onDeadlineKill?: () => void } const GIT_TERMINATION_BARRIER_FALLBACK_TIMEOUT_MS = 2_147_000_000 @@ -54,6 +56,9 @@ export async function execFileCaptureToTermination( ) { return { stdout, stderr } } + if (result.timedOut && !options.signal?.aborted) { + options.onDeadlineKill?.() + } const error = result.timedOut ? (options.createTimeoutError?.() ?? new Error(`${command} timed out.`)) : new Error( diff --git a/src/main/git/command-runner/gh-exec-file.ts b/src/main/git/command-runner/gh-exec-file.ts index e8308a8e4b4..9a92f1d59de 100644 --- a/src/main/git/command-runner/gh-exec-file.ts +++ b/src/main/git/command-runner/gh-exec-file.ts @@ -20,6 +20,7 @@ import { resolveHostGitHubCli } from './github-cli-host-fallback' import { execFileCaptureToTermination } from './exec-file-capture' +import { logHostedCliDeadlineKill } from './hosted-cli-deadline-log' import type { GitExecOptions } from './git-exec-options' import { argsLookIdempotent } from './gh-idempotency' import { applyGhHostToArgs, explicitGhHostname, explicitGhRepoHostname } from './gh-host-args' @@ -109,6 +110,7 @@ export async function ghExecFileAsync( // Why: scope by runtime and host so unrelated github.com, GHES, and WSL quotas cannot block each other. const rateLimitBucket = classifyGhRateLimitBucket(args) const rateLimitProbe = isGhRateLimitProbe(args) + const timeoutMs = options.timeout ?? defaultGhExecTimeoutMs(options.env) assertGhRateLimitScopeAvailable(args, options, resolved, rateLimitBucket, rateLimitProbe) let lastError: unknown let attemptedHostFallback = false @@ -128,9 +130,10 @@ export async function ghExecFileAsync( encoding: (options.encoding ?? 'utf-8') as BufferEncoding, maxBuffer: options.maxBuffer, // Why: bound gh so one stuck child fails visibly instead of wedging the IPC lane. - timeout: options.timeout ?? defaultGhExecTimeoutMs(options.env), + timeout: timeoutMs, env: nonInteractiveGhEnv(options.env), - signal: options.signal + signal: options.signal, + onDeadlineKill: () => logHostedCliDeadlineKill('gh', resolved.binary, args, timeoutMs) }, resolved.termination ) diff --git a/src/main/git/command-runner/glab-exec-file.ts b/src/main/git/command-runner/glab-exec-file.ts index 3257dd9e818..37aad697b95 100644 --- a/src/main/git/command-runner/glab-exec-file.ts +++ b/src/main/git/command-runner/glab-exec-file.ts @@ -3,6 +3,7 @@ import { extractExecError, parseRetryAfterMs } from '../exec-error' import { resolveCommand, resolveDefaultWslCli } from './wsl-command-resolution' import { isHostCommandMissing } from './github-cli-host-fallback' import { execFileCaptureToTermination } from './exec-file-capture' +import { logHostedCliDeadlineKill } from './hosted-cli-deadline-log' import type { GitExecOptions } from './git-exec-options' import { argsLookIdempotent } from './gh-idempotency' import { @@ -59,6 +60,7 @@ export async function glabExecFileAsync( ): Promise<{ stdout: string; stderr: string }> { ;({ args, options } = redirectPortedHostnameToEnv(args, options)) let resolved = resolveCommand('glab', args, options.cwd, options.wslDistro) + const timeoutMs = options.timeout ?? DEFAULT_GLAB_EXEC_TIMEOUT_MS let lastError: unknown let attemptedDefaultWslFallback = false for (let attempt = 0; attempt <= GH_RETRY_DELAYS_MS.length; attempt++) { @@ -72,9 +74,10 @@ export async function glabExecFileAsync( cwd: resolved.cwd, encoding: (options.encoding ?? 'utf-8') as BufferEncoding, maxBuffer: options.maxBuffer, - timeout: options.timeout ?? DEFAULT_GLAB_EXEC_TIMEOUT_MS, + timeout: timeoutMs, env: options.env, - signal: options.signal + signal: options.signal, + onDeadlineKill: () => logHostedCliDeadlineKill('glab', resolved.binary, args, timeoutMs) }, resolved.termination ) diff --git a/src/main/git/command-runner/hosted-cli-deadline-log.test.ts b/src/main/git/command-runner/hosted-cli-deadline-log.test.ts new file mode 100644 index 00000000000..06ad0e6261f --- /dev/null +++ b/src/main/git/command-runner/hosted-cli-deadline-log.test.ts @@ -0,0 +1,94 @@ +import { EventEmitter } from 'node:events' +import type { ChildProcess } from 'node:child_process' +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' + +const { spawnMock } = vi.hoisted(() => ({ spawnMock: vi.fn() })) + +vi.mock('node:child_process', async (importOriginal) => ({ + ...(await importOriginal()), + spawn: spawnMock +})) + +import { ghExecFileAsync } from './gh-exec-file' +import { logHostedCliDeadlineKill } from './hosted-cli-deadline-log' + +function mockChild(pid = 4321): ChildProcess { + const child = new EventEmitter() as EventEmitter & Record + child.pid = pid + child.kill = vi.fn(() => true) + child.stdin = Object.assign(new EventEmitter(), { end: vi.fn() }) + child.stdout = new EventEmitter() + child.stderr = new EventEmitter() + return child as unknown as ChildProcess +} + +/** + * #18234 took four rounds of strace/perf/proc spelunking from the reporter + * because a deadline kill produced no evidence at all. The resolved path is the + * fact that names a self-recursive wrapper. + */ +describe('hosted CLI deadline logging', () => { + let warn: ReturnType + + beforeEach(() => { + vi.useFakeTimers() + spawnMock.mockReset() + warn = vi.spyOn(console, 'warn').mockImplementation(() => {}) + vi.spyOn(process, 'kill').mockImplementation((() => true) as unknown as typeof process.kill) + }) + + afterEach(() => { + vi.useRealTimers() + vi.restoreAllMocks() + }) + + it('names the CLI, the deadline and the resolved path, and never the argv values', () => { + logHostedCliDeadlineKill( + 'gh', + '/home/user/.local/bin/gh', + ['api', '-H', 'Authorization: token ghp_secret'], + 15_000 + ) + + const line = warn.mock.calls[0][0] as string + expect(line).toContain('[gh]') + expect(line).toContain('15000ms') + expect(line).toContain('/home/user/.local/bin/gh') + expect(line).toContain('"api"') + expect(line).toContain('(3 args)') + expect(line).not.toContain('ghp_secret') + expect(line).not.toContain('Authorization') + }) + + it('logs once when gh is killed at its deadline', async () => { + spawnMock.mockReturnValue(mockChild()) + + const rejection = expect( + ghExecFileAsync(['api', '--include', 'user/starred/stablyai/orca'], { timeout: 15_000 }) + ).rejects.toThrow('timed out') + await vi.advanceTimersByTimeAsync(15_000) + await vi.advanceTimersByTimeAsync(15_000) + await rejection + + const deadlineLines = warn.mock.calls.filter((call) => String(call[0]).startsWith('[gh]')) + expect(deadlineLines).toHaveLength(1) + expect(String(deadlineLines[0][0])).toContain('wrapper script') + }) + + it('stays quiet when the caller aborted rather than the deadline firing', async () => { + const controller = new AbortController() + spawnMock.mockReturnValue(mockChild()) + + const rejection = expect( + ghExecFileAsync(['api', '--include', 'user/starred/stablyai/orca'], { + timeout: 15_000, + signal: controller.signal + }) + ).rejects.toThrow() + controller.abort() + await vi.advanceTimersByTimeAsync(15_000) + await rejection + + expect(warn.mock.calls.filter((call) => String(call[0]).startsWith('[gh]'))).toHaveLength(0) + }) +}) diff --git a/src/main/git/command-runner/hosted-cli-deadline-log.ts b/src/main/git/command-runner/hosted-cli-deadline-log.ts new file mode 100644 index 00000000000..3a8be660d14 --- /dev/null +++ b/src/main/git/command-runner/hosted-cli-deadline-log.ts @@ -0,0 +1,26 @@ +/** + * One log line when `gh`/`glab` is killed at its deadline without answering. + * + * Why this exists: the deadline kill was completely silent. In #18234 a user's + * `~/.local/bin/gh` wrapper (`exec mise x gh -- gh "$@"`) re-execed itself in + * place at 100% CPU on every invocation, and the only evidence Orca produced was + * that GitHub features quietly did nothing. Diagnosing it took the reporter four + * rounds of `strace`, `perf` and `/proc` spelunking. The resolved path below is + * the single most useful fact — it names the wrapper. + * + * Why not the full argv: `gh api` carries `-H Authorization: …` and `--field` + * bodies, so only the subcommand and an argument count are safe to print. + */ +export function logHostedCliDeadlineKill( + cli: string, + resolvedBinary: string, + args: readonly string[], + timeoutMs: number +): void { + const subcommand = args[0] ?? '(none)' + console.warn( + `[${cli}] killed at its ${timeoutMs}ms deadline without answering — ` + + `subcommand "${subcommand}" (${args.length} args), resolved to "${resolvedBinary}". ` + + `If that path is a wrapper script, check that it resolves the real ${cli} binary rather than itself.` + ) +} diff --git a/src/main/global-fetch-call-site-audit.test.ts b/src/main/global-fetch-call-site-audit.test.ts index e4c0539fbfb..7b4151a0293 100644 --- a/src/main/global-fetch-call-site-audit.test.ts +++ b/src/main/global-fetch-call-site-audit.test.ts @@ -23,6 +23,7 @@ const AUDITED_GLOBAL_FETCH_LINES = new Map([ ['main/orca-profiles/profile-cloud-client.ts', 1], ['main/orca-profiles/profile-cloud-org-members-client.ts', 1], ['main/rate-limits/codex-fetcher.ts', 3], + ['main/runtime/push/push-gateway-client.ts', 1], ['main/runtime/relay/relay-http-client.ts', 2], ['main/runtime/relay/relay-region-preference.ts', 3], ['main/source-control/hosted-review-api-request.ts', 1], diff --git a/src/main/ipc/minimax-credentials.test.ts b/src/main/ipc/minimax-credentials.test.ts index 77e6f592de0..242ee2217bc 100644 --- a/src/main/ipc/minimax-credentials.test.ts +++ b/src/main/ipc/minimax-credentials.test.ts @@ -16,6 +16,9 @@ const saveMiniMaxSessionCookieMock = vi.hoisted(() => vi.fn()) const clearMiniMaxSessionCookieMock = vi.hoisted(() => vi.fn()) const hasMiniMaxSessionCookieMock = vi.hoisted(() => vi.fn(() => false)) const clearMiniMaxSessionCookieJarMock = vi.hoisted(() => vi.fn(() => Promise.resolve())) +const saveMiniMaxApiKeyMock = vi.hoisted(() => vi.fn()) +const clearMiniMaxApiKeyMock = vi.hoisted(() => vi.fn()) +const hasMiniMaxApiKeyMock = vi.hoisted(() => vi.fn(() => false)) vi.mock('../minimax/minimax-cookie-store', () => ({ saveMiniMaxSessionCookie: saveMiniMaxSessionCookieMock, @@ -23,6 +26,12 @@ vi.mock('../minimax/minimax-cookie-store', () => ({ hasMiniMaxSessionCookie: hasMiniMaxSessionCookieMock })) +vi.mock('../minimax/minimax-api-key-store', () => ({ + saveMiniMaxApiKey: saveMiniMaxApiKeyMock, + clearMiniMaxApiKey: clearMiniMaxApiKeyMock, + hasMiniMaxApiKey: hasMiniMaxApiKeyMock +})) + vi.mock('../rate-limits/minimax-request-context', () => ({ clearMiniMaxSessionCookieJar: clearMiniMaxSessionCookieJarMock })) @@ -62,35 +71,65 @@ describe('registerMiniMaxCredentialsHandlers', () => { clearMiniMaxSessionCookieJarMock.mockResolvedValue(undefined) hasMiniMaxSessionCookieMock.mockReset() hasMiniMaxSessionCookieMock.mockReturnValue(false) + saveMiniMaxApiKeyMock.mockReset() + clearMiniMaxApiKeyMock.mockReset() + hasMiniMaxApiKeyMock.mockReset() + hasMiniMaxApiKeyMock.mockReturnValue(false) }) afterEach(() => { vi.restoreAllMocks() }) - it('registers the three MiniMax credential channels', () => { + it('registers all five MiniMax credential channels', () => { registerMiniMaxCredentialsHandlers(null) expect(ipcState.handleHandlers.has('minimaxCredentials:getStatus')).toBe(true) expect(ipcState.handleHandlers.has('minimaxCredentials:saveCookie')).toBe(true) expect(ipcState.handleHandlers.has('minimaxCredentials:clearCookie')).toBe(true) + expect(ipcState.handleHandlers.has('minimaxCredentials:saveApiKey')).toBe(true) + expect(ipcState.handleHandlers.has('minimaxCredentials:clearApiKey')).toBe(true) }) it('returns the configured state on getStatus from the cookie store', async () => { hasMiniMaxSessionCookieMock.mockReturnValue(true) registerMiniMaxCredentialsHandlers(null) - const status = await invoke<{ configured: boolean }>('minimaxCredentials:getStatus') - expect(status).toEqual({ configured: true }) + const status = await invoke<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }>('minimaxCredentials:getStatus') + expect(status).toEqual({ + configured: true, + cookieConfigured: true, + apiKeyConfigured: false + }) + }) + + it('returns apiKeyConfigured true on getStatus when the API key store has a key', async () => { + hasMiniMaxApiKeyMock.mockReturnValue(true) + registerMiniMaxCredentialsHandlers(null) + const status = await invoke<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }>('minimaxCredentials:getStatus') + expect(status).toEqual({ + configured: true, + cookieConfigured: false, + apiKeyConfigured: true + }) }) it('persists the cookie and reports configured after saveCookie', async () => { hasMiniMaxSessionCookieMock.mockReturnValueOnce(true) registerMiniMaxCredentialsHandlers(null) - const status = await invoke<{ configured: boolean }>( - 'minimaxCredentials:saveCookie', - '_token=abc; minimax_group_id_v2=42' - ) + const status = await invoke<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }>('minimaxCredentials:saveCookie', '_token=abc; minimax_group_id_v2=42') expect(saveMiniMaxSessionCookieMock).toHaveBeenCalledWith('_token=abc; minimax_group_id_v2=42') - expect(status).toEqual({ configured: true }) + expect(status).toMatchObject({ configured: true, cookieConfigured: true }) }) it('triggers a rate-limit refresh after saveCookie when a service is provided', async () => { @@ -113,11 +152,15 @@ describe('registerMiniMaxCredentialsHandlers', () => { const { refresh, invalidateMiniMaxCredentialState, service } = makeRefreshMock() hasMiniMaxSessionCookieMock.mockReturnValueOnce(false) registerMiniMaxCredentialsHandlers(service as RateLimitService) - const status = await invoke<{ configured: boolean }>('minimaxCredentials:clearCookie') + const status = await invoke<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }>('minimaxCredentials:clearCookie') expect(clearMiniMaxSessionCookieMock).toHaveBeenCalledTimes(1) expect(invalidateMiniMaxCredentialState).toHaveBeenCalledTimes(1) expect(clearMiniMaxSessionCookieJarMock).toHaveBeenCalledTimes(1) - expect(status).toEqual({ configured: false }) + expect(status).toMatchObject({ configured: false, cookieConfigured: false }) await new Promise((resolve) => setImmediate(resolve)) expect(refresh).toHaveBeenCalledTimes(1) }) @@ -129,12 +172,16 @@ describe('registerMiniMaxCredentialsHandlers', () => { hasMiniMaxSessionCookieMock.mockReturnValueOnce(false) registerMiniMaxCredentialsHandlers(service as RateLimitService) - const status = await invoke<{ configured: boolean }>('minimaxCredentials:clearCookie') + const status = await invoke<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }>('minimaxCredentials:clearCookie') expect(clearMiniMaxSessionCookieMock).toHaveBeenCalledTimes(1) expect(invalidateMiniMaxCredentialState).toHaveBeenCalledTimes(1) expect(clearMiniMaxSessionCookieJarMock).toHaveBeenCalledTimes(1) - expect(status).toEqual({ configured: false }) + expect(status).toMatchObject({ configured: false, cookieConfigured: false }) expect(errorSpy).toHaveBeenCalledWith( expect.stringContaining('failed to clear session cookie jar after credential clear'), expect.any(Error) @@ -159,4 +206,61 @@ describe('registerMiniMaxCredentialsHandlers', () => { expect.any(Error) ) }) + + it('persists the API key and reports apiKeyConfigured after saveApiKey', async () => { + hasMiniMaxApiKeyMock.mockReturnValueOnce(true) + registerMiniMaxCredentialsHandlers(null) + const status = await invoke<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }>('minimaxCredentials:saveApiKey', 'sk-test-1234567890') + expect(saveMiniMaxApiKeyMock).toHaveBeenCalledWith('sk-test-1234567890') + expect(status).toMatchObject({ configured: true, apiKeyConfigured: true }) + }) + + it('rejects non-string API keys on saveApiKey', async () => { + registerMiniMaxCredentialsHandlers(null) + await expect(invoke('minimaxCredentials:saveApiKey', 12345)).rejects.toThrow(/must be a string/) + expect(saveMiniMaxApiKeyMock).not.toHaveBeenCalled() + }) + + it('triggers a rate-limit refresh after saveApiKey when a service is provided', async () => { + const { refresh, invalidateMiniMaxCredentialState, service } = makeRefreshMock() + registerMiniMaxCredentialsHandlers(service as RateLimitService) + await invoke('minimaxCredentials:saveApiKey', 'sk-test-1234567890') + await new Promise((resolve) => setImmediate(resolve)) + expect(invalidateMiniMaxCredentialState).toHaveBeenCalledTimes(1) + expect(refresh).toHaveBeenCalledTimes(1) + }) + + it('clears the API key and triggers a refresh on clearApiKey', async () => { + const { refresh, invalidateMiniMaxCredentialState, service } = makeRefreshMock() + hasMiniMaxApiKeyMock.mockReturnValueOnce(false) + registerMiniMaxCredentialsHandlers(service as RateLimitService) + const status = await invoke<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }>('minimaxCredentials:clearApiKey') + expect(clearMiniMaxApiKeyMock).toHaveBeenCalledTimes(1) + expect(invalidateMiniMaxCredentialState).toHaveBeenCalledTimes(1) + expect(status).toMatchObject({ configured: false, apiKeyConfigured: false }) + await new Promise((resolve) => setImmediate(resolve)) + expect(refresh).toHaveBeenCalledTimes(1) + }) + + it('reports configured true when either cookie or API key is set', async () => { + hasMiniMaxSessionCookieMock.mockReturnValue(true) + hasMiniMaxApiKeyMock.mockReturnValue(true) + registerMiniMaxCredentialsHandlers(null) + const status = await invoke<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }>('minimaxCredentials:getStatus') + expect(status.configured).toBe(true) + expect(status.cookieConfigured).toBe(true) + expect(status.apiKeyConfigured).toBe(true) + }) }) diff --git a/src/main/ipc/minimax-credentials.ts b/src/main/ipc/minimax-credentials.ts index 97eefcd7116..bd0368a1c9e 100644 --- a/src/main/ipc/minimax-credentials.ts +++ b/src/main/ipc/minimax-credentials.ts @@ -4,18 +4,31 @@ import { hasMiniMaxSessionCookie, saveMiniMaxSessionCookie } from '../minimax/minimax-cookie-store' +import { + clearMiniMaxApiKey, + hasMiniMaxApiKey, + saveMiniMaxApiKey +} from '../minimax/minimax-api-key-store' import { clearMiniMaxSessionCookieJar } from '../rate-limits/minimax-request-context' import type { RateLimitService } from '../rate-limits/service' export type MiniMaxCredentialsStatus = { configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean } function getMiniMaxCredentialsStatus(): MiniMaxCredentialsStatus { - return { configured: hasMiniMaxSessionCookie() } + const cookieConfigured = hasMiniMaxSessionCookie() + const apiKeyConfigured = hasMiniMaxApiKey() + return { + configured: cookieConfigured || apiKeyConfigured, + cookieConfigured, + apiKeyConfigured + } } -// Why: fire-and-forget — callers get the persisted cookie status immediately; +// Why: fire-and-forget — callers get the persisted credential status immediately; // the rate-limit refresh runs in the background and only logs on failure. function refreshAfterMiniMaxCredentialChange( rateLimits: RateLimitService | null, @@ -49,4 +62,17 @@ export function registerMiniMaxCredentialsHandlers(rateLimits: RateLimitService refreshAfterMiniMaxCredentialChange(rateLimits, 'clear') return getMiniMaxCredentialsStatus() }) + ipcMain.handle('minimaxCredentials:saveApiKey', (_event, key: string) => { + if (typeof key !== 'string') { + throw new Error('MiniMax API key must be a string') + } + saveMiniMaxApiKey(key) + refreshAfterMiniMaxCredentialChange(rateLimits, 'save') + return getMiniMaxCredentialsStatus() + }) + ipcMain.handle('minimaxCredentials:clearApiKey', () => { + clearMiniMaxApiKey() + refreshAfterMiniMaxCredentialChange(rateLimits, 'clear') + return getMiniMaxCredentialsStatus() + }) } diff --git a/src/main/ipc/notification-burst-cooldown.ts b/src/main/ipc/notification-burst-cooldown.ts index e7616c57746..91e879a7e47 100644 --- a/src/main/ipc/notification-burst-cooldown.ts +++ b/src/main/ipc/notification-burst-cooldown.ts @@ -1,37 +1 @@ -const NOTIFICATION_COOLDOWN_MS = 5000 -const MAX_RECENT_NOTIFICATION_KEYS = 50 - -function pruneRecentNotifications(recentNotifications: Map, now: number): void { - if (recentNotifications.size <= MAX_RECENT_NOTIFICATION_KEYS) { - return - } - - for (const [key, ts] of recentNotifications) { - if (now - ts >= NOTIFICATION_COOLDOWN_MS) { - recentNotifications.delete(key) - } - } - - while (recentNotifications.size > MAX_RECENT_NOTIFICATION_KEYS) { - const oldest = recentNotifications.keys().next() - if (oldest.done) { - break - } - recentNotifications.delete(oldest.value) - } -} - -export function reserveNotificationCooldown( - recentNotifications: Map, - dedupeKey: string, - now: number -): boolean { - const lastSentAt = recentNotifications.get(dedupeKey) ?? 0 - if (now - lastSentAt < NOTIFICATION_COOLDOWN_MS) { - return false - } - recentNotifications.delete(dedupeKey) - recentNotifications.set(dedupeKey, now) - pruneRecentNotifications(recentNotifications, now) - return true -} +export { reserveNotificationCooldown } from '../../shared/notification-burst-cooldown' diff --git a/src/main/ipc/notification-options.ts b/src/main/ipc/notification-options.ts index a2553f05a3c..de05fc0c38a 100644 --- a/src/main/ipc/notification-options.ts +++ b/src/main/ipc/notification-options.ts @@ -57,12 +57,7 @@ function buildAgentTaskCompleteNotificationOptions( const agentLabel = formatNotificationAgentLabel(args.agentType) const worktreeContext = formatNotificationWorktreeContext(args) - const statusText = - args.agentState === 'blocked' || args.agentState === 'waiting' - ? 'needs input' - : args.agentState === 'done' && args.agentInterrupted - ? 'stopped' - : 'finished' + const statusText = formatAgentNotificationStatusText(args) return { title: `${worktreeContext} - ${agentLabel} ${statusText}`, @@ -70,6 +65,19 @@ function buildAgentTaskCompleteNotificationOptions( } } +// Why (#4375): a still-working agent must never be announced as finished. Only an +// explicit terminal state, or no state at all (the hook snapshot expired and the +// notification itself is the completion signal), may say "finished". +function formatAgentNotificationStatusText(args: NotificationDispatchRequest): string { + if (args.agentState === 'blocked' || args.agentState === 'waiting') { + return 'needs input' + } + if (args.agentState === 'working') { + return 'working' + } + return args.agentState === 'done' && args.agentInterrupted ? 'stopped' : 'finished' +} + function formatNotificationWorktreeContext(args: NotificationDispatchRequest): string { const worktreeLabel = normalizeNotificationText( args.worktreeLabel, diff --git a/src/main/ipc/notifications-message-formatting.test.ts b/src/main/ipc/notifications-message-formatting.test.ts index 4fcbc3e0b64..677c3131203 100644 --- a/src/main/ipc/notifications-message-formatting.test.ts +++ b/src/main/ipc/notifications-message-formatting.test.ts @@ -278,6 +278,73 @@ describe('registerNotificationHandlers', () => { expect(options.body.length).toBeLessThanOrEqual(180) }) + it.each([ + { agentState: 'working', expected: 'feat/notis - Claude working' }, + { agentState: 'blocked', expected: 'feat/notis - Claude needs input' }, + { agentState: 'waiting', expected: 'feat/notis - Claude needs input' }, + { agentState: 'done', expected: 'feat/notis - Claude finished' }, + { agentState: undefined, expected: 'feat/notis - Claude finished' } + ])('titles agentState $agentState without claiming a false finish', async (scenario) => { + registerNotificationHandlers({ + getSettings: () => ({ + notifications: { + enabled: true, + agentTaskComplete: true, + terminalBell: false, + suppressWhenFocused: true + } + }) + } as never) + + const handler = getDispatchHandler() + await handler( + {}, + { + source: 'agent-task-complete', + worktreeLabel: 'feat/notis', + agentType: 'claude', + ...(scenario.agentState ? { agentState: scenario.agentState } : {}), + agentLastAssistantMessage: 'Ran the suite.' + } + ) + + expect(notificationCtorMock).toHaveBeenCalledWith( + expectedNativeNotificationOptions({ title: scenario.expected, body: 'Ran the suite.' }) + ) + }) + + it('reports an interrupted finish as stopped', async () => { + registerNotificationHandlers({ + getSettings: () => ({ + notifications: { + enabled: true, + agentTaskComplete: true, + terminalBell: false, + suppressWhenFocused: true + } + }) + } as never) + + const handler = getDispatchHandler() + await handler( + {}, + { + source: 'agent-task-complete', + worktreeLabel: 'feat/notis', + agentType: 'claude', + agentState: 'done', + agentInterrupted: true + } + ) + + expect(notificationCtorMock).toHaveBeenCalledWith( + expectedNativeNotificationOptions({ + title: 'feat/notis - Claude stopped', + body: 'Claude stopped.' + }) + ) + }) + it('uses tool context before falling back when no prompt or assistant preview exists', async () => { registerNotificationHandlers({ getSettings: () => ({ @@ -308,7 +375,7 @@ describe('registerNotificationHandlers', () => { expect(notificationCtorMock).toHaveBeenCalledWith( expectedNativeNotificationOptions({ - title: 'feat/notis - Agent finished', + title: 'feat/notis - Agent working', body: 'Using Bash: pnpm test' }) ) diff --git a/src/main/ipc/notifications-mobile-fanout.test.ts b/src/main/ipc/notifications-mobile-fanout.test.ts index 94d2535a3cc..ab797293042 100644 --- a/src/main/ipc/notifications-mobile-fanout.test.ts +++ b/src/main/ipc/notifications-mobile-fanout.test.ts @@ -71,15 +71,17 @@ describe('registerNotificationHandlers', () => { expect(dispatchMobileNotification).toHaveBeenCalledWith({ type: 'notification', + emittedAt: expect.any(Number), source: 'agent-task-complete', title: 'feat/notis - Hermes finished', body: 'The diff updates notification formatting.', - worktreeId: 'repo::wt1' + worktreeId: 'repo::wt1', + agentState: 'done' }) expect(notificationCtorMock).not.toHaveBeenCalled() }) - it('does not dispatch mobile notifications when notifications are disabled', async () => { + it('offers disabled desktop events to independently configured phones', async () => { const dispatchMobileNotification = vi.fn() registerNotificationHandlers( { @@ -101,10 +103,12 @@ describe('registerNotificationHandlers', () => { reason: 'disabled' }) - expect(dispatchMobileNotification).not.toHaveBeenCalled() + expect(dispatchMobileNotification).toHaveBeenCalledWith( + expect.objectContaining({ desktopAllowed: false }) + ) }) - it('does not dispatch mobile notifications when the source is disabled', async () => { + it('marks a disabled desktop source for phones following desktop settings', async () => { const dispatchMobileNotification = vi.fn() registerNotificationHandlers( { @@ -126,7 +130,9 @@ describe('registerNotificationHandlers', () => { reason: 'source-disabled' }) - expect(dispatchMobileNotification).not.toHaveBeenCalled() + expect(dispatchMobileNotification).toHaveBeenCalledWith( + expect.objectContaining({ desktopAllowed: false }) + ) }) it('dispatches one mobile notification when the active worktree is focused on desktop', async () => { @@ -173,7 +179,7 @@ describe('registerNotificationHandlers', () => { expect(notificationCtorMock).not.toHaveBeenCalled() }) - it('does not dispatch mobile notifications for cooldown-suppressed bursts', async () => { + it('preserves different mobile event categories before per-phone burst suppression', async () => { const dispatchMobileNotification = vi.fn() registerNotificationHandlers( { @@ -198,7 +204,7 @@ describe('registerNotificationHandlers', () => { reason: 'cooldown' }) - expect(dispatchMobileNotification).toHaveBeenCalledTimes(1) + expect(dispatchMobileNotification).toHaveBeenCalledTimes(2) expect(dispatchMobileNotification).toHaveBeenCalledWith( expect.objectContaining({ source: 'agent-task-complete', worktreeId: 'repo::wt1' }) ) diff --git a/src/main/ipc/notifications.ts b/src/main/ipc/notifications.ts index 28f6bfd95e5..8d8098f538b 100644 --- a/src/main/ipc/notifications.ts +++ b/src/main/ipc/notifications.ts @@ -119,34 +119,43 @@ export function registerNotificationHandlers(store: Store, runtime?: OrcaRuntime } const settings = store.getSettings().notifications - if (!settings.enabled) { - return { delivered: false, reason: 'disabled' } - } - - if ( - (args.source === 'agent-task-complete' && !settings.agentTaskComplete) || - (args.source === 'terminal-bell' && !settings.terminalBell) - ) { - return { delivered: false, reason: 'source-disabled' } - } + const desktopAllowed = + settings.enabled && + (args.source !== 'agent-task-complete' || settings.agentTaskComplete) && + (args.source !== 'terminal-bell' || settings.terminalBell) const notificationOptions = buildNotificationOptions(args) // Why: desktop focus only means this computer sees the worktree; the paired phone may still need the alert. if (runtime && args.source !== 'test') { const dedupeKey = args.worktreeId ?? args.worktreeLabel ?? 'global' - if (reserveNotificationCooldown(recentMobileNotifications, dedupeKey, Date.now())) { + if ( + reserveNotificationCooldown( + recentMobileNotifications, + JSON.stringify([desktopAllowed, args.source, args.agentState, dedupeKey]), + Date.now() + ) + ) { runtime.dispatchMobileNotification({ type: 'notification', + emittedAt: Date.now(), source: args.source, + ...(!desktopAllowed ? { desktopAllowed: false } : {}), title: notificationOptions.title, body: notificationOptions.body, worktreeId: args.worktreeId, - ...(args.notificationId ? { notificationId: args.notificationId } : {}) + ...(args.notificationId ? { notificationId: args.notificationId } : {}), + // Why: background push needs the agent's real state to pick "needs input" + // vs "finished" — and to stay silent while the agent is still working. + ...(args.agentState ? { agentState: args.agentState } : {}) }) } } + if (!desktopAllowed) { + return { delivered: false, reason: settings.enabled ? 'source-disabled' : 'disabled' } + } + const browserWindow = BrowserWindow.getAllWindows().find((window) => !window.isDestroyed()) ?? null if ( diff --git a/src/main/ipc/pty/host-env/assembly.ts b/src/main/ipc/pty/host-env/assembly.ts index 908d0730221..3c65796b250 100644 --- a/src/main/ipc/pty/host-env/assembly.ts +++ b/src/main/ipc/pty/host-env/assembly.ts @@ -1,4 +1,3 @@ -import { join, delimiter } from 'node:path' import { resolveSetupAgentSequenceLaunchCommand } from '../../../../shared/setup-agent-sequencing' import { detectExplicitPiAgentKindFromCommand, @@ -10,13 +9,12 @@ import { mimoCodeHookService } from '../../../mimo/hook-service' import { agentHookServer } from '../../../agent-hooks/server' import { wslHookRelayManager } from '../../../agent-hooks/wsl-hook-relay-manager' import { piTitlebarExtensionService } from '../../../pi/titlebar-extension-service' -import { ensureLinuxTerminalOrcaCliShimDir } from '../../../cli/linux-terminal-orca-cli-shim' +import { prependOrcaCliDirToChildPath } from '../../../cli/orca-cli-child-path' import { stripLegacyTerminalShimEnv } from '../../../pty/legacy-terminal-shim-dir' -import { resolvePathEnvKey, mergePersistedWindowsPath } from '../../../pty/windows-environment-path' +import { mergePersistedWindowsPath } from '../../../pty/windows-environment-path' import { resolveCodexShellLaunchPreflightCommand } from '../../../pty/codex-shell-launch-preflight' import { buildConfiguredProxyEnv } from '../../../../shared/network-proxy' import type { BuildPtyHostEnvOptions } from './types' -import { readInheritedPath } from './path' import { stripInheritedOrcaCodexHomeOverride } from './codex-home' import { clearPiAgentShadowEnv, @@ -235,34 +233,11 @@ export function buildPtyHostEnv( } delete baseEnv.ORCA_CLI_COMMAND } - // Why: dev mode needs the launcher PATH override so `orca` resolves to the dev build instead of the production binary at /usr/local/bin/orca. - if (!opts.isPackaged) { - const devCliBin = join(opts.userDataPath, 'cli', 'bin') - const inheritedPath = readInheritedPath(baseEnv) - // Why: an empty PATH segment resolves as `.` in some shells (commands run from cwd); avoid a trailing delimiter. - baseEnv[resolvePathEnvKey(baseEnv, process.platform)] = inheritedPath - ? `${devCliBin}${delimiter}${inheritedPath}` - : devCliBin - } else if (process.platform === 'linux') { - // Why: bare-`orca` shim scoped to Orca PTYs — Linux CLI installs as `orca-ide` to avoid shadowing GNOME's /usr/bin/orca screen reader (stablyai/orca#7904). - const shimDir = ensureLinuxTerminalOrcaCliShimDir({ userDataPath: opts.userDataPath }) - if (shimDir) { - const inheritedEntries = readInheritedPath(baseEnv) - .split(delimiter) - .filter((entry) => entry.length > 0 && entry !== shimDir) - baseEnv.PATH = [shimDir, ...inheritedEntries].join(delimiter) - } - } else if ( - opts.resourcesPath && - (process.platform === 'darwin' || process.platform === 'win32') - ) { - // Why: global CLI registration is optional, but agents in Orca-managed PTYs must always reach this app's bundled CLI. - const bundledCliBin = join(opts.resourcesPath, 'bin') - const inheritedPath = readInheritedPath(baseEnv) - baseEnv[resolvePathEnvKey(baseEnv, process.platform)] = inheritedPath - ? `${bundledCliBin}${delimiter}${inheritedPath}` - : bundledCliBin - } + prependOrcaCliDirToChildPath(baseEnv, { + isPackaged: opts.isPackaged, + userDataPath: opts.userDataPath, + resourcesPath: opts.resourcesPath + }) if ( opts.routeBrowserOpensToClient === true && diff --git a/src/main/ipc/pty/host-env/path.ts b/src/main/ipc/pty/host-env/path.ts index 226ad8691fe..e9091864464 100644 --- a/src/main/ipc/pty/host-env/path.ts +++ b/src/main/ipc/pty/host-env/path.ts @@ -2,8 +2,11 @@ import { delimiter } from 'node:path' import { isLegacyTerminalShimPathEntry } from '../../../pty/legacy-terminal-shim-dir' import { resolvePathEnvKey } from '../../../pty/windows-environment-path' -export function readInheritedPath(baseEnv: Record): string { - const pathKey = resolvePathEnvKey(baseEnv, process.platform) +export function readInheritedPath( + baseEnv: Record, + platform: NodeJS.Platform = process.platform +): string { + const pathKey = resolvePathEnvKey(baseEnv, platform) return baseEnv[pathKey] ?? process.env[pathKey] ?? '' } diff --git a/src/main/ipc/worktrees-removal-recovery.test.ts b/src/main/ipc/worktrees-removal-recovery.test.ts index 49dc7b633bd..e8571449321 100644 --- a/src/main/ipc/worktrees-removal-recovery.test.ts +++ b/src/main/ipc/worktrees-removal-recovery.test.ts @@ -500,7 +500,9 @@ describe('registerWorktreeHandlers', () => { runtime: runtimeStub, resolvedWorktreeId: worktreeId, localProvider: ptyProvider, - onPtyStopped: clearProviderPtyStateMock + onPtyStopped: clearProviderPtyStateMock, + // Folder-workspace removal best-effort closes structured sessions the PTY sweeps cannot see. + closeStructuredSessions: true }) expect(killAllProcessesForWorktreeMock.mock.invocationCallOrder[0]).toBeLessThan( store.removeWorktreeMeta.mock.invocationCallOrder[0] @@ -564,7 +566,8 @@ describe('registerWorktreeHandlers', () => { localProvider: sshPtyProvider, onPtyStopped: clearProviderPtyStateMock, includeProviderInventory: true, - includeLocalRegistry: false + includeLocalRegistry: false, + closeStructuredSessions: true }) expect(store.removeWorktreeMeta).toHaveBeenCalledWith(worktreeId, 'ssh:conn-1') expect(advertisedUrlWatcherForgetWorktreeMock).not.toHaveBeenCalled() @@ -597,7 +600,8 @@ describe('registerWorktreeHandlers', () => { localProvider: runtimePtyProvider, onPtyStopped: clearProviderPtyStateMock, includeProviderInventory: false, - includeLocalRegistry: false + includeLocalRegistry: false, + closeStructuredSessions: true }) expect(getSshPtyProviderMock).not.toHaveBeenCalled() }) diff --git a/src/main/ipc/worktrees/removal/remove-folder-workspace.ts b/src/main/ipc/worktrees/removal/remove-folder-workspace.ts index c03b1449136..916f4574a92 100644 --- a/src/main/ipc/worktrees/removal/remove-folder-workspace.ts +++ b/src/main/ipc/worktrees/removal/remove-folder-workspace.ts @@ -43,6 +43,9 @@ export async function removeFolderWorkspace( : {}), localProvider: sshPtyProvider ?? getLocalPtyProvider(), onPtyStopped: clearProviderPtyState, + // A structured session is on no PTY surface, so the sweeps above leave it attached to a + // workspace this removal is about to forget. Closed best-effort, exactly like those PTYs. + closeStructuredSessions: true, ...(externalHost ? { includeProviderInventory: ownerHost?.kind === 'ssh' && Boolean(sshPtyProvider), diff --git a/src/main/minimax/minimax-api-key-store.test.ts b/src/main/minimax/minimax-api-key-store.test.ts new file mode 100644 index 00000000000..9dd3ebdea01 --- /dev/null +++ b/src/main/minimax/minimax-api-key-store.test.ts @@ -0,0 +1,184 @@ +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import type * as MiniMaxApiKeyStore from './minimax-api-key-store' + +const safeStorageMock = vi.hoisted(() => ({ + isEncryptionAvailable: vi.fn(() => true), + encryptString: vi.fn((value: string) => Buffer.from(value)), + decryptString: vi.fn((value: Buffer) => value.toString('utf8')) +})) + +const electronMock = vi.hoisted(() => ({ + safeStorage: safeStorageMock +})) + +vi.mock('electron', () => electronMock) + +const existsSyncMock = vi.fn() +const readFileSyncMock = vi.fn() +const rmSyncMock = vi.fn() +const hardenExistingSecureFileMock = vi.fn() +const writeSecureFileMock = vi.fn() +const homedirMock = vi.fn(() => '/home/test') + +vi.mock('node:fs', () => ({ + existsSync: existsSyncMock, + readFileSync: readFileSyncMock, + rmSync: rmSyncMock +})) + +vi.mock('node:os', () => ({ + homedir: homedirMock +})) + +vi.mock('node:path', () => ({ + join: (...parts: string[]) => parts.join('/') +})) + +vi.mock('../../shared/secure-file', () => ({ + hardenExistingSecureFile: hardenExistingSecureFileMock, + writeSecureFile: writeSecureFileMock +})) + +const storePath = '/home/test/.orca/minimax-api-key.enc' +const envelope = (kind: 'encrypted' | 'plaintext', value: string): string => + `orca-minimax-api-key:v1:${kind}:${Buffer.from(value, 'utf8').toString('base64')}` + +async function loadStore(): Promise { + return await import('./minimax-api-key-store') +} + +describe('minimax-api-key-store', () => { + beforeEach(() => { + existsSyncMock.mockReset() + readFileSyncMock.mockReset() + rmSyncMock.mockReset() + hardenExistingSecureFileMock.mockReset() + writeSecureFileMock.mockReset() + safeStorageMock.isEncryptionAvailable.mockReset() + safeStorageMock.encryptString.mockReset() + safeStorageMock.decryptString.mockReset() + safeStorageMock.isEncryptionAvailable.mockReturnValue(true) + safeStorageMock.encryptString.mockImplementation((value: string) => Buffer.from(value)) + safeStorageMock.decryptString.mockImplementation((value: Buffer) => value.toString('utf8')) + }) + + afterEach(() => { + vi.resetModules() + }) + + it('returns false when no file exists yet', async () => { + existsSyncMock.mockReturnValue(false) + const store = await loadStore() + expect(store.hasMiniMaxApiKey()).toBe(false) + expect(hardenExistingSecureFileMock).not.toHaveBeenCalled() + }) + + it('hardens the key file when checking status for an existing key', async () => { + existsSyncMock.mockReturnValue(true) + const store = await loadStore() + expect(store.hasMiniMaxApiKey()).toBe(true) + expect(hardenExistingSecureFileMock).toHaveBeenCalledWith(storePath) + }) + + it('still reports an existing key when status-path hardening fails', async () => { + const warn = vi.spyOn(console, 'warn').mockImplementation(() => undefined) + existsSyncMock.mockReturnValue(true) + hardenExistingSecureFileMock.mockImplementation(() => { + throw new Error('permission denied') + }) + const store = await loadStore() + expect(store.hasMiniMaxApiKey()).toBe(true) + expect(warn).toHaveBeenCalledWith( + expect.stringContaining('Failed to harden MiniMax API key file'), + expect.any(Error) + ) + warn.mockRestore() + }) + + it('writes the key using safeStorage when encryption is available', async () => { + existsSyncMock.mockReturnValue(false) + const store = await loadStore() + store.saveMiniMaxApiKey('sk-test-1234567890') + expect(safeStorageMock.encryptString).toHaveBeenCalledWith('sk-test-1234567890') + expect(writeSecureFileMock).toHaveBeenCalledWith( + storePath, + envelope('encrypted', 'sk-test-1234567890') + ) + }) + + it('warns and writes plaintext when safeStorage is unavailable', async () => { + const warn = vi.spyOn(console, 'warn').mockImplementation(() => undefined) + safeStorageMock.isEncryptionAvailable.mockReturnValue(false) + existsSyncMock.mockReturnValue(false) + const store = await loadStore() + store.saveMiniMaxApiKey('sk-test-1234567890') + expect(writeSecureFileMock).toHaveBeenCalledWith( + storePath, + envelope('plaintext', 'sk-test-1234567890') + ) + expect(warn).toHaveBeenCalledWith(expect.stringContaining('safeStorage encryption unavailable')) + warn.mockRestore() + }) + + it('refuses empty keys', async () => { + const store = await loadStore() + expect(() => store.saveMiniMaxApiKey(' ')).toThrow(/required/) + }) + + it('reads decrypted key from disk and caches it', async () => { + existsSyncMock.mockReturnValue(true) + readFileSyncMock.mockReturnValue(Buffer.from(envelope('encrypted', 'encrypted-payload'))) + safeStorageMock.decryptString.mockReturnValue('sk-cached-key') + const store = await loadStore() + const first = store.readMiniMaxApiKey() + const second = store.readMiniMaxApiKey() + expect(first).toBe('sk-cached-key') + expect(second).toBe(first) + expect(hardenExistingSecureFileMock).toHaveBeenCalledTimes(1) + expect(hardenExistingSecureFileMock).toHaveBeenCalledWith(storePath) + expect(safeStorageMock.decryptString).toHaveBeenCalledTimes(1) + expect(safeStorageMock.decryptString).toHaveBeenCalledWith(Buffer.from('encrypted-payload')) + }) + + it('returns null when no file exists', async () => { + existsSyncMock.mockReturnValue(false) + const store = await loadStore() + expect(store.readMiniMaxApiKey()).toBeNull() + }) + + it('throws for encrypted envelopes when safeStorage is unavailable', async () => { + safeStorageMock.isEncryptionAvailable.mockReturnValue(false) + existsSyncMock.mockReturnValue(true) + readFileSyncMock.mockReturnValue(Buffer.from(envelope('encrypted', 'encrypted-payload'))) + const store = await loadStore() + expect(() => store.readMiniMaxApiKey()).toThrow(/could not be decrypted/) + }) + + it('throws when decryption fails', async () => { + existsSyncMock.mockReturnValue(true) + readFileSyncMock.mockReturnValue(Buffer.from(envelope('encrypted', 'encrypted-payload'))) + safeStorageMock.decryptString.mockImplementation(() => { + throw new Error('boom') + }) + const store = await loadStore() + expect(() => store.readMiniMaxApiKey()).toThrow(/could not be decrypted/) + }) + + it('throws for non-envelope files (legacy safeStorage bytes with no prefix)', async () => { + existsSyncMock.mockReturnValue(true) + readFileSyncMock.mockReturnValue(Buffer.from('raw-bytes-without-envelope')) + const store = await loadStore() + expect(() => store.readMiniMaxApiKey()).toThrow(/could not be decrypted/) + }) + + it('clears the cached key and removes the file', async () => { + existsSyncMock.mockReturnValueOnce(true) + readFileSyncMock.mockReturnValueOnce(Buffer.from(envelope('encrypted', 'encrypted-payload'))) + safeStorageMock.decryptString.mockReturnValueOnce('sk-preclear') + const store = await loadStore() + expect(store.readMiniMaxApiKey()).toBe('sk-preclear') + store.clearMiniMaxApiKey() + expect(rmSyncMock).toHaveBeenCalledWith(storePath, { force: true }) + expect(store.readMiniMaxApiKey()).toBeNull() + }) +}) diff --git a/src/main/minimax/minimax-api-key-store.ts b/src/main/minimax/minimax-api-key-store.ts new file mode 100644 index 00000000000..efd0af65db9 --- /dev/null +++ b/src/main/minimax/minimax-api-key-store.ts @@ -0,0 +1,127 @@ +import { safeStorage } from 'electron' +import { existsSync, readFileSync, rmSync } from 'node:fs' +import { homedir } from 'node:os' +import { join } from 'node:path' +import { hardenExistingSecureFile, writeSecureFile } from '../../shared/secure-file' + +const MINIMAX_API_KEY_FILE = 'minimax-api-key.enc' +const API_KEY_ENVELOPE_PREFIX = 'orca-minimax-api-key:v1:' +let cachedMiniMaxApiKey: string | null = null +let warnedMiniMaxApiKeyStatusHardenFailure = false + +type MiniMaxApiKeyEnvelope = { + kind: 'encrypted' | 'plaintext' + payload: Buffer +} + +function getOrcaDir(): string { + return join(homedir(), '.orca') +} + +function getMiniMaxApiKeyPath(): string { + return join(getOrcaDir(), MINIMAX_API_KEY_FILE) +} + +function encodeApiKeyEnvelope(kind: MiniMaxApiKeyEnvelope['kind'], payload: Buffer): string { + return `${API_KEY_ENVELOPE_PREFIX}${kind}:${payload.toString('base64')}` +} + +function decodeApiKeyEnvelope(raw: Buffer): MiniMaxApiKeyEnvelope { + const text = raw.toString('utf8') + if (!text.startsWith(API_KEY_ENVELOPE_PREFIX)) { + throw new Error('MiniMax API key could not be decrypted') + } + const rest = text.slice(API_KEY_ENVELOPE_PREFIX.length) + const separator = rest.indexOf(':') + if (separator === -1) { + throw new Error('MiniMax API key could not be decrypted') + } + const kind = rest.slice(0, separator) + if (kind !== 'encrypted' && kind !== 'plaintext') { + throw new Error('MiniMax API key could not be decrypted') + } + return { + kind, + payload: Buffer.from(rest.slice(separator + 1), 'base64') + } +} + +function readEnvelope(envelope: MiniMaxApiKeyEnvelope): string { + if (envelope.kind === 'plaintext') { + return envelope.payload.toString('utf8') + } + if (!safeStorage.isEncryptionAvailable()) { + throw new Error('MiniMax API key could not be decrypted') + } + return safeStorage.decryptString(envelope.payload) +} + +export function hasMiniMaxApiKey(): boolean { + const keyPath = getMiniMaxApiKeyPath() + if (!existsSync(keyPath)) { + return false + } + try { + hardenExistingSecureFile(keyPath) + } catch (error) { + if (!warnedMiniMaxApiKeyStatusHardenFailure) { + warnedMiniMaxApiKeyStatusHardenFailure = true + console.warn('[minimax] Failed to harden MiniMax API key file while checking status', error) + } + } + return true +} + +export function saveMiniMaxApiKey(key: string): void { + const trimmed = key.trim() + if (!trimmed) { + throw new Error('MiniMax API key is required') + } + if (safeStorage.isEncryptionAvailable()) { + writeSecureFile( + getMiniMaxApiKeyPath(), + encodeApiKeyEnvelope('encrypted', safeStorage.encryptString(trimmed)) + ) + cachedMiniMaxApiKey = trimmed + return + } + console.warn( + '[minimax] safeStorage encryption unavailable — storing MiniMax API key in plaintext' + ) + writeSecureFile( + getMiniMaxApiKeyPath(), + encodeApiKeyEnvelope('plaintext', Buffer.from(trimmed, 'utf8')) + ) + cachedMiniMaxApiKey = trimmed +} + +export function readMiniMaxApiKey(): string | null { + if (cachedMiniMaxApiKey !== null) { + return cachedMiniMaxApiKey + } + const keyPath = getMiniMaxApiKeyPath() + if (!existsSync(keyPath)) { + return null + } + // Why: keep hardening out of the decode/decrypt try below so a chmod/ACL + // failure isn't misreported as a decrypt failure (matches hasMiniMaxApiKey). + try { + hardenExistingSecureFile(keyPath) + } catch (error) { + console.warn('[minimax] Failed to harden MiniMax API key file while reading', error) + } + try { + const raw = readFileSync(keyPath) + const envelope = decodeApiKeyEnvelope(raw) + cachedMiniMaxApiKey = readEnvelope(envelope) + return cachedMiniMaxApiKey + } catch (error) { + console.error('[minimax] failed to decode/decrypt API key', error) + throw new Error('MiniMax API key could not be decrypted') + } +} + +export function clearMiniMaxApiKey(): void { + cachedMiniMaxApiKey = null + rmSync(getMiniMaxApiKeyPath(), { force: true }) +} diff --git a/src/main/native-chat/agent-session-journal/journal-reducer.test.ts b/src/main/native-chat/agent-session-journal/journal-reducer.test.ts index 676a37f63a7..2dc0ac1aeb2 100644 --- a/src/main/native-chat/agent-session-journal/journal-reducer.test.ts +++ b/src/main/native-chat/agent-session-journal/journal-reducer.test.ts @@ -207,11 +207,46 @@ describe('submission and dispatch state machine', () => { const items = renderJournalState(state).items expect(items).toHaveLength(1) expect(items[0]?.itemId).toBe(agentJournalSubmissionKey('cm_1')) - // The echo updates content in place; the bubble keeps its original slot. + // The echo advances the revision; the submitted bubble keeps its original slot. expect(items[0]?.sequence).toBe(1) expect(items[0]?.revision).toBe(1) }) + it.each(['codex:thread-1:turn-1:0', 'claude:session-1:user-1'])( + 'preserves submitted text and attachments when %s is restored', + (providerItemId) => { + const body: AgentJournalMessageItem = { + kind: 'message', + role: 'user', + blocks: [ + { type: 'text', text: '/example-skill inspect this' }, + { type: 'image-ref', path: '/tmp/original.png' } + ] + } + const state = fold([ + { ...submission, body, payloadFingerprint: sendFingerprint(body) }, + { + kind: 'dispatch', + clientMessageId: 'cm_1', + state: 'accepted', + providerItemId, + reason: null, + ...base(2) + }, + { + kind: 'item', + itemId: providerItemId, + revision: 1, + body: userText('# Expanded skill instructions'), + ...base(3) + } + ]) + expect(renderJournalState(state).items).toEqual([ + expect.objectContaining({ itemId: agentJournalSubmissionKey('cm_1'), body, revision: 1 }) + ]) + } + ) + it('adopts a provider echo that arrives before dispatch settles', () => { const body = userText('early echo') const state = fold([ diff --git a/src/main/native-chat/agent-session-journal/journal-reducer.ts b/src/main/native-chat/agent-session-journal/journal-reducer.ts index 3dc4385d797..41625792aa0 100644 --- a/src/main/native-chat/agent-session-journal/journal-reducer.ts +++ b/src/main/native-chat/agent-session-journal/journal-reducer.ts @@ -26,6 +26,7 @@ export type JournalReducerState = { sessionId: string epoch: string lastSequence: number + lastActivityAt: number /** Lowest sequence still individually replayable; rows below it were compacted. */ oldestSequence: number highestFence: number @@ -45,6 +46,7 @@ export function createJournalReducerState(sessionId: string, epoch: string): Jou sessionId, epoch, lastSequence: 0, + lastActivityAt: 0, oldestSequence: 1, highestFence: 0, items: new Map(), @@ -62,6 +64,7 @@ export function applyJournalRow(state: JournalReducerState, row: JournalRow): vo if (row.kind === 'epoch') { return } + state.lastActivityAt = Math.max(state.lastActivityAt, row.ts) if (row.kind === 'item') { const itemId = resolveJournalItemId(state, row.itemId, row.body) upsertItem(state, itemId, row.revision, { @@ -190,7 +193,17 @@ function upsertItem( // so letting a revision advance it makes the row jump past everything that // landed in between — the provider's own echo of a send revises the submission // row, which relocated the user's bubble below later rows. - state.items.set(itemId, { ...next, sequence: existing.sequence, observedAt: existing.observedAt }) + const submitted = + existing.body.kind === 'message' && + existing.body.role === 'user' && + parseAgentJournalItemKey(itemId)?.provider === 'orca' + state.items.set(itemId, { + ...next, + // Provider history may normalize text or omit local attachments from the original send. + body: submitted ? existing.body : next.body, + sequence: existing.sequence, + observedAt: existing.observedAt + }) state.tombstones.delete(itemId) } diff --git a/src/main/native-chat/agent-session-journal/journal-store.ts b/src/main/native-chat/agent-session-journal/journal-store.ts index ab2715d0d86..e2936b2553d 100644 --- a/src/main/native-chat/agent-session-journal/journal-store.ts +++ b/src/main/native-chat/agent-session-journal/journal-store.ts @@ -162,6 +162,9 @@ export class AgentSessionJournal { snapshot = (): AgentJournalSnapshot => renderJournalState(this.state) + /** Includes revisions and completion tombstones, whose timestamps disappear from render items. */ + lastActivityAt = (): number => this.state.lastActivityAt + submissions = (): AgentJournalSubmission[] => [...this.state.submissions.values()] pendingSubmissions = (): AgentJournalSubmission[] => diff --git a/src/main/native-chat/agent-session-wire/provider-frame-activity.test.ts b/src/main/native-chat/agent-session-wire/provider-frame-activity.test.ts new file mode 100644 index 00000000000..79d9e4205cf --- /dev/null +++ b/src/main/native-chat/agent-session-wire/provider-frame-activity.test.ts @@ -0,0 +1,84 @@ +import { describe, expect, it } from 'vitest' +import { + MAX_PROVIDER_ACTIVITY_LENGTH, + claudeProviderFrameActivity, + codexProviderFrameActivity, + providerActivityText +} from './provider-frame-activity' + +describe('provider frame activity', () => { + it('derives bounded Codex activity without exposing item payloads or opcodes', () => { + expect( + codexProviderFrameActivity('item/started', { + item: { type: 'commandExecution', command: 'printenv SECRET_TOKEN' } + }) + ).toBe('Running a command') + expect( + codexProviderFrameActivity('item/mcpToolCall/progress', { + message: '**Indexing repository symbols**' + }) + ).toBe('Indexing repository symbols') + expect( + codexProviderFrameActivity( + 'item/reasoning/summaryTextDelta', + { delta: 'ignored-fragment' }, + 'Inspecting the session wire' + ) + ).toBe('Inspecting the session wire') + expect(codexProviderFrameActivity('item/reasoning/summaryPartAdded', {})).toBeNull() + }) + + it('uses Claude descriptions and safe semantic status without exposing tool labels', () => { + expect( + claudeProviderFrameActivity('message:system:task_started', { + description: 'Trace the activity channel' + }) + ).toBe('Working on: Trace the activity channel') + expect( + claudeProviderFrameActivity('message:system:task_progress', { + description: 'Reading tests', + summary: 'Checking remote compatibility' + }) + ).toBe('Checking remote compatibility') + expect( + claudeProviderFrameActivity('message:system:task_updated', { + patch: { description: 'Validating the renderer' } + }) + ).toBe('Validating the renderer') + expect(claudeProviderFrameActivity('message:system:status', { status: 'compacting' })).toBe( + 'Compacting the conversation' + ) + expect( + claudeProviderFrameActivity('message:system:control_request_progress', { + status: 'api_retry' + }) + ).toBe('Retrying a side question') + expect( + claudeProviderFrameActivity('message:tool_progress', { + tool_name: 'ReadSecretFile' + }) + ).toBeNull() + }) + + it('falls through on protocol noise and bounds long copy', () => { + expect(providerActivityText('codex · notification:warning')).toBeNull() + expect(providerActivityText('item/reasoning/summaryPartAdded')).toBeNull() + expect(providerActivityText('{"file":"contents"}')).toBeNull() + const bounded = providerActivityText(`Reviewing ${'long '.repeat(100)}`) + expect(Array.from(bounded ?? '').length).toBeLessThanOrEqual(MAX_PROVIDER_ACTIVITY_LENGTH) + expect(bounded?.endsWith('…')).toBe(true) + }) + + it('keeps only the reasoning headline and waits for an unterminated bold header', () => { + expect( + codexProviderFrameActivity( + 'item/reasoning/summaryTextDelta', + {}, + '**Inspecting the workspace**\n\nI am looking at notes.txt before answering.' + ) + ).toBe('Inspecting the workspace') + expect( + codexProviderFrameActivity('item/reasoning/summaryTextDelta', {}, '**Inspecting the wor') + ).toBeUndefined() + }) +}) diff --git a/src/main/native-chat/agent-session-wire/provider-frame-activity.ts b/src/main/native-chat/agent-session-wire/provider-frame-activity.ts new file mode 100644 index 00000000000..336170a4cfe --- /dev/null +++ b/src/main/native-chat/agent-session-wire/provider-frame-activity.ts @@ -0,0 +1,188 @@ +import { normalizeOptionalField } from '../../../shared/agent-status-field-normalization' + +export const MAX_PROVIDER_ACTIVITY_LENGTH = 160 + +type ActivityText = string | null | undefined + +function record(value: unknown): Record | null { + return typeof value === 'object' && value !== null && !Array.isArray(value) + ? (value as Record) + : null +} + +function stringField(source: Record | null, key: string): string | null { + const value = source?.[key] + return typeof value === 'string' && value.trim() ? value : null +} + +/** A reasoning summary streams as a bold headline plus body; only the headline is activity copy. */ +function reasoningHeadline(text: string | null | undefined): ActivityText { + const line = text?.split(/\r?\n/).find((candidate) => candidate.trim()) + if (!line) { + return null + } + // Hold the previous copy until the closing marker streams in; a half headline would flicker. + return /^\s*\*\*/.test(line) && !/\*\*.+\*\*/.test(line) ? undefined : line +} + +/** Keep only a short sentence-shaped preview from provider-declared display fields. */ +export function providerActivityText(value: unknown): string | null { + const normalized = normalizeOptionalField(value, MAX_PROVIDER_ACTIVITY_LENGTH + 1) + if (!normalized) { + return null + } + const unwrapped = normalized + .replace(/^(?:#{1,6}|[-+])\s+/, '') + .replace(/^\*\*(.+)\*\*$/, '$1') + .replace(/^`(.+)`$/, '$1') + .trim() + if ( + !unwrapped || + /^[{[]/.test(unwrapped) || + /^[\w.-]+\s*[·-]\s*(?:notification:|message:|item\/)/i.test(unwrapped) || + (/^[\w:./-]+$/.test(unwrapped) && /[:/]/.test(unwrapped)) || + !/\p{L}/u.test(unwrapped) + ) { + return null + } + const characters = Array.from(unwrapped) + if (characters.length <= MAX_PROVIDER_ACTIVITY_LENGTH) { + return unwrapped + } + const head = characters.slice(0, MAX_PROVIDER_ACTIVITY_LENGTH - 1).join('') + const boundary = head.lastIndexOf(' ') + const clipped = boundary >= MAX_PROVIDER_ACTIVITY_LENGTH * 0.6 ? head.slice(0, boundary) : head + return `${clipped.trimEnd()}…` +} + +const CODEX_ITEM_ACTIVITY: Readonly> = { + agentMessage: 'Drafting a response', + plan: 'Updating the plan', + reasoning: 'Thinking through the request', + commandExecution: 'Running a command', + fileChange: 'Editing files', + mcpToolCall: 'Using an external tool', + dynamicToolCall: 'Using an external tool', + functionCallOutput: 'Reviewing tool results', + collabAgentToolCall: 'Coordinating with another agent', + subAgentActivity: 'Coordinating with another agent', + webSearch: 'Searching the web', + imageView: 'Inspecting an image', + imageGeneration: 'Generating an image', + enteredReviewMode: 'Reviewing changes', + exitedReviewMode: 'Reviewing changes', + contextCompaction: 'Compacting the conversation', + sleep: 'Waiting briefly', + hookPrompt: 'Processing workspace guidance' +} + +export function codexProviderFrameActivity( + method: string, + payload: unknown, + reasoningText?: string | null +): ActivityText { + const source = record(payload) + if (method === 'item/mcpToolCall/progress') { + return providerActivityText(stringField(source, 'message')) + } + if (method === 'item/reasoning/summaryTextDelta') { + const headline = reasoningHeadline(reasoningText) + return headline === undefined ? undefined : providerActivityText(headline) + } + if (method === 'item/reasoning/summaryPartAdded') { + return null + } + if (method !== 'item/started') { + return undefined + } + const item = record(source?.item) + const itemType = stringField(item, 'type') + return itemType ? (CODEX_ITEM_ACTIVITY[itemType] ?? null) : null +} + +export function claudeProviderFrameActivity(kind: string, payload: unknown): ActivityText { + const source = record(payload) + if (kind === 'message:system:task_started') { + if (source?.ambient === true || source?.skip_transcript === true) { + return null + } + const description = providerActivityText(stringField(source, 'description')) + return description ? providerActivityText(`Working on: ${description}`) : null + } + if (kind === 'message:system:task_progress') { + return providerActivityText( + stringField(source, 'summary') ?? stringField(source, 'description') + ) + } + if (kind === 'message:system:task_updated') { + return providerActivityText(stringField(record(source?.patch), 'description')) + } + if (kind === 'message:system:status') { + const status = stringField(source, 'status') + return status === 'compacting' + ? 'Compacting the conversation' + : status === 'requesting' + ? 'Requesting a response' + : null + } + if (kind === 'message:system:control_request_progress') { + const status = stringField(source, 'status') + return status === 'started' + ? 'Exploring a side question' + : status === 'api_retry' + ? 'Retrying a side question' + : null + } + if (kind === 'message:tool_progress') { + return null + } + return undefined +} + +/** Retain only the current summary headline, never materialize the growing transcript. */ +export function createCodexProviderActivityReader(): ( + method: string, + payload: unknown +) => ActivityText { + let itemId: unknown + let summaryIndex: unknown + let headline = '' + let complete = false + const limit = MAX_PROVIDER_ACTIVITY_LENGTH * 2 + 16 + return (method, payload) => { + if ( + method !== 'item/reasoning/summaryTextDelta' && + method !== 'item/reasoning/summaryPartAdded' + ) { + return codexProviderFrameActivity(method, payload) + } + const source = record(payload) + if (!stringField(source, 'itemId')) { + return undefined + } + if ( + source?.itemId !== itemId || + source?.summaryIndex !== summaryIndex || + method === 'item/reasoning/summaryPartAdded' + ) { + itemId = source?.itemId + summaryIndex = source?.summaryIndex + headline = '' + complete = false + } + if (method === 'item/reasoning/summaryPartAdded') { + return null + } + if (complete || typeof source?.delta !== 'string') { + return undefined + } + headline += source.delta.slice(0, limit - headline.length) + const line = headline.trimStart().split(/\r?\n/, 1)[0] + complete = + headline.length === limit || /\r?\n/.test(headline.trimStart()) || /^\*\*.+\*\*/.test(line) + if (complete && line.startsWith('**') && !/\*\*.+\*\*/.test(line)) { + return providerActivityText(line.slice(2)) + } + return codexProviderFrameActivity(method, payload, line) + } +} diff --git a/src/main/native-chat/agent-session-wire/provider-turn-activity-routing.test.ts b/src/main/native-chat/agent-session-wire/provider-turn-activity-routing.test.ts new file mode 100644 index 00000000000..a66ac567a4d --- /dev/null +++ b/src/main/native-chat/agent-session-wire/provider-turn-activity-routing.test.ts @@ -0,0 +1,267 @@ +import { describe, expect, it, vi } from 'vitest' +import type { + AgentJournalItemBody, + AgentJournalItemIdentity +} from '../../../shared/agent-session-journal-types' +import type { AgentSessionTurnActivity } from '../../../shared/agent-session-wire' +import { createClaudeJournalTranslator } from '../../claude/claude-structured-journal-translation' +import { createCodexJournalTranslator } from '../../codex/codex-structured-journal-translation' +import type { CodexStructuredSessionEvent } from '../../codex/codex-structured-session-state' +import * as deltaCoalescer from './agent-session-delta-coalescer' +import type { StructuredAgentSessionEventSink } from './structured-agent-session-event-sink' + +const SESSION_ID = 'session-1' +const THREAD_ID = 'thread-1' +const TURN_ID = 'turn-1' + +function recordingSink() { + const rows: AgentJournalItemBody[] = [] + const tombstones: AgentJournalItemIdentity[] = [] + const activities: (AgentSessionTurnActivity | null)[] = [] + const sink: StructuredAgentSessionEventSink = { + appendItem: (_identity, body) => rows.push(body), + appendTombstone: (identity) => tombstones.push(identity), + publish: vi.fn(), + setActivity: (activity) => activities.push(activity) + } + return { sink, rows, tombstones, activities } +} + +function codexNotification(method: string, params: unknown): CodexStructuredSessionEvent { + return { type: 'notification', sessionId: SESSION_ID, threadId: THREAD_ID, method, params } +} + +function claudeMessage(message: Record) { + return { type: 'message' as const, sessionId: SESSION_ID, message } +} + +describe('provider turn activity routing', () => { + it('routes Codex activity without creating protocol rows', () => { + const state = recordingSink() + const translator = createCodexJournalTranslator({ + sink: state.sink, + primaryThreadId: () => THREAD_ID, + schedule: () => () => {} + }) + translator.handle(codexNotification('turn/started', { turn: { id: TURN_ID } })) + const lifecycleRows = state.rows.length + + translator.handle( + codexNotification('item/mcpToolCall/progress', { + turnId: TURN_ID, + itemId: 'mcp-1', + message: 'Reading the issue context' + }) + ) + expect(state.rows).toHaveLength(lifecycleRows) + expect(state.activities.at(-1)).toEqual({ + turnId: TURN_ID, + text: 'Reading the issue context' + }) + + translator.handle( + codexNotification('item/started', { + turnId: TURN_ID, + item: { type: 'reasoning', id: 'reasoning-1', summary: [], content: [] } + }) + ) + expect(state.rows).toHaveLength(lifecycleRows) + expect(state.activities.at(-1)?.text).toBe('Thinking through the request') + + translator.handle( + codexNotification('item/reasoning/summaryPartAdded', { + turnId: TURN_ID, + itemId: 'reasoning-1', + summaryIndex: 0 + }) + ) + expect(state.activities.at(-1)).toBeNull() + translator.handle( + codexNotification('item/reasoning/summaryTextDelta', { + turnId: TURN_ID, + itemId: 'reasoning-1', + summaryIndex: 0, + delta: 'Tracing the activity pipeline' + }) + ) + expect(state.rows).toHaveLength(lifecycleRows) + expect(state.activities.at(-1)?.text).toBe('Tracing the activity pipeline') + }) + + it('does not materialize full stream snapshots for activity on token deltas', () => { + const original = deltaCoalescer.createAgentSessionDeltaCoalescer + const snapshot = vi.fn() + const factory = vi + .spyOn(deltaCoalescer, 'createAgentSessionDeltaCoalescer') + .mockImplementation((deps) => { + const coalescer = original(deps) + return { + ...coalescer, + snapshot: (key) => { + snapshot() + return coalescer.snapshot(key) + } + } + }) + try { + const state = recordingSink() + const translator = createCodexJournalTranslator({ + sink: state.sink, + primaryThreadId: () => THREAD_ID, + schedule: () => () => {} + }) + translator.handle(codexNotification('turn/started', { turn: { id: TURN_ID } })) + for (const method of [ + 'item/agentMessage/delta', + 'item/commandExecution/outputDelta', + 'item/reasoning/summaryTextDelta' + ]) { + for (let index = 0; index < 100; index++) { + translator.handle( + codexNotification(method, { + turnId: TURN_ID, + itemId: method, + summaryIndex: 0, + delta: index === 0 ? '**Inspecting**\n' : 'more output' + }) + ) + } + } + expect(snapshot).not.toHaveBeenCalled() + translator.dispose() + } finally { + factory.mockRestore() + } + }) + + it('uses the newest summary part and stops republishing its body', () => { + const state = recordingSink() + const translator = createCodexJournalTranslator({ + sink: state.sink, + primaryThreadId: () => THREAD_ID, + schedule: () => () => {} + }) + translator.handle(codexNotification('turn/started', { turn: { id: TURN_ID } })) + const params = { turnId: TURN_ID, itemId: 'reasoning-1' } + for (const [summaryIndex, headline] of ['First headline', 'Newest headline'].entries()) { + translator.handle( + codexNotification('item/reasoning/summaryPartAdded', { ...params, summaryIndex }) + ) + translator.handle( + codexNotification('item/reasoning/summaryTextDelta', { + ...params, + summaryIndex, + delta: `**${headline}` + }) + ) + expect(state.activities.at(-1)).toBeNull() + translator.handle( + codexNotification('item/reasoning/summaryTextDelta', { + ...params, + summaryIndex, + delta: '**\n\nBody' + }) + ) + expect(state.activities.at(-1)?.text).toBe(headline) + } + const publications = state.activities.length + for (let index = 0; index < 100; index++) { + translator.handle( + codexNotification('item/reasoning/summaryTextDelta', { + ...params, + summaryIndex: 1, + delta: ' more body' + }) + ) + } + expect(state.activities).toHaveLength(publications) + translator.handle( + codexNotification('turn/completed', { turn: { id: TURN_ID, status: 'completed' } }) + ) + translator.handle(codexNotification('turn/started', { turn: { id: 'turn-2' } })) + translator.handle( + codexNotification('item/reasoning/summaryTextDelta', { + ...params, + turnId: 'turn-2', + summaryIndex: 1, + delta: '**Next turn**' + }) + ) + expect(state.activities.at(-1)).toEqual({ turnId: 'turn-2', text: 'Next turn' }) + translator.dispose() + }) + + it('keeps Codex tool rows singular and the activity free of tool labels', () => { + const state = recordingSink() + const translator = createCodexJournalTranslator({ + sink: state.sink, + primaryThreadId: () => THREAD_ID + }) + translator.handle(codexNotification('turn/started', { turn: { id: TURN_ID } })) + const lifecycleRows = state.rows.length + translator.handle( + codexNotification('item/started', { + turnId: TURN_ID, + item: { + type: 'commandExecution', + id: 'command-1', + command: 'pnpm test', + status: 'inProgress' + } + }) + ) + + expect(state.rows).toHaveLength(lifecycleRows + 1) + expect(state.rows.at(-1)).toMatchObject({ kind: 'tool-call', name: 'shell' }) + expect(state.activities.at(-1)).toEqual({ turnId: TURN_ID, text: 'Running a command' }) + expect(state.activities.at(-1)?.text).not.toContain('pnpm test') + }) + + it('routes Claude status frames without creating timeline rows and clears on settlement', () => { + const state = recordingSink() + const translator = createClaudeJournalTranslator({ sink: state.sink }) + translator.handle({ + ...claudeMessage({ + type: 'user', + uuid: TURN_ID, + session_id: 'claude-session', + parent_tool_use_id: null, + message: { role: 'user', content: [{ type: 'text', text: 'Investigate activity' }] } + }), + startsTurn: true + }) + const turnRows = state.rows.length + + translator.handle( + claudeMessage({ + type: 'system', + subtype: 'task_progress', + summary: 'Checking the renderer state' + }) + ) + translator.handle(claudeMessage({ type: 'system', subtype: 'status', status: 'compacting' })) + translator.handle( + claudeMessage({ + type: 'system', + subtype: 'control_request_progress', + status: 'started' + }) + ) + expect(state.rows).toHaveLength(turnRows) + expect(state.activities.slice(-3)).toEqual([ + { turnId: TURN_ID, text: 'Checking the renderer state' }, + { turnId: TURN_ID, text: 'Compacting the conversation' }, + { turnId: TURN_ID, text: 'Exploring a side question' } + ]) + + translator.handle(claudeMessage({ type: 'tool_progress', tool_name: 'SecretReader' })) + expect(state.rows).toHaveLength(turnRows) + expect(state.activities.at(-1)).toBeNull() + + translator.handle( + claudeMessage({ type: 'result', subtype: 'success', is_error: false, result: 'Done' }) + ) + expect(state.activities.at(-1)).toBeNull() + expect(state.tombstones).toHaveLength(1) + }) +}) diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-adapter-router.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-adapter-router.ts index 226b9c1aab5..2ac8a5570f0 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-adapter-router.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-adapter-router.ts @@ -60,6 +60,9 @@ export class StructuredAgentSessionAdapterRouter implements StructuredAgentSessi sessionId ) => this.owners.get(sessionId)?.backgroundTaskState?.(sessionId) + readCommands: NonNullable = (sessionId) => + this.owners.get(sessionId)?.readCommands?.(sessionId) + answerPrompt: StructuredAgentSessionAdapter['answerPrompt'] = (input) => this.owner(input.sessionId).answerPrompt(input) diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-adapter.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-adapter.ts index e6f8e478695..4872426d009 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-adapter.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-adapter.ts @@ -19,6 +19,7 @@ import type { import type { AgentSessionBackgroundTaskState, AgentSessionOptionsResult, + AgentSessionSlashCommand, AgentSessionWireRefusalCode } from '../../../shared/agent-session-wire' import type { StructuredAgentSessionEventSink } from './structured-agent-session-event-sink' @@ -143,6 +144,9 @@ export type StructuredAgentSessionAdapter = { taskId?: string }): Promise<{ cancelled: boolean }> backgroundTaskState?(sessionId: string): AgentSessionBackgroundTaskState | null | undefined + /** The `/` surface the running provider reports for itself. Undefined when the + * provider never reports one, which is what keeps the client on its catalog. */ + readCommands?(sessionId: string): AgentSessionSlashCommand[] | undefined /** Fires the provider callback for an approval or a question. The wire calls * this only after the durable compare-and-set won, so it runs exactly once. */ answerPrompt(input: { diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-attach-context.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-attach-context.ts index 7113be8d54b..412aa88025d 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-attach-context.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-attach-context.ts @@ -3,7 +3,10 @@ // Passing the host itself would let this quietly grow new dependencies; an explicit context makes // each one a deliberate addition and keeps the orchestration testable without constructing a host. -import type { AgentSessionWireRefusal } from '../../../shared/agent-session-wire' +import type { + AgentSessionTurnActivity, + AgentSessionWireRefusal +} from '../../../shared/agent-session-wire' import type { AgentJournalResetReason } from '../../../shared/agent-session-journal-types' import type { AgentSessionJournal } from '../agent-session-journal/journal-store' import type { @@ -25,7 +28,11 @@ export type StructuredAgentSessionAttachContext = { fence: number ) => void snapshot: (sessionId: string, journal: AgentSessionJournal, fence: number) => void - publish: (sessionId: string, journal: AgentSessionJournal) => void + publish: ( + sessionId: string, + journal: AgentSessionJournal, + activity?: AgentSessionTurnActivity | null + ) => void } tasks: StructuredAgentSessionTaskQueue reconcileLeases: (sessionId: string) => Promise diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-attach-orchestration.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-attach-orchestration.ts index a22bbdcbb3e..16e5593d27c 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-attach-orchestration.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-attach-orchestration.ts @@ -8,7 +8,8 @@ import { randomUUID } from 'node:crypto' import type { AgentSessionAttachResult, - AgentSessionMutationResult + AgentSessionMutationResult, + AgentSessionTurnActivity } from '../../../shared/agent-session-wire' import type { AgentSessionAttachParams } from './structured-agent-session-attach' import { performAttach } from './structured-agent-session-attach-flow' @@ -96,8 +97,8 @@ export function attachStructuredAgentSession( // Site 8: the provisional journal has no owner until the map takes it, // and the barrier below throws by design. try { - await bindAndDrain(eventSink, attached.journal, fence, () => - context.subscribers.publish(sessionId, attached.journal) + await bindAndDrain(eventSink, attached.journal, fence, (activity) => + context.subscribers.publish(sessionId, attached.journal, activity) ) } catch (error) { await agentSessionJournalCloseRetries.closeOrRetain(attached.journal) @@ -148,7 +149,7 @@ async function bindAndDrain( eventSink: DeferredStructuredAgentSessionEventSink, journal: AgentSessionJournal, fence: number, - publish: () => void + publish: (activity?: AgentSessionTurnActivity | null) => void ): Promise { eventSink.bind({ journal, fence, publish }) const barrier = await eventSink.drained() diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-command-publication.test.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-command-publication.test.ts new file mode 100644 index 00000000000..705baf182fa --- /dev/null +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-command-publication.test.ts @@ -0,0 +1,134 @@ +import { mkdtemp, rm } from 'node:fs/promises' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { expect, it, vi } from 'vitest' +import type { + AgentSessionSlashCommand, + AgentSessionSubscribeEvent +} from '../../../shared/agent-session-wire' +import { createStructuredAgentSessionEventCoalescer } from '../../../shared/structured-agent-session-coalescer' +import { + EMPTY_STRUCTURED_AGENT_SESSION, + reduceStructuredAgentSession +} from '../../../shared/structured-agent-session-reducer' +import { createTrackedJournalOpener } from '../agent-session-journal/journal-store-test-open' +import { AgentSessionSubscribers } from './structured-agent-session-subscribers' +import { + adapterFor, + fakeClaude, + identityFor, + PROVIDER_SESSION_ID +} from '../../claude/claude-structured-session-test-support' + +it('publishes idle provider reloads only when the actual command catalog changes', async () => { + const claude = fakeClaude() + const adapter = adapterFor(claude) + const publish = vi.fn() + await adapter.acquire({ + identity: identityFor(), + fence: 7, + spawnToken: 'commands', + events: { + appendItem: vi.fn(), + appendTombstone: vi.fn(), + publish + } + }) + publish.mockClear() + const message = { + type: 'system', + subtype: 'commands_changed', + session_id: PROVIDER_SESSION_ID, + commands: [{ name: 'new-skill', description: '', argumentHint: '' }] + } + claude.connections[0].handlers.onMessage?.(message) + expect(adapter.readCommands(identityFor().sessionId)).toEqual([ + { name: 'new-skill', kind: 'command', kindUnspecified: true } + ]) + expect(publish).toHaveBeenCalledTimes(1) + claude.connections[0].handlers.onMessage?.(message) + expect(publish).toHaveBeenCalledTimes(1) +}) + +it('delivers catalog changes through existing frames without resending them on ordinary output', async () => { + const root = await mkdtemp(join(tmpdir(), 'orca-command-publication-')) + const journals = createTrackedJournalOpener() + const events: AgentSessionSubscribeEvent[] = [] + let state = EMPTY_STRUCTURED_AGENT_SESSION + const coalescer = createStructuredAgentSessionEventCoalescer((event) => { + events.push(event) + state = reduceStructuredAgentSession(state, { type: 'event', event }) + }) + try { + const journal = await journals.open({ identity: identityFor(), journalDir: root }) + const sessionId = identityFor().sessionId + let commands: AgentSessionSlashCommand[] | undefined = [ + { name: 'loaded', kind: 'command', kindUnspecified: true } + ] + const subscribers = new AgentSessionSubscribers({ readCommands: () => commands }) + const close = subscribers.open({ + id: 'one', + sessionId, + journal, + fence: 7, + emit: coalescer.push + }) + expect(state.commands).toEqual(commands) + for (let i = 0; i < 25; i++) { + subscribers.handoff(sessionId, 7, { + owner: 'none', + direction: null, + phase: 'idle', + stage: null, + operationId: null + }) + } + coalescer.flush() + expect(events.filter((event) => 'commands' in event)).toHaveLength(1) + commands = [] + subscribers.publish(sessionId, journal) + subscribers.handoff(sessionId, 7, { + owner: 'none', + direction: null, + phase: 'idle', + stage: null, + operationId: null + }) + coalescer.flush() + expect(state.commands).toEqual([]) + expect(events.filter((event) => 'commands' in event)).toHaveLength(2) + commands = undefined + subscribers.publish(sessionId, journal) + coalescer.flush() + expect(state.commands).toBeNull() + close() + commands = [{ name: 'reconnected', kind: 'skill' }] + subscribers.open({ + id: 'two', + sessionId, + journal, + cursor: journal.cursor(), + fence: 8, + emit: coalescer.push + }) + coalescer.flush() + expect(state.commands).toEqual(commands) + subscribers.reset(sessionId, journal, 'epoch_changed', 8) + expect(state.commands).toEqual(commands) + commands = undefined + subscribers.open({ + id: 'three', + sessionId, + journal, + cursor: journal.cursor(), + fence: 9, + emit: coalescer.push + }) + coalescer.flush() + expect(state.commands).toBeNull() + } finally { + coalescer.dispose() + await journals.closeAll() + await rm(root, { recursive: true, force: true }) + } +}) diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-event-sink.test.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-event-sink.test.ts index c97161ce3dd..d1b7ea533a1 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-event-sink.test.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-event-sink.test.ts @@ -3,6 +3,7 @@ import type { AgentJournalItemBody, AgentJournalItemIdentity } from '../../../shared/agent-session-journal-types' +import type { AgentSessionTurnActivity } from '../../../shared/agent-session-wire' import type { AgentSessionJournal } from '../agent-session-journal/journal-store' import { createDeferredStructuredAgentSessionEventSink, @@ -20,7 +21,13 @@ function identity(ordinal: number): AgentJournalItemIdentity { return { provider: 'codex', threadId: 'thread-1', turnId: 'turn-1', ordinal } } -type Recorded = { call: string; fence?: number; ordinal?: number; settlementId?: string } +type Recorded = { + call: string + fence?: number + ordinal?: number + settlementId?: string + activity?: AgentSessionTurnActivity | null +} function target( fence: number, @@ -49,7 +56,12 @@ function target( return { epoch: 'e', sequence: 0 } }) } as unknown as AgentSessionJournal - return { journal, fence, publish: () => log.push({ call: 'publish', fence }) } + return { + journal, + fence, + publish: (activity) => + log.push({ call: 'publish', fence, ...(activity !== undefined ? { activity } : {}) }) + } } describe('deferred structured agent-session event sink', () => { @@ -317,4 +329,22 @@ describe('deferred structured agent-session event sink', () => { { call: 'appendItem', fence: 6, ordinal: 2 } ]) }) + + it('coalesces provider activity as a publication without a journal write', async () => { + const log: Recorded[] = [] + const deferred = createDeferredStructuredAgentSessionEventSink() + + deferred.sink.setActivity?.({ turnId: 'turn-1', text: 'Thinking' }) + deferred.sink.setActivity?.({ turnId: 'turn-1', text: 'Checking the result' }) + deferred.bind(target(6, log)) + await deferred.drained() + + expect(log).toEqual([ + { + call: 'publish', + fence: 6, + activity: { turnId: 'turn-1', text: 'Checking the result' } + } + ]) + }) }) diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-event-sink.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-event-sink.ts index 7e0192f179c..6952b6d93e6 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-event-sink.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-event-sink.ts @@ -3,6 +3,7 @@ import type { AgentJournalItemBody, AgentJournalItemIdentity } from '../../../shared/agent-session-journal-types' +import type { AgentSessionTurnActivity } from '../../../shared/agent-session-wire' import type { AgentSessionJournal } from '../agent-session-journal/journal-store' import type { JournalLifecycleMutationInput } from '../agent-session-journal/journal-row-builders' import { estimateStructuredAgentSessionItemBytes } from './structured-agent-session-event-sink-estimate' @@ -43,6 +44,7 @@ export type StructuredAgentSessionEventSink = { options?: StructuredAgentSessionAppendOptions ): StructuredAgentSessionSinkAdmission publish(options?: StructuredAgentSessionAppendOptions): void + setActivity?(activity: AgentSessionTurnActivity | null): void tryAppendItem?( identity: AgentJournalItemIdentity, body: AgentJournalItemBody, @@ -66,7 +68,7 @@ export type StructuredAgentSessionEventSink = { export type StructuredAgentSessionEventTarget = { journal: AgentSessionJournal fence: number - publish: () => void + publish: (activity?: AgentSessionTurnActivity | null) => void } export type DeferredStructuredAgentSessionEventSink = { @@ -209,6 +211,13 @@ export function createDeferredStructuredAgentSessionEventSink( publish: (options = {}) => { publish(options) }, + setActivity: (activity) => { + queue.submit({ + bytes: Buffer.byteLength(JSON.stringify(activity), 'utf8') + 64, + coalescingKey: 'turn-activity', + run: (bound) => bound.publish(activity) + }) + }, tryPublish: publish }, bind: (next) => queue.bind(next), diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-host-handoff.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-host-handoff.ts index 586df1476cf..abd2268c809 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-host-handoff.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-host-handoff.ts @@ -230,7 +230,7 @@ export async function acquireNativeHandoffOwner( eventSink.bind({ journal: session.journal, fence: proved.lease.runtimeFence, - publish: () => host.subscribers.publish(input.sessionId, session.journal) + publish: (activity) => host.subscribers.publish(input.sessionId, session.journal, activity) }) const acquiredBarrier = await eventSink.drained() if (!acquiredBarrier.ok) { diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-host-mutations.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-host-mutations.ts index f4a0244d0af..9a5f3e0c475 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-host-mutations.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-host-mutations.ts @@ -5,7 +5,10 @@ // they share one path here rather than five copies in the host. The host keeps attach, holds and // teardown; this is the surface that assumes those already happened. -import type { AgentJournalMessageItem } from '../../../shared/agent-session-journal-types' +import type { + AgentJournalItemIdentity, + AgentJournalMessageItem +} from '../../../shared/agent-session-journal-types' import type { AgentSessionCancelResult, AgentSessionMutationEnvelope, @@ -118,3 +121,26 @@ export function readStructuredAgentSessionOptions( return context.deps.adapter.readOptions({ sessionId, fence: session.fence }) }) } + +/** Settle provider-proven delivery independently of an in-flight client mutation. */ +export async function settleStructuredAgentSessionLateDispatch( + context: StructuredAgentSessionMutationContext, + input: { + sessionId: string + clientMessageId: string + providerIdentity: AgentJournalItemIdentity + } +): Promise { + const session = context.sessions.get(input.sessionId) + if (!session) { + return + } + // The journal queue drains before close; the host queue would defer this past teardown. + await session.journal.resolveDispatch({ + clientMessageId: input.clientMessageId, + state: 'accepted', + providerIdentity: input.providerIdentity, + fence: session.fence + }) + context.publish(input.sessionId, session.journal) +} diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-host-teardown.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-host-teardown.ts index 2b230db1357..b948ac08abd 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-host-teardown.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-host-teardown.ts @@ -6,6 +6,7 @@ // with the global runtime reference already cleared — the one state from which // nothing can ever close them. +import { withTimeout } from '../../../shared/promise-timeout-fallback' import { agentSessionJournalCloseRetries } from '../agent-session-journal/journal-close-retry' import type { StructuredAgentSessionHostSession } from './structured-agent-session-host-types' @@ -14,6 +15,39 @@ export type StructuredAgentSessionTeardownPhase = { run: () => Promise | void } +/** Quit must not wait indefinitely on an in-flight handoff; see `drain-handoffs` below. */ +const HANDOFF_DRAIN_TIMEOUT_MS = 5_000 + +/** + * The quit-path phase order, which is load-bearing rather than incidental. + * + * Handoffs drain BEFORE the session map is dropped: a flow left running writes rows into a + * journal this teardown is about to close, and publishes against a session it removed. That drain + * is bounded because a flow wedged in `launchTui` would otherwise hold the quit open forever; + * giving up merely restores the old orphaning, which the publish guard already makes survivable. + */ +export function structuredAgentSessionHostTeardownPhases(collaborators: { + holds: { dispose: () => Promise | void } + runtimeState: { + stopLeaseRenewal: () => void + flushAllEventSinks: () => Promise + } + handoffs: { stopTuiHistoryCatchup: () => void; drain: () => Promise } + tasks: { drainAttaches: () => Promise } +}): StructuredAgentSessionTeardownPhase[] { + return [ + { name: 'dispose-holds', run: () => collaborators.holds.dispose() }, + { name: 'stop-lease-renewal', run: () => collaborators.runtimeState.stopLeaseRenewal() }, + { name: 'stop-tui-catchup', run: () => collaborators.handoffs.stopTuiHistoryCatchup() }, + { + name: 'drain-handoffs', + run: () => withTimeout(collaborators.handoffs.drain(), HANDOFF_DRAIN_TIMEOUT_MS, undefined) + }, + { name: 'drain-attaches', run: () => collaborators.tasks.drainAttaches() }, + { name: 'flush-event-sinks', run: () => collaborators.runtimeState.flushAllEventSinks() } + ] +} + export async function tearDownStructuredAgentSessionHost(input: { phases: readonly StructuredAgentSessionTeardownPhase[] sessions: Map diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-host-types.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-host-types.ts index ea4b594ac44..7a321668c46 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-host-types.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-host-types.ts @@ -1,6 +1,7 @@ import type { AgentSessionOwnerProbe } from '../../../shared/agent-session-lease-adjudication' import type { AgentSessionProviderHandleLink } from '../../../shared/agent-session-provider-handle' import type { AgentSessionRecord } from '../../../shared/agent-session-record' +import type { AgentSessionStatusSummary } from '../../../shared/agent-session-wire' import type { AgentSessionRecordStore } from '../../runtime/agent-session-record-store' import type { AgentSessionSpawnTokenScan } from '../../runtime/agent-session-spawn-token-process-scan' import type { AgentSessionJournal } from '../agent-session-journal/journal-store' @@ -62,5 +63,11 @@ export type StructuredAgentSessionHostDeps = { /** How long a session outlives its last surface. Tests drive this; production takes the default. */ releaseGraceMs?: number onEventSinkError?: (input: { sessionId: string; error: unknown }) => void + /** Every status projection this host publishes. `replay` marks a re-projection of state the host + * already knew (restore, an arriving subscriber) rather than a fresh journal edge. */ + onSessionStatusChanged?: ( + summary: AgentSessionStatusSummary, + options: { replay: boolean } + ) => void handoffTransport?: StructuredAgentSessionHandoffTransport } diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-host.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-host.ts index 13b7d7b441e..898b5e2d4be 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-host.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-host.ts @@ -1,6 +1,7 @@ // Structured agent-session host: where the lease, journal, and provider adapter meet. // Mutations share one durable admission path and serialize per session. +import type { AgentJournalSnapshot } from '../../../shared/agent-session-journal-types' import type { AgentSessionExecutionLocation } from '../../../shared/agent-session-record' import type * as SessionWire from '../../../shared/agent-session-wire' import type { AgentSessionAttachParams } from './structured-agent-session-attach' @@ -38,9 +39,13 @@ import { respondToStructuredAgentSessionPrompt, sendStructuredAgentSessionTurn, setStructuredAgentSessionOption, + settleStructuredAgentSessionLateDispatch, type StructuredAgentSessionMutationContext } from './structured-agent-session-host-mutations' -import { tearDownStructuredAgentSessionHost } from './structured-agent-session-host-teardown' +import { + structuredAgentSessionHostTeardownPhases, + tearDownStructuredAgentSessionHost +} from './structured-agent-session-host-teardown' import type { StructuredAgentSessionCaller, StructuredAgentSessionHostDeps, @@ -50,19 +55,18 @@ import type { import { StructuredAgentSessionStatusFeed } from './structured-agent-session-status-feed' import { StructuredAgentSessionEventRecovery } from './structured-agent-session-event-recovery' import { StructuredAgentSessionBackgroundTaskChannel } from './structured-agent-session-background-task-channel' -import { withTimeout } from '../../../shared/promise-timeout-fallback' export type { StructuredAgentSessionHostDeps } from './structured-agent-session-host-types' -/** Quit must not wait indefinitely on an in-flight handoff; see the drain phase below. */ -const HANDOFF_DRAIN_TIMEOUT_MS = 5_000 export class StructuredAgentSessionHost { private readonly sessions = new Map() private readonly statusFeed = new StructuredAgentSessionStatusFeed({ sessions: this.sessions, getRecord: (sessionId) => this.deps.store.getRecord(sessionId), - now: () => this.now() + now: () => this.now(), + onStatusChanged: (summary, options) => this.deps.onSessionStatusChanged?.(summary, options) }) private readonly subscribers = new AgentSessionSubscribers({ + readCommands: (sessionId) => this.deps.adapter.readCommands?.(sessionId), onJournalPublished: (sessionId, journal) => this.statusFeed.publish(sessionId, journal) }) private readonly tasks = new StructuredAgentSessionTaskQueue() @@ -128,7 +132,7 @@ export class StructuredAgentSessionHost { // `hasSession` inside the same serialized step as this `set`. onReadable: (sessionId, restored) => { this.sessions.set(sessionId, restored) - this.statusFeed.publish(sessionId) + this.statusFeed.publish(sessionId, undefined, { replay: true }) }, restoreHandoff: (sessionId) => this.handoffs.restore(sessionId) }) @@ -179,14 +183,11 @@ export class StructuredAgentSessionHost { /** The host's half of attaching, named so it cannot grow dependencies unnoticed. */ private attachContext(): StructuredAgentSessionAttachContext { return { - deps: this.deps, - runtimeState: this.runtimeState, - sessions: this.sessions, + ...this.lifetimeContext(), subscribers: this.subscribers, tasks: this.tasks, reconcileLeases: (sessionId) => this.reconcileLeases(sessionId), - serialize: (sessionId, task) => this.serialize(sessionId, task), - now: () => this.now() + serialize: (sessionId, task) => this.serialize(sessionId, task) } } /** Releases a session's resources without ending the conversation: the record and journal stay @@ -253,22 +254,12 @@ export class StructuredAgentSessionHost { async flushAllStreamedEvents(): Promise { await tearDownStructuredAgentSessionHost({ - phases: [ - { name: 'dispose-holds', run: () => this.holds.dispose() }, - { name: 'stop-lease-renewal', run: () => this.runtimeState.stopLeaseRenewal() }, - { name: 'stop-tui-catchup', run: () => this.handoffs.stopTuiHistoryCatchup() }, - // Before the session map is dropped: a handoff flow left running writes rows into a - // journal this teardown is about to close, and publishes against a session it removed. - // Why bounded: this phase is on the app-quit path, and a flow wedged in `launchTui` would - // otherwise hold the quit open forever. Giving up merely restores the old orphaning, which - // the publish guard above already makes survivable. - { - name: 'drain-handoffs', - run: () => withTimeout(this.handoffs.drain(), HANDOFF_DRAIN_TIMEOUT_MS, undefined) - }, - { name: 'drain-attaches', run: () => this.tasks.drainAttaches() }, - { name: 'flush-event-sinks', run: () => this.runtimeState.flushAllEventSinks() } - ], + phases: structuredAgentSessionHostTeardownPhases({ + holds: this.holds, + runtimeState: this.runtimeState, + handoffs: this.handoffs, + tasks: this.tasks + }), sessions: this.sessions }) } @@ -317,6 +308,11 @@ export class StructuredAgentSessionHost { readOptions = (sessionId: string): Promise => readStructuredAgentSessionOptions(this.mutationContext(), sessionId) + /** Undefined means unavailable; an empty array is an authoritative catalog. */ + readCommands = (sessionId: string): SessionWire.AgentSessionCommandsResult => ({ + commands: this.deps.adapter.readCommands?.(sessionId) + }) + async handoffStatus(sessionId: string): Promise { this.requireSession(sessionId) return this.serialize(sessionId, () => @@ -324,23 +320,27 @@ export class StructuredAgentSessionHost { ) } - history = ( - request: SessionWire.AgentSessionHistoryRequest - ): SessionWire.AgentSessionHistoryResult => this.backgroundTasks.history(request) + history: StructuredAgentSessionBackgroundTaskChannel['history'] = (request) => + this.backgroundTasks.history(request) + + /** The fully reduced timeline, for readers that cannot tolerate a page's ambiguity — a settled + * turn is tombstoned, so an item's ABSENCE from a bounded page proves nothing. */ + journalSnapshot = (sessionId: string): AgentJournalSnapshot => + this.requireSession(sessionId).journal.snapshot() subscribe = (input: AgentSessionSubscribeInput): (() => void) => this.backgroundTasks.subscribe(input) - publishBackgroundTaskState: StructuredAgentSessionBackgroundTaskChannel['publish'] = ( - sessionId, - state - ) => this.backgroundTasks.publish(sessionId, state) + settleLateDispatch = (input: Parameters[1]) => + settleStructuredAgentSessionLateDispatch(this.mutationContext(), input) + + publishBackgroundTaskState: StructuredAgentSessionBackgroundTaskChannel['publish'] = (...args) => + this.backgroundTasks.publish(...args) unsubscribe = (sessionId: string, id: string): void => this.subscribers.close(sessionId, id) /** Every session's projected status for session lists; unlike `subscribe`, retains nothing. */ - subscribeStatus = ( - subscriber: Parameters[0] - ): (() => void) => this.statusFeed.subscribe(subscriber) + subscribeStatus: StructuredAgentSessionStatusFeed['subscribe'] = (subscriber) => + this.statusFeed.subscribe(subscriber) private requireSession(sessionId: string): StructuredAgentSessionHostSession { const session = this.sessions.get(sessionId) diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-late-settlement.test.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-late-settlement.test.ts new file mode 100644 index 00000000000..a75d2ea6512 --- /dev/null +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-late-settlement.test.ts @@ -0,0 +1,224 @@ +import { mkdtemp, rm } from 'node:fs/promises' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { afterEach, beforeEach, describe, expect, it, vi, type Mock } from 'vitest' +import { computeAgentSessionPayloadFingerprint } from '../../../shared/agent-session-mutation-envelope' +import type { + AgentSessionMutationEnvelope, + AgentSessionSubscribeEvent +} from '../../../shared/agent-session-wire' +import { AgentSessionRecordStore } from '../../runtime/agent-session-record-store' +import type { + AgentSessionDispatchOutcome, + StructuredAgentSessionAdapter +} from './structured-agent-session-adapter' +import { StructuredAgentSessionHost } from './structured-agent-session-host' +import { + HOST_TEST_NOW as NOW, + HOST_TEST_SESSION as SESSION, + HOST_TEST_THREAD as THREAD, + hostTestAttachParams, + hostTestMessage, + hostTestOperationId, + resetHostTestOperationIds +} from './structured-agent-session-host-test-data' + +const CALLER = { callerKey: 'client-1' } + +let root: string +let store: AgentSessionRecordStore +let host: StructuredAgentSessionHost +let dispatch: Mock +let closeSession: Mock> + +function accepted(): AgentSessionDispatchOutcome { + return { + state: 'accepted', + providerIdentity: { provider: 'codex', threadId: THREAD, turnId: 'turn-1', ordinal: 1 } + } +} + +function sendParams(text: string): { + envelope: AgentSessionMutationEnvelope + body: ReturnType +} { + const body = hostTestMessage(text) + return { + envelope: { + sessionId: SESSION, + clientOperationId: hostTestOperationId(), + expectedRuntimeFence: store.getRecord(SESSION)?.lease.runtimeFence ?? 1, + payloadFingerprint: computeAgentSessionPayloadFingerprint({ + method: 'agentSession.send', + sessionId: SESSION, + fields: { body } + }) + }, + body + } +} + +function submissions(): unknown { + const state = host.history({ sessionId: SESSION, direction: 'tail' }) + return state.ok ? state.page.submissions : null +} + +beforeEach(async () => { + root = await mkdtemp(join(tmpdir(), 'orca-wire-late-settle-')) + resetHostTestOperationIds() + dispatch = vi.fn(async () => accepted()) + closeSession = vi.fn(async () => true) + store = await AgentSessionRecordStore.open({ directory: join(root, 'store'), hostId: 'local' }) + host = new StructuredAgentSessionHost({ + store, + adapter: { + acquire: vi.fn(async ({ fence }) => ({ + process: { + hostId: 'local', + pid: 4242, + processStartTimeMs: 1_700_000_000_000, + spawnToken: store.getRecord(SESSION)?.lease.reservedSpawnToken ?? 'spawn-a' + }, + link: { + linkId: `link-${fence}`, + handle: { provider: 'codex' as const, threadId: THREAD }, + origin: 'created' as const, + mintedAtFence: fence, + observedAt: NOW + } + })), + releaseAcquisition: vi.fn(async () => true), + dispatch, + closeSession, + cancelTurn: vi.fn(async () => ({ cancelled: true })), + answerPrompt: vi.fn(async () => undefined), + setOption: vi.fn(async () => undefined) + }, + journalRoot: root, + claimKeyId: 'key-1', + mintSpawnToken: () => 'spawn-a', + now: () => NOW + }) + expect((await host.attach(CALLER, hostTestAttachParams(null))).ok).toBe(true) +}) + +afterEach(async () => { + await host.flushAllStreamedEvents() + await host.close(SESSION) + await rm(root, { recursive: true, force: true }) +}) + +describe('settling a send the provider proves it received after the ack window', () => { + it('publishes acceptance during a pending send and never reopens it for retry', async () => { + let finishDispatch!: (outcome: AgentSessionDispatchOutcome) => void + dispatch.mockImplementationOnce( + () => + new Promise((resolve) => { + finishDispatch = resolve + }) + ) + const events: AgentSessionSubscribeEvent[] = [] + const unsubscribe = host.subscribe({ + id: 'late-receipt', + sessionId: SESSION, + emit: (event) => events.push(event) + }) + const params = sendParams('echo before send completes') + const pending = host.send(CALLER, params) + await vi.waitFor(() => expect(dispatch).toHaveBeenCalledTimes(1)) + try { + await host.settleLateDispatch({ + sessionId: SESSION, + clientMessageId: params.envelope.clientOperationId, + providerIdentity: { provider: 'claude', sessionId: THREAD, uuid: 'early-echo' } + }) + expect(events.at(-1)).toMatchObject({ + type: 'batch', + batch: { + submissions: [ + { clientMessageId: params.envelope.clientOperationId, dispatchState: 'accepted' } + ] + } + }) + } finally { + finishDispatch({ state: 'unknown', reason: 'ack timeout' }) + unsubscribe() + } + await expect(pending).resolves.toMatchObject({ + ok: true, + value: { submission: { dispatchState: 'accepted' } } + }) + await expect(host.send(CALLER, { ...params, retryUnknown: true })).resolves.toMatchObject({ + ok: true, + value: { submission: { dispatchState: 'accepted' } } + }) + expect(dispatch).toHaveBeenCalledTimes(1) + }) + + it('persists an echo received while the provider is closing', async () => { + dispatch.mockResolvedValueOnce({ state: 'unknown', reason: 'ack timeout' }) + const params = sendParams('received just before shutdown') + await host.send(CALLER, params) + let settlement: Promise | undefined + closeSession.mockImplementationOnce(async () => { + settlement = host.settleLateDispatch({ + sessionId: SESSION, + clientMessageId: params.envelope.clientOperationId, + providerIdentity: { provider: 'claude', sessionId: THREAD, uuid: 'closing-echo' } + }) + void settlement.catch(() => undefined) + return true + }) + + await host.close(SESSION) + await expect(settlement).resolves.toBeUndefined() + await host.revealSession(SESSION) + expect(submissions()).toMatchObject([{ dispatchState: 'accepted' }]) + expect(dispatch).toHaveBeenCalledTimes(1) + }) + + it('moves a durable unknown to accepted so nothing offers to send it again', async () => { + dispatch.mockRejectedValueOnce(new Error('socket closed')) + const params = sendParams('sent while a turn was running') + const first = await host.send(CALLER, params) + expect(first).toMatchObject({ ok: true, value: { submission: { dispatchState: 'unknown' } } }) + + await host.settleLateDispatch({ + sessionId: SESSION, + clientMessageId: params.envelope.clientOperationId, + providerIdentity: { provider: 'claude', sessionId: THREAD, uuid: 'late-uuid' } + }) + + expect(submissions()).toMatchObject([ + { clientMessageId: params.envelope.clientOperationId, dispatchState: 'accepted' } + ]) + // The point of the fix: the client stops rendering Retry, and Retry is what + // was delivering the message to the agent a second time. + expect(dispatch).toHaveBeenCalledTimes(1) + }) + + it('leaves an already accepted send alone', async () => { + const params = sendParams('ordinary send') + await host.send(CALLER, params) + + await host.settleLateDispatch({ + sessionId: SESSION, + clientMessageId: params.envelope.clientOperationId, + providerIdentity: { provider: 'claude', sessionId: THREAD, uuid: 'a-different-uuid' } + }) + + expect(submissions()).toMatchObject([ + { clientMessageId: params.envelope.clientOperationId, dispatchState: 'accepted' } + ]) + }) + + it('ignores a session this host is not holding', async () => { + await expect( + host.settleLateDispatch({ + sessionId: 'session-that-is-not-attached', + clientMessageId: 'whatever', + providerIdentity: { provider: 'claude', sessionId: THREAD, uuid: 'x' } + }) + ).resolves.toBeUndefined() + }) +}) diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-option-settlement.test.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-option-settlement.test.ts index 41054d10ece..5b4f0556915 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-option-settlement.test.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-option-settlement.test.ts @@ -3,7 +3,10 @@ import { tmpdir } from 'node:os' import { join } from 'node:path' import { afterEach, beforeEach, describe, expect, it, vi, type Mock } from 'vitest' import { computeAgentSessionPayloadFingerprint } from '../../../shared/agent-session-mutation-envelope' -import type { AgentSessionMutationEnvelope } from '../../../shared/agent-session-wire' +import type { + AgentSessionMutationEnvelope, + AgentSessionStatusEvent +} from '../../../shared/agent-session-wire' import { AgentSessionRecordStore } from '../../runtime/agent-session-record-store' import type { StructuredAgentSessionAdapter } from './structured-agent-session-adapter' import { AgentSessionOptionRejectedError } from './structured-agent-session-option-error' @@ -212,6 +215,39 @@ afterEach(async () => { }) describe('structured session options and close', () => { + it('publishes an acknowledged model without waiting for journal traffic', async () => { + const body = { + kind: 'message' as const, + role: 'user' as const, + blocks: [{ type: 'text' as const, text: 'first task' }] + } + await host.send(CALLER, { + envelope: envelope('agentSession.send', { body }), + body + }) + const events: AgentSessionStatusEvent[] = [] + host.subscribeStatus({ id: 'session-list', emit: (event) => events.push(event) }) + expect(events).toEqual([ + { + type: 'snapshot', + sessions: [expect.objectContaining({ status: 'idle', model: DEFAULT_MODEL })] + } + ]) + const fields = { key: 'model', value: PICKED_MODEL } + + await host.setOption(CALLER, { + envelope: envelope('agentSession.setOption', fields), + ...fields + }) + + expect(events.slice(1)).toEqual([ + { + type: 'status', + session: expect.objectContaining({ sessionId: SESSION, model: PICKED_MODEL }) + } + ]) + }) + it('settles a pre-mutation rejection so a fresh retry can succeed', async () => { optionFailure = new AgentSessionOptionRejectedError('model list unavailable') const fields = { key: 'model', value: PICKED_MODEL } diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-status-feed.test.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-status-feed.test.ts index 7efd147c420..c1b52f879d4 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-status-feed.test.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-status-feed.test.ts @@ -2,9 +2,16 @@ import { mkdtemp, rm } from 'node:fs/promises' import { tmpdir } from 'node:os' import { join } from 'node:path' import { afterEach, beforeEach, describe, expect, it } from 'vitest' +import type { AgentSessionRecord } from '../../../shared/agent-session-record' import type { AgentSessionStatusEvent } from '../../../shared/agent-session-wire' +import { createClaudeJournalTranslator } from '../../claude/claude-structured-journal-translation' +import { publishCodexTurnLifecycle } from '../../codex/codex-structured-journal-translation-turns' +import { createDeferredStructuredAgentSessionEventSink } from './structured-agent-session-event-sink' import { createTrackedJournalOpener } from '../agent-session-journal/journal-store-test-open' -import { StructuredAgentSessionStatusFeed } from './structured-agent-session-status-feed' +import { + StructuredAgentSessionStatusFeed, + type StructuredAgentSessionStatusFeedDeps +} from './structured-agent-session-status-feed' const SESSION = 'status-session' const TURN_IDENTITY = { @@ -32,7 +39,7 @@ afterEach(async () => { await rm(root, { recursive: true, force: true }) }) -async function openJournal(sessionId = SESSION) { +async function openJournal(sessionId = SESSION, now?: () => number) { return journals.open({ identity: { sessionId, @@ -41,6 +48,7 @@ async function openJournal(sessionId = SESSION) { agent: 'codex', providerHandle: { kind: 'codex', threadId: 'thread-1' } }, + now, journalDir: join(root, sessionId) }) } @@ -52,9 +60,14 @@ function indexed(session: { journal: Awaited> }) } } -function feedFor(sessions: Map> }>) { +function feedFor( + sessions: Map> }>, + record: Partial | null = null, + onStatusChanged?: StructuredAgentSessionStatusFeedDeps['onStatusChanged'] +) { let now = 1_000 const feed = new StructuredAgentSessionStatusFeed({ + ...(onStatusChanged ? { onStatusChanged } : {}), sessions: { get: (sessionId: string) => { const session = sessions.get(sessionId) @@ -66,7 +79,7 @@ function feedFor(sessions: Map>, - getRecord: () => null, + getRecord: () => record as AgentSessionRecord | null, now: () => (now += 1) }) const events: AgentSessionStatusEvent[] = [] @@ -132,6 +145,145 @@ describe('StructuredAgentSessionStatusFeed', () => { expect(events).toHaveLength(3) }) + it('preserves the completion tombstone time when the journal and host reopen', async () => { + let now = 100 + const journal = await openJournal(SESSION, () => now) + await journal.appendItem( + USER_IDENTITY, + { kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'hello' }] }, + { fence: 1 } + ) + await journal.appendItem( + TURN_IDENTITY, + { kind: 'status', text: 'Working', turnLifecycle: { turnId: 'turn-1', state: 'running' } }, + { fence: 1 } + ) + const { feed, events } = feedFor(new Map([[SESSION, { journal }]])) + now = 200 + await journal.appendTombstone(TURN_IDENTITY, { fence: 1 }) + feed.publish(SESSION) + expect(events.at(-1)).toMatchObject({ + type: 'status', + session: { status: 'idle', updatedAt: 200 } + }) + await journal.close() + now = 900 + const reopened = await openJournal(SESSION, () => now) + const restored = feedFor(new Map([[SESSION, { journal: reopened }]])) + expect(restored.events[0]).toMatchObject({ + type: 'snapshot', + sessions: [{ status: 'idle', updatedAt: 200 }] + }) + }) + + it('publishes settled activity revisions and restores the same age after reopening', async () => { + let now = 100 + const journal = await openJournal(SESSION, () => now) + await journal.appendItem( + USER_IDENTITY, + { kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'hello' }] }, + { fence: 1 } + ) + const assistant = { ...USER_IDENTITY, ordinal: 2 } + await journal.appendItem( + assistant, + { kind: 'message', role: 'assistant', blocks: [{ type: 'text', text: 'first' }] }, + { fence: 1 } + ) + const { feed, events } = feedFor(new Map([[SESSION, { journal }]])) + now = 200 + await journal.appendItem( + assistant, + { kind: 'message', role: 'assistant', blocks: [{ type: 'text', text: 'finished' }] }, + { fence: 1 } + ) + feed.publish(SESSION) + expect(events.at(-1)).toMatchObject({ + type: 'status', + session: { status: 'idle', updatedAt: 200 } + }) + feed.publish(SESSION) + expect(events).toHaveLength(2) + await journal.close() + const reopened = await openJournal(SESSION, () => 900) + const restored = feedFor(new Map([[SESSION, { journal: reopened }]])) + expect(restored.events[0]).toMatchObject({ + type: 'snapshot', + sessions: [{ status: 'idle', updatedAt: 200 }] + }) + }) + + it('does not publish timestamp-only revisions while a turn is working', async () => { + let now = 100 + const journal = await openJournal(SESSION, () => now) + await journal.appendItem( + USER_IDENTITY, + { kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'hello' }] }, + { fence: 1 } + ) + await journal.appendItem( + TURN_IDENTITY, + { kind: 'status', text: 'Working', turnLifecycle: { turnId: 'turn-1', state: 'running' } }, + { fence: 1 } + ) + const { feed, events } = feedFor(new Map([[SESSION, { journal }]])) + for (let revision = 1; revision <= 20; revision += 1) { + now += 1 + await journal.appendItem( + TURN_IDENTITY, + { kind: 'status', text: 'Working', turnLifecycle: { turnId: 'turn-1', state: 'running' } }, + { fence: 1 } + ) + feed.publish(SESSION) + } + expect(events).toHaveLength(1) + now = 200 + await journal.appendTombstone(TURN_IDENTITY, { fence: 1 }) + feed.publish(SESSION) + expect(events).toHaveLength(2) + expect(events.at(-1)).toMatchObject({ + type: 'status', + session: { status: 'idle', updatedAt: 200 } + }) + }) + + it('carries the record model and the running tool line the sidebar row shows', async () => { + const journal = await openJournal() + const { feed, events } = feedFor(new Map([[SESSION, { journal }]]), { + options: { model: 'gpt-5-codex' }, + providerHandleChain: [] + }) + await journal.appendItem( + USER_IDENTITY, + { kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'run the tests' }] }, + { fence: 1 } + ) + await journal.appendItem( + TURN_IDENTITY, + { kind: 'status', text: 'Working', turnLifecycle: { turnId: 'turn-1', state: 'running' } }, + { fence: 1 } + ) + feed.publish(SESSION) + expect(events.at(-1)).toEqual({ + type: 'status', + session: expect.objectContaining({ status: 'working', model: 'gpt-5-codex' }) + }) + + await journal.appendItem( + { ...USER_IDENTITY, ordinal: 2 }, + { kind: 'tool-call', name: 'shell', input: { command: 'pnpm test' }, state: 'running' }, + { fence: 1 } + ) + feed.publish(SESSION) + + // A tool boundary changes nothing else about the session, so only comparing the new + // fields keeps it from being deduped away as an unchanged projection. + expect(events.at(-1)).toEqual({ + type: 'status', + session: expect.objectContaining({ toolName: 'shell', toolInput: 'pnpm test' }) + }) + }) + it('reports a pending approval as attention', async () => { const journal = await openJournal() const { feed, events } = feedFor(new Map([[SESSION, { journal }]])) @@ -239,4 +391,135 @@ describe('StructuredAgentSessionStatusFeed', () => { session: expect.objectContaining({ status: 'idle' }) }) }) + it('reports each projection change to the host observer, marking re-projections as replay', async () => { + const journal = await openJournal() + const seen: { status: string | null; prompt: string; replay: boolean }[] = [] + const { feed } = feedFor(new Map([[SESSION, { journal }]]), null, (summary, options) => + seen.push({ status: summary.status, prompt: summary.latestPrompt, replay: options.replay }) + ) + await journal.appendItem( + USER_IDENTITY, + { kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'fix the auth bug' }] }, + { fence: 1 } + ) + await journal.appendItem( + TURN_IDENTITY, + { kind: 'status', text: 'Working', turnLifecycle: { turnId: 'turn-1', state: 'running' } }, + { fence: 1 } + ) + + feed.publish(SESSION, journal) + // A second identical publication is deduped, so the observer only ever sees changes. + feed.publish(SESSION, journal) + // seen[0] is the opening projection the harness's own subscriber triggered. + expect(seen.slice(1)).toEqual([ + { status: 'working', prompt: 'fix the auth bug', replay: false } + ]) + + // An arriving subscriber re-projects state the host already knew. + await journal.appendTombstone(TURN_IDENTITY, { fence: 1 }) + feed.subscribe({ id: 'list-2', emit: () => undefined }) + expect(seen.at(-1)).toEqual({ status: 'idle', prompt: 'fix the auth bug', replay: true }) + }) + + it.each(['claude', 'codex'] as const)( + 'observes a fast %s turn even when start and finish queue before persistence', + async (agent) => { + const journal = await openJournal() + await journal.appendItem( + USER_IDENTITY, + { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: 'Fix auth' }] + }, + { fence: 1 } + ) + const seen: (string | null)[] = [] + const { feed } = feedFor(new Map([[SESSION, { journal }]]), null, (summary) => + seen.push(summary.status) + ) + const deferred = createDeferredStructuredAgentSessionEventSink() + if (agent === 'claude') { + const translator = createClaudeJournalTranslator({ sink: deferred.sink }) + translator.handle({ + type: 'message', + sessionId: SESSION, + startsTurn: true, + message: { + type: 'user', + uuid: 'prompt-1', + session_id: 'claude-session', + parent_tool_use_id: null, + message: { role: 'user', content: [{ type: 'text', text: 'Fix auth' }] } + } + }) + translator.handle({ + type: 'message', + sessionId: SESSION, + message: { + type: 'result', + subtype: 'success', + session_id: 'claude-session', + uuid: 'result-1', + result: 'Done' + } + }) + translator.dispose() + } else { + for (const state of ['running', 'completed'] as const) { + publishCodexTurnLifecycle({ + sink: deferred.sink, + primaryThreadId: 'thread-1', + sessionId: SESSION, + threadId: 'thread-1', + turnId: 'turn-1', + state + }) + } + } + for (let index = 0; index < 100; index++) { + deferred.sink.publish() + } + // This queue is also reached while a previous asynchronous journal write is pending. + let publications = 0 + let activityPublications = 0 + deferred.bind({ + journal, + fence: 1, + publish: (activity) => { + if (activity === undefined) { + publications += 1 + } else { + activityPublications += 1 + } + feed.publish(SESSION, journal) + } + }) + expect(await deferred.drained()).toEqual({ ok: true }) + expect(seen).toEqual(['idle', 'working', 'idle']) + expect(publications).toBe(2) + expect(activityPublications).toBe(agent === 'claude' ? 1 : 0) + expect(deferred.state()).toMatchObject({ queuedBytes: 0, queuedOperations: 0 }) + deferred.close() + } + ) + + it('keeps publishing to subscribers when the host observer throws', async () => { + const journal = await openJournal() + const { feed, events } = feedFor(new Map([[SESSION, { journal }]]), null, () => { + throw new Error('observer exploded') + }) + await journal.appendItem( + USER_IDENTITY, + { kind: 'message', role: 'user', blocks: [{ type: 'text', text: 'hello' }] }, + { fence: 1 } + ) + + expect(() => feed.publish(SESSION, journal)).not.toThrow() + expect(events.at(-1)).toEqual({ + type: 'status', + session: expect.objectContaining({ status: 'idle', latestPrompt: 'hello' }) + }) + }) }) diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-status-feed.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-status-feed.ts index 5ad494cf830..348b5aebc21 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-status-feed.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-status-feed.ts @@ -12,6 +12,8 @@ import { agentProviderSessionsEqual } from '../../../shared/agent-session-resume' import type { AgentSessionRecord } from '../../../shared/agent-session-record' +import { normalizeOptionalField } from '../../../shared/agent-status-field-normalization' +import { AGENT_MODEL_MAX_LENGTH } from '../../../shared/agent-status-types' import type { AgentSessionStatusEvent, AgentSessionStatusSummary @@ -34,6 +36,9 @@ export type StructuredAgentSessionStatusFeedDeps = { sessions: ReadonlyMap getRecord: (sessionId: string) => AgentSessionRecord | null now: () => number + /** Every projection change, whether or not anyone is subscribed. `replay` marks a re-projection + * of state the host already knew (restore, an arriving subscriber) rather than a journal edge. */ + onStatusChanged?: (summary: AgentSessionStatusSummary, options: { replay: boolean }) => void } function summariesEqual(a: AgentSessionStatusSummary, b: AgentSessionStatusSummary): boolean { @@ -41,7 +46,13 @@ function summariesEqual(a: AgentSessionStatusSummary, b: AgentSessionStatusSumma a.workspaceId === b.workspaceId && a.agent === b.agent && a.status === b.status && + // Settled activity changes ranking; streaming active turns must stay quiet. + (a.status !== 'idle' || a.updatedAt === b.updatedAt) && a.latestPrompt === b.latestPrompt && + a.model === b.model && + a.toolName === b.toolName && + a.toolInput === b.toolInput && + a.lastAssistantMessage === b.lastAssistantMessage && agentProviderSessionsEqual(undefined, a.providerSession, b.providerSession) ) } @@ -57,7 +68,7 @@ export class StructuredAgentSessionStatusFeed { // Re-project before registering: a change found here has to reach the subscribers that // already read the old value, and the arriving one carries it in its snapshot instead. for (const [sessionId] of this.deps.sessions) { - this.publish(sessionId) + this.publish(sessionId, undefined, { replay: true }) } this.subscribers.set(subscriber.id, subscriber) this.emit(subscriber, { type: 'snapshot', sessions: [...this.published.values()] }) @@ -78,7 +89,7 @@ export class StructuredAgentSessionStatusFeed { } /** Re-projects one session after its journal changed; equal projections are not re-sent. */ - publish(sessionId: string, journal?: AgentSessionJournal): void { + publish(sessionId: string, journal?: AgentSessionJournal, options?: { replay?: boolean }): void { const session = this.deps.sessions.get(sessionId) if (!session) { return @@ -90,6 +101,12 @@ export class StructuredAgentSessionStatusFeed { } this.published.set(sessionId, summary) this.broadcast({ type: 'status', session: summary }) + try { + this.deps.onStatusChanged?.(summary, { replay: options?.replay === true }) + } catch (error) { + // An observer must never cost the subscribers their status event. + console.warn('[structured-session-status] status observer failed', error) + } } private summaryFor( @@ -99,16 +116,19 @@ export class StructuredAgentSessionStatusFeed { ): AgentSessionStatusSummary { // An unreadable journal projects as "no turn": the chat itself shows the reset. const items = journal.isReadOnly ? [] : journal.snapshot().items - const providerSession = structuredAgentSessionProviderSessionMetadata( - this.deps.getRecord(sessionId) - ) + const record = this.deps.getRecord(sessionId) + const providerSession = structuredAgentSessionProviderSessionMetadata(record) + // The journal has no model: the record's acknowledged options are where an owner + // handoff or a mid-session switch lands, so the row follows whichever is in force. + const model = normalizeOptionalField(record?.options?.model, AGENT_MODEL_MAX_LENGTH) return { sessionId, workspaceId: session.params.location.workspaceId, agent: session.params.provider, ...projectStructuredAgentSessionStatusSummary(items), + ...(model ? { model } : {}), ...(providerSession ? { providerSession } : {}), - updatedAt: this.deps.now() + updatedAt: journal.lastActivityAt() || this.deps.now() } } diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-subscribers.test.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-subscribers.test.ts index 81bcfa82b40..a32db8786d4 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-subscribers.test.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-subscribers.test.ts @@ -67,11 +67,54 @@ describe('AgentSessionSubscribers', () => { removedItemIds: [], submissions: [] }, - fence: 7 + fence: 7, + activity: null } ]) }) + it('includes catalogs on reconnect and sends an idle checkpoint without journal work', async () => { + const journal = await journals.open({ + identity: { + sessionId: SESSION, + workspaceId: 'workspace-1', + hostId: 'local', + agent: 'codex', + providerHandle: { kind: 'codex', threadId: 'thread-1' } + }, + journalDir: join(root, 'catalog-journal') + }) + let commands = [{ name: 'first', kind: 'skill' as const }] + const events: AgentSessionSubscribeEvent[] = [] + const subscribers = new AgentSessionSubscribers({ readCommands: () => commands }) + subscribers.open({ + id: 'one', + sessionId: SESSION, + journal, + fence: 7, + emit: (event) => events.push(event) + }) + expect(events[0]).toMatchObject({ type: 'snapshot', commands }) + commands = [{ name: 'second', kind: 'skill' as const }] + subscribers.publish(SESSION, journal) + expect(events[1]).toEqual({ + type: 'batch', + sessionId: SESSION, + fence: 7, + commands, + batch: { cursor: journal.cursor(), items: [], removedItemIds: [], submissions: [] } + }) + subscribers.open({ + id: 'two', + sessionId: SESSION, + journal, + cursor: journal.cursor(), + fence: 7, + emit: (event) => events.push(event) + }) + expect(events[2]).toMatchObject({ type: 'batch', commands }) + }) + it('reports every content publication to the journal hook, subscribed or not', async () => { const journal = await journals.open({ identity: { @@ -254,6 +297,56 @@ describe('AgentSessionSubscribers', () => { expect(events.at(-1)).toMatchObject({ type: 'batch', fence: 2 }) }) + it('publishes latest turn activity without advancing or adding journal rows', async () => { + const journal = await journals.open({ + identity: { + sessionId: SESSION, + workspaceId: 'workspace-1', + hostId: 'local', + agent: 'codex', + providerHandle: { kind: 'codex', threadId: 'thread-1' } + }, + journalDir: join(root, 'activity-journal') + }) + const subscribers = new AgentSessionSubscribers() + const events: AgentSessionSubscribeEvent[] = [] + subscribers.open({ + id: 'subscriber-1', + sessionId: SESSION, + journal, + fence: 1, + emit: (event) => events.push(event) + }) + const cursor = journal.cursor() + + subscribers.publish(SESSION, journal, { + turnId: 'turn-1', + text: 'Inspecting the session wire' + }) + + expect(journal.cursor()).toEqual(cursor) + expect(events.at(-1)).toEqual({ + type: 'batch', + sessionId: SESSION, + batch: { cursor, items: [], removedItemIds: [], submissions: [] }, + fence: 1, + activity: { turnId: 'turn-1', text: 'Inspecting the session wire' } + }) + + subscribers.close(SESSION, 'subscriber-1') + subscribers.publish(SESSION, journal, null) + subscribers.open({ + id: 'reconnected', + sessionId: SESSION, + journal, + fence: 1, + cursor, + emit: (event) => events.push(event) + }) + expect(journal.cursor()).toEqual(cursor) + expect(events.at(-1)).toMatchObject({ activity: null }) + }) + it('catches a subscriber up past a pre-existing unsendable removal with a bounded reset', async () => { const journalDir = join(root, 'oversized-removal-journal') const seeded = await journals.open({ diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-subscribers.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-subscribers.ts index 29dffa6a687..ba439e6427b 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-subscribers.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-subscribers.ts @@ -11,8 +11,10 @@ import type { import { AGENT_SESSION_HISTORY_MAX_LIMIT, type AgentSessionBackgroundTaskState, + type AgentSessionSlashCommand, type AgentSessionHandoffStatus, - type AgentSessionSubscribeEvent + type AgentSessionSubscribeEvent, + type AgentSessionTurnActivity } from '../../../shared/agent-session-wire' import type { AgentSessionJournal } from '../agent-session-journal/journal-store' import { @@ -34,9 +36,11 @@ type Subscriber = { emit: AgentSessionSubscriberEmit cursor: AgentJournalCursor fence: number + commands?: AgentSessionSlashCommand[] | null } export type AgentSessionSubscribersHooks = { + readCommands?: (sessionId: string) => AgentSessionSlashCommand[] | undefined /** Fires after any publication that can change journal content, whether or not anyone * is subscribed to the transcript: session lists project status from this same edge. */ onJournalPublished?: (sessionId: string, journal: AgentSessionJournal) => void @@ -44,6 +48,7 @@ export type AgentSessionSubscribersHooks = { export class AgentSessionSubscribers { private readonly bySession = new Map>() + private readonly activityBySession = new Map() constructor(private readonly hooks: AgentSessionSubscribersHooks = {}) {} @@ -81,7 +86,8 @@ export class AgentSessionSubscribers { page, fence: input.fence, ...(input.handoff ? { handoff: input.handoff } : {}), - ...(input.backgroundTasks !== undefined ? { backgroundTasks: input.backgroundTasks } : {}) + ...(input.backgroundTasks !== undefined ? { backgroundTasks: input.backgroundTasks } : {}), + ...this.activityField(input.sessionId) }) subscriber.cursor = page.liveCursor ?? page.window.nextCursor } @@ -103,11 +109,24 @@ export class AgentSessionSubscribers { } /** Fan out whatever each subscriber has not yet seen. */ - publish(sessionId: string, journal: AgentSessionJournal): void { + publish( + sessionId: string, + journal: AgentSessionJournal, + activity?: AgentSessionTurnActivity | null + ): void { + if (activity !== undefined) { + if (activity) { + this.activityBySession.set(sessionId, activity) + } else { + this.activityBySession.delete(sessionId) + } + } for (const subscriber of this.subscribers(sessionId)) { - this.deliver(subscriber, journal) + this.deliver(subscriber, journal, undefined, false, undefined, activity) + } + if (activity === undefined) { + this.hooks.onJournalPublished?.(sessionId, journal) } - this.hooks.onJournalPublished?.(sessionId, journal) } /** Force every subscriber back to a bounded tail page — recovery, epoch @@ -127,7 +146,8 @@ export class AgentSessionSubscribers { reset: reason, page, fence, - ...(backgroundTasks !== undefined ? { backgroundTasks } : {}) + ...(backgroundTasks !== undefined ? { backgroundTasks } : {}), + ...this.activityField(sessionId) }) subscriber.cursor = page.liveCursor ?? page.window.nextCursor subscriber.fence = fence @@ -148,7 +168,8 @@ export class AgentSessionSubscribers { sessionId, page, fence, - ...(backgroundTasks !== undefined ? { backgroundTasks } : {}) + ...(backgroundTasks !== undefined ? { backgroundTasks } : {}), + ...this.activityField(sessionId) }) subscriber.cursor = page.liveCursor ?? page.window.nextCursor subscriber.fence = fence @@ -205,8 +226,15 @@ export class AgentSessionSubscribers { journal: AgentSessionJournal, handoff?: AgentSessionHandoffStatus, emitCheckpoint = false, - backgroundTasks?: AgentSessionBackgroundTaskState | null + backgroundTasks?: AgentSessionBackgroundTaskState | null, + activity?: AgentSessionTurnActivity | null ): void { + const publishedActivity = + activity !== undefined + ? activity + : emitCheckpoint + ? (this.activityBySession.get(subscriber.sessionId) ?? null) + : undefined while (true) { const result = readAgentSessionHistory(journal, { sessionId: subscriber.sessionId, @@ -223,7 +251,8 @@ export class AgentSessionSubscribers { page, fence: subscriber.fence, ...(handoff ? { handoff } : {}), - ...(backgroundTasks !== undefined ? { backgroundTasks } : {}) + ...(backgroundTasks !== undefined ? { backgroundTasks } : {}), + ...(publishedActivity !== undefined ? { activity: publishedActivity } : {}) }) subscriber.cursor = page.liveCursor ?? page.window.nextCursor return @@ -231,7 +260,10 @@ export class AgentSessionSubscribers { const page = result.page const advanced = page.window.nextCursor.sequence > subscriber.cursor.sequence if (!advanced) { - if (handoff || emitCheckpoint) { + const commandsChanged = + this.hooks.readCommands !== undefined && + (this.hooks.readCommands(subscriber.sessionId) ?? null) !== subscriber.commands + if (handoff || emitCheckpoint || publishedActivity !== undefined || commandsChanged) { this.emit(subscriber, { type: 'batch', sessionId: subscriber.sessionId, @@ -243,7 +275,8 @@ export class AgentSessionSubscribers { }, fence: subscriber.fence, ...(handoff ? { handoff } : {}), - ...(backgroundTasks !== undefined ? { backgroundTasks } : {}) + ...(backgroundTasks !== undefined ? { backgroundTasks } : {}), + ...(publishedActivity !== undefined ? { activity: publishedActivity } : {}) }) } return @@ -259,7 +292,8 @@ export class AgentSessionSubscribers { }, fence: subscriber.fence, ...(handoff ? { handoff } : {}), - ...(backgroundTasks !== undefined ? { backgroundTasks } : {}) + ...(backgroundTasks !== undefined ? { backgroundTasks } : {}), + ...(publishedActivity !== undefined ? { activity: publishedActivity } : {}) }) subscriber.cursor = page.window.nextCursor if (!page.hasNewer || !this.isActive(subscriber)) { @@ -268,15 +302,20 @@ export class AgentSessionSubscribers { } } - private isActive(subscriber: Subscriber): boolean { - return this.bySession.get(subscriber.sessionId)?.get(subscriber.id) === subscriber - } + private isActive = (subscriber: Subscriber): boolean => + this.bySession.get(subscriber.sessionId)?.get(subscriber.id) === subscriber /** A dead transport cannot be allowed to turn a durable mutation into an * unknown outcome or poison every later publication. */ private emit(subscriber: Subscriber, event: AgentSessionSubscribeEvent): void { try { - subscriber.emit(event) + const commands = this.hooks.readCommands?.(subscriber.sessionId) ?? null + const includeCommands = + this.hooks.readCommands !== undefined && + event.type !== 'end' && + (event.type !== 'batch' || commands !== subscriber.commands) + subscriber.emit(includeCommands ? { ...event, commands: commands ?? null } : event) + subscriber.commands = commands } catch { this.drop(subscriber) } @@ -289,4 +328,8 @@ export class AgentSessionSubscribers { this.bySession.delete(subscriber.sessionId) } } + + private activityField(sessionId: string): { activity: AgentSessionTurnActivity | null } { + return { activity: this.activityBySession.get(sessionId) ?? null } + } } diff --git a/src/main/native-chat/agent-session-wire/structured-agent-session-turns-options.ts b/src/main/native-chat/agent-session-wire/structured-agent-session-turns-options.ts index 68e5b290847..cb1685405a1 100644 --- a/src/main/native-chat/agent-session-wire/structured-agent-session-turns-options.ts +++ b/src/main/native-chat/agent-session-wire/structured-agent-session-turns-options.ts @@ -23,5 +23,6 @@ export async function performSetOption( throw error } await ctx.persistOptions(applied ?? { [input.key]: input.value }) + ctx.publish() return { ok: true, value: { ...input, ...(applied ? { options: { ...applied } } : {}) } } } diff --git a/src/main/native-chat/transcript-line-decoders-codex-skill-context.test.ts b/src/main/native-chat/transcript-line-decoders-codex-skill-context.test.ts new file mode 100644 index 00000000000..3c7e2e50759 --- /dev/null +++ b/src/main/native-chat/transcript-line-decoders-codex-skill-context.test.ts @@ -0,0 +1,83 @@ +import { describe, expect, it } from 'vitest' +import { decodeCodexTranscriptLine } from './transcript-line-decoders-codex' + +describe('Codex transcript skill context', () => { + it.each(['message', 'response_item'])( + 'preserves prompt text and images beside a skill expansion in %s', + (type) => { + const message = { + type: 'message', + role: 'user', + content: [ + { type: 'text', text: 'Inspect this image' }, + { type: 'text', text: '\nInstructions\n' }, + { type: 'image', url: 'https://example.test/image.png' } + ] + } + const record = type === 'message' ? message : { type, payload: message } + expect(decodeCodexTranscriptLine(JSON.stringify(record), 'mixed')?.blocks).toEqual([ + { type: 'text', text: 'Inspect this image' }, + { type: 'image-ref', url: 'https://example.test/image.png' } + ]) + } + ) + + it('preserves an authoritative user event containing a literal skill wrapper', () => { + const text = 'Explain this XML' + expect( + decodeCodexTranscriptLine( + JSON.stringify({ type: 'event_msg', payload: { type: 'user_message', message: text } }), + 'submitted' + )?.blocks + ).toEqual([{ type: 'text', text }]) + }) + + it.each(['', ' \n'])( + 'drops expanded skill response items beginning with %j', + (prefix) => { + const message = { + type: 'message', + role: 'user', + content: [{ type: 'text', text: `${prefix}\nexample\nInstructions\n` }] + } + for (const record of [message, { type: 'response_item', payload: message }]) { + expect(decodeCodexTranscriptLine(JSON.stringify(record), 'context')).toBeNull() + } + } + ) + + it.each(['$example', 'Explain tags', 'user XML'])( + 'preserves the actual user prompt %j', + (text) => { + expect( + decodeCodexTranscriptLine( + JSON.stringify({ + type: 'response_item', + payload: { + type: 'message', + role: 'user', + content: [{ type: 'text', text }] + } + }), + 'user' + )?.blocks + ).toEqual([{ type: 'text', text }]) + } + ) + + it('preserves assistant explanations containing the skill wrapper', () => { + expect( + decodeCodexTranscriptLine( + JSON.stringify({ + type: 'response_item', + payload: { + type: 'message', + role: 'assistant', + content: [{ type: 'text', text: 'example' }] + } + }), + 'assistant' + )?.role + ).toBe('assistant') + }) +}) diff --git a/src/main/native-chat/transcript-line-decoders-codex.ts b/src/main/native-chat/transcript-line-decoders-codex.ts index d70bd7a7c80..229ace2a461 100644 --- a/src/main/native-chat/transcript-line-decoders-codex.ts +++ b/src/main/native-chat/transcript-line-decoders-codex.ts @@ -48,7 +48,9 @@ function codexUnwrappedResponseItem( return codexResponseItem(record, id, timestamp) } const role = record.role === 'assistant' ? 'assistant' : record.role === 'user' ? 'user' : null - const blocks = codexTurnItemBlocks(record.content) + const decodedBlocks = codexTurnItemBlocks(record.content) + const blocks = + role === 'user' ? decodedBlocks.filter((block) => !isSkillContext(block)) : decodedBlocks return role && blocks.length > 0 ? { id, role, blocks, timestamp, source: 'transcript' } : null } @@ -63,7 +65,9 @@ function codexResponseItem( if (!role) { return null } - const blocks = claudeContentBlocks(payload.content) + const decodedBlocks = claudeContentBlocks(payload.content) + const blocks = + role === 'user' ? decodedBlocks.filter((block) => !isSkillContext(block)) : decodedBlocks if (blocks.length === 0) { return null } @@ -108,6 +112,11 @@ function codexResponseItem( return null } +// Explicit skill expansions are model context, not the user's recorded prompt. +function isSkillContext(block: NativeChatBlock): boolean { + return block.type === 'text' && block.text.trimStart().slice(0, 7).toLowerCase() === '' +} + function codexEventMessage( payload: Record, id: string, diff --git a/src/main/orca-profiles/profile-cloud-auth-config.ts b/src/main/orca-profiles/profile-cloud-auth-config.ts index 09cfd8dfc6b..f6e56058935 100644 --- a/src/main/orca-profiles/profile-cloud-auth-config.ts +++ b/src/main/orca-profiles/profile-cloud-auth-config.ts @@ -19,6 +19,7 @@ const DEFAULT_SCOPE = 'openid profile email offline_access' const PRODUCTION_API_BASE_URL = 'https://login.onorca.dev' const PRODUCTION_CLIENT_ID = 'orca-desktop' const PRODUCTION_RELAY_DIRECTOR_URL = 'https://relay.onorca.dev' +const PRODUCTION_PUSH_GATEWAY_URL = 'https://push.onorca.dev' // Why: packaged main bundles never define NODE_ENV, so packaged-ness is the // only reliable production signal for gating dev-only auth escape hatches. @@ -124,6 +125,18 @@ export function getOrcaCloudAuthConfig( } } +/** + * Where the host registers phones for background push. Deliberately outside + * OrcaCloudAuthConfig: the push gateway authenticates with the host keypair, so an + * accountless host reaches it on exactly the same path as a signed-in one. + */ +export function getOrcaPushGatewayUrl( + env: NodeJS.ProcessEnv = process.env, + packaged: boolean = isPackagedOrcaBuild() +): string { + return cleanOrigin(env.ORCA_PUSH_GATEWAY_URL, !packaged) ?? PRODUCTION_PUSH_GATEWAY_URL +} + export function allowsPlaintextOrcaCloudSession( env: NodeJS.ProcessEnv = process.env, packaged: boolean = isPackagedOrcaBuild() diff --git a/src/main/persistence-deregistered-repo-residue.test.ts b/src/main/persistence-deregistered-repo-residue.test.ts index 3a7a3372b3c..a7fb4d7353f 100644 --- a/src/main/persistence-deregistered-repo-residue.test.ts +++ b/src/main/persistence-deregistered-repo-residue.test.ts @@ -1,7 +1,5 @@ // Why this file exists: deregistering a project used to strand every row it owned. No sweeper could -// reach them -- the missing-directory prune is gated on the repo still being registered, and a -// paired client's mirror of a remote host's rows is keyed by ids that client never registers, so the -// owning host's removal never reached it (#17776). +// reach them because the missing-directory prune is gated on the repo still being registered. import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest' import { rmSync, mkdtempSync } from 'node:fs' import { join } from 'node:path' @@ -103,7 +101,7 @@ describe('deregistered repo residue', () => { expect(session.sleepingAgentSessionsByPaneKey ?? {}).toEqual({}) }) - it("sweeps a remote host's session partition the owning host's removal can never reach", async () => { + it('keeps a remote session whose repo is not registered on the desktop', async () => { writeDataFile({ schemaVersion: 1, repos: [makeRepo({ id: LIVE_REPO, path: '/workspace/live' })], @@ -117,8 +115,10 @@ describe('deregistered repo residue', () => { store.flush() const partition = store.getWorkspaceSession(RUNTIME_HOST) - expect(partition.tabsByWorktree).toEqual({}) - expect(partition.activeTabTypeByWorktree).toEqual({}) + expect(partition.tabsByWorktree[GONE_WORKTREE]).toHaveLength(1) + expect(partition.activeTabTypeByWorktree).toEqual( + sessionFor(GONE_WORKTREE).activeTabTypeByWorktree + ) }) it('keeps rows for every registered repo, on any execution host', async () => { @@ -204,15 +204,13 @@ describe('deregistered repo residue', () => { schemaVersion: 1, repos: [makeRepo({ id: LIVE_REPO, path: '/workspace/live' })], worktreeMeta: {}, - workspaceSessionsByHostId: { - [RUNTIME_HOST]: { ...getDefaultWorkspaceSession(), ...session } - } + workspaceSession: { ...getDefaultWorkspaceSession(), ...session } }) const store = await createStore() store.flush() - const partition = store.getWorkspaceSession(RUNTIME_HOST) + const partition = store.getWorkspaceSession() expect(partition.activeWorktreeId ?? null).toBeNull() expect(partition.activeWorkspaceKey ?? null).toBeNull() expect(partition.activeWorktreeIdsOnShutdown ?? []).toEqual([]) diff --git a/src/main/persistence-remote-session-startup.test.ts b/src/main/persistence-remote-session-startup.test.ts new file mode 100644 index 00000000000..ddb3f57827b --- /dev/null +++ b/src/main/persistence-remote-session-startup.test.ts @@ -0,0 +1,109 @@ +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { mkdtempSync, rmSync } from 'node:fs' +import { join } from 'node:path' +import { tmpdir } from 'node:os' +import { getDefaultWorkspaceSession } from '../shared/constants' +import type { BrowserPage, BrowserWorkspace } from '../shared/browser-workspace-types' +import { createStore, makeRepo, testState } from './persistence-test-harness' + +vi.mock('./ssh/ssh-config-parser', () => ({ + loadUserSshConfig: vi.fn(), + sshConfigHostsToTargets: vi.fn() +})) +vi.mock('electron', () => ({ + app: { getPath: () => testState.dir }, + safeStorage: { isEncryptionAvailable: () => false } +})) +vi.mock('./telemetry/client', () => ({ track: vi.fn() })) +vi.mock('./telemetry/cohort-classifier', () => ({ getCohortAtEmit: vi.fn().mockReturnValue({}) })) + +const HOST = 'runtime:paired-host' +const REPO = 'remote-repo' +const WORKTREE = `${REPO}::/remote/project` +const PAGE: BrowserPage = { + id: 'page-1', + workspaceId: 'browser-1', + worktreeId: WORKTREE, + url: 'https://example.test/moved', + title: 'Moved page', + loading: false, + canGoBack: true, + canGoForward: false, + faviconUrl: null, + loadError: null, + createdAt: 1, + browserRuntimeEnvironmentId: 'paired-host', + remoteBrowserPageId: 'remote-page-1', + remoteBrowserPageClientHosted: true +} +const BROWSER: BrowserWorkspace = { + id: PAGE.workspaceId, + worktreeId: WORKTREE, + sessionProfileId: null, + activePageId: PAGE.id, + pageIds: [PAGE.id], + url: PAGE.url, + title: PAGE.title, + loading: false, + faviconUrl: null, + canGoBack: true, + canGoForward: false, + loadError: null, + createdAt: 1 +} + +function browserSession() { + return { + ...getDefaultWorkspaceSession(), + browserTabsByWorktree: { [WORKTREE]: [BROWSER] }, + browserPagesByWorkspace: { [BROWSER.id]: [PAGE] }, + activeBrowserTabIdByWorktree: { [WORKTREE]: BROWSER.id } + } +} + +describe('remote session startup ownership', () => { + beforeEach(() => { + testState.dir = mkdtempSync(join(tmpdir(), 'orca-remote-session-')) + }) + afterEach(() => { + rmSync(testState.dir, { recursive: true, force: true }) + }) + + it('keeps a paired browser row and its hosting identity across two Store reloads', () => { + const seed = createStore() + seed.addRepo(makeRepo({ id: 'local-repo', path: join(testState.dir, 'local') })) + seed.setWorkspaceSession(browserSession(), HOST) + seed.flush() + + for (let i = 0; i < 2; i += 1) { + const reloaded = createStore() + expect(reloaded.getWorkspaceSession(HOST).browserPagesByWorkspace).toEqual({ + [BROWSER.id]: [PAGE] + }) + expect(reloaded.sweepDeregisteredRepoResidue()).toEqual([]) + reloaded.flush() + } + }) + + it('retains remote metadata when no session or local catalog row names its repo', () => { + const seed = createStore() + seed.setWorktreeMetaForHost(WORKTREE, HOST, { displayName: 'Remote work' }) + seed.flush() + const reloaded = createStore() + expect(reloaded.getWorktreeMeta(WORKTREE)).toMatchObject({ displayName: 'Remote work' }) + expect(reloaded.sweepDeregisteredRepoResidue()).toEqual([]) + reloaded.flush() + }) + + it('still applies an explicit remote project removal', () => { + const seed = createStore() + seed.setWorkspaceSession(browserSession(), HOST) + seed.flush() + const reloaded = createStore() + // Assert the row survived load first, or an empty partition below would prove nothing. + expect(reloaded.getWorkspaceSession(HOST).browserPagesByWorkspace).not.toEqual({}) + reloaded.removeProjectForHost(REPO, HOST) + reloaded.flush() + expect(createStore().getWorkspaceSession(HOST).browserPagesByWorkspace).toEqual({}) + }) +}) diff --git a/src/main/persistence/loading-store/repo-lifecycle-operations.ts b/src/main/persistence/loading-store/repo-lifecycle-operations.ts index 535108845e1..e5c75c1ff01 100644 --- a/src/main/persistence/loading-store/repo-lifecycle-operations.ts +++ b/src/main/persistence/loading-store/repo-lifecycle-operations.ts @@ -136,11 +136,9 @@ export class RepoLifecycleOperations { /** * Drop every persisted row owned by a repo id that is no longer registered. * - * Runs at load because no removal path can: `removeProject` only fires while the repo is still in - * `state.repos`, and a paired client's mirror of a remote host's rows is keyed by ids that client - * never registers, so the owning host's removal never reaches it (#17776). An orphan has no owner - * that could object, so this ignores the session-ownership and local-execution-host gates the - * missing-directory sweeper needs. + * Runs at load to reach leftover local rows after deregistration. Rows owned by a `runtime:*` + * host are exempt: this runs before pairing, so their absence from the local catalog cannot + * establish deletion. Only an explicit `removeProjectForHost` retires them. */ sweepDeregisteredRepoResidue(): string[] { const state = this[repoLifecycleOperationsContext].runtime.state diff --git a/src/main/persistence/restoring-sessions/session-worktree-ownership.ts b/src/main/persistence/restoring-sessions/session-worktree-ownership.ts index e06e39dc5c5..5c54af92975 100644 --- a/src/main/persistence/restoring-sessions/session-worktree-ownership.ts +++ b/src/main/persistence/restoring-sessions/session-worktree-ownership.ts @@ -209,7 +209,7 @@ export function collectWorkspaceSessionWorktreeOwners( return owners } -function addWorkspaceSessionWorktreeOwners( +export function addWorkspaceSessionWorktreeOwners( session: WorkspaceSessionState, collector: WorktreeOwnerCandidateCollector ): void { diff --git a/src/main/persistence/tracking-repos/deregistered-repo-residue.ts b/src/main/persistence/tracking-repos/deregistered-repo-residue.ts index c52bddbb712..cf97adc11ae 100644 --- a/src/main/persistence/tracking-repos/deregistered-repo-residue.ts +++ b/src/main/persistence/tracking-repos/deregistered-repo-residue.ts @@ -1,19 +1,61 @@ import type { PersistedState } from '../../../shared/persisted-state-types' -import { getWorktreeIdFromHostIdentity } from '../../../shared/worktree/host-qualified-identity' +import { + getExecutionHostIdFromWorktreeHostIdentity, + getWorktreeIdFromHostIdentity +} from '../../../shared/worktree/host-qualified-identity' +import { parseExecutionHostId } from '../../../shared/execution-host' +import { addWorkspaceSessionWorktreeOwners } from '../restoring-sessions/session-worktree-ownership' import { splitWorktreeId } from '../../../shared/worktree/id' import type { WorkspaceSessionState } from '../../../shared/workspace-session-state-types' import { SESSION_FIELDS_PRUNED_BY_OWNER_KEY } from '../../orca-profiles/profile-project-session-field-disposition' import { ownerKeyWorktreeIds } from '../../orca-profiles/profile-project-worktree-identity' +/** A `runtime:*` host addresses a paired Orca desktop's rows, whose catalog lives on that host. */ +const isPairedHost = (hostId: string | null | undefined): boolean => + parseExecutionHostId(hostId)?.kind === 'runtime' + +/** Repo ids an owner key can name, across both readings (see `ownerKeyWorktreeIds`). */ +function ownerKeyRepoIds(ownerKey: string | null | undefined): string[] { + return ownerKey + ? ownerKeyWorktreeIds(ownerKey).flatMap((worktreeId) => { + const repoId = splitWorktreeId(worktreeId)?.repoId + return repoId ? [repoId] : [] + }) + : [] +} + /** * Repo ids that still own persisted rows but no longer appear in `state.repos`. * - * Why nothing else finds them: every other sweeper is gated on the repo still being registered, so - * deregistering a project stranded the rows it owned permanently — including a paired client's - * mirror of a remote host's session partition, which no local repo removal can reach (#17776). + * Rows owned by a `runtime:*` host are held live instead of swept: a paired client mirrors that + * host's sessions without ever registering its repos, and this runs in the Store constructor, + * before pairing, so catalog absence there proves nothing (#17776 read it as proof and deleted + * live sessions). The cost is that residue outliving a removal is no longer swept for those hosts. */ export function collectDeregisteredRepoIds(state: PersistedState): Set { const liveRepoIds = new Set(state.repos.map((repo) => repo.id)) + const retainOwner = (ownerKey: string | null | undefined): void => { + for (const repoId of ownerKeyRepoIds(ownerKey)) { + liveRepoIds.add(repoId) + } + } + // `owners` goes unread: the walker only ever calls `addOwner`. + const retainCollector = { owners: new Set(), addOwner: retainOwner } + for (const [hostId, session] of Object.entries(state.workspaceSessionsByHostId ?? {})) { + if (session && isPairedHost(hostId)) { + addWorkspaceSessionWorktreeOwners(session, retainCollector) + } + } + for (const [worktreeId, meta] of Object.entries(state.worktreeMeta)) { + if (isPairedHost(meta.hostId)) { + retainOwner(worktreeId) + } + } + for (const alias of Object.keys(state.worktreeIdentityAliases ?? {})) { + if (isPairedHost(getExecutionHostIdFromWorktreeHostIdentity(alias))) { + retainOwner(getWorktreeIdFromHostIdentity(alias)) + } + } const orphanRepoIds = new Set() // Only a full `::` locator seeds the set. A bare key -- a folder workspace id, a // repo-keyed topology revision, a test-shaped locator -- cannot be told apart from a repo id, and @@ -30,10 +72,7 @@ export function collectDeregisteredRepoIds(state: PersistedState): Set { * other reading would hand the removal pass -- which accepts either -- a live row to delete. */ const addOwnerKey = (ownerKey: string): void => { - const repoIds = ownerKeyWorktreeIds(ownerKey).flatMap((worktreeId) => { - const repoId = splitWorktreeId(worktreeId)?.repoId - return repoId ? [repoId] : [] - }) + const repoIds = ownerKeyRepoIds(ownerKey) if (repoIds.length > 0 && repoIds.every((repoId) => !liveRepoIds.has(repoId))) { for (const repoId of repoIds) { orphanRepoIds.add(repoId) diff --git a/src/main/providers/agent-foreground-process.test.ts b/src/main/providers/agent-foreground-process.test.ts index fb883d2166c..91b1e992c06 100644 --- a/src/main/providers/agent-foreground-process.test.ts +++ b/src/main/providers/agent-foreground-process.test.ts @@ -221,13 +221,13 @@ describe('resolveAgentForegroundProcess', () => { }) it('confirms a quoted login shell only when its fresh PTY tree contains shells', async () => { - mockPs(['100 99 Ss+ "/bin/zsh" -l', '101 100 S+ /bin/bash'].join('\n')) + mockPs(['100 99 100 100 Ss+ "/bin/zsh" -l', '101 100 101 100 S+ /bin/bash'].join('\n')) await expect(confirmShellForegroundProcess(100, 'zsh')).resolves.toBe(true) }) it('uses spawned-shell identity instead of a lagging foreground child label', async () => { - mockPs(['100 99 Ss+ /bin/zsh -l'].join('\n')) + mockPs(['100 99 100 100 Ss+ /bin/zsh -l'].join('\n')) await expect(confirmShellForegroundProcess(100, '/bin/zsh')).resolves.toBe(true) }) @@ -235,11 +235,11 @@ describe('resolveAgentForegroundProcess', () => { it('confirms the spawned shell behind a login wrapper while prompt hooks run', async () => { mockPs( [ - '100 99 Ss /usr/bin/login -pfl developer /bin/zsh', - '101 100 S+ -zsh', - '102 101 S+ (zsh)', - '103 102 S+ (sed)', - '104 102 R+ (git)' + '100 99 100 101 Ss /usr/bin/login -pfl developer /bin/zsh', + '101 100 101 101 S+ -zsh', + '102 101 101 101 S+ (zsh)', + '103 102 101 101 S+ (sed)', + '104 102 101 101 R+ (git)' ].join('\n') ) @@ -249,10 +249,10 @@ describe('resolveAgentForegroundProcess', () => { it('rejects a foreground nested shell while the spawned shell remains suspended', async () => { mockPs( [ - '100 99 Ss /usr/bin/login -pfl developer /bin/zsh', - '101 100 S -zsh', - '102 101 S+ agent-tui', - '103 102 S+ /bin/zsh -i' + '100 99 100 102 Ss /usr/bin/login -pfl developer /bin/zsh', + '101 100 101 102 S -zsh', + '102 101 102 102 S+ agent-tui', + '103 102 102 102 S+ /bin/zsh -i' ].join('\n') ) @@ -262,9 +262,9 @@ describe('resolveAgentForegroundProcess', () => { it('rejects shell ownership while a TUI and its nested shell remain in the PTY tree', async () => { mockPs( [ - '100 99 Ss /bin/zsh -l', - '101 100 S+ /usr/local/bin/agent-tui', - '102 101 S+ /bin/bash -i' + '100 99 100 101 Ss /bin/zsh -l', + '101 100 101 101 S+ /usr/local/bin/agent-tui', + '102 101 101 101 S+ /bin/bash -i' ].join('\n') ) @@ -272,7 +272,7 @@ describe('resolveAgentForegroundProcess', () => { }) it('rejects shell ownership while a stopped TUI remains resumable', async () => { - mockPs(['100 99 Ss+ /bin/zsh -l', '101 100 T /usr/local/bin/agent-tui'].join('\n')) + mockPs(['100 99 100 100 Ss+ /bin/zsh -l', '101 100 101 100 T agent-tui'].join('\n')) await expect(confirmShellForegroundProcess(100, 'zsh')).resolves.toBe(false) }) diff --git a/src/main/providers/agent-foreground-process.ts b/src/main/providers/agent-foreground-process.ts index 7fface8941e..69276098369 100644 --- a/src/main/providers/agent-foreground-process.ts +++ b/src/main/providers/agent-foreground-process.ts @@ -3,6 +3,7 @@ import { resolveOuterWrapperForegroundProcess } from '../../shared/foreground-wr import type { ProcessTableRow } from '../../shared/process-table-snapshot' import { getFreshProcessTableSnapshot, + getFreshShellForegroundSnapshot, getProcessTableSnapshot } from '../../shared/process-table-snapshot-reader' import { collectDescendantsFromIndex, getProcessTableIndex } from '../../shared/process-table-index' @@ -76,7 +77,7 @@ export async function confirmShellForegroundProcess( } } try { - const index = getProcessTableIndex(await getFreshProcessTableSnapshot()) + const index = getProcessTableIndex(await getFreshShellForegroundSnapshot()) const root = index.byPid.get(shellPid) if (!root) { return false diff --git a/src/main/rate-limits/grok-fetcher.test.ts b/src/main/rate-limits/grok-fetcher.test.ts index 63d7bc5cfcd..e5037f68dd4 100644 --- a/src/main/rate-limits/grok-fetcher.test.ts +++ b/src/main/rate-limits/grok-fetcher.test.ts @@ -100,6 +100,9 @@ describe('fetchGrokRateLimits', () => { ) }) + // Why: this payload emits NO usage scalars, so the omitted percent really is + // the dropped protobuf zero (#9214/#9219). #15740's payload does emit them — + // keep the two shapes apart. it('maps an omitted protobuf percentage as zero for a weekly credits period', async () => { authState.file = freshAuthJson() netFetchMock.mockResolvedValueOnce( @@ -125,6 +128,138 @@ describe('fetchGrokRateLimits', () => { expect(netFetchMock).toHaveBeenCalledTimes(1) }) + // Why: #15740 — an absent creditUsagePercent alongside explicitly-emitted zero + // credit fields means "not reported", never 0%. + it('reports usage as unavailable when the credits view omits the percent but emits explicit zero credit fields', async () => { + authState.file = freshAuthJson() + netFetchMock + .mockResolvedValueOnce( + jsonResponse({ + config: { + currentPeriod: { + type: 'USAGE_PERIOD_TYPE_WEEKLY', + start: '2026-08-16T12:54:39.515635+00:00', + end: '2026-08-23T12:54:39.515635+00:00' + }, + onDemandCap: { val: 100 }, + onDemandUsed: { val: 0 }, + isUnifiedBillingUser: true, + prepaidBalance: { val: 0 }, + topUpMethod: 'TOP_UP_METHOD_SAVED_PAYMENT_METHOD', + billingPeriodStart: '2026-08-16T12:54:39.515635+00:00', + billingPeriodEnd: '2026-08-23T12:54:39.515635+00:00' + } + }) + ) + .mockResolvedValueOnce( + jsonResponse({ + config: { + monthlyLimit: { val: 0 }, + used: { val: 37.5 }, + billingPeriodStart: '2026-08-16T12:54:39.515635+00:00', + billingPeriodEnd: '2026-08-23T12:54:39.515635+00:00' + } + }) + ) + + const result = await fetchGrokRateLimits() + expect(result.status).toBe('unavailable') + expect(result.weekly).toBeNull() + expect(result.monthly).toBeUndefined() + expect(result.error).toMatch(/did not report a usage percentage/i) + expect(netFetchMock).toHaveBeenCalledTimes(2) + }) + + // Why: the monthly budget pair is a monthly window wherever it arrives — the + // credits view must not relabel it 'Weekly credits'. + it('publishes a credits-view monthly budget pair as a monthly window without a second request', async () => { + authState.file = freshAuthJson() + netFetchMock.mockResolvedValueOnce( + jsonResponse({ + config: { + currentPeriod: { + type: 'USAGE_PERIOD_TYPE_WEEKLY', + start: '2026-08-16T12:54:39.515635+00:00', + end: '2026-08-23T12:54:39.515635+00:00' + }, + billingPeriodStart: '2026-08-16T12:54:39.515635+00:00', + billingPeriodEnd: '2026-08-23T12:54:39.515635+00:00', + monthlyLimit: { val: 100 }, + used: { val: 25 } + } + }) + ) + + const result = await fetchGrokRateLimits() + expect(result.status).toBe('ok') + expect(result.weekly).toBeNull() + expect(result.monthly?.usedPercent).toBe(25) + expect(result.monthly?.windowMinutes).toBe(43_200) + expect(netFetchMock).toHaveBeenCalledTimes(1) + }) + + // Why: #9214/#9219 — non-zero money fields never prove the encoder emits + // default zeros, so the omitted percent still reads as the dropped zero. + it('still reads an omitted percentage as zero when the payload carries only non-zero money fields', async () => { + authState.file = freshAuthJson() + netFetchMock.mockResolvedValueOnce( + jsonResponse({ + config: { + currentPeriod: { + type: 'USAGE_PERIOD_TYPE_WEEKLY', + start: '2026-07-17T19:38:56.948570+00:00', + end: '2026-07-24T19:38:56.948570+00:00' + }, + billingPeriodStart: '2026-07-17T19:38:56.948570+00:00', + billingPeriodEnd: '2026-07-24T19:38:56.948570+00:00', + onDemandCap: { val: 100 }, + prepaidBalance: { val: 25 }, + isUnifiedBillingUser: true + } + }) + ) + + const result = await fetchGrokRateLimits() + expect(result.status).toBe('ok') + expect(result.weekly?.usedPercent).toBe(0) + expect(result.weekly?.windowMinutes).toBe(10_080) + expect(netFetchMock).toHaveBeenCalledTimes(1) + }) + + it.each([{ val: 0 }, { val: '0' }])( + 'does not divide by a zero monthly limit (%o)', + async (monthlyLimit) => { + authState.file = freshAuthJson() + netFetchMock + .mockResolvedValueOnce(jsonResponse({ config: { isUnifiedBillingUser: true } })) + .mockResolvedValueOnce(jsonResponse({ config: { monthlyLimit, used: { val: 12 } } })) + + const result = await fetchGrokRateLimits() + expect(result.status).toBe('unavailable') + expect(result.weekly).toBeNull() + expect(result.monthly).toBeUndefined() + expect(result.error).toMatch(/did not report a usage percentage/i) + } + ) + + it('reads a flat billing payload that carries usage fields but no percent', async () => { + authState.file = freshAuthJson() + netFetchMock.mockResolvedValueOnce( + jsonResponse({ + monthlyLimit: { val: 200 }, + used: { val: 50 }, + billingPeriodEnd: '2026-09-01T00:00:00+00:00' + }) + ) + + const result = await fetchGrokRateLimits() + expect(result.status).toBe('ok') + expect(result.weekly).toBeNull() + expect(result.monthly?.usedPercent).toBe(25) + expect(result.monthly?.windowMinutes).toBe(43_200) + expect(netFetchMock).toHaveBeenCalledTimes(1) + }) + it('returns unavailable when not signed in even if a token-less auth file exists', async () => { authState.file = JSON.stringify({}) const result = await fetchGrokRateLimits() diff --git a/src/main/rate-limits/grok-fetcher.ts b/src/main/rate-limits/grok-fetcher.ts index 366c75c3835..33c4fac9aec 100644 --- a/src/main/rate-limits/grok-fetcher.ts +++ b/src/main/rate-limits/grok-fetcher.ts @@ -89,8 +89,9 @@ function timestampsMatch(left: string | undefined, right: string | undefined): b function hasConfirmedWeeklyPeriod(config: GrokBillingConfig): boolean { const period = config.currentPeriod - // Why: monthly unified-billing responses can also carry a weekly currentPeriod; - // matching billing bounds identify Grok's omitted protobuf zero unambiguously. + // Why: matching billing bounds only prove the current period IS the billing + // period; they say nothing about consumption (#15740), so resolveWeeklyPercent + // rules out the other consumption evidence before trusting this. return ( period?.type === 'USAGE_PERIOD_TYPE_WEEKLY' && timestampsMatch(period.start, config.billingPeriodStart) && @@ -98,12 +99,49 @@ function hasConfirmedWeeklyPeriod(config: GrokBillingConfig): boolean { ) } +function usageScalars(config: GrokBillingConfig): (GrokMoneyVal | undefined)[] { + return [ + config.onDemandCap, + config.onDemandUsed, + config.prepaidBalance, + config.monthlyLimit, + config.used + ] +} + +// Why: proto3 JSON drops default zeros, so an omitted percent can mean zero — +// but only an explicitly-emitted zero proves this encoder keeps them. #15740 +// ships `onDemandUsed: {val: 0}`, so there the omission means "not reported" +// and must never render as 0%. Non-zero money fields prove nothing either way, +// so #9214/#9219 accounts that carry only those keep their genuine 0%. +function emitsExplicitZeroScalar(config: GrokBillingConfig): boolean { + return usageScalars(config).some((value) => parseMoneyVal(value) === 0) +} + +function reportsAnyUsageScalar(config: GrokBillingConfig): boolean { + return usageScalars(config).some((value) => parseMoneyVal(value) !== null) +} + +function resolveWeeklyPercent(config: GrokBillingConfig): number | null { + const reported = config.creditUsagePercent + if (typeof reported === 'number' && Number.isFinite(reported)) { + return reported + } + if (reported !== undefined) { + return null + } + // Why: infer the dropped zero only when nothing else in the payload speaks + // for consumption — an explicit zero proves the encoder keeps defaults, and a + // computable budget pair is a real monthly number this must not shadow. + if (emitsExplicitZeroScalar(config) || mapMonthlyUsage(config) !== null) { + return null + } + return hasConfirmedWeeklyPeriod(config) ? 0 : null +} + function mapWeeklyCredits(config: GrokBillingConfig): RateLimitWindow | null { - const usedPercent = - config.creditUsagePercent === undefined && hasConfirmedWeeklyPeriod(config) - ? 0 - : config.creditUsagePercent - if (typeof usedPercent !== 'number' || !Number.isFinite(usedPercent)) { + const usedPercent = resolveWeeklyPercent(config) + if (usedPercent === null) { return null } const periodEnd = config.currentPeriod?.end ?? config.billingPeriodEnd @@ -125,13 +163,16 @@ function parseMoneyVal(value: GrokMoneyVal | undefined): number | null { function mapMonthlyUsage(config: GrokBillingConfig): RateLimitWindow | null { const limit = parseMoneyVal(config.monthlyLimit) const used = parseMoneyVal(config.used) + // Why: a zero, missing or unparseable denominator yields no window rather + // than NaN/Infinity or a fabricated 0%. if (limit === null || used === null || limit <= 0) { return null } + const usedPercent = Math.min(100, Math.max(0, (used / limit) * 100)) const periodEnd = config.currentPeriod?.end ?? config.billingPeriodEnd const resetsAt = periodEnd ? Date.parse(periodEnd) : null return { - usedPercent: Math.min(100, Math.max(0, (used / limit) * 100)), + usedPercent, windowMinutes: MONTHLY_WINDOW_MINUTES, resetsAt: resetsAt !== null && Number.isFinite(resetsAt) ? resetsAt : null, resetDescription: parseResetDescription(periodEnd) @@ -150,14 +191,26 @@ function grokRequestHeaders(session: GrokAuthSession): Record { return headers } +// Why: a flat response can carry monthly/on-demand fields and no percent at all +// (#15740); keying only on creditUsagePercent misreported those as "no config". +const FLAT_BILLING_FIELDS: readonly (keyof GrokBillingConfig)[] = [ + 'creditUsagePercent', + 'currentPeriod', + 'billingPeriodStart', + 'billingPeriodEnd', + 'subscriptionTier', + 'monthlyLimit', + 'used', + 'onDemandCap', + 'onDemandUsed', + 'prepaidBalance' +] + function resolveBillingConfig(data: GrokBillingResponse): GrokBillingConfig | null { if (data.config) { return data.config } - if (typeof data.creditUsagePercent === 'number') { - return data - } - return null + return FLAT_BILLING_FIELDS.some((field) => data[field] !== undefined) ? data : null } function billingUsageResult( @@ -219,7 +272,7 @@ async function fetchBillingData( } type GrokMonthlyFallbackOutcome = - | { kind: 'window'; window: RateLimitWindow | null } + | { kind: 'window'; window: RateLimitWindow | null; config: GrokBillingConfig } | { kind: 'result'; result: ProviderRateLimits } // Why: request failures propagate as 'error' (thrown errors reach the caller's @@ -235,7 +288,7 @@ async function fetchMonthlyUsageFallback( return outcome } const config = outcome.data.config ?? outcome.data - return { kind: 'window', window: mapMonthlyUsage(config) } + return { kind: 'window', window: mapMonthlyUsage(config), config } } // Why: Orca never runs grok login; it only reads the session file the CLI updates. @@ -277,6 +330,13 @@ export async function fetchGrokRateLimits( if (weekly) { return billingUsageResult({ weekly }, config, session) } + // Why: the credits view can already carry the monthly budget pair; that pair + // is a monthly window, so publish it as one rather than mislabelling it + // weekly — and skip the redundant second request. + const creditsMonthly = mapMonthlyUsage(config) + if (creditsMonthly) { + return billingUsageResult({ monthly: creditsMonthly }, config, session) + } // Why: some unified-billing accounts expose only a monthly included budget; // their credits view omits creditUsagePercent, so read the default view. const fallback = await fetchMonthlyUsageFallback(session, options.signal) @@ -286,7 +346,11 @@ export async function fetchGrokRateLimits( if (fallback.window) { return billingUsageResult({ monthly: fallback.window }, config, session) } - return result('unavailable', 'Grok billing response did not include credit usage') + // Why: an account that reports spend fields but no computable percentage is + // not a quota-less plan — say the usage is unknown instead of implying zero. + return reportsAnyUsageScalar(config) || reportsAnyUsageScalar(fallback.config) + ? result('unavailable', 'Grok did not report a usage percentage for this account') + : result('unavailable', 'Grok billing response did not include credit usage') } catch (err) { return result('error', err instanceof Error ? err.message : 'Grok usage request failed') } diff --git a/src/main/rate-limits/minimax-fetcher-data.ts b/src/main/rate-limits/minimax-fetcher-data.ts new file mode 100644 index 00000000000..b2c6eddad6e --- /dev/null +++ b/src/main/rate-limits/minimax-fetcher-data.ts @@ -0,0 +1,149 @@ +import type { ProviderRateLimits, RateLimitWindow } from '../../shared/rate-limit-types' + +// Why: pure data-shape helpers for the MiniMax Coding Plan API. Lives in its +// own file so both minimax-fetcher.ts (transport) and minimax-fetcher-parse.ts +// (response handling) can import without creating a dependency cycle. + +export type MiniMaxUsageItem = { + model_name?: unknown + current_interval_remaining_percent?: unknown + start_time?: unknown + end_time?: unknown + remains_time?: unknown + // Why: Coding Plan also reports a separate 7-day quota. The API returns the + // raw remaining percent against the un-boosted base; the opencode-tku + // equivalent uses the value as-is. weekly_boost_permille exists in the + // payload but is intentionally not parsed yet (see handleMiniMaxWeeklyBoost). + current_weekly_remaining_percent?: unknown + weekly_remains_time?: unknown + weekly_boost_permille?: unknown +} + +export type MiniMaxUsageSnapshot = { + modelName: string + // Why: weekly may be absent if the API omits it (older schema, mid-migration + // window). Session is required (matches the existing parseUsageItem contract). + session: RateLimitWindow + weekly: RateLimitWindow | null +} + +export type MiniMaxModelList = string | readonly string[] | null | undefined + +export function makeMiniMaxUnavailable(error: string): ProviderRateLimits { + return { + provider: 'minimax', + session: null, + weekly: null, + updatedAt: Date.now(), + error, + status: 'unavailable', + usageMetadata: { failureKind: 'missing-credentials', source: 'web' } + } +} + +export function makeMiniMaxError( + error: string, + failureKind: NonNullable['failureKind'] +): ProviderRateLimits { + return { + provider: 'minimax', + session: null, + weekly: null, + updatedAt: Date.now(), + error, + status: 'error', + usageMetadata: { failureKind, source: 'web' } + } +} + +function clampPercent(value: number): number { + return Math.max(0, Math.min(100, Math.round(value))) +} + +function asNumber(value: unknown): number | null { + if (typeof value === 'number' && Number.isFinite(value)) { + return value + } + if (typeof value === 'string' && value.trim()) { + const parsed = Number(value) + return Number.isFinite(parsed) ? parsed : null + } + return null +} + +export function parseMiniMaxModels(models: MiniMaxModelList): string[] { + if (Array.isArray(models)) { + const parsed = models.map((model) => model.trim()).filter(Boolean) + return parsed.length > 0 ? parsed : ['general'] + } + if (typeof models === 'string') { + const parsed = models + .split(',') + .map((model) => model.trim()) + .filter(Boolean) + return parsed.length > 0 ? parsed : ['general'] + } + return ['general'] +} + +// Why: MiniMax's API returns `end_time - start_time` that can drift below the +// 5-hour bucket (e.g. 4h or 295 min). The UI labels must reflect the contracted +// session — a fixed 5-hour window — so the status bar reads "5h" regardless of +// what the API reports. Mirrors how Codex always reports 300/10080 minutes. +const MINIMAX_SESSION_WINDOW_MINUTES = 300 +// Why: 7-day window. The API doesn't expose a `weekly_end_time` analog of the +// session's end_time, so we label the chip via windowMinutes + a relative +// `resetsAt` derived from `weekly_remains_time` + now. +const MINIMAX_WEEKLY_WINDOW_MINUTES = 10080 + +export function parseMiniMaxUsageItem(value: unknown): MiniMaxUsageSnapshot | null { + if (!value || typeof value !== 'object' || Array.isArray(value)) { + return null + } + const item: MiniMaxUsageItem = value + const modelName = typeof item.model_name === 'string' ? item.model_name : null + const remainingPercent = asNumber(item.current_interval_remaining_percent) + const startTime = asNumber(item.start_time) + const endTime = asNumber(item.end_time) + if (!modelName || remainingPercent === null || startTime === null || endTime === null) { + return null + } + const session: RateLimitWindow = { + usedPercent: clampPercent(100 - remainingPercent), + windowMinutes: MINIMAX_SESSION_WINDOW_MINUTES, + resetsAt: endTime, + resetDescription: null + } + const weekly = parseMiniMaxWeeklyWindow(item) + return { modelName, session, weekly } +} + +export function parseMiniMaxWeeklyWindow(item: MiniMaxUsageItem): RateLimitWindow | null { + const weeklyRemaining = asNumber(item.current_weekly_remaining_percent) + if (weeklyRemaining === null) { + return null + } + // Why: `weekly_remains_time` is a duration (matches `remains_time` units + // for the 5h window). Anchor to `Date.now()` so the status bar's + // countdown stays in lockstep with the session window shape. + const weeklyRemainsMs = asNumber(item.weekly_remains_time) + return { + usedPercent: clampPercent(100 - weeklyRemaining), + windowMinutes: MINIMAX_WEEKLY_WINDOW_MINUTES, + resetsAt: weeklyRemainsMs != null ? Date.now() + weeklyRemainsMs : null, + resetDescription: null + } +} + +export function selectMiniMaxSnapshot( + snapshots: MiniMaxUsageSnapshot[], + preferredModels: string[] +): MiniMaxUsageSnapshot | null { + for (const model of preferredModels) { + const match = snapshots.find((snapshot) => snapshot.modelName === model) + if (match) { + return match + } + } + return snapshots.length === 1 ? snapshots[0] : null +} diff --git a/src/main/rate-limits/minimax-fetcher-parse.ts b/src/main/rate-limits/minimax-fetcher-parse.ts new file mode 100644 index 00000000000..98efb88cd6d --- /dev/null +++ b/src/main/rate-limits/minimax-fetcher-parse.ts @@ -0,0 +1,134 @@ +import type { ProviderRateLimits } from '../../shared/rate-limit-types' +import { + logMiniMaxFetchFailure, + redactMiniMaxSecret, + type MiniMaxFetchResponse +} from './minimax-request-context' +import { + makeMiniMaxError, + parseMiniMaxModels, + parseMiniMaxUsageItem, + selectMiniMaxSnapshot, + type MiniMaxModelList, + type MiniMaxUsageSnapshot +} from './minimax-fetcher-data' + +// Why: split out of minimax-fetcher.ts so the transport + routing file +// stays under the 300-line cap (AGENTS.md disallows max-lines disables). +// Pure data-shape → ProviderRateLimits translation; no I/O. + +export type MiniMaxUsageResponse = { + base_resp?: { + status_code?: unknown + status_msg?: unknown + } + model_remains?: { + model_name?: unknown + current_interval_remaining_percent?: unknown + start_time?: unknown + end_time?: unknown + remains_time?: unknown + current_weekly_remaining_percent?: unknown + weekly_remains_time?: unknown + weekly_boost_permille?: unknown + }[] +} + +function handleMiniMaxHttpError(fetchResult: MiniMaxFetchResponse): ProviderRateLimits | null { + const { response } = fetchResult + if (response.status === 401 || response.status === 403) { + logMiniMaxFetchFailure({ + transport: fetchResult.transport, + responseStatus: response.status, + cookieNames: fetchResult.cookieNames, + requestHeaderNames: fetchResult.requestHeaderNames + }) + const credentialLabel = fetchResult.transport === 'api-key' ? 'API key' : 'session cookie' + return makeMiniMaxError( + `MiniMax ${credentialLabel} expired. Replace it in Settings.`, + 'stale-token' + ) + } + if (!response.ok) { + logMiniMaxFetchFailure({ + transport: fetchResult.transport, + responseStatus: response.status, + cookieNames: fetchResult.cookieNames, + requestHeaderNames: fetchResult.requestHeaderNames + }) + return makeMiniMaxError(`MiniMax usage fetch failed (${response.status})`, 'server') + } + return null +} + +function handleMiniMaxPayloadError( + fetchResult: MiniMaxFetchResponse, + payload: MiniMaxUsageResponse +): ProviderRateLimits | null { + const statusCode = payload.base_resp?.status_code + if (statusCode === undefined || statusCode === 0) { + return null + } + logMiniMaxFetchFailure({ + transport: fetchResult.transport, + responseStatus: fetchResult.response.status, + statusCode, + statusMsg: payload.base_resp?.status_msg, + cookieNames: fetchResult.cookieNames, + requestHeaderNames: fetchResult.requestHeaderNames + }) + const message = + typeof payload.base_resp?.status_msg === 'string' + ? payload.base_resp.status_msg + : 'MiniMax returned an error' + return makeMiniMaxError(redactMiniMaxSecret(message), 'usage-unavailable') +} + +export async function parseMiniMaxUsageResponse( + fetchResult: MiniMaxFetchResponse, + models: MiniMaxModelList +): Promise { + const httpError = handleMiniMaxHttpError(fetchResult) + if (httpError) { + return httpError + } + let payload: MiniMaxUsageResponse + try { + const value: unknown = await fetchResult.response.json() + if (!value || typeof value !== 'object' || Array.isArray(value)) { + return makeMiniMaxError('Invalid MiniMax usage response', 'parse') + } + payload = value + } catch (error) { + const message = error instanceof Error ? error.message : 'Invalid MiniMax usage response' + return makeMiniMaxError(redactMiniMaxSecret(message), 'parse') + } + const payloadError = handleMiniMaxPayloadError(fetchResult, payload) + if (payloadError) { + return payloadError + } + // Why: a non-array `model_remains` (object / string) throws inside `.map` + // and surfaces as a 'network' error rather than 'parse'. Treat any + // non-array as an empty list and let the snapshot selection flag the + // missing usage. + const rawItems = Array.isArray(payload.model_remains) ? payload.model_remains : [] + const snapshots = rawItems + .map(parseMiniMaxUsageItem) + .filter((snapshot): snapshot is MiniMaxUsageSnapshot => snapshot !== null) + const selected = selectMiniMaxSnapshot(snapshots, parseMiniMaxModels(models)) + if (!selected) { + return makeMiniMaxError( + 'MiniMax usage data for the configured model was not found', + 'usage-unavailable' + ) + } + return { + provider: 'minimax', + session: selected.session, + weekly: selected.weekly, + updatedAt: Date.now(), + error: null, + status: 'ok', + usageMetadata: { source: 'web' } + } +} diff --git a/src/main/rate-limits/minimax-fetcher.test.ts b/src/main/rate-limits/minimax-fetcher.test.ts index 2336825e512..f3ca0709aba 100644 --- a/src/main/rate-limits/minimax-fetcher.test.ts +++ b/src/main/rate-limits/minimax-fetcher.test.ts @@ -83,6 +83,35 @@ describe('fetchMiniMaxRateLimits', () => { vi.restoreAllMocks() }) + it.each([null, [], 'invalid', 42])( + 'rejects invalid payload %j as a parse error', + async (payload) => { + netFetchMock.mockResolvedValueOnce(makeResponse(payload)) + const result = await fetchMiniMaxRateLimits({ cookie: FULL_COOKIE }) + expect(result.usageMetadata?.failureKind).toBe('parse') + } + ) + + it('skips malformed usage entries without losing valid usage', async () => { + netFetchMock.mockResolvedValueOnce( + makeResponse({ + model_remains: [ + null, + 'invalid', + { + model_name: 'general', + current_interval_remaining_percent: 25, + start_time: Date.now(), + end_time: Date.now() + 300 * 60_000 + } + ] + }) + ) + const result = await fetchMiniMaxRateLimits({ cookie: FULL_COOKIE }) + expect(result.status).toBe('ok') + expect(result.session?.usedPercent).toBe(75) + }) + it('returns unavailable when cookie is empty', async () => { const result = await fetchMiniMaxRateLimits({ cookie: '' }) expect(result.status).toBe('unavailable') @@ -116,7 +145,7 @@ describe('fetchMiniMaxRateLimits', () => { const result = await fetchMiniMaxRateLimits({ cookie: FULL_COOKIE }) expect(result.status).toBe('error') expect(result.usageMetadata?.failureKind).toBe('stale-token') - expect(result.error).toMatch(/session expired/i) + expect(result.error).toMatch(/session cookie expired/i) }) it('classifies 403 as stale-token', async () => { @@ -450,6 +479,146 @@ describe('fetchMiniMaxRateLimits', () => { }) ) }) + + it('routes CN + API key to the bearer transport and hits www.minimaxi.com', async () => { + netFetchMock.mockResolvedValueOnce(makeResponse(makeOkPayload(72))) + const result = await fetchMiniMaxRateLimits({ + cookie: FULL_COOKIE, + apiKey: 'sk-test-1234567890', + endpointMode: 'cn' + }) + expect(result.status).toBe('ok') + expect(result.session?.usedPercent).toBe(28) + const [url, init] = netFetchMock.mock.calls[0] + expect(url).toBe('https://www.minimaxi.com/v1/api/openplatform/coding_plan/remains') + expect(init.headers.Authorization).toBe('Bearer sk-test-1234567890') + // Why: the API key path must not set browser-shaped headers — they're + // there to defeat hotlink protection on the cookie path and would only + // look like scraping on the API key path. + expect(init.headers.Referer).toBeUndefined() + expect(init.headers['User-Agent']).toBeUndefined() + expect(init.headers.Cookie).toBeUndefined() + // Why: cookie jar / session partition must not be touched on the bearer + // path, since the user is not using cookies for this endpoint mode. + expect(cookiesSetMock).not.toHaveBeenCalled() + expect(sessionFromPartitionMock).not.toHaveBeenCalled() + }) + + it('falls back to the cookie transport when no API key is provided', async () => { + netFetchMock.mockResolvedValueOnce(makeResponse(makeOkPayload(60))) + const result = await fetchMiniMaxRateLimits({ + cookie: FULL_COOKIE, + endpointMode: 'cn' + }) + expect(result.status).toBe('ok') + const [url, init] = netFetchMock.mock.calls[0] + expect(url).toBe('https://www.minimaxi.com/v1/api/openplatform/coding_plan/remains') + expect(init.headers.Authorization).toBeUndefined() + expect(init.headers.Cookie).toBeUndefined() + // Why: cookie transport relies on the session jar — verify it was + // populated for the CN host so the .io-only Referer mock doesn't crash. + expect(cookiesSetMock).toHaveBeenCalled() + }) + + it('routes to API key on the overseas endpoint when both are configured', async () => { + // Why: the API key path is a Bearer header — endpoint-agnostic, the + // endpoint only picks the host URL. Either auth works on either endpoint. + netFetchMock.mockResolvedValueOnce(makeResponse(makeOkPayload(50))) + const result = await fetchMiniMaxRateLimits({ + cookie: FULL_COOKIE, + apiKey: 'sk-overseas-key', + endpointMode: 'overseas' + }) + expect(result.status).toBe('ok') + const [url, init] = netFetchMock.mock.calls[0] + expect(url).toBe('https://platform.minimax.io/v1/api/openplatform/coding_plan/remains') + expect(init.headers.Authorization).toBe('Bearer sk-overseas-key') + // Why: API key path skips the cookie jar — cookiesSetMock is never called. + expect(cookiesSetMock).not.toHaveBeenCalled() + }) + + it('classifies a 401 on the API key path as a stale API key error', async () => { + netFetchMock.mockResolvedValueOnce(makeResponse({}, 401)) + const result = await fetchMiniMaxRateLimits({ + apiKey: 'sk-stale', + endpointMode: 'cn' + }) + expect(result.status).toBe('error') + expect(result.usageMetadata?.failureKind).toBe('stale-token') + expect(result.error).toMatch(/API key expired/i) + }) + + it('parses the 7-day weekly window alongside the 5-hour session', async () => { + const now = Date.now() + netFetchMock.mockResolvedValueOnce( + makeResponse({ + base_resp: { status_code: 0, status_msg: 'ok' }, + model_remains: [ + { + model_name: 'general', + current_interval_remaining_percent: 80, + start_time: now - 60_000, + end_time: now + 5 * 60 * 60 * 1000, + remains_time: 5 * 60 * 60 * 1000, + current_weekly_remaining_percent: 45, + weekly_remains_time: 3 * 24 * 60 * 60 * 1000, + weekly_boost_permille: 1500 + } + ] + }) + ) + const result = await fetchMiniMaxRateLimits({ cookie: FULL_COOKIE }) + expect(result.status).toBe('ok') + expect(result.session?.usedPercent).toBe(20) + expect(result.session?.windowMinutes).toBe(300) + expect(result.weekly).not.toBeNull() + expect(result.weekly?.usedPercent).toBe(55) + expect(result.weekly?.windowMinutes).toBe(10080) + // Why: resetsAt is anchored to now + the API's reported duration, so the + // status-bar countdown stays consistent with the session shape. + const weeklyResets = result.weekly?.resetsAt ?? 0 + expect(weeklyResets).toBeGreaterThan(now + 3 * 24 * 60 * 60 * 1000 - 5_000) + expect(weeklyResets).toBeLessThan(now + 3 * 24 * 60 * 60 * 1000 + 5_000) + }) + + it('returns weekly = null when the API omits weekly fields', async () => { + // Why: matches the existing makeOkPayload (no weekly fields) — guards + // against the case where the upstream rolls back to the 5h-only schema. + netFetchMock.mockResolvedValueOnce(makeResponse(makeOkPayload(50))) + const result = await fetchMiniMaxRateLimits({ cookie: FULL_COOKIE }) + expect(result.status).toBe('ok') + expect(result.session?.usedPercent).toBe(50) + expect(result.weekly).toBeNull() + }) + + it('parses weekly usedPercent when weekly_remains_time is missing', async () => { + // Why: some accounts report the percent without a reset duration; the + // status bar falls back to the "wk" label via formatWindowLabel in that + // case. Don't drop the percent just because resetsAt is unknown. + const now = Date.now() + netFetchMock.mockResolvedValueOnce( + makeResponse({ + base_resp: { status_code: 0, status_msg: 'ok' }, + model_remains: [ + { + model_name: 'general', + current_interval_remaining_percent: 90, + start_time: now - 60_000, + end_time: now + 5 * 60 * 60 * 1000, + remains_time: 5 * 60 * 60 * 1000, + current_weekly_remaining_percent: 70 + } + ] + }) + ) + const result = await fetchMiniMaxRateLimits({ cookie: FULL_COOKIE }) + expect(result.status).toBe('ok') + expect(result.weekly).toMatchObject({ + usedPercent: 30, + windowMinutes: 10080, + resetsAt: null + }) + }) }) describe('normalizeMiniMaxCookieHeader', () => { diff --git a/src/main/rate-limits/minimax-fetcher.ts b/src/main/rate-limits/minimax-fetcher.ts index 20348feee72..a56edb2d257 100644 --- a/src/main/rate-limits/minimax-fetcher.ts +++ b/src/main/rate-limits/minimax-fetcher.ts @@ -1,16 +1,24 @@ -import type { ProviderRateLimits, RateLimitWindow } from '../../shared/rate-limit-types' +import type { ProviderRateLimits } from '../../shared/rate-limit-types' +import type { MiniMaxEndpoint } from '../../shared/global-settings-types' import { extractMiniMaxCookieValue, + fetchMiniMaxWithApiKey, fetchMiniMaxWithManualCookieHeader, fetchMiniMaxWithSessionCookieJar, + getMiniMaxEndpointUrl, getUniqueMiniMaxCookieNames, - logMiniMaxFetchFailure, makeMiniMaxRequestHeaders, - MINIMAX_USAGE_ENDPOINT, + MINIMAX_API_KEY_TIMEOUT_MS, normalizeMiniMaxCookieHeader, redactMiniMaxSecret, type MiniMaxFetchResponse } from './minimax-request-context' +import { parseMiniMaxUsageResponse } from './minimax-fetcher-parse' +import { + makeMiniMaxError, + makeMiniMaxUnavailable, + type MiniMaxModelList +} from './minimax-fetcher-data' export { extractMiniMaxCookieValue, @@ -20,133 +28,20 @@ export { const API_TIMEOUT_MS = 15_000 -type MiniMaxUsageItem = { - model_name?: unknown - current_interval_remaining_percent?: unknown - start_time?: unknown - end_time?: unknown - remains_time?: unknown -} - -type MiniMaxUsageResponse = { - base_resp?: { - status_code?: unknown - status_msg?: unknown - } - model_remains?: MiniMaxUsageItem[] -} - -type MiniMaxUsageSnapshot = { - modelName: string - window: RateLimitWindow -} - export type FetchMiniMaxRateLimitsOptions = { - cookie: string + cookie?: string groupId?: string | null - models?: string | readonly string[] | null + models?: MiniMaxModelList endpoint?: string + endpointMode?: MiniMaxEndpoint + apiKey?: string | null } -function clampPercent(value: number): number { - return Math.max(0, Math.min(100, Math.round(value))) -} - -function makeUnavailable(error: string): ProviderRateLimits { - return { - provider: 'minimax', - session: null, - weekly: null, - updatedAt: Date.now(), - error, - status: 'unavailable', - usageMetadata: { failureKind: 'missing-credentials', source: 'web' } - } -} - -function makeError( - error: string, - failureKind: NonNullable['failureKind'] -): ProviderRateLimits { - return { - provider: 'minimax', - session: null, - weekly: null, - updatedAt: Date.now(), - error, - status: 'error', - usageMetadata: { failureKind, source: 'web' } - } -} - -function parseModels(models: FetchMiniMaxRateLimitsOptions['models']): string[] { - if (Array.isArray(models)) { - const parsed = models.map((model) => model.trim()).filter(Boolean) - return parsed.length > 0 ? parsed : ['general'] - } - if (typeof models === 'string') { - const parsed = models - .split(',') - .map((model) => model.trim()) - .filter(Boolean) - return parsed.length > 0 ? parsed : ['general'] - } - return ['general'] -} - -function asNumber(value: unknown): number | null { - if (typeof value === 'number' && Number.isFinite(value)) { - return value - } - if (typeof value === 'string' && value.trim()) { - const parsed = Number(value) - return Number.isFinite(parsed) ? parsed : null - } - return null -} - -// Why: MiniMax's API returns `end_time - start_time` that can drift below the -// 5-hour bucket (e.g. 4h or 295 min). The UI labels must reflect the contracted -// session — a fixed 5-hour window — so the status bar reads "5h" regardless of -// what the API reports. Mirrors how Codex always reports 300/10080 minutes. -const MINIMAX_SESSION_WINDOW_MINUTES = 300 - -function parseUsageItem(item: MiniMaxUsageItem): MiniMaxUsageSnapshot | null { - const modelName = typeof item.model_name === 'string' ? item.model_name : null - const remainingPercent = asNumber(item.current_interval_remaining_percent) - const startTime = asNumber(item.start_time) - const endTime = asNumber(item.end_time) - if (!modelName || remainingPercent === null || startTime === null || endTime === null) { - return null - } - return { - modelName, - window: { - usedPercent: clampPercent(100 - remainingPercent), - windowMinutes: MINIMAX_SESSION_WINDOW_MINUTES, - resetsAt: endTime, - resetDescription: null - } - } -} - -function selectSnapshot( - snapshots: MiniMaxUsageSnapshot[], - preferredModels: string[] -): MiniMaxUsageSnapshot | null { - for (const model of preferredModels) { - const match = snapshots.find((snapshot) => snapshot.modelName === model) - if (match) { - return match - } - } - return snapshots.length === 1 ? snapshots[0] : null -} - -async function fetchMiniMaxResponse(args: { +async function fetchMiniMaxResponseWithCookie(args: { cookie: string endpoint: string groupId: string | null + endpointMode: MiniMaxEndpoint signal: AbortSignal }): Promise { try { @@ -159,72 +54,37 @@ async function fetchMiniMaxResponse(args: { { error: redactMiniMaxSecret(message), cookieNames: getUniqueMiniMaxCookieNames(args.cookie), - requestHeaderNames: Object.keys(makeMiniMaxRequestHeaders(args.groupId)) + requestHeaderNames: Object.keys(makeMiniMaxRequestHeaders(args.groupId, args.endpointMode)) } ) return await fetchMiniMaxWithManualCookieHeader(args) } } -function handleMiniMaxHttpError(fetchResult: MiniMaxFetchResponse): ProviderRateLimits | null { - const { response } = fetchResult - if (response.status === 401 || response.status === 403) { - logMiniMaxFetchFailure({ - transport: fetchResult.transport, - responseStatus: response.status, - cookieNames: fetchResult.cookieNames, - requestHeaderNames: fetchResult.requestHeaderNames - }) - return makeError( - 'MiniMax session expired. Replace the MiniMax cookie in Settings.', - 'stale-token' - ) - } - if (!response.ok) { - logMiniMaxFetchFailure({ - transport: fetchResult.transport, - responseStatus: response.status, - cookieNames: fetchResult.cookieNames, - requestHeaderNames: fetchResult.requestHeaderNames - }) - return makeError(`MiniMax usage fetch failed (${response.status})`, 'server') - } - return null -} - -function handleMiniMaxPayloadError( - fetchResult: MiniMaxFetchResponse, - payload: MiniMaxUsageResponse -): ProviderRateLimits | null { - const statusCode = payload.base_resp?.status_code - if (statusCode === undefined || statusCode === 0) { - return null - } - logMiniMaxFetchFailure({ - transport: fetchResult.transport, - responseStatus: fetchResult.response.status, - statusCode, - statusMsg: payload.base_resp?.status_msg, - cookieNames: fetchResult.cookieNames, - requestHeaderNames: fetchResult.requestHeaderNames - }) - const message = - typeof payload.base_resp?.status_msg === 'string' - ? payload.base_resp.status_msg - : 'MiniMax returned an error' - return makeError(redactMiniMaxSecret(message), 'usage-unavailable') -} - export async function fetchMiniMaxRateLimits( options: FetchMiniMaxRateLimitsOptions ): Promise { - const rawCookie = options.cookie.trim() + const rawCookie = options.cookie?.trim() ?? '' + const rawApiKey = options.apiKey?.trim() ?? '' + const endpointMode: MiniMaxEndpoint = options.endpointMode ?? 'overseas' + const endpoint = options.endpoint ?? getMiniMaxEndpointUrl(endpointMode) + + const useApiKey = rawApiKey.length > 0 + + if (useApiKey) { + return await fetchMiniMaxWithApiKeyFlow({ + apiKey: rawApiKey, + endpoint, + models: options.models + }) + } + if (!rawCookie) { - return makeUnavailable('MiniMax session cookie not configured') + return makeMiniMaxUnavailable('MiniMax session cookie not configured') } const cookie = normalizeMiniMaxCookieHeader(rawCookie) if (!extractMiniMaxCookieValue(cookie, '_token')) { - return makeError( + return makeMiniMaxError( 'MiniMax auth cookie not found — paste a Cookie header with _token', 'missing-credentials' ) @@ -232,48 +92,34 @@ export async function fetchMiniMaxRateLimits( const groupId = options.groupId?.trim() || extractMiniMaxCookieValue(cookie, 'minimax_group_id_v2') try { - const fetchResult = await fetchMiniMaxResponse({ + const fetchResult = await fetchMiniMaxResponseWithCookie({ cookie, - endpoint: options.endpoint ?? MINIMAX_USAGE_ENDPOINT, + endpoint, groupId, + endpointMode, signal: AbortSignal.timeout(API_TIMEOUT_MS) }) - const httpError = handleMiniMaxHttpError(fetchResult) - if (httpError) { - return httpError - } - let payload: MiniMaxUsageResponse - try { - payload = (await fetchResult.response.json()) as MiniMaxUsageResponse - } catch (error) { - const message = error instanceof Error ? error.message : 'Invalid MiniMax usage response' - return makeError(redactMiniMaxSecret(message), 'parse') - } - const payloadError = handleMiniMaxPayloadError(fetchResult, payload) - if (payloadError) { - return payloadError - } - const snapshots = (payload.model_remains ?? []) - .map(parseUsageItem) - .filter((snapshot): snapshot is MiniMaxUsageSnapshot => snapshot !== null) - const selected = selectSnapshot(snapshots, parseModels(options.models)) - if (!selected) { - return makeError( - 'MiniMax usage data for the configured model was not found', - 'usage-unavailable' - ) - } - return { - provider: 'minimax', - session: selected.window, - weekly: null, - updatedAt: Date.now(), - error: null, - status: 'ok', - usageMetadata: { source: 'web' } - } + return await parseMiniMaxUsageResponse(fetchResult, options.models) } catch (error) { const message = error instanceof Error ? error.message : 'Unknown MiniMax usage error' - return makeError(redactMiniMaxSecret(message), 'network') + return makeMiniMaxError(redactMiniMaxSecret(message), 'network') + } +} + +async function fetchMiniMaxWithApiKeyFlow(args: { + apiKey: string + endpoint: string + models: MiniMaxModelList +}): Promise { + try { + const fetchResult = await fetchMiniMaxWithApiKey({ + apiKey: args.apiKey, + endpoint: args.endpoint, + signal: AbortSignal.timeout(MINIMAX_API_KEY_TIMEOUT_MS) + }) + return await parseMiniMaxUsageResponse(fetchResult, args.models) + } catch (error) { + const message = error instanceof Error ? error.message : 'Unknown MiniMax API key error' + return makeMiniMaxError(redactMiniMaxSecret(message), 'network') } } diff --git a/src/main/rate-limits/minimax-request-context.test.ts b/src/main/rate-limits/minimax-request-context.test.ts index 9b01b4ff013..42e98c74588 100644 --- a/src/main/rate-limits/minimax-request-context.test.ts +++ b/src/main/rate-limits/minimax-request-context.test.ts @@ -22,8 +22,10 @@ vi.mock('electron', () => ({ import { clearMiniMaxSessionCookieJar, extractMiniMaxCookieValue, + fetchMiniMaxWithApiKey, fetchMiniMaxWithManualCookieHeader, fetchMiniMaxWithSessionCookieJar, + getMiniMaxEndpointUrl, getUniqueMiniMaxCookieNames, logMiniMaxFetchFailure, makeMiniMaxRequestHeaders, @@ -137,7 +139,7 @@ describe('redactMiniMaxSecret', () => { describe('makeMiniMaxRequestHeaders', () => { it('always includes browser-like Accept, Accept-Language, Referer, and User-Agent', () => { - const headers = makeMiniMaxRequestHeaders(null) + const headers = makeMiniMaxRequestHeaders(null, 'overseas') expect(headers.Accept).toMatch(/application\/json/) expect(headers['Accept-Language']).toBe('en-US,en;q=0.9') expect(headers.Referer).toBe('https://platform.minimax.io/console/usage') @@ -145,18 +147,23 @@ describe('makeMiniMaxRequestHeaders', () => { expect(headers['User-Agent']).not.toContain('orca-minimax-usage') }) + it('switches the Referer to the CN console when endpointMode is "cn" (#14264)', () => { + const headers = makeMiniMaxRequestHeaders(null, 'cn') + expect(headers.Referer).toBe('https://platform.minimaxi.com/console/usage') + }) + it('omits X-Group-Id when groupId is null', () => { - const headers = makeMiniMaxRequestHeaders(null) + const headers = makeMiniMaxRequestHeaders(null, 'overseas') expect(headers['X-Group-Id']).toBeUndefined() }) it('omits X-Group-Id when groupId is empty string', () => { - const headers = makeMiniMaxRequestHeaders('') + const headers = makeMiniMaxRequestHeaders('', 'overseas') expect(headers['X-Group-Id']).toBeUndefined() }) it('includes X-Group-Id when groupId is provided', () => { - const headers = makeMiniMaxRequestHeaders('2034972027806299092') + const headers = makeMiniMaxRequestHeaders('2034972027806299092', 'overseas') expect(headers['X-Group-Id']).toBe('2034972027806299092') }) }) @@ -189,7 +196,8 @@ describe('fetchMiniMaxWithSessionCookieJar', () => { cookie: FULL_COOKIE, endpoint: MINIMAX_USAGE_ENDPOINT, groupId: '12345', - signal: controller.signal + signal: controller.signal, + endpointMode: 'overseas' }) expect(sessionFromPartitionMock).toHaveBeenCalledWith('orca-minimax-rate-limit-fetch') expect(clearStorageDataMock).toHaveBeenCalledTimes(2) @@ -215,7 +223,8 @@ describe('fetchMiniMaxWithSessionCookieJar', () => { cookie: FULL_COOKIE, endpoint: MINIMAX_USAGE_ENDPOINT, groupId: '12345', - signal: controller.signal + signal: controller.signal, + endpointMode: 'overseas' }) ).rejects.toThrow('pre-clear boom') @@ -242,6 +251,18 @@ describe('fetchMiniMaxWithSessionCookieJar', () => { }) }) + it('clears both overseas and CN origins on demand (#14264)', async () => { + await clearMiniMaxSessionCookieJar() + expect(clearStorageDataMock).toHaveBeenNthCalledWith(1, { + origin: 'https://platform.minimax.io', + storages: ['cookies'] + }) + expect(clearStorageDataMock).toHaveBeenNthCalledWith(2, { + origin: 'https://www.minimaxi.com', + storages: ['cookies'] + }) + }) + it('sets every cookie pair onto the session jar with secure + path /', async () => { netFetchMock.mockResolvedValueOnce({ ok: true, @@ -253,7 +274,8 @@ describe('fetchMiniMaxWithSessionCookieJar', () => { cookie: '_token=tok; ak_bmsc=ak; minimax_group_id_v2=42', endpoint: MINIMAX_USAGE_ENDPOINT, groupId: null, - signal: controller.signal + signal: controller.signal, + endpointMode: 'overseas' }) expect(cookiesSetMock).toHaveBeenCalledTimes(3) const setDetails = cookiesSetMock.mock.calls.map((call) => { @@ -287,7 +309,8 @@ describe('fetchMiniMaxWithSessionCookieJar', () => { cookie: FULL_COOKIE, endpoint: MINIMAX_USAGE_ENDPOINT, groupId: '12345', - signal: controller.signal + signal: controller.signal, + endpointMode: 'overseas' }) expect(result.transport).toBe('session-cookie-jar') expect(result.cookieNames).toEqual([ @@ -328,7 +351,8 @@ describe('fetchMiniMaxWithManualCookieHeader', () => { cookie: FULL_COOKIE, endpoint: MINIMAX_USAGE_ENDPOINT, groupId: '12345', - signal: controller.signal + signal: controller.signal, + endpointMode: 'overseas' }) expect(result.transport).toBe('manual-cookie-header') expect(sessionFromPartitionMock).toHaveBeenCalledWith('orca-minimax-rate-limit-fetch') @@ -353,7 +377,8 @@ describe('fetchMiniMaxWithManualCookieHeader', () => { cookie: FULL_COOKIE, endpoint: MINIMAX_USAGE_ENDPOINT, groupId: null, - signal: controller.signal + signal: controller.signal, + endpointMode: 'overseas' }) const [, init] = netFetchMock.mock.calls[0] expect(init.headers['X-Group-Id']).toBeUndefined() @@ -370,7 +395,8 @@ describe('fetchMiniMaxWithManualCookieHeader', () => { cookie: 'Cookie: _token=tok; minimax_group_id_v2=42; _twpid:"tw"', endpoint: MINIMAX_USAGE_ENDPOINT, groupId: null, - signal: controller.signal + signal: controller.signal, + endpointMode: 'overseas' }) const [, init] = netFetchMock.mock.calls[0] expect(init.headers.Cookie).toBe('_token=tok; minimax_group_id_v2=42; _twpid=tw') @@ -388,7 +414,8 @@ describe('fetchMiniMaxWithManualCookieHeader', () => { cookie: FULL_COOKIE, endpoint: MINIMAX_USAGE_ENDPOINT, groupId: '12345', - signal: controller.signal + signal: controller.signal, + endpointMode: 'overseas' }) ).rejects.toThrow('manual pre-clear boom') @@ -398,6 +425,83 @@ describe('fetchMiniMaxWithManualCookieHeader', () => { }) }) +describe('getMiniMaxEndpointUrl', () => { + it('returns the overseas .io endpoint by default', () => { + expect(getMiniMaxEndpointUrl('overseas')).toBe(MINIMAX_USAGE_ENDPOINT) + expect(getMiniMaxEndpointUrl('overseas')).toBe( + 'https://platform.minimax.io/v1/api/openplatform/coding_plan/remains' + ) + }) + + it('returns the CN www.minimaxi.com endpoint when requested', () => { + expect(getMiniMaxEndpointUrl('cn')).toBe( + 'https://www.minimaxi.com/v1/api/openplatform/coding_plan/remains' + ) + }) + + it('uses the same usage path on both endpoints so response parsing stays uniform', () => { + const overseas = new URL(getMiniMaxEndpointUrl('overseas')) + const cn = new URL(getMiniMaxEndpointUrl('cn')) + expect(overseas.pathname).toBe(cn.pathname) + }) +}) + +describe('fetchMiniMaxWithApiKey', () => { + beforeEach(() => { + clearStorageDataMock.mockClear() + cookiesSetMock.mockClear() + netFetchMock.mockReset() + sessionFromPartitionMock.mockClear() + }) + + afterEach(() => { + vi.restoreAllMocks() + }) + + it('sends only the Authorization Bearer header and accepts JSON', async () => { + netFetchMock.mockResolvedValueOnce({ + ok: true, + status: 200, + json: async () => ({ base_resp: { status_code: 0 }, model_remains: [] }) + }) + const controller = new AbortController() + const result = await fetchMiniMaxWithApiKey({ + apiKey: 'sk-test-1234567890', + endpoint: 'https://www.minimaxi.com/v1/api/openplatform/coding_plan/remains', + signal: controller.signal + }) + expect(result.transport).toBe('api-key') + expect(result.cookieNames).toEqual([]) + expect(result.requestHeaderNames).toEqual(['Authorization', 'Accept']) + expect(netFetchMock).toHaveBeenCalledTimes(1) + const [url, init] = netFetchMock.mock.calls[0] + expect(url).toBe('https://www.minimaxi.com/v1/api/openplatform/coding_plan/remains') + expect(init.method).toBe('GET') + expect(init.headers.Authorization).toBe('Bearer sk-test-1234567890') + expect(init.headers.Accept).toBe('application/json') + expect(init.headers.Cookie).toBeUndefined() + expect(init.headers.Referer).toBeUndefined() + expect(init.headers['User-Agent']).toBeUndefined() + }) + + it('does not touch the session cookie jar', async () => { + netFetchMock.mockResolvedValueOnce({ + ok: true, + status: 200, + json: async () => ({ base_resp: { status_code: 0 }, model_remains: [] }) + }) + const controller = new AbortController() + await fetchMiniMaxWithApiKey({ + apiKey: 'sk-test', + endpoint: 'https://www.minimaxi.com/v1/api/openplatform/coding_plan/remains', + signal: controller.signal + }) + expect(sessionFromPartitionMock).not.toHaveBeenCalled() + expect(clearStorageDataMock).not.toHaveBeenCalled() + expect(cookiesSetMock).not.toHaveBeenCalled() + }) +}) + describe('logMiniMaxFetchFailure', () => { let warn: ReturnType @@ -450,4 +554,34 @@ describe('logMiniMaxFetchFailure', () => { }) ) }) + + it('stores cookies under the CN origin when endpointMode is "cn" (#14264 repro)', async () => { + // Why: the previous hardcoded origin (platform.minimax.io) caused CN + // users' cookies to be sent against the wrong host, so Electron's + // session never attached them. Cookies must be stored under + // www.minimaxi.com for a CN fetch to actually carry the auth. + netFetchMock.mockResolvedValueOnce({ + ok: true, + status: 200, + json: async () => ({ base_resp: { status_code: 0 }, model_remains: [] }) + }) + await fetchMiniMaxWithSessionCookieJar({ + cookie: FULL_COOKIE, + endpoint: 'https://www.minimaxi.com/v1/api/openplatform/coding_plan/remains', + groupId: '12345', + signal: new AbortController().signal, + endpointMode: 'cn' + }) + // The cookies must be stored under the CN origin, not overseas. + const cnWrites = cookiesSetMock.mock.calls.filter((call) => { + const [details] = call as unknown as [{ url: string }] + return details.url === 'https://www.minimaxi.com' + }) + expect(cnWrites.length).toBeGreaterThan(0) + const overseasWrites = cookiesSetMock.mock.calls.filter((call) => { + const [details] = call as unknown as [{ url: string }] + return details.url === 'https://platform.minimax.io' + }) + expect(overseasWrites.length).toBe(0) + }) }) diff --git a/src/main/rate-limits/minimax-request-context.ts b/src/main/rate-limits/minimax-request-context.ts index ecda8329a2e..10a1ea07b91 100644 --- a/src/main/rate-limits/minimax-request-context.ts +++ b/src/main/rate-limits/minimax-request-context.ts @@ -1,10 +1,38 @@ -import { session, type Session } from 'electron' +import { net, session, type Session } from 'electron' +import type { MiniMaxEndpoint } from '../../shared/global-settings-types' -export const MINIMAX_USAGE_ENDPOINT = - 'https://platform.minimax.io/v1/api/openplatform/coding_plan/remains' +const MINIMAX_USAGE_PATH = '/v1/api/openplatform/coding_plan/remains' +const MINIMAX_OVERSEAS_BASE = 'https://platform.minimax.io' +const MINIMAX_CN_BASE = 'https://www.minimaxi.com' + +export function getMiniMaxEndpointUrl(endpoint: MiniMaxEndpoint): string { + if (endpoint === 'cn') { + return `${MINIMAX_CN_BASE}${MINIMAX_USAGE_PATH}` + } + return `${MINIMAX_OVERSEAS_BASE}${MINIMAX_USAGE_PATH}` +} + +/** + * @deprecated Prefer `getMiniMaxEndpointUrl('overseas')`. Kept for the + * status-bar copy and any older callers that still compare against the + * hardcoded URL string. + */ +export const MINIMAX_USAGE_ENDPOINT = getMiniMaxEndpointUrl('overseas') + +// Why: each endpoint has its own origin and console URL. The cookie jar +// keys cookies by origin, so a CN request must store cookies under +// https://www.minimaxi.com — otherwise Electron's session won't send them +// to the CN host. Computing these from the endpoint URL keeps auth, jar, +// and Referer in lockstep. +function getMiniMaxOrigin(endpoint: MiniMaxEndpoint): string { + return endpoint === 'cn' ? MINIMAX_CN_BASE : MINIMAX_OVERSEAS_BASE +} + +function getMiniMaxReferer(endpoint: MiniMaxEndpoint): string { + const consoleOrigin = endpoint === 'cn' ? 'https://platform.minimaxi.com' : MINIMAX_OVERSEAS_BASE + return `${consoleOrigin}/console/usage` +} -const MINIMAX_ORIGIN = 'https://platform.minimax.io' -const MINIMAX_REFERER = 'https://platform.minimax.io/console/usage' const MINIMAX_SESSION_PARTITION = 'orca-minimax-rate-limit-fetch' const SENSITIVE_COOKIE_NAMES = new Set([ '_token', @@ -17,7 +45,9 @@ const SENSITIVE_COOKIE_NAMES = new Set([ 'minimax_group_id_v2' ]) -export type MiniMaxFetchTransport = 'session-cookie-jar' | 'manual-cookie-header' +const MINIMAX_API_KEY_TIMEOUT_MS = 10_000 + +export type MiniMaxFetchTransport = 'session-cookie-jar' | 'manual-cookie-header' | 'api-key' export type MiniMaxFetchResponse = { response: Response @@ -91,11 +121,14 @@ export function redactMiniMaxSecret(value: string): string { return redacted } -export function makeMiniMaxRequestHeaders(groupId: string | null): Record { +export function makeMiniMaxRequestHeaders( + groupId: string | null, + endpoint: MiniMaxEndpoint +): Record { const headers: Record = { Accept: 'application/json, text/plain, */*', 'Accept-Language': 'en-US,en;q=0.9', - Referer: MINIMAX_REFERER, + Referer: getMiniMaxReferer(endpoint), 'User-Agent': getMiniMaxBrowserUserAgent() } if (groupId) { @@ -104,28 +137,40 @@ export function makeMiniMaxRequestHeaders(groupId: string | null): Record { - await miniMaxSession.clearStorageData({ origin: MINIMAX_ORIGIN, storages: ['cookies'] }) +async function clearMiniMaxSessionCookieJarForSession( + miniMaxSession: Session, + origin: string +): Promise { + await miniMaxSession.clearStorageData({ origin, storages: ['cookies'] }) } export async function clearMiniMaxSessionCookieJar(): Promise { - await clearMiniMaxSessionCookieJarForSession(session.fromPartition(MINIMAX_SESSION_PARTITION)) + // Why: clear cookies under both origins so a user who switches endpoint + // (overseas -> CN or vice versa) does not leave stale cookies that the + // next request might pick up against the wrong host. + const miniMaxSession = session.fromPartition(MINIMAX_SESSION_PARTITION) + await Promise.all([ + clearMiniMaxSessionCookieJarForSession(miniMaxSession, getMiniMaxOrigin('overseas')), + clearMiniMaxSessionCookieJarForSession(miniMaxSession, getMiniMaxOrigin('cn')) + ]) } export async function fetchMiniMaxWithSessionCookieJar(args: { cookie: string endpoint: string groupId: string | null + endpointMode: MiniMaxEndpoint signal: AbortSignal }): Promise { const miniMaxSession = session.fromPartition(MINIMAX_SESSION_PARTITION) const cookiePairs = parseCookiePairs(args.cookie) + const origin = getMiniMaxOrigin(args.endpointMode) try { - await clearMiniMaxSessionCookieJarForSession(miniMaxSession) + await clearMiniMaxSessionCookieJarForSession(miniMaxSession, origin) await Promise.all( cookiePairs.map((pair) => miniMaxSession.cookies.set({ - url: MINIMAX_ORIGIN, + url: origin, name: pair.name, value: pair.value, secure: true, @@ -133,7 +178,7 @@ export async function fetchMiniMaxWithSessionCookieJar(args: { }) ) ) - const headers = makeMiniMaxRequestHeaders(args.groupId) + const headers = makeMiniMaxRequestHeaders(args.groupId, args.endpointMode) return { response: await miniMaxSession.fetch(args.endpoint, { method: 'GET', @@ -145,7 +190,7 @@ export async function fetchMiniMaxWithSessionCookieJar(args: { transport: 'session-cookie-jar' } } finally { - await clearMiniMaxSessionCookieJarForSession(miniMaxSession).catch((error: unknown) => { + await clearMiniMaxSessionCookieJarForSession(miniMaxSession, origin).catch((error: unknown) => { console.warn('[minimax] failed to clear session cookie jar after fetch', error) }) } @@ -155,13 +200,15 @@ export async function fetchMiniMaxWithManualCookieHeader(args: { cookie: string endpoint: string groupId: string | null + endpointMode: MiniMaxEndpoint signal: AbortSignal }): Promise { const miniMaxSession = session.fromPartition(MINIMAX_SESSION_PARTITION) + const origin = getMiniMaxOrigin(args.endpointMode) try { - await clearMiniMaxSessionCookieJarForSession(miniMaxSession) + await clearMiniMaxSessionCookieJarForSession(miniMaxSession, origin) const headers = { - ...makeMiniMaxRequestHeaders(args.groupId), + ...makeMiniMaxRequestHeaders(args.groupId, args.endpointMode), Cookie: normalizeMiniMaxCookieHeader(args.cookie) } return { @@ -175,12 +222,38 @@ export async function fetchMiniMaxWithManualCookieHeader(args: { transport: 'manual-cookie-header' } } finally { - await clearMiniMaxSessionCookieJarForSession(miniMaxSession).catch((error: unknown) => { + await clearMiniMaxSessionCookieJarForSession(miniMaxSession, origin).catch((error: unknown) => { console.warn('[minimax] failed to clear session cookie jar after fetch', error) }) } } +export async function fetchMiniMaxWithApiKey(args: { + apiKey: string + endpoint: string + signal: AbortSignal +}): Promise { + // Why: net.fetch routes through Electron's URL stack, matching the cookie + // transport's surface area and avoiding Node's TLS quirks for CN routing. + const headers: Record = { + Authorization: `Bearer ${args.apiKey}`, + Accept: 'application/json' + } + const response = await net.fetch(args.endpoint, { + method: 'GET', + headers, + signal: args.signal + }) + return { + response, + requestHeaderNames: Object.keys(headers), + cookieNames: [], + transport: 'api-key' + } +} + +export { MINIMAX_API_KEY_TIMEOUT_MS } + export function logMiniMaxFetchFailure(details: { transport: MiniMaxFetchTransport responseStatus?: number diff --git a/src/main/rate-limits/service-minimax-usage.test.ts b/src/main/rate-limits/service-minimax-usage.test.ts index 737ca0ff03d..7c3db21c63e 100644 --- a/src/main/rate-limits/service-minimax-usage.test.ts +++ b/src/main/rate-limits/service-minimax-usage.test.ts @@ -49,6 +49,10 @@ vi.mock('../minimax/minimax-cookie-store', () => ({ hasMiniMaxSessionCookie: vi.fn(() => false) })) +vi.mock('../minimax/minimax-api-key-store', () => ({ + hasMiniMaxApiKey: vi.fn(() => false) +})) + describe('RateLimitService', () => { beforeEach(() => { resetRateLimitProviderMocks() @@ -64,7 +68,9 @@ describe('RateLimitService', () => { service.setMiniMaxConfigResolver(() => ({ sessionCookie: '_token=abc; minimax_group_id_v2=42', groupId: '', - models: 'general' + models: 'general', + endpoint: 'overseas', + apiKey: '' })) vi.mocked(hasMiniMaxSessionCookie).mockReturnValue(true) vi.mocked(fetchMiniMaxRateLimits).mockResolvedValueOnce(okProvider('minimax', 50, Date.now())) @@ -75,7 +81,9 @@ describe('RateLimitService', () => { expect(fetchMiniMaxRateLimits).toHaveBeenCalledWith({ cookie: '_token=abc; minimax_group_id_v2=42', groupId: '', - models: 'general' + models: 'general', + endpointMode: 'overseas', + apiKey: '' }) const state = service.getState() @@ -96,7 +104,9 @@ describe('RateLimitService', () => { service.setMiniMaxConfigResolver(() => ({ sessionCookie: '_token=abc', groupId: '', - models + models, + endpoint: 'overseas', + apiKey: '' })) vi.mocked(hasMiniMaxSessionCookie).mockReturnValue(true) vi.mocked(fetchMiniMaxRateLimits) @@ -114,6 +124,33 @@ describe('RateLimitService', () => { expect(state.minimax?.session?.usedPercent).toBe(10) }) + it('clears the old quota when replacing a non-empty API key and the refresh fails', async () => { + const service = new RateLimitService() + let apiKey = 'sk-account-a' + service.setMiniMaxConfigResolver(() => ({ + sessionCookie: '', + groupId: '', + models: 'general', + endpoint: 'cn', + apiKey + })) + vi.mocked(fetchMiniMaxRateLimits) + .mockResolvedValueOnce(okProvider('minimax', 40, Date.now())) + .mockRejectedValueOnce(new Error('MiniMax unavailable')) + + await service.refresh() + expect(service.getState().minimax?.session?.usedPercent).toBe(40) + + apiKey = 'sk-account-b' + await service.refresh() + + expect(service.getState().minimax?.status).toBe('error') + expect(service.getState().minimax?.session).toBeNull() + expect(fetchMiniMaxRateLimits).toHaveBeenLastCalledWith( + expect.objectContaining({ apiKey: 'sk-account-b' }) + ) + }) + it('does not apply an in-flight MiniMax result after credential invalidation', async () => { const service = new RateLimitService() const firstMiniMax = deferred() @@ -121,7 +158,9 @@ describe('RateLimitService', () => { service.setMiniMaxConfigResolver(() => ({ sessionCookie: '_token=abc', groupId: '', - models: 'general' + models: 'general', + endpoint: 'overseas', + apiKey: '' })) vi.mocked(fetchMiniMaxRateLimits) .mockImplementationOnce(() => firstMiniMax.promise) @@ -155,7 +194,9 @@ describe('RateLimitService', () => { service.setMiniMaxConfigResolver(() => ({ sessionCookie: '_token=abc', groupId: '', - models: 'general' + models: 'general', + endpoint: 'overseas', + apiKey: '' })) vi.mocked(fetchMiniMaxRateLimits).mockRejectedValueOnce(new Error('minimax down')) vi.mocked(fetchClaudeRateLimits).mockResolvedValueOnce(okProvider('claude', 10, Date.now())) @@ -183,4 +224,57 @@ describe('RateLimitService', () => { expect(state.minimax?.error).toBe('MiniMax session cookie could not be decrypted') expect(state.claude?.status).toBe('ok') }) + + it('passes the CN endpoint and API key to the fetcher when the resolver selects CN', async () => { + const service = new RateLimitService() + service.setMiniMaxConfigResolver(() => ({ + sessionCookie: '', + groupId: '', + models: 'general', + endpoint: 'cn', + apiKey: 'sk-cn-key-9876' + })) + vi.mocked(fetchMiniMaxRateLimits).mockResolvedValueOnce(okProvider('minimax', 33, Date.now())) + + await service.refresh() + + expect(fetchMiniMaxRateLimits).toHaveBeenCalledWith({ + cookie: '', + groupId: '', + models: 'general', + endpointMode: 'cn', + apiKey: 'sk-cn-key-9876' + }) + }) + + it('bumps the MiniMax fetch generation when the endpoint or API key changes', async () => { + const service = new RateLimitService() + let endpointMode: 'overseas' | 'cn' = 'overseas' + let apiKey = '' + service.setMiniMaxConfigResolver(() => ({ + sessionCookie: '_token=abc', + groupId: '', + models: 'general', + endpoint: endpointMode, + apiKey + })) + vi.mocked(fetchMiniMaxRateLimits) + .mockResolvedValueOnce(okProvider('minimax', 10, Date.now())) + .mockResolvedValueOnce(okProvider('minimax', 20, Date.now())) + .mockResolvedValueOnce(okProvider('minimax', 30, Date.now())) + + await service.refresh() + expect(service.getState().minimax?.session?.usedPercent).toBe(10) + + // Why: changing only the endpoint must invalidate the previous snapshot — + // the response shape and host differ, so the old data is misleading. + endpointMode = 'cn' + await service.refresh() + expect(service.getState().minimax?.session?.usedPercent).toBe(20) + + // Why: adding an API key while staying on CN must also force a refresh. + apiKey = 'sk-cn-key-9876' + await service.refresh() + expect(service.getState().minimax?.session?.usedPercent).toBe(30) + }) }) diff --git a/src/main/rate-limits/service/service-configuration.ts b/src/main/rate-limits/service/service-configuration.ts index 52aa02ccddd..655ba9b93d8 100644 --- a/src/main/rate-limits/service/service-configuration.ts +++ b/src/main/rate-limits/service/service-configuration.ts @@ -1,5 +1,6 @@ import type { BrowserWindow } from 'electron' import { hasMiniMaxSessionCookie } from '../../minimax/minimax-cookie-store' +import { hasMiniMaxApiKey } from '../../minimax/minimax-api-key-store' import { RateLimitServiceAccountRefresh } from './service-account-refresh' import { type CodexAccountSelectionTarget, @@ -123,6 +124,7 @@ export abstract class RateLimitServiceConfiguration extends RateLimitServiceAcco ...this.state, // Why: the cookie lives on the filesystem, not GlobalSettings; surface its presence so the renderer keeps the MiniMax bar across reloads. minimaxCookieConfigured: hasMiniMaxSessionCookie(), + minimaxApiKeyConfigured: hasMiniMaxApiKey(), grokAuthConfigured: this.grokAuthConfigured, claudeTarget: this.claudeFetchTarget, codexTarget: this.codexFetchTarget, diff --git a/src/main/rate-limits/service/service-fetch-targets.ts b/src/main/rate-limits/service/service-fetch-targets.ts index 2d316294dbc..09274ad3d82 100644 --- a/src/main/rate-limits/service/service-fetch-targets.ts +++ b/src/main/rate-limits/service/service-fetch-targets.ts @@ -152,7 +152,9 @@ export abstract class RateLimitServiceFetchTargets extends RateLimitServiceResul config: this.miniMaxConfigResolver?.() ?? { sessionCookie: '', groupId: '', - models: 'general' + models: 'general', + endpoint: 'overseas', + apiKey: '' }, error: null } @@ -162,7 +164,9 @@ export abstract class RateLimitServiceFetchTargets extends RateLimitServiceResul config: { sessionCookie: '', groupId: '', - models: 'general' + models: 'general', + endpoint: 'overseas', + apiKey: '' }, error: toErrorMessage(error) } diff --git a/src/main/rate-limits/service/service-full-cycle-preparation.ts b/src/main/rate-limits/service/service-full-cycle-preparation.ts index 1bbf3bf497d..c5bf533bc86 100644 --- a/src/main/rate-limits/service/service-full-cycle-preparation.ts +++ b/src/main/rate-limits/service/service-full-cycle-preparation.ts @@ -80,6 +80,8 @@ export abstract class RateLimitServiceFullCyclePreparation extends RateLimitServ const miniMaxCookie = miniMaxConfigResult.config.sessionCookie const miniMaxGroupId = miniMaxConfigResult.config.groupId const miniMaxModels = miniMaxConfigResult.config.models + const miniMaxEndpoint = miniMaxConfigResult.config.endpoint + const miniMaxApiKey = miniMaxConfigResult.config.apiKey const geminiCliOAuthEnabled = this.geminiCliOAuthEnabledResolver?.() ?? false // Why: getState() is hot (renderer pushes + mobile snapshots); keep Grok's sync auth-file probe on fetch cycles instead. const grokAuthReadResult = readGrokAuthSession() @@ -94,7 +96,7 @@ export abstract class RateLimitServiceFullCyclePreparation extends RateLimitServ } const opencodeGeneration = this.opencodeFetchGeneration - const currentMiniMaxConfigHash = `${miniMaxCookie}|${miniMaxGroupId}|${miniMaxModels}|${miniMaxConfigResult.error ?? ''}` + const currentMiniMaxConfigHash = `${miniMaxCookie}|${miniMaxGroupId}|${miniMaxModels}|${miniMaxEndpoint}|${miniMaxApiKey}|${miniMaxConfigResult.error ?? ''}` const miniMaxConfigChanged = currentMiniMaxConfigHash !== this.lastMiniMaxConfigHash if (miniMaxConfigChanged) { this.lastMiniMaxConfigHash = currentMiniMaxConfigHash @@ -167,7 +169,9 @@ export abstract class RateLimitServiceFullCyclePreparation extends RateLimitServ : fetchMiniMaxRateLimits({ cookie: miniMaxCookie, groupId: miniMaxGroupId, - models: miniMaxModels + models: miniMaxModels, + endpointMode: miniMaxEndpoint, + apiKey: miniMaxApiKey }) ]) diff --git a/src/main/rate-limits/service/service-types.ts b/src/main/rate-limits/service/service-types.ts index 0b38bd39433..414fa9b8dfd 100644 --- a/src/main/rate-limits/service/service-types.ts +++ b/src/main/rate-limits/service/service-types.ts @@ -50,6 +50,8 @@ export type MiniMaxRateLimitConfig = { sessionCookie: string groupId: string models: string + endpoint: 'overseas' | 'cn' + apiKey: string } export type MiniMaxResolvedConfig = { diff --git a/src/main/runtime/browser-session-tab-selection-snapshot.test.ts b/src/main/runtime/browser-session-tab-selection-snapshot.test.ts index a96a5617d3b..f553650323c 100644 --- a/src/main/runtime/browser-session-tab-selection-snapshot.test.ts +++ b/src/main/runtime/browser-session-tab-selection-snapshot.test.ts @@ -2,8 +2,6 @@ import { describe, expect, it } from 'vitest' import type { RuntimeMobileSessionTabsSnapshot } from '../../shared/runtime-types' import { applyBrowserSessionTabSelection } from './browser-session-tab-selection-snapshot' -const EPOCH = 'headless:test' - function makeSnapshot(): RuntimeMobileSessionTabsSnapshot { return { worktree: 'wt-1', @@ -46,8 +44,7 @@ function select(overrides: { focusesHost: boolean; targetGroupId?: string }) { snapshot: makeSnapshot(), tabId: 'page-new', focusesHost: overrides.focusesHost, - ...(overrides.targetGroupId ? { targetGroupId: overrides.targetGroupId } : {}), - publicationEpoch: EPOCH + ...(overrides.targetGroupId ? { targetGroupId: overrides.targetGroupId } : {}) }) } @@ -115,10 +112,11 @@ describe('applyBrowserSessionTabSelection', () => { expect(snapshot.activeGroupId).toBe('group-left') }) - it('republishes under a fresh epoch and a newer version either way', () => { + // Rotating the epoch here retires the renderer's own publication client-side. + it('keeps the publication epoch and advances the version either way', () => { for (const focusesHost of [true, false]) { const { snapshot } = select({ focusesHost }) - expect(snapshot.publicationEpoch).toBe(EPOCH) + expect(snapshot.publicationEpoch).toBe('headless:before') expect(snapshot.snapshotVersion).toBe(5) } }) diff --git a/src/main/runtime/browser-session-tab-selection-snapshot.ts b/src/main/runtime/browser-session-tab-selection-snapshot.ts index aa95c9f2685..e5862e1926d 100644 --- a/src/main/runtime/browser-session-tab-selection-snapshot.ts +++ b/src/main/runtime/browser-session-tab-selection-snapshot.ts @@ -22,7 +22,6 @@ export function applyBrowserSessionTabSelection(args: { tabId: string targetGroupId?: string focusesHost: boolean - publicationEpoch: string }): BrowserSessionTabSelectionResult { const { snapshot, tabId, targetGroupId, focusesHost } = args const groups = snapshot.tabGroups ?? [] @@ -56,7 +55,6 @@ export function applyBrowserSessionTabSelection(args: { placedInTargetGroup, snapshot: { ...snapshot, - publicationEpoch: args.publicationEpoch, snapshotVersion: snapshot.snapshotVersion + 1, ...(placedInTargetGroup && focusesHost ? { activeGroupId: targetGroupId } : {}), ...(focusesHost diff --git a/src/main/runtime/device-registry.ts b/src/main/runtime/device-registry.ts index b2d5de8ef41..e3d848405f0 100644 --- a/src/main/runtime/device-registry.ts +++ b/src/main/runtime/device-registry.ts @@ -15,6 +15,10 @@ import { DEVICE_REGISTRY_FILENAME } from './mobile-pairing-files' import type { RelayDeviceBinding } from './relay/relay-revoke-outbox' import type { MobilePairingConnectionMode } from '../../shared/mobile-pairing-connection-mode' import type { RuntimePairingReach } from '../../shared/runtime-pairing-reach' +import { + parseMobilePushRegistration, + type MobilePushRegistration +} from '../../shared/mobile-push-contract' export type { DeviceScope } @@ -30,6 +34,9 @@ export type DeviceEntry = { // Why: STA-2370 — a grant minted for "This computer only" proves nothing about off-host reach when its // client connects, so the bind decision must be able to tell it apart from a LAN/phone grant. pairingReach?: RuntimePairingReach + // Why: survives a desktop restart so the host can keep pushing without the phone + // re-registering. Absent on every registry written before background push existed. + pushRegistration?: MobilePushRegistration } function validRelayBinding(value: unknown, deviceId: string): RelayDeviceBinding | undefined { @@ -179,6 +186,26 @@ export class DeviceRegistry { return true } + /** Passing null clears the registration (unregister, or a token the gateway reported dead). */ + setPushRegistration(deviceId: string, registration: MobilePushRegistration | null): boolean { + const index = this.devices.findIndex((candidate) => candidate.deviceId === deviceId) + if (index === -1 || this.devices[index]?.scope !== 'mobile') { + return false + } + const nextDevices = this.devices.map((device, candidateIndex) => { + if (candidateIndex !== index) { + return device + } + const { pushRegistration: _dropped, ...rest } = device + return registration ? { ...rest, pushRegistration: registration } : rest + }) + // Why: persist before the memory swap so a failed write cannot leave the dispatcher + // pushing to a registration disk says is gone (or vice versa on reload). + this.save(nextDevices) + this.devices = nextDevices + return true + } + setMobilePairingConnectionMode(deviceId: string, mode: MobilePairingConnectionMode): boolean { const index = this.devices.findIndex((candidate) => candidate.deviceId === deviceId) if (index === -1 || this.devices[index]?.scope !== 'mobile') { @@ -297,7 +324,10 @@ export class DeviceRegistry { device.mobilePairingConnectionMode === 'local-only' ? 'local-only' : 'automatic', // Why: registries written before this field existed only ever held network-reach grants (phones and // LAN links), so a missing value must keep binding every interface on reconnect. - pairingReach: device.pairingReach === 'this-computer' ? 'this-computer' : 'network' + pairingReach: device.pairingReach === 'this-computer' ? 'this-computer' : 'network', + // Why: a malformed row must degrade to "no background push", never fail the load + // and strand every paired device. + pushRegistration: parseMobilePushRegistration(device.pushRegistration) })) this.registryUnreadable = false } catch (error) { diff --git a/src/main/runtime/folder-workspace-pty-teardown.ts b/src/main/runtime/folder-workspace-pty-teardown.ts index f5a5b7df4b3..e54c9165c46 100644 --- a/src/main/runtime/folder-workspace-pty-teardown.ts +++ b/src/main/runtime/folder-workspace-pty-teardown.ts @@ -29,6 +29,9 @@ export async function teardownFolderWorkspacePtys( ...(connectionId ? { resolvedConnectionId: connectionId } : {}), localProvider: ptyProvider, onPtyStopped: deps.onPtyStopped ?? undefined, + // A structured session is on no PTY surface, so the sweeps above leave it attached to a + // workspace this removal is about to forget. Closed best-effort, exactly like those PTYs. + closeStructuredSessions: true, ...(connectionId ? { includeProviderInventory: Boolean(sshPtyProvider), includeLocalRegistry: false } : {}) diff --git a/src/main/runtime/hidden-output-restored-provider-snapshot.test.ts b/src/main/runtime/hidden-output-restored-provider-snapshot.test.ts new file mode 100644 index 00000000000..bcca4491fca --- /dev/null +++ b/src/main/runtime/hidden-output-restored-provider-snapshot.test.ts @@ -0,0 +1,79 @@ +import { describe, expect, it, vi } from 'vitest' +import { createRuntime, syncSinglePty } from './orca-runtime-test-fixtures.spec' + +describe('hidden-output recovery after provider reattach', () => { + it('uses retained provider modes instead of the pre-attach redraw suffix', async () => { + const runtime = createRuntime() + const serializeProviderBuffer = vi.fn(async () => ({ + data: '\x1b[?1049hRetained TUI', + cols: 100, + rows: 30, + seq: 1000, + source: 'headless' as const, + alternateScreen: true + })) + runtime.setPtyController({ + write: () => true, + kill: () => true, + getForegroundProcess: async () => null, + serializeProviderBuffer + }) + syncSinglePty(runtime, 'pty-1') + runtime.onPtyData('pty-1', '\x1b[HRedraw without the original alternate-screen entry', 60) + runtime.synchronizePtyOutputSequenceFromProvider('pty-1', { + value: 1000, + generation: 'continued' + }) + + const snapshot = await runtime.serializeHiddenOutputRecoveryBuffer('pty-1', { + scrollbackRows: 5000 + }) + + expect(snapshot).toMatchObject({ data: '\x1b[?1049hRetained TUI', alternateScreen: true }) + expect(serializeProviderBuffer).toHaveBeenCalledWith('pty-1', { scrollbackRows: 5000 }) + }) + + it('keeps the renderer fallback for providers without retained snapshots', async () => { + const runtime = createRuntime() + runtime.onPtyData('pty-1', 'partial redraw', 14) + runtime.synchronizePtyOutputSequenceFromProvider('pty-1', { + value: 1000, + generation: 'continued' + }) + const serializeBuffer = vi.fn(async () => ({ + data: '\x1b[?1049hRenderer TUI', + cols: 100, + rows: 30 + })) + runtime.setPtyController({ + write: () => true, + kill: () => true, + getForegroundProcess: async () => null, + hasRendererSerializer: () => true, + serializeBuffer + }) + + await expect(runtime.serializeHiddenOutputRecoveryBuffer('pty-1')).resolves.toMatchObject({ + data: '\x1b[?1049hRenderer TUI', + source: 'renderer' + }) + }) + + it('keeps an authoritative main model without polling the provider', async () => { + const runtime = createRuntime() + const serializeProviderBuffer = vi.fn(async () => null) + runtime.setPtyController({ + write: () => true, + kill: () => true, + getForegroundProcess: async () => null, + serializeProviderBuffer + }) + runtime.onPtyData('pty-1', '\x1b[?1049hLive TUI', 20) + + await expect(runtime.serializeHiddenOutputRecoveryBuffer('pty-1')).resolves.toMatchObject({ + alternateScreen: true, + source: 'headless' + }) + expect(serializeProviderBuffer).not.toHaveBeenCalled() + }) +}) diff --git a/src/main/runtime/host-challenge-envelope.ts b/src/main/runtime/host-challenge-envelope.ts new file mode 100644 index 00000000000..6a00381c158 --- /dev/null +++ b/src/main/runtime/host-challenge-envelope.ts @@ -0,0 +1,139 @@ +// Why: the relay and the push gateway both authenticate this host with the same +// sealed-box challenge shape (the host keypair is X25519, so it cannot sign). +// Only the domain strings and the transcript fields differ, so the envelope +// handling lives here and each protocol owns its own field validation. +import { createHmac, timingSafeEqual } from 'node:crypto' +import nacl from 'tweetnacl' + +const textEncoder = new TextEncoder() +const textDecoder = new TextDecoder() + +export function decodeCanonicalBase64(value: string, expectedBytes: number): Uint8Array | null { + if (!/^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/.test(value)) { + return null + } + const decoded = Buffer.from(value, 'base64') + return decoded.byteLength === expectedBytes && decoded.toString('base64') === value + ? decoded + : null +} + +export function encodeUint64(value: number): Uint8Array { + const bytes = new Uint8Array(8) + new DataView(bytes.buffer).setBigUint64(0, BigInt(value), false) + return bytes +} + +export function equalBytes(left: Uint8Array | undefined, right: Uint8Array): boolean { + return Boolean(left && left.byteLength === right.byteLength && timingSafeEqual(left, right)) +} + +export function encodeText(value: string): Uint8Array { + return textEncoder.encode(value) +} + +/** Length-prefixed field map: u32be(len(name)) || name || u32be(len(value)) || value. */ +export function parseHostChallengeTranscript( + transcript: Uint8Array +): Map | null { + const fields = new Map() + const view = new DataView(transcript.buffer, transcript.byteOffset, transcript.byteLength) + let offset = 0 + try { + while (offset < transcript.byteLength) { + const nameLength = view.getUint32(offset, false) + offset += 4 + const name = textDecoder.decode(transcript.slice(offset, offset + nameLength)) + offset += nameLength + const valueLength = view.getUint32(offset, false) + offset += 4 + if (fields.has(name) || offset + valueLength > transcript.byteLength) { + return null + } + fields.set(name, transcript.slice(offset, offset + valueLength)) + offset += valueLength + } + } catch { + return null + } + return offset === transcript.byteLength ? fields : null +} + +export function readTranscriptUint64(value: Uint8Array | undefined): number | null { + if (!value || value.byteLength !== 8) { + return null + } + const parsed = new DataView(value.buffer, value.byteOffset, value.byteLength).getBigUint64( + 0, + false + ) + return parsed <= BigInt(Number.MAX_SAFE_INTEGER) ? Number(parsed) : null +} + +export type HostChallengeEnvelope = { + transcript: Uint8Array + secret: Uint8Array + peerEphemeralPublicKey: Uint8Array + nonce: Uint8Array +} + +/** + * Opens the sealed challenge and splits out the transcript and the 32-byte secret. + * Returns null for any malformed or undecryptable challenge; the caller still has + * to validate the transcript's fields before answering. + */ +export function openHostChallengeEnvelope(input: { + peerEphemeralPublicKeyB64: string + nonceB64: string + ciphertextB64: string + hostSecretKey: Uint8Array + plaintextDomain: string + /** Reports the failing check by name only; never receives field values. */ + onInvalid?: (reason: string) => void +}): HostChallengeEnvelope | null { + const peerKey = decodeCanonicalBase64(input.peerEphemeralPublicKeyB64, 32) + const nonce = decodeCanonicalBase64(input.nonceB64, 24) + const ciphertext = Buffer.from(input.ciphertextB64, 'base64') + if (!peerKey || !nonce || ciphertext.toString('base64') !== input.ciphertextB64) { + return null + } + const plaintext = nacl.box.open(ciphertext, nonce, peerKey, input.hostSecretKey) + if (!plaintext) { + input.onInvalid?.('challenge-box-open') + return null + } + const domain = textEncoder.encode(`${input.plaintextDomain}\0`) + if ( + !equalBytes(plaintext.slice(0, domain.byteLength), domain) || + plaintext.byteLength < domain.byteLength + 36 + ) { + return null + } + const transcriptLength = new DataView( + plaintext.buffer, + plaintext.byteOffset + domain.byteLength, + 4 + ).getUint32(0, false) + const transcriptStart = domain.byteLength + 4 + const secretStart = transcriptStart + transcriptLength + if (secretStart + 32 !== plaintext.byteLength) { + return null + } + return { + transcript: plaintext.slice(transcriptStart, secretStart), + secret: plaintext.slice(secretStart), + peerEphemeralPublicKey: peerKey, + nonce + } +} + +export function hostChallengeAckProof(input: { + secret: Uint8Array + transcript: Uint8Array + proofDomain: string +}): string { + return createHmac('sha256', input.secret) + .update(textEncoder.encode(`${input.proofDomain}\0ack\0`)) + .update(input.transcript) + .digest('base64') +} diff --git a/src/main/runtime/keyed-trailing-edge-coalescer.ts b/src/main/runtime/keyed-trailing-edge-coalescer.ts new file mode 100644 index 00000000000..b08d02a10fc --- /dev/null +++ b/src/main/runtime/keyed-trailing-edge-coalescer.ts @@ -0,0 +1,105 @@ +/** + * Per-key trailing-edge coalescing with a starvation cap. + * + * A burst of edges for one key collapses into a single `emit(key)` once the key has been quiet for + * `flushMs`; under sustained churn the cap forces an emit every `maxWaitMs` so a key that never + * goes quiet still makes progress. `emit` is expected to read the latest state itself, so the + * intermediate edges it never sees carry no information. + * + * Extracted from the session.tabs notify coalescer so the orchestration redrive edge coalesces on + * the same mechanism rather than a second timer layer with its own bugs. The windows stay per + * caller: 50ms is right for a spinner-driven title, and much too tight for a journal stream. + */ + +export type KeyedTrailingEdgeCoalescer = { + /** Schedule a coalesced emit for a key. */ + schedule: (key: string) => void + /** Drop a key's pending emit without firing. Use when it has been superseded or the key is gone. */ + cancel: (key: string) => void + /** Fire a key's pending emit now, if it has one. */ + flush: (key: string) => void + /** Fire every pending emit now. */ + flushAll: () => void + /** Drop all pending state without emitting (teardown). */ + dispose: () => void +} + +export type KeyedTrailingEdgeCoalescerOptions = { + /** Quiet window before a coalesced emit fires. */ + flushMs: number + /** Longest a key may be held back under sustained churn. */ + maxWaitMs: number +} + +type PendingEmit = { + timer: ReturnType + firstScheduledAt: number +} + +export function createKeyedTrailingEdgeCoalescer( + emit: (key: string) => void, + options: KeyedTrailingEdgeCoalescerOptions +): KeyedTrailingEdgeCoalescer { + const pending = new Map() + + const clear = (key: string): void => { + const entry = pending.get(key) + if (!entry) { + return + } + clearTimeout(entry.timer) + pending.delete(key) + } + + const fire = (key: string): void => { + clear(key) + emit(key) + } + + const arm = (key: string): ReturnType => { + const timer = setTimeout(() => fire(key), options.flushMs) + if (typeof timer.unref === 'function') { + timer.unref() + } + return timer + } + + return { + schedule(key: string): void { + const now = Date.now() + const existing = pending.get(key) + if (existing) { + // Cap total delay so sustained churn can't starve the emit forever. + if (now - existing.firstScheduledAt >= options.maxWaitMs) { + fire(key) + return + } + clearTimeout(existing.timer) + existing.timer = arm(key) + return + } + pending.set(key, { timer: arm(key), firstScheduledAt: now }) + }, + cancel(key: string): void { + clear(key) + }, + flush(key: string): void { + if (pending.has(key)) { + fire(key) + } + }, + flushAll(): void { + // Snapshot keys first: fire() deletes from `pending`, and emit may schedule new work, so + // mutating the live map mid-iteration is unsafe. + for (const key of Array.from(pending.keys())) { + fire(key) + } + }, + dispose(): void { + for (const entry of pending.values()) { + clearTimeout(entry.timer) + } + pending.clear() + } + } +} diff --git a/src/main/runtime/mobile-rpc-allowlist.test.ts b/src/main/runtime/mobile-rpc-allowlist.test.ts index 1a27de18d55..1f196400ea3 100644 --- a/src/main/runtime/mobile-rpc-allowlist.test.ts +++ b/src/main/runtime/mobile-rpc-allowlist.test.ts @@ -167,6 +167,7 @@ describe('mobile RPC allowlist', () => { 'agentSession.setOption', 'agentSession.handoffStatus', 'agentSession.options', + 'agentSession.commands', 'agentSession.history', 'agentSession.subscribe', 'agentSession.unsubscribe', diff --git a/src/main/runtime/mobile-session-tabs-notify-coalescer.ts b/src/main/runtime/mobile-session-tabs-notify-coalescer.ts index ac6608b71fb..4bb66b41b2b 100644 --- a/src/main/runtime/mobile-session-tabs-notify-coalescer.ts +++ b/src/main/runtime/mobile-session-tabs-notify-coalescer.ts @@ -7,6 +7,11 @@ // safe. Structural changes (tab added/removed/activated) bypass this via an // immediate flush so they still propagate promptly. +import { + createKeyedTrailingEdgeCoalescer, + type KeyedTrailingEdgeCoalescer +} from './keyed-trailing-edge-coalescer' + // Trailing-edge window: title/status is latency-sensitive UI, so this is // tighter than files.watch's 150ms but looser than native-chat's 40ms. const SESSION_TABS_FLUSH_MS = 50 @@ -14,25 +19,8 @@ const SESSION_TABS_FLUSH_MS = 50 // keeps spinning never starves the emit indefinitely. const SESSION_TABS_MAX_WAIT_MS = 250 -export type MobileSessionTabsNotifyCoalescer = { - // Schedule a coalesced (trailing-edge) notify for a worktree. - schedule: (worktreeId: string) => void - // Cancel any pending notify for a worktree without emitting. Use when an - // immediate emit has already superseded the pending state, or the worktree - // was removed and a stale notify must not fire. - cancel: (worktreeId: string) => void - // Flush a worktree's pending notify now (emit if one is pending). - flush: (worktreeId: string) => void - // Flush every pending worktree now. - flushAll: () => void - // Drop all pending state without emitting (runtime teardown). - dispose: () => void -} - -type PendingNotify = { - timer: ReturnType - firstScheduledAt: number -} +/** Keys are worktree ids; `emit` reads the latest snapshot for the worktree itself. */ +export type MobileSessionTabsNotifyCoalescer = KeyedTrailingEdgeCoalescer /** * Coalesces per-worktree session.tabs notifications on a short trailing-edge @@ -43,67 +31,8 @@ type PendingNotify = { export function createMobileSessionTabsNotifyCoalescer( emit: (worktreeId: string) => void ): MobileSessionTabsNotifyCoalescer { - const pending = new Map() - - const clear = (worktreeId: string): void => { - const entry = pending.get(worktreeId) - if (!entry) { - return - } - clearTimeout(entry.timer) - pending.delete(worktreeId) - } - - const fire = (worktreeId: string): void => { - clear(worktreeId) - emit(worktreeId) - } - - const arm = (worktreeId: string): ReturnType => { - const timer = setTimeout(() => fire(worktreeId), SESSION_TABS_FLUSH_MS) - if (typeof timer.unref === 'function') { - timer.unref() - } - return timer - } - - return { - schedule(worktreeId: string): void { - const now = Date.now() - const existing = pending.get(worktreeId) - if (existing) { - // Cap total delay so sustained churn can't starve the emit forever. - if (now - existing.firstScheduledAt >= SESSION_TABS_MAX_WAIT_MS) { - fire(worktreeId) - return - } - clearTimeout(existing.timer) - existing.timer = arm(worktreeId) - return - } - pending.set(worktreeId, { timer: arm(worktreeId), firstScheduledAt: now }) - }, - cancel(worktreeId: string): void { - clear(worktreeId) - }, - flush(worktreeId: string): void { - if (pending.has(worktreeId)) { - fire(worktreeId) - } - }, - flushAll(): void { - // Snapshot keys first: fire() deletes from `pending`, and emit may - // schedule new work, so mutating the live map mid-iteration is unsafe. - const worktreeIds = Array.from(pending.keys()) - for (const worktreeId of worktreeIds) { - fire(worktreeId) - } - }, - dispose(): void { - for (const entry of pending.values()) { - clearTimeout(entry.timer) - } - pending.clear() - } - } + return createKeyedTrailingEdgeCoalescer(emit, { + flushMs: SESSION_TABS_FLUSH_MS, + maxWaitMs: SESSION_TABS_MAX_WAIT_MS + }) } diff --git a/src/main/runtime/orca-runtime-adopt-terminal-orphans-from-inventory.ts b/src/main/runtime/orca-runtime-adopt-terminal-orphans-from-inventory.ts index 3e006916396..d062870c4d8 100644 --- a/src/main/runtime/orca-runtime-adopt-terminal-orphans-from-inventory.ts +++ b/src/main/runtime/orca-runtime-adopt-terminal-orphans-from-inventory.ts @@ -1,4 +1,9 @@ // @ts-nocheck -- mechanically split from OrcaRuntimeService; behavior is covered by AST equivalence and characterization tests. +import { + observeStructuredWorker, + resolveStructuredWorkerAuthority +} from './structured-worker-authority' +import type { RuntimeLeafRecord } from './runtime-terminal-state-records' import { OrcaRuntimeWithSubscribeToTerminalResize } from './orca-runtime-subscribe-to-terminal-resize' import type { RuntimeMobileSessionTabsResult, @@ -82,7 +87,10 @@ export class OrcaRuntimeWithAdoptTerminalOrphansFromInventory extends OrcaRuntim // Why: when --terminal is omitted, the CLI auto-resolves to the active // terminal in the current worktree — matching browser's implicit active tab. - async resolveActiveTerminal(worktreeSelector?: string): Promise { + async resolveActiveTerminal( + worktreeSelector?: string, + options: { requireUnambiguous?: boolean } = {} + ): Promise { if (this.graphStatus !== 'ready') { const targetWorktreeId = worktreeSelector ? (await this.resolveWorktreeSelector(worktreeSelector)).id @@ -90,7 +98,9 @@ export class OrcaRuntimeWithAdoptTerminalOrphansFromInventory extends OrcaRuntim const snapshots = targetWorktreeId ? [this.getMobileSessionTabsForWorktree(targetWorktreeId)] : await this.listAllMobileSessionTabs() - for (const snapshot of snapshots) { + // Skipped for an identity claim for the same reason as the ready path below: the active tab + // is where the user last looked, which says nothing about which terminal the CALLER is. + for (const snapshot of options.requireUnambiguous ? [] : snapshots) { const activeTerminal = snapshot.tabs.find( (tab) => tab.type === 'terminal' && @@ -105,6 +115,10 @@ export class OrcaRuntimeWithAdoptTerminalOrphansFromInventory extends OrcaRuntim const listed = await this.listTerminals(worktreeSelector, undefined, { includeVisualLayouts: false }) + // Same arbitrary pick, same misattribution: refuse for callers claiming their own identity. + if (options.requireUnambiguous && listed.terminals.length > 1) { + throw new Error('no_active_terminal') + } const first = listed.terminals[0]?.handle if (first) { return first @@ -117,8 +131,11 @@ export class OrcaRuntimeWithAdoptTerminalOrphansFromInventory extends OrcaRuntim ? (await this.resolveWorktreeSelector(worktreeSelector)).id : null - // Prefer the tab's activeLeafId — this is the pane the user last focused - for (const tab of this.tabs.values()) { + // Prefer the tab's activeLeafId — this is the pane the user last focused. + // + // Skipped entirely for an identity claim: which pane the user last looked at says nothing + // about which terminal the CALLER is, so preferring it is still a guess. + for (const tab of options.requireUnambiguous ? [] : this.tabs.values()) { if (targetWorktreeId && tab.worktreeId !== targetWorktreeId) { continue } @@ -132,12 +149,37 @@ export class OrcaRuntimeWithAdoptTerminalOrphansFromInventory extends OrcaRuntim } } - // Fallback: any leaf in the target worktree + // Fallback: any leaf in the target worktree. + // + // `requireUnambiguous` callers are asking "which terminal AM I" — today the implicit `--from` + // sender — and an arbitrary iteration-order pick answers that with someone else's pane: a bare + // `send --type worker_done` then settles a SIBLING's context-only dispatch, a tier that has no + // capability token to reject on, and every message it sends is attributed to that sibling. + // Refusing is the only safe answer when more than one leaf could be meant. + // + // `check` resolves through the `--terminal` scope, which still guesses, and a DISPATCHED + // structured worker is covered by the `ORCA_TERMINAL_HANDLE` its child is spawned with. That + // was once written as covering structured sessions generally, and it never did: an ordinary + // structured chat session is not in the worker registry, so it is spawned with no handle at + // all, and the guess below handed it a sibling's pane — which a destructive `check` then + // consumed. `requireUnambiguous` does not save it either, because with exactly one terminal + // pane the guess resolves. Such a child now carries `ORCA_STRUCTURED_SESSION` and the CLI + // refuses before reaching here (`shared/structured-session-marker.ts`). + const candidates: RuntimeLeafRecord[] = [] for (const leaf of this.leaves.values()) { if (targetWorktreeId && leaf.worktreeId !== targetWorktreeId) { continue } - return this.issueHandle(leaf) + if (!options.requireUnambiguous) { + return this.issueHandle(leaf) + } + candidates.push(leaf) + if (candidates.length > 1) { + break + } + } + if (candidates.length === 1) { + return this.issueHandle(candidates[0]!) } throw new Error('no_active_terminal') @@ -147,10 +189,25 @@ export class OrcaRuntimeWithAdoptTerminalOrphansFromInventory extends OrcaRuntim // identity at dispatch time; null (best-effort) rather than throwing so // dispatch still works for handles without a resolvable pane. getTerminalPaneKey(handle: string): string | null { - return this.getPaneKeyForTerminalHandle(handle) + return ( + resolveStructuredWorkerAuthority(handle, this.getOrchestrationDbIfAvailable?.() ?? null) + ?.identity.paneKey ?? this.getPaneKeyForTerminalHandle(handle) + ) } getLiveTerminalPaneKey(handle: string): string | null { + const structured = resolveStructuredWorkerAuthority( + handle, + this.getOrchestrationDbIfAvailable?.() ?? null + ) + if (structured) { + // `resolveBareOrchestrationRecipient` routes direct mail through this, not through + // getTerminalPaneKey. The connected-gate below exists so mail is never routed to a corpse, + // so the structured answer needs a real liveness proof too, not just a registry hit. + return observeStructuredWorker(structured.identity).status === 'live' + ? structured.identity.paneKey + : null + } const runtimePty = this.getLivePtyForHandle(handle) if (runtimePty) { return runtimePty.pty.connected ? (runtimePty.pty.paneKey ?? null) : null diff --git a/src/main/runtime/orca-runtime-build-pty-terminal-summary.ts b/src/main/runtime/orca-runtime-build-pty-terminal-summary.ts index b9835be212c..d781bd2ae90 100644 --- a/src/main/runtime/orca-runtime-build-pty-terminal-summary.ts +++ b/src/main/runtime/orca-runtime-build-pty-terminal-summary.ts @@ -7,6 +7,7 @@ import { getLatestPtyTitle } from './runtime-worktree-status-projection' import { parsePaneKey } from '../../shared/stable-pane-id' import type { TerminalHandleRecord } from './runtime-terminal-contracts' import { readTerminalTail } from './terminal-tail-read' +import { structuredWorkerTerminalRefusal } from './structured-worker-terminal-refusal' import { randomUUID } from 'node:crypto' export class OrcaRuntimeWithBuildPtyTerminalSummary extends OrcaRuntimeWithGetPtyRecordForPaneKey { @@ -52,7 +53,11 @@ export class OrcaRuntimeWithBuildPtyTerminalSummary extends OrcaRuntimeWithGetPt this.assertGraphReady() const record = this.handles.get(handle) if (!record || record.runtimeId !== this.runtimeId) { - throw new Error('terminal_handle_stale') + // A structured worker's handle is not stale — nothing went dead. It names a live agent + // session that simply has no terminal, and saying `terminal_handle_stale` sent callers + // hunting for a remint that will never exist. Read paths (`terminal read`, + // `isTerminalRunningAgent`, the identity probe) answer for it BEFORE reaching here. + throw structuredWorkerTerminalRefusal(handle, this._orchestrationDb) } if (record.rendererGraphEpoch !== this.rendererGraphEpoch) { throw new Error('terminal_handle_stale') diff --git a/src/main/runtime/orca-runtime-close-mobile-session-tab.ts b/src/main/runtime/orca-runtime-close-mobile-session-tab.ts index a0a02474911..42f17c533e3 100644 --- a/src/main/runtime/orca-runtime-close-mobile-session-tab.ts +++ b/src/main/runtime/orca-runtime-close-mobile-session-tab.ts @@ -292,7 +292,7 @@ export class OrcaRuntimeWithCloseMobileSessionTab extends OrcaRuntimeWithRefuseU } } } - await this.closeStructuredAgentSessionTab(worktreeId, snapshot, tab) + await this.closeStructuredAgentSessionTab(tab) } else { if (!this.notifier?.closeSessionTab) { throw new Error('runtime_unavailable') diff --git a/src/main/runtime/orca-runtime-close-structured-agent-session-tab.ts b/src/main/runtime/orca-runtime-close-structured-agent-session-tab.ts index bd282b6575d..568083bf177 100644 --- a/src/main/runtime/orca-runtime-close-structured-agent-session-tab.ts +++ b/src/main/runtime/orca-runtime-close-structured-agent-session-tab.ts @@ -13,42 +13,48 @@ import type { BrowserSessionTabSelectionOptions } from './browser-tab-create-pub import { getRuntimeBrowserPageRegistry } from './runtime-browser-page-registry' import { applyBrowserSessionTabSelection } from './browser-session-tab-selection-snapshot' import { getStructuredAgentSessionHost } from '../native-chat/agent-session-wire/structured-agent-session-registry' +import { retireStructuredAgentSessionTabFrom } from './structured-agent-session-tab-retirement' export class OrcaRuntimeWithCloseStructuredAgentSessionTab extends OrcaRuntimeWithCloseMobileSessionTab { - protected async closeStructuredAgentSessionTab( - worktreeId: string, - snapshot: RuntimeMobileSessionTabsSnapshot, - tab: RuntimeMobileSessionAgentTab - ): Promise { + protected async closeStructuredAgentSessionTab(tab: RuntimeMobileSessionAgentTab): Promise { const host = getStructuredAgentSessionHost() if (host) { if (typeof host.setSessionTabVisibility === 'function') { await host.setSessionTabVisibility(tab.sessionId, false) } } - const nextTabs = snapshot.tabs.filter((candidate) => candidate.id !== tab.id) - const active = nextTabs.find((candidate) => candidate.isActive) ?? nextTabs[0] ?? null - const nextSnapshot: RuntimeMobileSessionTabsSnapshot = { - ...snapshot, - snapshotVersion: snapshot.snapshotVersion + 1, - activeTabId: active?.id ?? null, - activeTabType: active?.type ?? null, - tabGroups: (snapshot.tabGroups ?? []).map((group) => ({ - ...group, - tabOrder: group.tabOrder.filter((id) => id !== tab.id), - activeTabId: group.activeTabId === tab.id ? null : group.activeTabId, - recentTabIds: group.recentTabIds?.filter((id) => id !== tab.id) - })), - tabs: nextTabs - } - this.storeMobileSessionSnapshot(worktreeId, nextSnapshot) - this.emitMobileSessionTabsSnapshot(nextSnapshot) // Retire durable visibility and the runtime snapshot before stopping the provider. + this.retireStructuredAgentSessionTabFromSnapshot(tab.sessionId) if (typeof host?.close === 'function') { await host.close(tab.sessionId) } } + /** + * Prunes a structured session's chat tab from whichever worktree snapshot still carries it. + * + * Public because orchestration settles structured workers outside the tab surface: stop, release + * and the half-started discard all prove their own close and then have to retire the tab that + * `publishStructuredAgentSessionTab` put on screen. `setSessionTabVisibility(false)` only clears + * the DURABLE restore index, so without this the dead chat tab survives for the rest of the app + * session and re-attaches the released session when opened. + * + * Snapshot-only and renderer-free: it never asks the renderer to close anything, so it is safe on + * the startup release reconciler where no renderer exists. + */ + retireStructuredAgentSessionTabFromSnapshot(sessionId: string): boolean { + for (const [worktreeId, snapshot] of this.mobileSessionTabsByWorktree) { + const nextSnapshot = retireStructuredAgentSessionTabFrom(snapshot, sessionId) + if (!nextSnapshot) { + continue + } + this.storeMobileSessionSnapshot(worktreeId, nextSnapshot) + this.emitMobileSessionTabsSnapshot(nextSnapshot) + return true + } + return false + } + // Why: a refused echoed close means the echoing client already pruned its // local mirror. Bump the version and emit the unchanged snapshot so clients // that dedupe by snapshotVersion re-add and re-attach the still-live tab. @@ -206,8 +212,7 @@ export class OrcaRuntimeWithCloseStructuredAgentSessionTab extends OrcaRuntimeWi snapshot, tabId: tab.id, ...(targetGroupId !== undefined ? { targetGroupId } : {}), - focusesHost, - publicationEpoch: `headless:${Date.now().toString(36)}` + focusesHost }) this.storeMobileSessionSnapshot(worktreeId, nextSnapshot) // Why: browser group membership is otherwise live-only; persist it so a diff --git a/src/main/runtime/orca-runtime-create-runtime-owned-mobile-session-terminal.ts b/src/main/runtime/orca-runtime-create-runtime-owned-mobile-session-terminal.ts index dd74cfbfb7a..31fb8a90ab0 100644 --- a/src/main/runtime/orca-runtime-create-runtime-owned-mobile-session-terminal.ts +++ b/src/main/runtime/orca-runtime-create-runtime-owned-mobile-session-terminal.ts @@ -120,7 +120,8 @@ export class OrcaRuntimeWithCreateRuntimeOwnedMobileSessionTerminal extends Orca } const next: RuntimeMobileSessionTabsSnapshot = { worktree: worktreeId, - publicationEpoch: `headless:${Date.now().toString(36)}`, + // Why: a fresh epoch retires the current publisher, so clients drop its later tab updates. + publicationEpoch: existing?.publicationEpoch ?? `headless:${Date.now().toString(36)}`, snapshotVersion: (existing?.snapshotVersion ?? 0) + 1, // Why: activating the new tab also focuses its group, so a "+" targeting a specific split group makes that group active too. activeGroupId: diff --git a/src/main/runtime/orca-runtime-get-orchestration-dispatch-authority.ts b/src/main/runtime/orca-runtime-get-orchestration-dispatch-authority.ts index 76308c2df08..14345d10a77 100644 --- a/src/main/runtime/orca-runtime-get-orchestration-dispatch-authority.ts +++ b/src/main/runtime/orca-runtime-get-orchestration-dispatch-authority.ts @@ -16,6 +16,7 @@ import { import { getAppEnvironment } from '../../shared/app-environment' import type { FleetAgentStatusEvidence } from '../../shared/orchestration-fleet-agent-status-evidence' import { readOrchestrationFleetAgentStatusSnapshot } from './orchestration-fleet-agent-status-snapshot' +import { resolveStructuredWorkerAuthority } from './structured-worker-authority' export class OrcaRuntimeWithGetOrchestrationDispatchAuthority extends OrcaRuntimeWithVerifyOrchestrationCompatibilityCaller { /** Every pane key this PTY could be addressed by, including restored receipts. */ @@ -40,6 +41,26 @@ export class OrcaRuntimeWithGetOrchestrationDispatchAuthority extends OrcaRuntim getOrchestrationDispatchAuthority( terminalHandle: string ): OrchestrationCompatibilityTerminalAuthority | null { + const structured = resolveStructuredWorkerAuthority( + terminalHandle, + this.getOrchestrationDbIfAvailable?.() ?? null + ) + if (structured) { + return { + runtimeId: this.runtimeId, + terminalHandle, + // Both EMPTY on purpose. `verifyOrchestrationCompatibilityCaller` falls back to the + // restored-authority receipt keyed by ptyId when there is no launch token, so filling + // either of these in would silently open hook attestation to a session that has no PTY, + // no launch secret, and no hook to attest with. + ptyId: '', + worktreeId: structured.identity.worktreeId, + processIncarnation: structured.identity.processIncarnation, + paneKey: structured.identity.paneKey, + launchTokenHash: null, + hostScope: structured.identity.hostScope + } + } let ptyId: string | null try { ptyId = diff --git a/src/main/runtime/orca-runtime-get-pty-record-for-pane-key.ts b/src/main/runtime/orca-runtime-get-pty-record-for-pane-key.ts index 8e34244b3d0..0c41b176d34 100644 --- a/src/main/runtime/orca-runtime-get-pty-record-for-pane-key.ts +++ b/src/main/runtime/orca-runtime-get-pty-record-for-pane-key.ts @@ -4,6 +4,15 @@ import type { RuntimeLeafRecord, RuntimePtyWorktreeRecord } from './runtime-term import { isTerminalLeafId, makePaneKey, parsePaneKey } from '../../shared/stable-pane-id' import { detectAgentStatusFromTitle, isClaudeManagementTitle } from '../../shared/agent-detection' import { recognizeAgentProcess } from '../../shared/agent-process-recognition' +import { agentSessionPtyWriteGate } from './agent-session-pty-write-gate' +import { resolveStructuredWorkerAuthority } from './structured-worker-authority' +import { structuredWorkerIdentities } from './structured-worker-identity' +import { isSettledNativeOwner } from './orchestration/structured-session-pointer-delivery' +import type { StructuredPointerTarget } from './orchestration/structured-mailbox-pointer-delivery' +import { + resolveTerminalIdentityFromProbes, + type RuntimeTerminalIdentity +} from './terminal-identity-probe' export class OrcaRuntimeWithGetPtyRecordForPaneKey extends OrcaRuntimeWithPruneMobileSessionTabGroupLayout { protected getPtyRecordForPaneKey(paneKey: string): RuntimePtyWorktreeRecord | null { @@ -140,10 +149,151 @@ export class OrcaRuntimeWithGetPtyRecordForPaneKey extends OrcaRuntimeWithPruneM } } + /** + * The identity seam: whether this handle still names a live agent identity, in either lane. + * + * Read-only by construction — a handle and a boolean — so it can serve the CLI's sender + * validation without `terminal.show`'s writable-looking pane payload. + */ + resolveTerminalIdentity(handle: string): RuntimeTerminalIdentity { + return resolveTerminalIdentityFromProbes(handle, { + isLiveStructuredWorker: () => + Boolean(resolveStructuredWorkerAuthority(handle, this._orchestrationDb)), + hasLivePty: () => Boolean(this.getLivePtyForHandle(handle)), + assertLiveLeaf: () => { + this.getLiveLeafForHandle(handle) + } + }) + } + + /** + * A structured worker's own pane key, for callers that can only name the session. + * + * Resolved HERE rather than published: the pane key is a random identity credential — anyone + * holding it can read and consume that worker's mailbox, and session ids are embedded in tab ids + * — so it must never travel to a renderer to be echoed back. + */ + getStructuredWorkerPaneKeyForSession(sessionId: string): string | null { + const identity = structuredWorkerIdentities.getBySessionId(sessionId) + return identity && resolveStructuredWorkerAuthority(identity.handle, this._orchestrationDb) + ? identity.paneKey + : null + } + deliverPendingMessagesForHandle(handle: string, reservedTypes?: ReadonlySet): void { this.orchestrationMailboxNotifications.deliverForHandle(handle, reservedTypes) } + /** The structured idle edge: any journal movement is a chance to redrive parked mail. */ + notifyStructuredSessionJournalActivity(sessionId: string): void { + this.orchestrationStructuredMailboxPointerDelivery.onJournalActivity(sessionId) + } + + /** Settlement drops anything parked for the session; nothing will ever redrive it again. */ + forgetStructuredSessionMail(sessionId: string): void { + this.orchestrationStructuredMailboxPointerDelivery.forgetSession(sessionId) + } + + /** + * The session a mailbox must be nudged through, or null when a live PTY can take the bytes. + * + * All THREE address forms a structured session can own resolve here — its `dispatch:` address, + * its `run:` mailbox when it coordinates, and its own bearer handle for peer mail outside a + * dispatch. `run:` was the one that fell in a hole: the PTY lane declines because the owner is + * structured, and this lane used to decline anything that was not `dispatch:`, so each half + * believed the other owned it and a structured coordinator was never nudged. + */ + protected resolveStructuredMailboxTarget(mailboxHandle: string): StructuredPointerTarget | null { + if (mailboxHandle.startsWith('run:')) { + return this.resolveStructuredCoordinatorMailboxTarget(mailboxHandle.slice('run:'.length)) + } + if (!mailboxHandle.startsWith('dispatch:')) { + return this.resolveStructuredWorkerDirectMailboxTarget(mailboxHandle) + } + const dispatchId = mailboxHandle.slice('dispatch:'.length) + const assignee = this._orchestrationDb?.getDispatchContextById?.(dispatchId)?.assignee_handle + if (!assignee) { + return null + } + const identity = resolveStructuredWorkerAuthority(assignee, this._orchestrationDb)?.identity + if (identity) { + return { sessionId: identity.sessionId, dispatchId } + } + return this.resolveAdoptedStructuredMailboxTarget(assignee, dispatchId) + } + + /** + * A Run's own mailbox, when the coordinator holding it is a structured session. + * + * A structured coordinator does NOT block in `check --wait` the way a PTY one does — it is a + * chat session, and its turn ends — so the waiter that used to preempt pointer delivery is not + * there to cover for the missing nudge. Session-scoped: a coordinator's run mailbox has no + * dispatch, and needs none, since the ledger bucket is all a dispatch id ever supplied. + */ + protected resolveStructuredCoordinatorMailboxTarget( + runId: string + ): StructuredPointerTarget | null { + const coordinator = this._orchestrationDb?.getRun?.(runId)?.coordinator_handle + if (!coordinator) { + return null + } + const identity = resolveStructuredWorkerAuthority(coordinator, this._orchestrationDb)?.identity + return identity ? { sessionId: identity.sessionId, dispatchId: null } : null + } + + /** + * Direct peer mail, addressed to the worker's own handle rather than to a dispatch. + * + * Nothing else can serve it: the PTY lane refuses a structured handle outright, so without this + * the send stores durably, reports success, and no lane ever nudges the worker — the sender sees + * success and the peer waiting on a reply hangs. + * + * The worker's ACTIVE dispatch is preferred when it has one, so peer and coordinator nudges share + * one operation-ledger budget and one set of retain rules. A worker BETWEEN dispatches is still + * nudged, under a session-scoped budget: the mail is durable, the session is live, and a dispatch + * says nothing about whether delivery is safe — the idle gate and the lease fence do that. + */ + protected resolveStructuredWorkerDirectMailboxTarget( + handle: string + ): StructuredPointerTarget | null { + const db = this._orchestrationDb + // Answers null for anything that is not a live structured worker of THIS runtime, so `run:` + // and PTY handles fall through to the PTY lane exactly as before. + const identity = resolveStructuredWorkerAuthority(handle, db)?.identity + if (!identity) { + return null + } + const dispatchId = db?.findActiveDispatchForAssignee?.(handle, identity.paneKey)?.id ?? null + return { sessionId: identity.sessionId, dispatchId } + } + + /** + * A PTY-born worker whose pane was since adopted by native chat. + * + * Its bytes cannot land — every runtime write path re-admits through the same gate — so the + * pointer has to travel as a session turn instead. Only a SETTLED native owner qualifies: a + * mid-handoff lease may become a TUI again, and redirecting there races the takeover. + */ + protected resolveAdoptedStructuredMailboxTarget( + assignee: string, + dispatchId: string + ): StructuredPointerTarget | null { + let ptyId: string | null | undefined + try { + ptyId = this.getLiveLeafForHandle(assignee).leaf.ptyId + } catch { + return null + } + if (!ptyId) { + return null + } + const admission = agentSessionPtyWriteGate.admit(ptyId) + if (admission.admitted || !isSettledNativeOwner(admission.refusal)) { + return null + } + return { sessionId: admission.refusal.sessionId, dispatchId, refusal: admission.refusal } + } + protected scheduleRestoredMessageRepoints(): void { let handles: Set try { diff --git a/src/main/runtime/orca-runtime-get-terminal-interactive-wait.ts b/src/main/runtime/orca-runtime-get-terminal-interactive-wait.ts index 059f4ffd7b3..a87f6cc03f2 100644 --- a/src/main/runtime/orca-runtime-get-terminal-interactive-wait.ts +++ b/src/main/runtime/orca-runtime-get-terminal-interactive-wait.ts @@ -1,4 +1,5 @@ // @ts-nocheck -- mechanically split from OrcaRuntimeService; behavior is covered by AST equivalence and characterization tests. +import { resolveStructuredWorkerAuthority } from './structured-worker-authority' import { OrcaRuntimeWithAdoptTerminalOrphansFromInventory } from './orca-runtime-adopt-terminal-orphans-from-inventory' import type { RuntimeTerminalAgentStatus, @@ -119,6 +120,13 @@ export class OrcaRuntimeWithGetTerminalInteractiveWait extends OrcaRuntimeWithAd } getTerminalProcessIncarnation(handle: string): string | null { + const structured = resolveStructuredWorkerAuthority( + handle, + this.getOrchestrationDbIfAvailable?.() ?? null + ) + if (structured) { + return structured.identity.processIncarnation + } const live = this.getLivePtyForHandle(handle) const record = live?.record ?? this.handles.get(handle) if (!record?.ptyId) { diff --git a/src/main/runtime/orca-runtime-get-worktree-ps.ts b/src/main/runtime/orca-runtime-get-worktree-ps.ts index 06391e2ae83..240156d93c6 100644 --- a/src/main/runtime/orca-runtime-get-worktree-ps.ts +++ b/src/main/runtime/orca-runtime-get-worktree-ps.ts @@ -14,6 +14,8 @@ import type { AgentSessionRecord } from '../../shared/agent-session-record' import type { Repo } from '../../shared/repo-types' import { enrichMissingRepoGitRemoteIdentities } from '../repo-git-remote-identity-enrichment' import { ensureStructuredAgentSessionHost as installStructuredAgentSessionHost } from './structured-agent-session-runtime' +import { maybeAutoRenameWorkspaceOnFirstStructuredTurn } from '../agent-hooks/first-work-structured-session-rename' +import { firstWorkRenameDeps } from '../agent-hooks/first-work-rename-runtime' import { getProfileUserDataPath } from '../orca-profiles/profile-storage-paths' import { LOCAL_EXECUTION_HOST_ID } from '../../shared/execution-host' import { buildWorktreeListingPage } from './worktree-listing-host-scope' @@ -156,6 +158,15 @@ export class OrcaRuntimeWithGetWorktreePs extends OrcaRuntimeWithStructuredAgent claudeStructuredAuthPolicyForSettings(this.requireStore().getSettings()), // Same gate and same settings as agentSession.createSupport, re-read on every acquisition. getClaudeManagedAccountGateSettings: () => this.requireStore().getSettings(), + // Structured chat has no agent CLI hooks, so this projection is what the first-work + // workspace rename listens to instead of `agentStatus:set`. + onSessionStatusChanged: (summary, options) => { + void maybeAutoRenameWorkspaceOnFirstStructuredTurn( + summary, + options, + firstWorkRenameDeps(this.requireStore(), this) + ) + }, handoffTransport: this.createStructuredAgentSessionHandoffTransport() }) } diff --git a/src/main/runtime/orca-runtime-process-incarnation-liveness.test.ts b/src/main/runtime/orca-runtime-process-incarnation-liveness.test.ts index 3066367d7e8..9eb72c03a6b 100644 --- a/src/main/runtime/orca-runtime-process-incarnation-liveness.test.ts +++ b/src/main/runtime/orca-runtime-process-incarnation-liveness.test.ts @@ -1,6 +1,8 @@ -import { describe, expect, it, vi } from 'vitest' +import { afterEach, describe, expect, it, vi } from 'vitest' import type { PtyProcessInfo } from '../providers/pty-process-info' +import { setStructuredAgentSessionHost } from '../native-chat/agent-session-wire/structured-agent-session-registry' import { OrcaRuntimeService } from './orca-runtime' +import { structuredWorkerIdentities } from './structured-worker-identity' const SSH_SCOPE = JSON.stringify({ kind: 'ssh', targetId: 'ssh-1' }) const PROCESS_INCARNATION = 'remote:ssh-1:pty-1:inc-1' @@ -87,3 +89,64 @@ describe('terminal process incarnation liveness', () => { expect(listProcesses).toHaveBeenCalledWith(connectionId) }) }) + +describe('structured worker incarnation liveness', () => { + const SESSION = '11111111-1111-4111-a111-111111111111' + const INCARNATION = `structured:${SESSION}` + const LOCAL_SCOPE = JSON.stringify({ kind: 'local', hostId: 'local' }) + + function installHost(lease: Record): void { + setStructuredAgentSessionHost({ + hasSession: () => false, + deps: { + store: { + getRecord: () => ({ location: { executionHostId: 'local', wslDistro: null }, lease }) + } + } + } as never) + } + + afterEach(() => { + setStructuredAgentSessionHost(null) + structuredWorkerIdentities.clear() + }) + + it('settles a stopped worker as exited after its identity was forgotten', async () => { + // Settlement forgets the in-memory identity. Gating on one left the durable resource answering + // `unverifiable` forever, so its row never reconciled out of `worker-list --terminalState + // retained` for the life of the DB. The durable agent-session record is what actually knows. + installHost({ + runtimeKind: 'native', + claimStatus: 'released', + deathEvidence: { kind: 'exit-observed', detail: 'closed', observedAt: 1 }, + runtimeFence: 2 + }) + const runtime = new OrcaRuntimeService() + expect(structuredWorkerIdentities.getBySessionId(SESSION)).toBeNull() + + await expect( + runtime.inspectTerminalProcessIncarnationLiveness(INCARNATION, LOCAL_SCOPE) + ).resolves.toBe('exited') + }) + + it('never answers exited from a record that proves no death', async () => { + installHost({ + runtimeKind: 'native', + claimStatus: 'live', + deathEvidence: null, + runtimeFence: 2 + }) + const runtime = new OrcaRuntimeService() + + await expect( + runtime.inspectTerminalProcessIncarnationLiveness(INCARNATION, LOCAL_SCOPE) + ).resolves.toBe('unverifiable') + }) + + it('stays unverifiable when no structured host is installed to look with', async () => { + const runtime = new OrcaRuntimeService() + await expect( + runtime.inspectTerminalProcessIncarnationLiveness(INCARNATION, LOCAL_SCOPE) + ).resolves.toBe('unverifiable') + }) +}) diff --git a/src/main/runtime/orca-runtime-prune-mobile-session-tab-group-layout.ts b/src/main/runtime/orca-runtime-prune-mobile-session-tab-group-layout.ts index feb36204ff0..c2240de5e8d 100644 --- a/src/main/runtime/orca-runtime-prune-mobile-session-tab-group-layout.ts +++ b/src/main/runtime/orca-runtime-prune-mobile-session-tab-group-layout.ts @@ -24,6 +24,8 @@ import { FIRST_PANE_ID } from '../../shared/pane-key' import { isTerminalLeafId, makePaneKey, parsePaneKey } from '../../shared/stable-pane-id' import type { SleepingAgentLaunchConfig } from '../../shared/agent-session-resume' import { copySleepingAgentLaunchConfig } from './runtime-agent-launch-resolution' +import { resolveStructuredWorkerAuthority } from './structured-worker-authority' +import { structuredWorkerAgentStatus } from './orchestration/structured-worker-group-addressing' export class OrcaRuntimeWithPruneMobileSessionTabGroupLayout extends OrcaRuntimeWithScheduleMobileSessionTabsChanged { protected pruneMobileSessionTabGroupLayout( @@ -189,6 +191,12 @@ export class OrcaRuntimeWithPruneMobileSessionTabGroupLayout extends OrcaRuntime // Why: group address resolution (Section 4.5) queries per-handle status and must not throw on stale handles; return null on any error. getAgentStatusForHandle(handle: string): string | null { + // A structured worker has no pane and no title, so every PTY probe below answers null and + // `@idle` would enumerate it and then silently drop it. Its status is the journal's. + const structured = resolveStructuredWorkerAuthority(handle, this._orchestrationDb) + if (structured) { + return structuredWorkerAgentStatus(structured.identity.sessionId) + } try { const ptyId = this.getTerminalAgentStatusPtyId(handle) return this.getTerminalAgentStatusSnapshot(handle, ptyId).titleStatus @@ -204,7 +212,7 @@ export class OrcaRuntimeWithPruneMobileSessionTabGroupLayout extends OrcaRuntime if (!handle) { return undefined } - return this.agentOrchestrationProjection.getForHandle(handle) + return this.agentOrchestrationProjection.getForHandle(handle, undefined, { paneKey }) } getAgentStatusTerminalHandleForPaneKey(paneKey: string): string | undefined { diff --git a/src/main/runtime/orca-runtime-pty-foreground-process-reads.ts b/src/main/runtime/orca-runtime-pty-foreground-process-reads.ts index 9aeb1a54c79..a7e10247fed 100644 --- a/src/main/runtime/orca-runtime-pty-foreground-process-reads.ts +++ b/src/main/runtime/orca-runtime-pty-foreground-process-reads.ts @@ -19,6 +19,7 @@ import type { FeatureInteractionId } from '../../shared/feature-interactions' import type { RuntimeClientSettingsUpdate } from './runtime-client-settings' import type { TerminalQuickCommand } from '../../shared/terminal-quick-command-types' import type { TerminalQuickCommandMutation } from '../../shared/terminal-quick-commands' +import type { NativeChatSessionOptionSettingsMutation } from '../../shared/native-chat-session-options' import type { Automation } from '../../shared/automations-types' export class OrcaRuntimeWithPtyForegroundProcessReads extends OrcaRuntimeWithStateFields { @@ -214,6 +215,10 @@ export class OrcaRuntimeWithPtyForegroundProcessReads extends OrcaRuntimeWithSta return this.clientSettings.updatePRBotAuthorOverride(args) } + updateClientNativeChatSessionOptions(mutation: NativeChatSessionOptionSettingsMutation): void { + this.clientSettings.updateNativeChatSessionOptions(mutation) + } + listAutomations(): Automation[] { return this.automation.list() } diff --git a/src/main/runtime/orca-runtime-reconcile-headless-mobile-session-browser-tabs.ts b/src/main/runtime/orca-runtime-reconcile-headless-mobile-session-browser-tabs.ts index a9f377e5a7b..ddb74df4dc8 100644 --- a/src/main/runtime/orca-runtime-reconcile-headless-mobile-session-browser-tabs.ts +++ b/src/main/runtime/orca-runtime-reconcile-headless-mobile-session-browser-tabs.ts @@ -25,11 +25,28 @@ export class OrcaRuntimeWithReconcileHeadlessMobileSessionBrowserTabs extends Or worktreeId: string, existing: RuntimeMobileSessionTabsSnapshot ): void { - const liveBrowserTabs = this.buildHeadlessMobileSessionBrowserTabs(worktreeId) - const liveIds = liveBrowserTabs.map((tab) => tab.id) const existingBrowserTabs = existing.tabs.filter( (tab): tab is RuntimeMobileSessionBrowserTab => tab.type === 'browser' ) + const publishedBrowserTabs = this.buildHeadlessMobileSessionBrowserTabs(worktreeId) + // An attached renderer owns its browser rows; the client-page registry cannot retire them. + const rendererBrowserTabs = + this.getAvailableAuthoritativeWindow() && !this.offscreenBrowserBackend + ? existingBrowserTabs.filter((tab) => tab.placement?.kind !== 'client') + : [] + // Keyed by id so no row can publish twice whatever the two sources overlap on; a freshly + // built row wins over the retained one it replaces. + const liveById = new Map( + [...rendererBrowserTabs, ...publishedBrowserTabs].map((tab) => [tab.id, tab]) + ) + // Emit in the order the snapshot already had, because the equality check below compares by + // index: rebuilding renderer-first would read a pure reordering as a change and republish. + const retainedInOrder = existingBrowserTabs.flatMap((tab) => { + const live = liveById.get(tab.id) + return live && liveById.delete(tab.id) ? [live] : [] + }) + const liveBrowserTabs = [...retainedInOrder, ...liveById.values()] + const liveIds = liveBrowserTabs.map((tab) => tab.id) const existingBrowserIds = existingBrowserTabs.map((tab) => tab.id) if (headlessBrowserTabsUnchanged(liveBrowserTabs, existingBrowserTabs)) { return @@ -53,7 +70,6 @@ export class OrcaRuntimeWithReconcileHeadlessMobileSessionBrowserTabs extends Or : (nextTabs.find((tab) => tab.isActive) ?? nextTabs[0] ?? null) this.storeMobileSessionSnapshot(worktreeId, { ...existing, - publicationEpoch: `headless-hydrated:${Date.now().toString(36)}`, snapshotVersion: existing.snapshotVersion + 1, ...(activeStillPresent ? {} diff --git a/src/main/runtime/orca-runtime-remove-orphan-or-folder-worktree.ts b/src/main/runtime/orca-runtime-remove-orphan-or-folder-worktree.ts index 01f4f305803..6c011014b6c 100644 --- a/src/main/runtime/orca-runtime-remove-orphan-or-folder-worktree.ts +++ b/src/main/runtime/orca-runtime-remove-orphan-or-folder-worktree.ts @@ -44,6 +44,9 @@ export async function removeOrphanOrFolderWorktree({ : {}), localProvider: ptyProvider, onPtyStopped: runtime.onPtyStopped ?? undefined, + // A structured session is on no PTY surface, so the sweeps above leave it attached to a + // workspace this removal is about to forget. Closed best-effort, exactly like those PTYs. + closeStructuredSessions: true, ...(externalOrphanHost ? { includeProviderInventory: orphanHost?.kind === 'ssh' && Boolean(sshPtyProvider), diff --git a/src/main/runtime/orca-runtime-resolve-terminal-pane.ts b/src/main/runtime/orca-runtime-resolve-terminal-pane.ts index 64ce6c4df12..51e81ee3d5d 100644 --- a/src/main/runtime/orca-runtime-resolve-terminal-pane.ts +++ b/src/main/runtime/orca-runtime-resolve-terminal-pane.ts @@ -13,6 +13,7 @@ import { readTerminalTail } from './terminal-tail-read' import { getTerminalState } from './terminal-wait-results' +import { readStructuredWorkerTerminal } from './structured-worker-terminal-read' export class OrcaRuntimeWithResolveTerminalPane extends OrcaRuntimeWithGetTerminalInteractiveWait { resolveTerminalPane(paneKey: string, expectedWorktreeId?: string): RuntimeTerminalResolvePane { @@ -181,6 +182,18 @@ export class OrcaRuntimeWithResolveTerminalPane extends OrcaRuntimeWithGetTermin opts: { cursor?: number; limit?: number; screen?: boolean } = {}, providerSnapshot: RuntimeProviderSnapshotReadOptions = {} ): Promise { + // Before the PTY lookup, because a structured worker has no PTY and no leaf: without this the + // only peer read verb answers `terminal_handle_stale` for a perfectly live worker. + const structured = readStructuredWorkerTerminal({ + handle, + db: this.getOrchestrationDbIfAvailable?.() ?? null, + ...(opts.cursor === undefined ? {} : { cursor: opts.cursor }), + ...(opts.limit === undefined ? {} : { limit: opts.limit }) + }) + if (structured) { + // `screen` asks for a rendered grid; there is none, and the journal is the whole record. + return { ...structured, source: opts.screen ? 'screen-unavailable' : 'stream' } + } const pty = this.getLivePtyForHandle(handle) if (pty) { const read = this.readPtyTerminal(handle, pty.pty, opts) diff --git a/src/main/runtime/orca-runtime-restore-structured-agent-session-tabs-once.ts b/src/main/runtime/orca-runtime-restore-structured-agent-session-tabs-once.ts index 5b2c160f2c6..0d7b00fe1b7 100644 --- a/src/main/runtime/orca-runtime-restore-structured-agent-session-tabs-once.ts +++ b/src/main/runtime/orca-runtime-restore-structured-agent-session-tabs-once.ts @@ -76,6 +76,8 @@ export class OrcaRuntimeWithRestoreStructuredAgentSessionTabsOnce extends OrcaRu const existing = this.mobileSessionTabsByWorktree.get(input.workspaceId) const id = `agent-session:${input.sessionId}` if (existing?.tabs.some((tab) => tab.id === id)) { + // A background re-publish is a no-op — no store write, no emit — so it cannot re-surface a + // client whose mirror lost the tab; healing one needs `activate` or an explicit republish. if (!input.activate) { return } diff --git a/src/main/runtime/orca-runtime-serialize-main-terminal-buffer.ts b/src/main/runtime/orca-runtime-serialize-main-terminal-buffer.ts index ed777c1f7d1..6559dbfd349 100644 --- a/src/main/runtime/orca-runtime-serialize-main-terminal-buffer.ts +++ b/src/main/runtime/orca-runtime-serialize-main-terminal-buffer.ts @@ -47,6 +47,10 @@ export class OrcaRuntimeWithSerializeMainTerminalBuffer extends OrcaRuntimeWithA pendingEscapeTailAnsi?: string terminalOwner?: 'shell' } | null> { + const restoredSnapshot = await this.serializePreferredRestoredTerminalBuffer(ptyId, opts) + if (restoredSnapshot) { + return restoredSnapshot + } const headlessSnapshot = await this.serializeHeadlessTerminalBuffer(ptyId, { ...opts, includeEmpty: true diff --git a/src/main/runtime/orca-runtime-serialize-terminal-buffer-from-available-state.ts b/src/main/runtime/orca-runtime-serialize-terminal-buffer-from-available-state.ts index e031dc1b6f5..8969841359b 100644 --- a/src/main/runtime/orca-runtime-serialize-terminal-buffer-from-available-state.ts +++ b/src/main/runtime/orca-runtime-serialize-terminal-buffer-from-available-state.ts @@ -23,18 +23,9 @@ export class OrcaRuntimeWithSerializeTerminalBufferFromAvailableState extends Or kittyKeyboardFlags?: number terminalOwner?: 'shell' } | null> { - if (this.providerSnapshotPreferredPtys.has(ptyId)) { - // Why: pre-attach stream bytes only form a suffix of restored state. A - // sequenced provider snapshot safely reconciles live bytes; renderer is - // the fallback when an older provider cannot expose that boundary. - const providerSnapshot = await this.serializeProviderTerminalBuffer(ptyId, opts) - if (providerSnapshot) { - return providerSnapshot - } - const rendererSnapshot = await this.serializeRendererTerminalBuffer(ptyId, opts) - if (rendererSnapshot) { - return rendererSnapshot - } + const restoredSnapshot = await this.serializePreferredRestoredTerminalBuffer(ptyId, opts) + if (restoredSnapshot) { + return restoredSnapshot } const headlessSnapshot = await this.serializeHeadlessTerminalBuffer(ptyId, opts) if (headlessSnapshot) { @@ -58,6 +49,20 @@ export class OrcaRuntimeWithSerializeTerminalBufferFromAvailableState extends Or : rendererSnapshot } + protected async serializePreferredRestoredTerminalBuffer( + ptyId: string, + opts: { scrollbackRows?: number } = {} + ) { + if (!this.providerSnapshotPreferredPtys.has(ptyId)) { + return null + } + // Pre-attach bytes are only a suffix; older providers can fall back to the renderer. + return ( + (await this.serializeProviderTerminalBuffer(ptyId, opts)) ?? + (await this.serializeRendererTerminalBuffer(ptyId, opts)) + ) + } + async serializeRendererTerminalBuffer( ptyId: string, opts: { scrollbackRows?: number } = {} diff --git a/src/main/runtime/orca-runtime-stop-requested-pty-ids.ts b/src/main/runtime/orca-runtime-stop-requested-pty-ids.ts index 8954c64b379..a3785adb2bb 100644 --- a/src/main/runtime/orca-runtime-stop-requested-pty-ids.ts +++ b/src/main/runtime/orca-runtime-stop-requested-pty-ids.ts @@ -1,4 +1,8 @@ // @ts-nocheck -- mechanically split from OrcaRuntimeService; behavior is covered by AST equivalence and characterization tests. +import { OrchestrationStructuredMailboxPointerDelivery } from './orchestration/structured-mailbox-pointer-delivery' +import { createStructuredMailboxPointerHost } from './orchestration/structured-mailbox-pointer-host' +import { isStructuredWorkerHandle } from './structured-worker-identity' +import { resolveStructuredWorkerAuthority } from './structured-worker-authority' import { OrcaRuntimeWithRuntimeId } from './orca-runtime-runtime-id' import { RuntimeTerminalAgentPresence } from './runtime-terminal-agent-presence' import type { RuntimeNotifier } from './runtime-notifier-contract' @@ -37,6 +41,8 @@ export class OrcaRuntimeWithStopRequestedPtyIds extends OrcaRuntimeWithRuntimeId protected readonly ptyExitListenersByPtyId = new Map void>>() protected readonly terminalAgentPresence = new RuntimeTerminalAgentPresence({ + isLiveStructuredAgent: (handle) => + Boolean(resolveStructuredWorkerAuthority(handle, this._orchestrationDb)), getLivePty: (handle) => this.getLivePtyForHandle(handle)?.pty ?? null, getLiveLeaf: (handle) => this.getLiveLeafForHandle(handle).leaf, getPrimaryLeaf: (ptyId) => this.getLeavesForPty(ptyId)[0] ?? null, @@ -184,6 +190,7 @@ export class OrcaRuntimeWithStopRequestedPtyIds extends OrcaRuntimeWithRuntimeId getDb: () => this._orchestrationDb, getTerminalHandleForPaneKey: (paneKey) => this.getTerminalHandleForPaneKey(paneKey), hasTerminalHandle: (handle) => this.handles.has(handle), + isStructuredWorkerHandle: (handle) => isStructuredWorkerHandle(handle), canProbePtyLiveness: () => Boolean(this.ptyController?.probePtyLiveness), controllerKnowsPtyIsLive: (ptyId) => this.controllerKnowsPtyIsLive(ptyId), isLeafPtyProvenAbsent: (ptyId) => this.isLeafPtyProvenAbsent(ptyId) @@ -207,10 +214,20 @@ export class OrcaRuntimeWithStopRequestedPtyIds extends OrcaRuntimeWithRuntimeId writePty: (ptyId, data) => this.writeOrchestrationPointerPty(ptyId, data) }) + protected readonly orchestrationStructuredMailboxPointerDelivery = + new OrchestrationStructuredMailboxPointerDelivery({ + getDb: () => this._orchestrationDb, + getMessageWaiters: (mailboxHandle) => this.messageWaiters.get(mailboxHandle), + resolveStructuredTarget: (mailboxHandle) => + this.resolveStructuredMailboxTarget(mailboxHandle), + host: createStructuredMailboxPointerHost() + }) + protected readonly orchestrationMailboxNotifications = new OrchestrationMailboxNotificationCoordinator({ mailboxOwner: this.orchestrationMailboxOwner, pointerDelivery: this.orchestrationMailboxPointerDelivery, + structuredPointerDelivery: this.orchestrationStructuredMailboxPointerDelivery, getDb: () => this._orchestrationDb, getLiveLeafForHandle: (handle) => this.getLiveLeafForHandle(handle).leaf, getPaneKeyForHandle: (handle) => { diff --git a/src/main/runtime/orca-runtime-subscribe-to-terminal-resize.ts b/src/main/runtime/orca-runtime-subscribe-to-terminal-resize.ts index a169b1efd69..d610dbb235f 100644 --- a/src/main/runtime/orca-runtime-subscribe-to-terminal-resize.ts +++ b/src/main/runtime/orca-runtime-subscribe-to-terminal-resize.ts @@ -1,4 +1,6 @@ // @ts-nocheck -- mechanically split from OrcaRuntimeService; behavior is covered by AST equivalence and characterization tests. +import { sessionIdFromStructuredWorkerIncarnation } from './structured-worker-identity' +import { observeStructuredWorker } from './rpc/methods/orchestration-structured-worker-lifecycle' import { OrcaRuntimeWithApplyMobileDisplayMode } from './orca-runtime-apply-mobile-display-mode' import { addListenerToMap } from './orca-runtime-core' import { notifyRuntimeListeners, withTimeoutResult } from './runtime-async-boundaries' @@ -171,6 +173,16 @@ export class OrcaRuntimeWithSubscribeToTerminalResize extends OrcaRuntimeWithApp processIncarnation: string, serializedHostScope: string | null ): Promise<'live' | 'exited' | 'unverifiable'> { + const structuredSessionId = sessionIdFromStructuredWorkerIncarnation(processIncarnation) + if (structuredSessionId) { + // A structured session has no PTY, so the process table can only ever fail to find it — + // answering `exited` from that absence would release a running provider child. The durable + // agent-session record is asked directly rather than through the in-memory identity + // registry: settlement forgets the registry entry, so gating on one made a stopped worker's + // resource answer `unverifiable` forever and stay in `worker-list --terminalState retained` + // for the life of the DB. + return observeStructuredWorker({ sessionId: structuredSessionId }).status + } const hostScope = parseWorkerTerminalHostScope(serializedHostScope) if (!hostScope || !this.ptyController?.listProcesses) { return 'unverifiable' diff --git a/src/main/runtime/orca-runtime-tests/lineage-and-scan-cache-part-07.spec.ts b/src/main/runtime/orca-runtime-tests/lineage-and-scan-cache-part-07.spec.ts new file mode 100644 index 00000000000..49177a9af24 --- /dev/null +++ b/src/main/runtime/orca-runtime-tests/lineage-and-scan-cache-part-07.spec.ts @@ -0,0 +1,219 @@ +import { describe, expect, it } from 'vitest' +import { + OrcaRuntimeService, + OrchestrationDb, + createRootDispatch, + makePaneKey +} from '../orca-runtime-test-mocks.spec' +import { TEST_WORKTREE_ID, store } from '../orca-runtime-test-fixtures.spec' + +type RestartTerminal = { + name: string + leafId: string + tabId: string + ptyId: string + paneRuntimeId: number +} + +function makeTerminals(): RestartTerminal[] { + return [ + { name: 'coordinator', leafId: '11111111-1111-4111-8111-111111111111' }, + { name: 'worker', leafId: '22222222-2222-4222-8222-222222222222' }, + { name: 'nested-worker', leafId: '33333333-3333-4333-8333-333333333333' } + ].map((terminal, index) => ({ + ...terminal, + tabId: `tab-${terminal.name}`, + ptyId: `pty-${terminal.name}`, + paneRuntimeId: index + 1 + })) +} + +function makeGraph(terminals: readonly RestartTerminal[]) { + return { + tabs: terminals.map((terminal) => ({ + tabId: terminal.tabId, + worktreeId: TEST_WORKTREE_ID, + title: terminal.name, + activeLeafId: terminal.leafId, + layout: null + })), + leaves: terminals.map((terminal) => ({ + tabId: terminal.tabId, + worktreeId: TEST_WORKTREE_ID, + leafId: terminal.leafId, + paneRuntimeId: terminal.paneRuntimeId, + ptyId: terminal.ptyId, + paneTitle: null + })) + } +} + +/** + * Restart shape: the renderer graph (tab ids, leaf ids, pty ids) is persisted and comes back + * identical, but every terminal handle is minted per process. The daemon keeps the WORKER's + * ORCA_TERMINAL_HANDLE alive so its dispatch still resolves; the coordinator's handle in + * `runs.coordinator_handle` is only ever rebound by a later orchestration command. + */ +/** Attention is projected from liveness facts, not lineage; exact equality is on the rest. */ +function lineageOf( + context: T | undefined +): Omit | undefined { + if (!context) { + return undefined + } + const { attention: _attention, ...lineage } = context + return lineage +} + +describe('OrcaRuntimeService orchestration lineage across restart', () => { + it('projects the coordinator pane key as the worker parent after the handles are reminted', () => { + const terminals = makeTerminals() + const paneKey = (name: string): string => { + const terminal = terminals.find((entry) => entry.name === name) as RestartTerminal + return makePaneKey(terminal.tabId, terminal.leafId) + } + const db = new OrchestrationDb(':memory:') + const before = new OrcaRuntimeService(store) + try { + const beforeHandles = Object.fromEntries( + terminals.map((terminal) => [terminal.name, before.preAllocateHandleForPty(terminal.ptyId)]) + ) + before.setOrchestrationDb(db) + before.attachWindow(1) + before.syncWindowGraph(1, makeGraph(terminals)) + const coordinatorAuthority = before.getOrchestrationDispatchAuthority( + beforeHandles.coordinator + ) + expect(coordinatorAuthority?.processIncarnation).toBeTruthy() + const run = db.createRun({ + objective: 'survive a restart', + coordinatorHandle: beforeHandles.coordinator, + coordinatorPaneKey: paneKey('coordinator') + }) + const workerTask = db.createTask({ + spec: 'worker task', + runId: run.id, + createdByTerminalHandle: beforeHandles.coordinator, + createdByPaneKey: paneKey('coordinator'), + createdByProcessIncarnation: coordinatorAuthority?.processIncarnation ?? undefined, + createdByRunGeneration: run.consumer_generation + }) + const workerAuthority = before.getOrchestrationDispatchAuthority(beforeHandles.worker) + const workerDispatch = createRootDispatch( + db, + workerTask.id, + beforeHandles.worker, + paneKey('worker'), + undefined, + workerAuthority?.processIncarnation ?? undefined + ) + const nestedTask = db.createTask({ + spec: 'nested task', + runId: run.id, + createdByTerminalHandle: beforeHandles.worker, + createdByPaneKey: paneKey('worker'), + createdByProcessIncarnation: workerAuthority?.processIncarnation ?? undefined, + createdByRunGeneration: run.consumer_generation + }) + const nestedDispatch = createRootDispatch( + db, + nestedTask.id, + beforeHandles['nested-worker'], + paneKey('nested-worker') + ) + expect( + before.syncWindowGraph(1, makeGraph(terminals)).agentOrchestrationByPaneKey + ).toMatchObject({ + [paneKey('worker')]: { + parentTerminalHandle: beforeHandles.coordinator, + parentPaneKey: paneKey('coordinator') + }, + [paneKey('nested-worker')]: { + parentTerminalHandle: beforeHandles.worker, + parentPaneKey: paneKey('worker') + } + }) + + // Restart: a fresh runtime, same persisted graph, and the daemon-retained worker handles + // (ORCA_TERMINAL_HANDLE) re-adopted for the still-live worker PTYs. The coordinator did not + // run an orchestration command yet, so its handle is fresh and the Run still names the old one. + const after = new OrcaRuntimeService(store) + after.registerPreAllocatedHandleForPty('pty-worker', beforeHandles.worker) + after.registerPreAllocatedHandleForPty('pty-nested-worker', beforeHandles['nested-worker']) + const freshCoordinatorHandle = after.preAllocateHandleForPty('pty-coordinator') + expect(freshCoordinatorHandle).not.toBe(beforeHandles.coordinator) + after.setOrchestrationDb(db) + after.attachWindow(1) + const contexts = after.syncWindowGraph(1, makeGraph(terminals)).agentOrchestrationByPaneKey + + expect(db.getRun(run.id)?.coordinator_handle).toBe(beforeHandles.coordinator) + expect(lineageOf(contexts?.[paneKey('worker')])).toEqual({ + taskId: workerTask.id, + dispatchId: workerDispatch.id, + dispatchStatus: 'dispatched', + taskTitle: 'worker task', + displayName: 'worker task', + parentTerminalHandle: freshCoordinatorHandle, + parentPaneKey: paneKey('coordinator'), + coordinatorHandle: freshCoordinatorHandle, + orchestrationRunId: run.id + }) + // The nested worker's creator (the worker) kept its daemon handle, but its authority is + // gated on the process incarnation the task was created under; it must still nest under + // the worker pane by durable pane key, never fall through to the coordinator. + expect(lineageOf(contexts?.[paneKey('nested-worker')])).toEqual({ + taskId: nestedTask.id, + dispatchId: nestedDispatch.id, + dispatchStatus: 'dispatched', + taskTitle: 'nested task', + displayName: 'nested task', + parentTerminalHandle: beforeHandles.worker, + parentPaneKey: paneKey('worker'), + coordinatorHandle: freshCoordinatorHandle, + orchestrationRunId: run.id + }) + } finally { + db.close() + } + }) + + it('omits a stale coordinator handle when no live pane owns the coordinator pane key', () => { + const terminals = makeTerminals().filter((terminal) => terminal.name === 'worker') + const workerPaneKey = makePaneKey('tab-worker', terminals[0]!.leafId) + const coordinatorPaneKey = makePaneKey( + 'tab-coordinator', + '11111111-1111-4111-8111-111111111111' + ) + const db = new OrchestrationDb(':memory:') + const runtime = new OrcaRuntimeService(store) + try { + const workerHandle = runtime.preAllocateHandleForPty('pty-worker') + runtime.setOrchestrationDb(db) + runtime.attachWindow(1) + const run = db.createRun({ + objective: 'coordinator pane closed before restart', + coordinatorHandle: 'term_stale-coordinator', + coordinatorPaneKey: coordinatorPaneKey + }) + const task = db.createTask({ spec: 'orphaned worker', runId: run.id }) + const dispatch = createRootDispatch(db, task.id, workerHandle, workerPaneKey) + + const context = runtime.syncWindowGraph(1, makeGraph(terminals)) + .agentOrchestrationByPaneKey?.[workerPaneKey] + + // Why: a handle no live row carries must not reach the renderer, and the durable pane key + // is still published so the row nests again the moment that pane is restored. + expect(lineageOf(context)).toEqual({ + taskId: task.id, + dispatchId: dispatch.id, + dispatchStatus: 'dispatched', + taskTitle: 'orphaned worker', + displayName: 'orphaned worker', + parentPaneKey: coordinatorPaneKey, + orchestrationRunId: run.id + }) + } finally { + db.close() + } + }) +}) diff --git a/src/main/runtime/orca-runtime-tests/paired-settings.spec.ts b/src/main/runtime/orca-runtime-tests/paired-settings.spec.ts index ac3fdd4154d..e5cb14671e3 100644 --- a/src/main/runtime/orca-runtime-tests/paired-settings.spec.ts +++ b/src/main/runtime/orca-runtime-tests/paired-settings.spec.ts @@ -84,6 +84,78 @@ describe('OrcaRuntimeService', () => { expect(runtime.getClientSettings()).not.toHaveProperty('terminalQuickCommands') }) + it('applies native-chat option deltas atomically on the runtime host', () => { + let settings = { + ...store.getSettings(), + nativeChatSessionOptions: { + claude: { model: 'opus', valuesByModel: { opus: { effort: 'high' } } } + } + } + const updateSettings = vi.fn((updates: Partial) => { + settings = { ...settings, ...updates } + }) + const runtime = new OrcaRuntimeService({ + ...store, + getSettings: () => settings, + updateSettings + } as never) + + runtime.updateClientNativeChatSessionOptions({ + type: 'apply-picks', + agent: 'codex', + picks: [ + { modelId: 'gpt-fast', optionId: 'model', value: 'gpt-fast' }, + { modelId: 'gpt-fast', optionId: 'effort', value: 'low' } + ] + }) + runtime.updateClientNativeChatSessionOptions({ + type: 'apply-picks', + agent: 'claude', + picks: [{ modelId: 'sonnet', optionId: 'model', value: 'sonnet' }] + }) + + expect(settings.nativeChatSessionOptions).toEqual({ + claude: { + model: 'sonnet', + valuesByModel: { opus: { effort: 'high' } } + }, + codex: { + model: 'gpt-fast', + valuesByModel: { 'gpt-fast': { effort: 'low' } } + } + }) + expect(updateSettings).toHaveBeenCalledTimes(2) + }) + + it('compares retired models against the host record at mutation time', () => { + let settings = { + ...store.getSettings(), + nativeChatSessionOptions: { grok: { model: 'grok-5' } } + } + const updateSettings = vi.fn((updates: Partial) => { + settings = { ...settings, ...updates } + }) + const runtime = new OrcaRuntimeService({ + ...store, + getSettings: () => settings, + updateSettings + } as never) + + runtime.updateClientNativeChatSessionOptions({ + type: 'clear-model-if-missing', + agent: 'grok', + availableModelIds: ['grok-5'] + }) + expect(updateSettings).not.toHaveBeenCalled() + + runtime.updateClientNativeChatSessionOptions({ + type: 'clear-model-if-missing', + agent: 'grok', + availableModelIds: ['grok-4.5'] + }) + expect(settings.nativeChatSessionOptions).toEqual({ grok: {} }) + }) + it('rejects a concurrent add after the quick command limit is reached', () => { const terminalQuickCommands = Array.from({ length: MAX_QUICK_COMMANDS }, (_, index) => ({ id: `command-${index}`, diff --git a/src/main/runtime/orca-runtime.test.ts b/src/main/runtime/orca-runtime.test.ts index 4dd03c27e18..829c2ca321e 100644 --- a/src/main/runtime/orca-runtime.test.ts +++ b/src/main/runtime/orca-runtime.test.ts @@ -95,6 +95,7 @@ await import('./orca-runtime-tests/lineage-and-scan-cache-part-04.spec') await import('./orca-runtime-tests/lineage-and-scan-cache-part-05.spec') await import('./orca-runtime-tests/orchestration-attention-batching.spec') await import('./orca-runtime-tests/lineage-and-scan-cache-part-06.spec') +await import('./orca-runtime-tests/lineage-and-scan-cache-part-07.spec') await import('./orca-runtime-tests/worktree-setup-and-startup.spec') await import('./orca-runtime-tests/worktree-setup-and-startup-part-02.spec') await import('./orca-runtime-tests/worktree-setup-and-startup-part-03.spec') diff --git a/src/main/runtime/orchestration/adopted-structured-pointer-delivery.test.ts b/src/main/runtime/orchestration/adopted-structured-pointer-delivery.test.ts new file mode 100644 index 00000000000..6eb8398cc60 --- /dev/null +++ b/src/main/runtime/orchestration/adopted-structured-pointer-delivery.test.ts @@ -0,0 +1,129 @@ +import type { WriteSettlement } from '../../../shared/pty-write-settlement' +import { afterEach, describe, expect, it, vi } from 'vitest' +import { agentSessionPtyWriteGate } from '../agent-session-pty-write-gate' +import { OrcaRuntimeWithWriteOrchestrationPointerPty } from '../orca-runtime-write-orchestration-pointer-pty' +import { OrcaRuntimeWithGetPtyRecordForPaneKey } from '../orca-runtime-get-pty-record-for-pane-key' +import type { AgentSessionRecord } from '../../../shared/agent-session-record' + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' +const PTY_ID = 'pty_adopted' + +// Both methods are protected, and a subclass is the sanctioned way to reach them. The probes +// borrow the REAL implementations through the real prototype chain; a re-declared copy would pin +// nothing. +class PointerWriteProbe extends OrcaRuntimeWithWriteOrchestrationPointerPty { + probeWritePointer(ptyId: string, data: string): WriteSettlement | Promise { + return this.writeOrchestrationPointerPty(ptyId, data) + } +} + +class MailboxTargetProbe extends OrcaRuntimeWithGetPtyRecordForPaneKey { + probeResolveTarget(mailboxHandle: string): unknown { + return this.resolveStructuredMailboxTarget(mailboxHandle) + } +} + +/** A probe instance whose prototype chain is the real class, with only its state stubbed. */ +function probe( + prototype: TProbe, + state: TState +): TProbe & TState { + return Object.assign(Object.create(prototype), state) as TProbe & TState +} + +/** A pane bound to a session a settled NATIVE owner holds — the adopted-TUI state. */ +function bindNativeOwnedPane(overrides: Partial = {}): void { + agentSessionPtyWriteGate.attachRecordLookup( + (sessionId) => + ({ + sessionId, + location: { executionHostId: 'local', wslDistro: null }, + lease: { + sessionId, + runtimeKind: 'native', + claimStatus: 'live', + handoffStage: null, + unreconciled: false, + ownerProcess: { pid: 4242 }, + runtimeFence: 7, + ...overrides + } + }) as unknown as AgentSessionRecord + ) + agentSessionPtyWriteGate.bindPty(PTY_ID, SESSION_ID) +} + +afterEach(() => { + agentSessionPtyWriteGate.detachRecordLookup() + vi.restoreAllMocks() +}) + +describe('an orchestration pointer aimed at an adopted pane', () => { + it('reaches no provider and never reports a write failure to the renderer', () => { + bindNativeOwnedPane() + const write = vi.fn(() => true) + const writeWithSettlement = vi.fn(async () => true) + const stub = { + orchestrationPointerAdmissionByPtyId: new Map(), + ptyController: { write, writeWithSettlement } + } + // Zero bytes: the controller path would re-admit, refuse again, and fire + // `pty:writeUnavailable`, whose renderer handler runs transport RECOVERY on a healthy pane. + // A proven refusal, not a bare false: the settlement vocabulary keeps "declined before any + // byte moved" distinct from "we lost track", which is what a durable reservation reads. + expect( + probe(PointerWriteProbe.prototype, stub).probeWritePointer( + PTY_ID, + 'You have 1 orchestration message.' + ) + ).toEqual({ outcome: 'refused', reason: 'write_gate_denied' }) + expect(write).not.toHaveBeenCalled() + expect(writeWithSettlement).not.toHaveBeenCalled() + }) + + it('still writes through when nothing owns the pane', () => { + const write = vi.fn(() => true) + // The gate admits an unbound pane, so the bytes reach the provider and its own settlement is + // what the caller gets back. + const writeWithSettlement = vi.fn(() => ({ outcome: 'accepted' }) as const) + const stub = { + orchestrationPointerAdmissionByPtyId: new Map(), + ptyController: { write, writeWithSettlement } + } + expect( + probe(PointerWriteProbe.prototype, stub).probeWritePointer('pty_unbound', 'pointer') + ).toEqual({ outcome: 'accepted' }) + expect(writeWithSettlement).toHaveBeenCalledTimes(1) + }) +}) + +describe('the mailbox target for an adopted pane', () => { + function targetStub() { + return probe(MailboxTargetProbe.prototype, { + _orchestrationDb: { + getDispatchContextById: () => ({ assignee_handle: 'term_adopted' }) + }, + getLiveLeafForHandle: () => ({ leaf: { ptyId: PTY_ID } }) + }) + } + + it('routes the mailbox to the owning session so the nudge travels as a turn', () => { + bindNativeOwnedPane() + const target = targetStub().probeResolveTarget('dispatch:d1') as { + sessionId: string + dispatchId: string + refusal?: { ownerRuntimeKind: string } + } | null + expect(target).toMatchObject({ sessionId: SESSION_ID, dispatchId: 'd1' }) + expect(target?.refusal?.ownerRuntimeKind).toBe('native') + }) + + it('leaves a mid-handoff lease to the PTY lane', () => { + bindNativeOwnedPane({ handoffStage: 'preparing' }) + expect(targetStub().probeResolveTarget('dispatch:d1')).toBeNull() + }) + + it('leaves an unowned pane to the PTY lane', () => { + expect(targetStub().probeResolveTarget('dispatch:d1')).toBeNull() + }) +}) diff --git a/src/main/runtime/orchestration/db/attach-orchestration-db-methods.ts b/src/main/runtime/orchestration/db/attach-orchestration-db-methods.ts index 74c5fc00110..135dcc5ae3f 100644 --- a/src/main/runtime/orchestration/db/attach-orchestration-db-methods.ts +++ b/src/main/runtime/orchestration/db/attach-orchestration-db-methods.ts @@ -34,6 +34,7 @@ import { attachMailboxPointerEnterState } from './messages/mailbox-pointer-enter import { attachMessageInbox } from './messages/message-inbox' import { attachMessageInsert } from './messages/message-insert' import { attachRoleMailboxDelivery } from './messages/role-mailbox-delivery' +import { attachStructuredPointerOperationStore } from './messages/structured-pointer-operation-store' import { attachMutationReceiptStore } from './mutation-receipts/mutation-receipt-store' import { attachLifecycleTransition } from './lifecycle-transition' import { attachQuestionThreads } from './questions/question-threads' @@ -94,6 +95,7 @@ export function attachOrchestrationDbMethods(ctor: { prototype: object }): void attachRunDelivery(ctor) attachMessageInsert(ctor) attachRoleMailboxDelivery(ctor) + attachStructuredPointerOperationStore(ctor) attachMessageInbox(ctor) attachMailboxPointerEnterState(ctor) attachDirectMailboxRouting(ctor) diff --git a/src/main/runtime/orchestration/db/contract-constants.ts b/src/main/runtime/orchestration/db/contract-constants.ts index 4f9975529e4..287ce565d5b 100644 --- a/src/main/runtime/orchestration/db/contract-constants.ts +++ b/src/main/runtime/orchestration/db/contract-constants.ts @@ -6,5 +6,5 @@ export const LEGACY_RUN_ID = ORCHESTRATION_LEGACY_RUN_ID export const LEGACY_CONTRACT_VERSION = 0 export const CURRENT_CONTRACT_VERSION = ORCHESTRATION_CONTRACT_VERSION -// Schema versions: v2 'heartbeat'+last_heartbeat_at, v3 delivered_at, v4 task-creator terminal, v5 task_title/display_name, v6 pane identity, v7 lightweight Runs, v8 crash-safe Run deliveries, v9 durable question threads, v10 Dispatch capabilities, v11 durable mutation receipts, v12 composed worker state, v18 post-v6 version-skew repair, v19 adopted legacy Runs and compatibility receipts, v20 legacy question backfill, v21 legacy scheduler-loss provenance, v22 dispatch assignee lookup, v23 worker terminal resource ownership, v24 creator-incarnation authority, v25 active Dispatch handle lookup, v26 indexed mutation receipt capacity, v27 durable federation acknowledgments, v28 durable local mutation caller identity, v31 dispatch/resource identity links, v32 bounded worker-terminal recovery metadata, v33 durable mailbox pointer Enter state, v34 role-addressed mailbox deliveries, v35 mailbox delivery default and index-predicate repair, v36 dispatch mailbox consumer generation, v37 recorded dispatch creator identity. -export const SCHEMA_VERSION = 38 +// Schema versions: v2 'heartbeat'+last_heartbeat_at, v3 delivered_at, v4 task-creator terminal, v5 task_title/display_name, v6 pane identity, v7 lightweight Runs, v8 crash-safe Run deliveries, v9 durable question threads, v10 Dispatch capabilities, v11 durable mutation receipts, v12 composed worker state, v18 post-v6 version-skew repair, v19 adopted legacy Runs and compatibility receipts, v20 legacy question backfill, v21 legacy scheduler-loss provenance, v22 dispatch assignee lookup, v23 worker terminal resource ownership, v24 creator-incarnation authority, v25 active Dispatch handle lookup, v26 indexed mutation receipt capacity, v27 durable federation acknowledgments, v28 durable local mutation caller identity, v31 dispatch/resource identity links, v32 bounded worker-terminal recovery metadata, v33 durable mailbox pointer Enter state, v34 role-addressed mailbox deliveries, v35 mailbox delivery default and index-predicate repair, v36 dispatch mailbox consumer generation, v37 recorded dispatch creator identity, v39 structured session journal archives. +export const SCHEMA_VERSION = 39 diff --git a/src/main/runtime/orchestration/db/dispatch-row-writer-boundary.test.ts b/src/main/runtime/orchestration/db/dispatch-row-writer-boundary.test.ts index 9b7581d6acd..d755b2bd688 100644 --- a/src/main/runtime/orchestration/db/dispatch-row-writer-boundary.test.ts +++ b/src/main/runtime/orchestration/db/dispatch-row-writer-boundary.test.ts @@ -100,6 +100,7 @@ describe('live-worker row insert boundary', () => { const exempt = [ 'src/main/runtime/orchestration/db/schema/create-graph-tables-sql.ts', 'src/main/runtime/orchestration/db/schema/migrate-v13-v30.ts', + 'src/main/runtime/orchestration/db/schema/migrate-v39.ts', 'src/main/runtime/orchestration/db/reset/orchestration-reset.ts' ] for (const rel of exempt) { diff --git a/src/main/runtime/orchestration/db/messages/structured-pointer-operation-store.ts b/src/main/runtime/orchestration/db/messages/structured-pointer-operation-store.ts new file mode 100644 index 00000000000..54b51b18e6e --- /dev/null +++ b/src/main/runtime/orchestration/db/messages/structured-pointer-operation-store.ts @@ -0,0 +1,64 @@ +import type { OrchestrationDb } from '../orchestration-db' + +/** The live agent-session operation id backing one structured worker mailbox's pointer send. */ +export type StructuredPointerOperationRow = { + mailbox_handle: string + session_id: string + operation_id: string + batch_fingerprint: string + minted_at_ms: number +} + +export function getStructuredPointerOperation( + this: OrchestrationDb, + mailboxHandle: string +): StructuredPointerOperationRow | undefined { + return this.db + .prepare('SELECT * FROM structured_pointer_operations WHERE mailbox_handle = ?') + .get(mailboxHandle) as StructuredPointerOperationRow | undefined +} + +export function putStructuredPointerOperation( + this: OrchestrationDb, + row: StructuredPointerOperationRow +): void { + this.db + .prepare( + `INSERT INTO structured_pointer_operations + (mailbox_handle, session_id, operation_id, batch_fingerprint, minted_at_ms) + VALUES (?, ?, ?, ?, ?) + ON CONFLICT(mailbox_handle) DO UPDATE SET + session_id = excluded.session_id, operation_id = excluded.operation_id, + batch_fingerprint = excluded.batch_fingerprint, minted_at_ms = excluded.minted_at_ms` + ) + .run( + row.mailbox_handle, + row.session_id, + row.operation_id, + row.batch_fingerprint, + row.minted_at_ms + ) +} + +export function deleteStructuredPointerOperation( + this: OrchestrationDb, + mailboxHandle: string +): void { + this.db + .prepare('DELETE FROM structured_pointer_operations WHERE mailbox_handle = ?') + .run(mailboxHandle) +} + +export type StructuredPointerOperationStoreMethods = { + getStructuredPointerOperation: typeof getStructuredPointerOperation + putStructuredPointerOperation: typeof putStructuredPointerOperation + deleteStructuredPointerOperation: typeof deleteStructuredPointerOperation +} + +export function attachStructuredPointerOperationStore(ctor: { prototype: object }): void { + Object.assign(ctor.prototype, { + getStructuredPointerOperation, + putStructuredPointerOperation, + deleteStructuredPointerOperation + }) +} diff --git a/src/main/runtime/orchestration/db/orchestration-db-methods.ts b/src/main/runtime/orchestration/db/orchestration-db-methods.ts index b63a1a0f6a5..55dcddaf44c 100644 --- a/src/main/runtime/orchestration/db/orchestration-db-methods.ts +++ b/src/main/runtime/orchestration/db/orchestration-db-methods.ts @@ -63,6 +63,7 @@ import type { WorkerTerminalRecoveryMethods } from './worker-dispatch/worker-ter import type { WorkerTerminalArchiveMethods } from './worker-terminal/worker-terminal-archive' import type { WorkerTerminalListingMethods } from './worker-terminal/worker-terminal-listing' import type { WorkerTerminalReleaseMethods } from './worker-terminal/worker-terminal-release' +import type { StructuredPointerOperationStoreMethods } from './messages/structured-pointer-operation-store' import type { WorkerTerminalResourceStoreMethods } from './worker-terminal/worker-terminal-resource-store' import type { WorkerTerminalTransferMethods } from './worker-terminal/worker-terminal-transfer' @@ -119,6 +120,7 @@ export type OrchestrationDbMethods = AttemptObservationStoreMethods & FederationRelayImportMethods & RemoteQuestionStoreMethods & FederationRelayItemMethods & + StructuredPointerOperationStoreMethods & WorkerTerminalResourceStoreMethods & WorkerTerminalTransferMethods & WorkerTerminalReleaseMethods & diff --git a/src/main/runtime/orchestration/db/reset/orchestration-reset.ts b/src/main/runtime/orchestration/db/reset/orchestration-reset.ts index f5532a2a1ae..dad2d3003f7 100644 --- a/src/main/runtime/orchestration/db/reset/orchestration-reset.ts +++ b/src/main/runtime/orchestration/db/reset/orchestration-reset.ts @@ -36,6 +36,7 @@ export function resetAll(this: OrchestrationDb): void { DELETE FROM worker_terminal_archives; DELETE FROM worker_terminal_resources; DELETE FROM attempt_observation_facts; + DELETE FROM structured_pointer_operations; DELETE FROM worker_dispatches; DELETE FROM dispatch_contexts; DELETE FROM tasks; @@ -68,6 +69,7 @@ export function resetTasks(this: OrchestrationDb): void { DELETE FROM worker_terminal_archives; DELETE FROM worker_terminal_resources; DELETE FROM attempt_observation_facts; + DELETE FROM structured_pointer_operations; DELETE FROM worker_dispatches; DELETE FROM dispatch_contexts; DELETE FROM tasks; @@ -77,11 +79,14 @@ export function resetTasks(this: OrchestrationDb): void { export function resetMessages(this: OrchestrationDb): void { // Why: federation_relay_items is deliberately kept — relay rows carry contiguous cross-server cursors, not just inbox history. + // Why structured_pointer_operations goes: the row is one nudge's idempotency key over a batch of + // messages this deletes, so keeping it would suppress the re-mint for a batch that no longer exists. this.runResetTransaction(` DELETE FROM legacy_mail_receipts; DELETE FROM question_threads; DELETE FROM deliveries; DELETE FROM messages; + DELETE FROM structured_pointer_operations; `) } diff --git a/src/main/runtime/orchestration/db/schema/create-core-tables-sql.ts b/src/main/runtime/orchestration/db/schema/create-core-tables-sql.ts index 92d56063763..f02e7e8c15a 100644 --- a/src/main/runtime/orchestration/db/schema/create-core-tables-sql.ts +++ b/src/main/runtime/orchestration/db/schema/create-core-tables-sql.ts @@ -192,10 +192,21 @@ CREATE INDEX IF NOT EXISTS idx_worker_terminal_resources_identity CREATE INDEX IF NOT EXISTS idx_worker_terminal_resources_release ON worker_terminal_resources(release_state); +-- One live agent-session operation id per structured worker mailbox. Persisted because the id is +-- the send's idempotency key: re-minting it after a restart would re-deliver an already-queued +-- pointer as a second turn. +CREATE TABLE IF NOT EXISTS structured_pointer_operations ( + mailbox_handle TEXT PRIMARY KEY, + session_id TEXT NOT NULL, + operation_id TEXT NOT NULL, + batch_fingerprint TEXT NOT NULL, + minted_at_ms INTEGER NOT NULL +); + CREATE TABLE IF NOT EXISTS worker_terminal_archives ( dispatch_id TEXT PRIMARY KEY, resource_id TEXT NOT NULL, - kind TEXT NOT NULL CHECK(kind IN ('transcript_pin', 'terminal_tail')), + kind TEXT NOT NULL CHECK(kind IN ('transcript_pin', 'terminal_tail', 'structured_journal')), content TEXT NOT NULL, created_at TEXT NOT NULL DEFAULT (datetime('now')) ); diff --git a/src/main/runtime/orchestration/db/schema/migrate-v39.ts b/src/main/runtime/orchestration/db/schema/migrate-v39.ts new file mode 100644 index 00000000000..66df30398ff --- /dev/null +++ b/src/main/runtime/orchestration/db/schema/migrate-v39.ts @@ -0,0 +1,31 @@ +import type { OrchestrationDb } from '../orchestration-db' + +/** + * Admits a structured session's journal into the worker archive. + * + * A CHECK constraint cannot be widened in place, so the table is rebuilt and copied forward. This + * is the one part of the structured-session schema that a fresh `createTables` cannot supply to an + * existing database: `IF NOT EXISTS` leaves an already-created table's narrower CHECK untouched. + * + * `structured_pointer_operations` is deliberately not created here — `createTables` runs + * unconditionally on every open, ahead of migration, and already declares it. + */ +export function migrateV39(this: OrchestrationDb, current: number): void { + if (current >= 39) { + return + } + this.db.exec(` + CREATE TABLE IF NOT EXISTS worker_terminal_archives_v39 ( + dispatch_id TEXT PRIMARY KEY, + resource_id TEXT NOT NULL, + kind TEXT NOT NULL CHECK(kind IN ('transcript_pin', 'terminal_tail', 'structured_journal')), + content TEXT NOT NULL, + created_at TEXT NOT NULL DEFAULT (datetime('now')) + ); + INSERT OR REPLACE INTO worker_terminal_archives_v39 + (dispatch_id, resource_id, kind, content, created_at) + SELECT dispatch_id, resource_id, kind, content, created_at FROM worker_terminal_archives; + DROP TABLE worker_terminal_archives; + ALTER TABLE worker_terminal_archives_v39 RENAME TO worker_terminal_archives; + `) +} diff --git a/src/main/runtime/orchestration/db/schema/migrate.ts b/src/main/runtime/orchestration/db/schema/migrate.ts index fade2bf15e4..9cdc9544da1 100644 --- a/src/main/runtime/orchestration/db/schema/migrate.ts +++ b/src/main/runtime/orchestration/db/schema/migrate.ts @@ -9,6 +9,7 @@ import { migrateV35 } from './migrate-v35' import { migrateV36 } from './migrate-v36' import { migrateV37 } from './migrate-v37' import { migrateV38 } from './migrate-v38' +import { migrateV39 } from './migrate-v39' // Why: CREATE TABLE IF NOT EXISTS won't alter existing DBs; migrate in a txn that bumps user_version only on success (atomic all-or-nothing). export function migrate(this: OrchestrationDb): void { @@ -28,6 +29,7 @@ export function migrate(this: OrchestrationDb): void { migrateV36.call(this, current) migrateV37.call(this, current) migrateV38.call(this, current) + migrateV39.call(this, current) this.db.pragma(`user_version = ${SCHEMA_VERSION}`) this.db.exec('COMMIT') } catch (err) { diff --git a/src/main/runtime/orchestration/db/schema/structured-pointer-schema-migration.test.ts b/src/main/runtime/orchestration/db/schema/structured-pointer-schema-migration.test.ts new file mode 100644 index 00000000000..2e549e2d8e7 --- /dev/null +++ b/src/main/runtime/orchestration/db/schema/structured-pointer-schema-migration.test.ts @@ -0,0 +1,89 @@ +import { mkdtempSync, rmSync } from 'node:fs' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import Database from '../../../../sqlite/sync-database' +import { afterEach, describe, expect, it } from 'vitest' +import { OrchestrationDb } from '../orchestration-db' +import { SCHEMA_VERSION } from '../contract-constants' + +/** + * A pre-v39 database: the narrow archive CHECK, stamped at 38 so ONLY v39 runs. + * + * Seeding lower would still pass while exercising the whole v13->v39 chain instead, which + * would mask a broken rebuild. `createTables` supplies every other table before migration, so + * a 38 stamp survives the completeness check and the migration start resolves to 38. + */ +function seedLegacyDatabase(path: string): void { + const db = new Database(path) + db.exec(` + CREATE TABLE worker_terminal_archives ( + dispatch_id TEXT PRIMARY KEY, + resource_id TEXT NOT NULL, + kind TEXT NOT NULL CHECK(kind IN ('transcript_pin', 'terminal_tail')), + content TEXT NOT NULL, + created_at TEXT NOT NULL DEFAULT (datetime('now')) + ); + INSERT INTO worker_terminal_archives (dispatch_id, resource_id, kind, content, created_at) + VALUES ('d_old', 'res_old', 'terminal_tail', '{"lines":["kept"]}', '2026-01-01 00:00:00'); + `) + db.pragma('user_version = 38') + db.close() +} + +describe('structured pointer schema migration', () => { + // Why a real temp dir and a teardown: `$TMPDIR` is unset on Windows CI, so the interpolated + // `/tmp/...` opened as `SQLITE_CANTOPEN`, and nothing removed the file on the platforms where it + // did open. + const tempRoots: string[] = [] + + afterEach(() => { + while (tempRoots.length > 0) { + rmSync(tempRoots.pop() as string, { recursive: true, force: true }) + } + }) + + it('admits the structured archive kind and keeps existing rows', () => { + const root = mkdtempSync(join(tmpdir(), 'orca-structured-migration-')) + tempRoots.push(root) + const path = join(root, 'orchestration.db') + seedLegacyDatabase(path) + const db = new OrchestrationDb(path) + try { + expect(db.db.pragma('user_version', { simple: true })).toBe(SCHEMA_VERSION) + const kept = db.db + .prepare('SELECT content FROM worker_terminal_archives WHERE dispatch_id = ?') + .get('d_old') as { content: string } + expect(kept.content).toContain('kept') + db.storeWorkerTerminalArchive({ + dispatchId: 'd_new', + resourceId: 'res_new', + kind: 'structured_journal', + content: '{"version":1}' + }) + expect(db.getWorkerTerminalArchive('d_new')?.kind).toBe('structured_journal') + } finally { + db.close() + } + }) + + it('creates the structured pointer operation store', () => { + const db = new OrchestrationDb(':memory:') + try { + expect(db.getStructuredPointerOperation('dispatch:d1')).toBeUndefined() + db.putStructuredPointerOperation({ + mailbox_handle: 'dispatch:d1', + session_id: 's1', + operation_id: '1757030400000-0123456789abcdef0123456789abcdef', + batch_fingerprint: 'fp', + minted_at_ms: 1_757_030_400_000 + }) + expect(db.getStructuredPointerOperation('dispatch:d1')?.operation_id).toBe( + '1757030400000-0123456789abcdef0123456789abcdef' + ) + db.deleteStructuredPointerOperation('dispatch:d1') + expect(db.getStructuredPointerOperation('dispatch:d1')).toBeUndefined() + } finally { + db.close() + } + }) +}) diff --git a/src/main/runtime/orchestration/db/worker-terminal/worker-terminal-archive.ts b/src/main/runtime/orchestration/db/worker-terminal/worker-terminal-archive.ts index 988f3b46f11..4bb4fcff823 100644 --- a/src/main/runtime/orchestration/db/worker-terminal/worker-terminal-archive.ts +++ b/src/main/runtime/orchestration/db/worker-terminal/worker-terminal-archive.ts @@ -1,4 +1,5 @@ import type { + WorkerTerminalArchiveKind, WorkerTerminalResourceRow, WorkerTerminalArchiveRow, WorkerTerminalArchiveStatus, @@ -12,7 +13,7 @@ export function storeWorkerTerminalArchive( params: { dispatchId: string resourceId: string - kind: 'transcript_pin' | 'terminal_tail' + kind: WorkerTerminalArchiveKind content: string } ): void { @@ -31,7 +32,7 @@ export function commitWorkerTerminalArchiveForRelease( params: { dispatchId: string resourceId: string - kind?: 'transcript_pin' | 'terminal_tail' + kind?: WorkerTerminalArchiveKind content?: string archiveSource: 'transcript' | 'terminal' archiveStatus: Extract diff --git a/src/main/runtime/orchestration/db/worker-terminal/worker-terminal-resource-store.ts b/src/main/runtime/orchestration/db/worker-terminal/worker-terminal-resource-store.ts index dc211f01a08..1d7232107b1 100644 --- a/src/main/runtime/orchestration/db/worker-terminal/worker-terminal-resource-store.ts +++ b/src/main/runtime/orchestration/db/worker-terminal/worker-terminal-resource-store.ts @@ -110,6 +110,18 @@ export function getWorkerTerminalResourceByOwner( .get(dispatchId) as WorkerTerminalResourceRow | undefined } +export function getWorkerTerminalResourceByHandle( + this: OrchestrationDb, + terminalHandle: string +): WorkerTerminalResourceRow | undefined { + return this.db + .prepare( + `SELECT * FROM worker_terminal_resources + WHERE terminal_handle = ? ORDER BY updated_at DESC LIMIT 1` + ) + .get(terminalHandle) as WorkerTerminalResourceRow | undefined +} + export function getWorkerTerminalResourceFormerlyOwnedBy( this: OrchestrationDb, dispatchId: string @@ -193,6 +205,7 @@ export type WorkerTerminalResourceStoreMethods = { backfillWorkerTerminalResources: typeof backfillWorkerTerminalResources createWorkerTerminalResourceStatement: typeof createWorkerTerminalResourceStatement getWorkerTerminalResource: typeof getWorkerTerminalResource + getWorkerTerminalResourceByHandle: typeof getWorkerTerminalResourceByHandle getWorkerTerminalResourceByOwner: typeof getWorkerTerminalResourceByOwner getWorkerTerminalResourceFormerlyOwnedBy: typeof getWorkerTerminalResourceFormerlyOwnedBy recordWorkerTerminalRecoveryAttempt: typeof recordWorkerTerminalRecoveryAttempt @@ -204,6 +217,7 @@ export function attachWorkerTerminalResourceStore(ctor: { prototype: object }): backfillWorkerTerminalResources, createWorkerTerminalResourceStatement, getWorkerTerminalResource, + getWorkerTerminalResourceByHandle, getWorkerTerminalResourceByOwner, getWorkerTerminalResourceFormerlyOwnedBy, recordWorkerTerminalRecoveryAttempt, diff --git a/src/main/runtime/orchestration/groups.ts b/src/main/runtime/orchestration/groups.ts index 64c0900bfb9..a4e548f06d0 100644 --- a/src/main/runtime/orchestration/groups.ts +++ b/src/main/runtime/orchestration/groups.ts @@ -1,5 +1,5 @@ -import type { RuntimeTerminalSummary } from '../../../shared/runtime-types' import type { TuiAgent } from '../../../shared/tui-agent' +import type { OrchestrationAddressableAgent } from './structured-worker-group-addressing' // Why: group addresses enable broadcast messaging to logical groups of agents. // Resolution is done at send-time: one message record per recipient, same thread_id, @@ -51,14 +51,17 @@ const GROUP_AGENT_IDS: Record = { * delivering is visible and recoverable — the sender sees no recipients; delivering to the wrong * agent is neither. */ -function terminalIsAgent(terminal: RuntimeTerminalSummary, agentName: AgentNameGroup): boolean { +function terminalIsAgent( + terminal: OrchestrationAddressableAgent, + agentName: AgentNameGroup +): boolean { return terminal.agentIdentity === GROUP_AGENT_IDS[agentName] } export function resolveGroupAddress( to: string, senderHandle: string, - terminals: RuntimeTerminalSummary[], + terminals: readonly OrchestrationAddressableAgent[], getAgentStatus: (handle: string) => string | null ): string[] { if (!isGroupAddress(to)) { diff --git a/src/main/runtime/orchestration/mailbox-delivery-target.ts b/src/main/runtime/orchestration/mailbox-delivery-target.ts index 3e47b020ff0..5d412bb12e8 100644 --- a/src/main/runtime/orchestration/mailbox-delivery-target.ts +++ b/src/main/runtime/orchestration/mailbox-delivery-target.ts @@ -5,6 +5,8 @@ type OrchestrationMailboxDeliveryTargetDependencies = { getDb: () => OrchestrationDb | null getTerminalHandleForPaneKey: (paneKey: string) => string | null hasTerminalHandle: (handle: string) => boolean + /** A structured worker has no PTY handle; its own lane delivers, so this must not claim it. */ + isStructuredWorkerHandle: (handle: string) => boolean canProbePtyLiveness: () => boolean controllerKnowsPtyIsLive: (ptyId: string) => boolean isLeafPtyProvenAbsent: (ptyId: string) => Promise @@ -19,6 +21,9 @@ export class OrchestrationMailboxDeliveryTarget { if (this.deps.hasTerminalHandle(handle)) { return handle } + if (this.deps.isStructuredWorkerHandle(handle)) { + return null + } const db = this.deps.getDb() const runId = handle.startsWith('run:') ? handle.slice('run:'.length) : '' const dispatchId = handle.startsWith('dispatch:') ? handle.slice('dispatch:'.length) : '' @@ -31,7 +36,23 @@ export class OrchestrationMailboxDeliveryTarget { : ((paneKey ? this.deps.getTerminalHandleForPaneKey(paneKey) : null) ?? dispatch?.assignee_handle ?? remote?.terminal_handle) - return ownerHandle && this.deps.hasTerminalHandle(ownerHandle) ? ownerHandle : null + if (!ownerHandle) { + return null + } + if (this.deps.isStructuredWorkerHandle(ownerHandle)) { + // The structured lane owns this mailbox; nothing here can type into it. + return null + } + if (!this.deps.hasTerminalHandle(ownerHandle)) { + // Why logged rather than silent: an unroutable owner is the shape of a lost mailbox, and a + // silent null is indistinguishable from "no mail". + console.warn('[orchestration] mailbox owner resolved to an unknown terminal', { + mailboxHandle: handle, + ownerHandle + }) + return null + } + return ownerHandle } deferForAbsenceProbe( diff --git a/src/main/runtime/orchestration/mailbox-notification-coordinator.ts b/src/main/runtime/orchestration/mailbox-notification-coordinator.ts index 4856a08843b..07cf3b5b7dc 100644 --- a/src/main/runtime/orchestration/mailbox-notification-coordinator.ts +++ b/src/main/runtime/orchestration/mailbox-notification-coordinator.ts @@ -8,10 +8,13 @@ import type { OrchestrationMailboxPointerDelivery, OrchestrationMessageWaiter } from './mailbox-pointer-delivery' +import type { OrchestrationStructuredMailboxPointerDelivery } from './structured-mailbox-pointer-delivery' type NotificationCoordinatorDependencies = { mailboxOwner: OrchestrationMailboxOwner pointerDelivery: OrchestrationMailboxPointerDelivery + /** Sibling lane for workers that ARE a structured session; it has no PTY to type into. */ + structuredPointerDelivery?: OrchestrationStructuredMailboxPointerDelivery getDb: () => OrchestrationDb | null getLiveLeafForHandle: (handle: string) => OrchestrationMailboxLeaf getPaneKeyForHandle: (handle: string) => string | undefined @@ -28,6 +31,9 @@ export class OrchestrationMailboxNotificationCoordinator< constructor(private readonly deps: NotificationCoordinatorDependencies) {} deliverForHandle(handle: string, reservedTypes?: ReadonlySet): void { + if (this.deps.structuredPointerDelivery?.deliverForHandle(handle, reservedTypes)) { + return + } this.deps.pointerDelivery.deliverForHandle(handle, reservedTypes) } diff --git a/src/main/runtime/orchestration/mailbox-pointer-delivery.ts b/src/main/runtime/orchestration/mailbox-pointer-delivery.ts index 5ddc9f443c8..11fbc08229d 100644 --- a/src/main/runtime/orchestration/mailbox-pointer-delivery.ts +++ b/src/main/runtime/orchestration/mailbox-pointer-delivery.ts @@ -1,7 +1,7 @@ -import { ORCHESTRATION_DELIVERY_BATCH_LIMIT } from './db' import type { PointerDeliveryDependencies } from './mailbox-pointer-delivery-contract' import { hasUnfilteredOrchestrationWaiter, + selectOrchestrationPointerBatch, type OrchestrationMessageWaiter } from './mailbox-pointer-eligibility' import type { OrchestrationMailboxLeaf } from './mailbox-owner' @@ -104,16 +104,11 @@ export class OrchestrationMailboxPointerDelivery { + return new Set(filters.map((typeFilter) => ({ typeFilter }))) +} + +describe('selectOrchestrationPointerBatch', () => { + it('excludes a waiter-claimed type', () => { + const db = seeded() + try { + const batch = selectOrchestrationPointerBatch({ + db, + mailboxHandle: MAILBOX, + waiters: waiters(['question']), + reservedTypes: undefined + }) + expect(batch.map((m) => m.type)).toEqual(['status', 'status']) + } finally { + db.close() + } + }) + + it('excludes a reserved type', () => { + const db = seeded() + try { + const batch = selectOrchestrationPointerBatch({ + db, + mailboxHandle: MAILBOX, + waiters: undefined, + reservedTypes: new Set(['status']) + }) + expect(batch.map((m) => m.type)).toEqual(['question']) + } finally { + db.close() + } + }) + + it('unions reserved types with every waiter filter', () => { + const db = seeded() + try { + expect( + selectOrchestrationPointerBatch({ + db, + mailboxHandle: MAILBOX, + waiters: waiters(['question']), + reservedTypes: new Set(['status']) + }) + ).toEqual([]) + } finally { + db.close() + } + }) + + // An unfiltered waiter owns the mailbox: a caller blocked in `check --wait` preempts delivery. + it('yields nothing when any waiter is unfiltered', () => { + const db = seeded() + try { + expect( + selectOrchestrationPointerBatch({ + db, + mailboxHandle: MAILBOX, + waiters: waiters(['question'], undefined), + reservedTypes: undefined + }) + ).toEqual([]) + } finally { + db.close() + } + }) +}) diff --git a/src/main/runtime/orchestration/mailbox-pointer-eligibility.ts b/src/main/runtime/orchestration/mailbox-pointer-eligibility.ts index 9d4e0b87f48..b095df34099 100644 --- a/src/main/runtime/orchestration/mailbox-pointer-eligibility.ts +++ b/src/main/runtime/orchestration/mailbox-pointer-eligibility.ts @@ -1,4 +1,4 @@ -import type { OrchestrationDb } from './db' +import { ORCHESTRATION_DELIVERY_BATCH_LIMIT, type MessageRow, type OrchestrationDb } from './db' export type OrchestrationMessageWaiter = { typeFilter: string[] | undefined } @@ -25,6 +25,40 @@ export function hasUnfilteredOrchestrationWaiter( return false } +/** + * The rows a pointer may carry right now. + * + * Both delivery lanes — bytes into a PTY, a turn into a structured session — select their batch + * identically and differ only in what they do with it, so the selection lives here rather than + * being kept in step in two copies. An unfiltered waiter owns the whole mailbox and yields an + * empty batch: a caller blocked in `check --wait` preempts pointer delivery entirely. + * + * Type exclusion is exact in SQL and no post-filter is owed. The unfiltered case returns above, so + * every remaining waiter contributes a concrete type list, and `messages.type` is TEXT with no + * NOCASE collation — `NOT IN` is the same byte-exact test JS would repeat. Selection is synchronous + * throughout, so no waiter can register partway through it either. + */ +export function selectOrchestrationPointerBatch(input: { + db: OrchestrationDb + mailboxHandle: string + waiters: ReadonlySet | undefined + reservedTypes: ReadonlySet | undefined +}): MessageRow[] { + if (hasUnfilteredOrchestrationWaiter(input.waiters)) { + return [] + } + const excludedTypes = new Set(input.reservedTypes) + for (const waiter of input.waiters ?? []) { + for (const type of waiter.typeFilter ?? []) { + excludedTypes.add(type) + } + } + return input.db.getUndeliveredUnreadMessages(input.mailboxHandle, undefined, { + excludeTypes: [...excludedTypes], + limit: ORCHESTRATION_DELIVERY_BATCH_LIMIT + }) +} + export function shouldReleaseOrchestrationPointer( db: OrchestrationDb | null, mailboxHandle: string, diff --git a/src/main/runtime/orchestration/orchestration-legacy-worker-terminal-recovery.ts b/src/main/runtime/orchestration/orchestration-legacy-worker-terminal-recovery.ts index d675acd1bf1..8b1426cb07e 100644 --- a/src/main/runtime/orchestration/orchestration-legacy-worker-terminal-recovery.ts +++ b/src/main/runtime/orchestration/orchestration-legacy-worker-terminal-recovery.ts @@ -1,3 +1,4 @@ +import { sessionIdFromStructuredWorkerIncarnation } from '../structured-worker-identity' import { isPtyIncarnationId, type PtyIncarnationId } from '../../../shared/pty-incarnation' import { parsePaneKey } from '../../../shared/stable-pane-id' import type { LegacyWorkerTerminalRecoveryRow } from './types' @@ -41,6 +42,11 @@ function parseProcessIncarnation( } const ptyId = value.slice(0, separator) const incarnationId = value.slice(separator + 1) + // A structured worker's incarnation names a session lineage, not a PTY; adopting it as one + // would hand a live chat session's dispatch to the PTY recovery path. + if (sessionIdFromStructuredWorkerIncarnation(value)) { + return null + } return ptyId && isPtyIncarnationId(incarnationId) ? { ptyId, incarnationId } : null } diff --git a/src/main/runtime/orchestration/orchestration-reset-db.test.ts b/src/main/runtime/orchestration/orchestration-reset-db.test.ts index 4b3e033a019..6d82f14e823 100644 --- a/src/main/runtime/orchestration/orchestration-reset-db.test.ts +++ b/src/main/runtime/orchestration/orchestration-reset-db.test.ts @@ -53,6 +53,13 @@ describe('OrchestrationDb reset scopes', () => { messageId: 'question_1', remoteQuestion: true }) + db.putStructuredPointerOperation({ + mailbox_handle: `dispatch:${started.dispatch.id}`, + session_id: 'session_1', + operation_id: '1700000000000-00112233445566778899aabbccddeeff', + batch_fingerprint: 'fingerprint_1', + minted_at_ms: 1_700_000_000_000 + }) return { run, task, started, message, localQuestion } } @@ -78,6 +85,9 @@ describe('OrchestrationDb reset scopes', () => { afterSequence: 0 }) ).toEqual([]) + expect( + db!.getStructuredPointerOperation(`dispatch:${state.started.dispatch.id}`) + ).toBeUndefined() }) it('resetTasks preserves Runs and messages while clearing every worker attachment', () => { @@ -104,6 +114,10 @@ describe('OrchestrationDb reset scopes', () => { body: 'Yes' }) ).toThrowError(expect.objectContaining({ code: 'dispatch_inactive' })) + // The dispatch its send was keyed to is gone; the pointer operation must not outlive it. + expect( + db!.getStructuredPointerOperation(`dispatch:${state.started.dispatch.id}`) + ).toBeUndefined() }) it('resetMessages preserves active relay cursors while clearing the Run inbox', () => { @@ -121,5 +135,9 @@ describe('OrchestrationDb reset scopes', () => { afterSequence: 0 }) ).toHaveLength(1) + // The row is one nudge's idempotency key over messages this scope deletes. + expect( + db!.getStructuredPointerOperation(`dispatch:${state.started.dispatch.id}`) + ).toBeUndefined() }) }) diff --git a/src/main/runtime/orchestration/structured-mailbox-pointer-delivery.test.ts b/src/main/runtime/orchestration/structured-mailbox-pointer-delivery.test.ts new file mode 100644 index 00000000000..fd495ff1dbb --- /dev/null +++ b/src/main/runtime/orchestration/structured-mailbox-pointer-delivery.test.ts @@ -0,0 +1,430 @@ +import { describe, expect, it, vi } from 'vitest' +import type { AgentJournalRenderItem } from '../../../shared/agent-session-journal-types' +import type { AgentSessionPtyWriteRefusal } from '../../../shared/agent-session-pty-write-admission' +import { + OrchestrationStructuredMailboxPointerDelivery, + type StructuredMailboxPointerHost +} from './structured-mailbox-pointer-delivery' +import { structuredSessionGateFacts } from './structured-session-pointer-delivery' +import type { StructuredWorkerIdentity } from '../structured-worker-identity' + +const IDENTITY: StructuredWorkerIdentity = { + handle: 'structworker_1', + sessionId: 'session-1', + agent: 'claude', + paneKey: 'structured-agent-session-session-1:11111111-1111-4111-a111-111111111111', + processIncarnation: 'structured:session-1', + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } +} + +function idleJournal(): AgentJournalRenderItem[] { + return [ + { + itemId: 'i1', + observedAt: 1, + body: { kind: 'status', text: 'done', turnLifecycle: { state: 'completed', turnId: 't1' } } + } as unknown as AgentJournalRenderItem + ] +} + +function runningJournal(): AgentJournalRenderItem[] { + return [ + { + itemId: 'i1', + observedAt: 1, + body: { kind: 'status', text: 'working', turnLifecycle: { state: 'running', turnId: 't1' } } + } as unknown as AgentJournalRenderItem + ] +} + +/** What a worker's journal looks like once it has finished a substantial turn: history, and no + * turnLifecycle row anywhere, because settlement tombstones it. */ +function settledLongJournal(): AgentJournalRenderItem[] { + return Array.from( + { length: 120 }, + (_unused, index) => + ({ + itemId: `tool-${index}`, + observedAt: index, + body: { kind: 'tool-call', name: 'Bash', input: {}, state: 'completed' } + }) as unknown as AgentJournalRenderItem + ) +} + +/** A prompt raised at the very start of a long turn, far outside any bounded tail window. */ +function staleAttentionJournal(): AgentJournalRenderItem[] { + return [...attentionJournal(), ...settledLongJournal()] +} + +function attentionJournal(): AgentJournalRenderItem[] { + return [ + { + itemId: 'i1', + observedAt: 1, + body: { + kind: 'question', + question: 'which?', + options: [], + resolution: { state: 'pending' } + } + } as unknown as AgentJournalRenderItem + ] +} + +function harness(options: { + journal: AgentJournalRenderItem[] | null + dispatchState?: 'accepted' | 'rejected' | 'unknown' + refusal?: AgentSessionPtyWriteRefusal + /** The coordinator of this worker's Run is mid-batch: it checked and has not acked yet. */ + outstandingRunDelivery?: boolean + outstandingOwnDelivery?: boolean + /** The mailbox this worker owns; its own handle for direct peer mail outside a dispatch. */ + mailbox?: string + dispatchId?: string | null +}) { + const mailbox = options.mailbox ?? 'dispatch:d1' + const dispatchId = options.dispatchId === undefined ? 'd1' : options.dispatchId + let journal = options.journal + const markAsDelivered = vi.fn() + const send: StructuredMailboxPointerHost['send'] = vi.fn(async () => ({ + kind: 'sent' as const, + state: options.dispatchState ?? ('accepted' as const) + })) + const sendMock = vi.mocked(send) + const stored = new Map() + const db = { + getDispatchContextById: () => ({ run_id: 'run_1' }), + hasOutstandingMailboxDelivery: (handle: string) => + ((options.outstandingRunDelivery ?? false) && handle.startsWith('run:')) || + ((options.outstandingOwnDelivery ?? false) && !handle.startsWith('run:')), + getUndeliveredUnreadMessages: () => [{ id: 'm1', type: 'status', sequence: 3 }], + markAsDelivered, + getStructuredPointerOperation: (key: string) => stored.get(key), + putStructuredPointerOperation: (row: { mailbox_handle: string }) => + stored.set(row.mailbox_handle, row), + deleteStructuredPointerOperation: (key: string) => stored.delete(key) + } + const delivery = new OrchestrationStructuredMailboxPointerDelivery({ + getDb: () => db as never, + getMessageWaiters: () => undefined, + resolveStructuredTarget: (mailboxHandle) => + mailboxHandle === mailbox + ? { + sessionId: IDENTITY.sessionId, + dispatchId, + ...(options.refusal ? { refusal: options.refusal } : {}) + } + : null, + host: { + readGateFacts: () => (journal === null ? null : structuredSessionGateFacts(journal)), + currentFence: () => 4, + send + } + }) + return { + delivery, + markAsDelivered, + send: sendMock, + stored, + setJournal: (next: AgentJournalRenderItem[] | null) => { + journal = next + } + } +} + +const flush = () => new Promise((resolve) => setTimeout(resolve, 0)) + +describe('structured mailbox pointer delivery', () => { + it('claims only mailboxes whose assignee is a structured worker', () => { + const { delivery } = harness({ journal: idleJournal() }) + expect(delivery.deliverForHandle('dispatch:d1')).toBe(true) + expect(delivery.deliverForHandle('run:run_1')).toBe(false) + }) + + it('sends the pointer as a turn and consumes mail on an accepted dispatch', async () => { + const { delivery, markAsDelivered, send } = harness({ journal: idleJournal() }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).toHaveBeenCalledTimes(1) + expect(send.mock.calls[0]![0].operationId).toMatch(/^\d{13}-[0-9a-f]{32}$/) + expect(markAsDelivered).toHaveBeenCalledWith(['m1']) + }) + + it('nudges through the worker`s own handle for direct peer mail outside a dispatch', async () => { + const { delivery, send, markAsDelivered } = harness({ + journal: idleJournal(), + mailbox: IDENTITY.handle, + dispatchId: null + }) + expect(delivery.deliverForHandle(IDENTITY.handle)).toBe(true) + await flush() + expect(send).toHaveBeenCalledTimes(1) + expect(send.mock.calls[0]![0].dispatchId).toBeNull() + // A plain `check`, with no `--run`: the worker resolves its OWN mailbox by identity, and for a + // worker outside a dispatch that is the direct mailbox this mail is sitting in. Pointing it at + // a run would send it to read a coordinator mailbox that has nothing waiting. + expect(send.mock.calls[0]![0].body.blocks[0]).toMatchObject({ + text: expect.not.stringContaining('--run') + }) + expect(markAsDelivered).toHaveBeenCalledWith(['m1']) + }) + + it('retains mail when the dispatch settles unknown', async () => { + const { delivery, markAsDelivered } = harness({ + journal: idleJournal(), + dispatchState: 'unknown' + }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(markAsDelivered).not.toHaveBeenCalled() + }) + + it('retains mail while a turn is running', async () => { + const { delivery, send, markAsDelivered } = harness({ journal: runningJournal() }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).not.toHaveBeenCalled() + expect(markAsDelivered).not.toHaveBeenCalled() + }) + + it('retains mail while a prompt is waiting for a human', async () => { + const { delivery, send } = harness({ journal: attentionJournal() }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).not.toHaveBeenCalled() + }) + + it('delivers to a worker whose finished turn left a long history and no lifecycle row', async () => { + // The steady state after a worker's first substantial turn. Gating on a bounded tail page read + // this as permanently busy, so every later nudge parked forever and the worker went unnudged. + const { delivery, send, markAsDelivered } = harness({ journal: settledLongJournal() }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).toHaveBeenCalledTimes(1) + expect(markAsDelivered).toHaveBeenCalledWith(['m1']) + }) + + it('retains mail for a prompt that scrolled out of the tail window', async () => { + const { delivery, send } = harness({ journal: staleAttentionJournal() }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).not.toHaveBeenCalled() + }) + + it('retains mail when the session is not attached', async () => { + const { delivery, send } = harness({ journal: null }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).not.toHaveBeenCalled() + }) + + it('redrives a detached session when the journal replays on re-attach', async () => { + // A transient detach parks nothing to be woken unless `session-not-attached` waits for the + // journal edge, and the dispatch preamble tells the worker not to poll. + const { delivery, send, setJournal, markAsDelivered } = harness({ journal: null }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).not.toHaveBeenCalled() + setJournal(idleJournal()) + delivery.onJournalActivity('session-1') + await flush() + expect(send).toHaveBeenCalledTimes(1) + expect(markAsDelivered).toHaveBeenCalledWith(['m1']) + }) + + it('retries a parked pointer when the journal moves', async () => { + const { delivery, send, setJournal, markAsDelivered } = harness({ journal: runningJournal() }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).not.toHaveBeenCalled() + setJournal(idleJournal()) + delivery.onJournalActivity('session-1') + await flush() + expect(send).toHaveBeenCalledTimes(1) + expect(markAsDelivered).toHaveBeenCalledWith(['m1']) + }) + + it('nudges the worker while its coordinator holds an unacked Run delivery', async () => { + // The exact window in which a coordinator replies to its workers: it checked, is acting on the + // batch, and has not acked yet. The gate is keyed on the handle being nudged, so the + // coordinator's `run:` delivery is invisible here — gating the WORKER's dispatch mailbox on it + // dropped the nudge with nothing parked, and the worker sat idle on mail it was never told of. + const { delivery, send, markAsDelivered } = harness({ + journal: idleJournal(), + outstandingRunDelivery: true + }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).toHaveBeenCalledTimes(1) + expect(markAsDelivered).toHaveBeenCalledWith(['m1']) + }) + + it('does not re-nudge a mailbox still holding its own unacked batch', async () => { + // The other half of the same gate: the consumer already has this batch, so a second nudge + // spends a whole provider turn telling it something it was told. + const { delivery, send } = harness({ journal: idleJournal(), outstandingOwnDelivery: true }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).not.toHaveBeenCalled() + }) + + it('retries a rejected nudge on the next journal edge', async () => { + // A rejection consumes no mail and nothing else redrives this mailbox, so leaving it unparked + // stranded the worker until unrelated mail happened to arrive. + const { delivery, send, markAsDelivered } = harness({ + journal: idleJournal(), + dispatchState: 'rejected' + }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).toHaveBeenCalledTimes(1) + expect(markAsDelivered).not.toHaveBeenCalled() + delivery.onJournalActivity('session-1') + await flush() + expect(send).toHaveBeenCalledTimes(2) + }) + + it('reuses one operation id for the same batch and re-mints when it grows', async () => { + const { delivery, send, stored } = harness({ + journal: idleJournal(), + dispatchState: 'unknown' + }) + delivery.deliverForHandle('dispatch:d1') + await flush() + const first = send.mock.calls[0]![0].operationId + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send.mock.calls[1]![0].operationId).toBe(first) + stored.clear() + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send.mock.calls[2]![0].operationId).not.toBe(first) + }) +}) + +describe('an adopted pane is redirected through its native owner', () => { + const settled: AgentSessionPtyWriteRefusal = { + code: 'agent_session_conflict', + sessionId: 'session-1', + ownerRuntimeKind: 'native', + handoffStage: null, + ownerPid: 4242, + runtimeFence: 7 + } + + it('sends through the session when the refusal names a settled native owner', async () => { + const { delivery, send, markAsDelivered } = harness({ + journal: idleJournal(), + refusal: settled + }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).toHaveBeenCalledTimes(1) + expect(markAsDelivered).toHaveBeenCalledWith(['m1']) + }) + + it('retains rather than redirecting into a lease that is handing back to a TUI', async () => { + // Re-checked at SEND time: the owner can settle differently between resolve and send, and + // redirecting into a mid-handoff lease races the takeover. + const { delivery, send, markAsDelivered } = harness({ + journal: idleJournal(), + refusal: { ...settled, handoffStage: 'preparing' } + }) + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).not.toHaveBeenCalled() + expect(markAsDelivered).not.toHaveBeenCalled() + }) +}) + +describe('forgetting one settled worker', () => { + /** Two workers, each mid-turn and so each parked on its OWN session's journal edge. */ + function twoWorkerHarness() { + let resolves = true + let journal = runningJournal() + const sessionByMailbox: Record = { + 'dispatch:d1': 'session-1', + 'dispatch:d2': 'session-2' + } + const send: StructuredMailboxPointerHost['send'] = vi.fn(async () => ({ + kind: 'sent' as const, + state: 'accepted' as const + })) + const db = { + getDispatchContextById: () => ({ run_id: 'run_1' }), + hasOutstandingMailboxDelivery: () => false, + getUndeliveredUnreadMessages: () => [{ id: 'm1', type: 'status', sequence: 3 }], + markAsDelivered: vi.fn(), + getStructuredPointerOperation: () => undefined, + putStructuredPointerOperation: () => {}, + deleteStructuredPointerOperation: () => {} + } + const delivery = new OrchestrationStructuredMailboxPointerDelivery({ + getDb: () => db as never, + getMessageWaiters: () => undefined, + resolveStructuredTarget: (mailboxHandle) => { + const sessionId = sessionByMailbox[mailboxHandle] + return resolves && sessionId + ? { sessionId, dispatchId: mailboxHandle.slice('dispatch:'.length) } + : null + }, + host: { + readGateFacts: () => structuredSessionGateFacts(journal), + currentFence: () => 4, + send + } + }) + return { + delivery, + send: vi.mocked(send), + goIdle: () => { + journal = idleJournal() + }, + stopResolving: () => { + resolves = false + }, + resumeResolving: () => { + resolves = true + } + } + } + + it("keeps a sibling worker's wake-up edge when the target cannot be resolved", async () => { + // The bug: `forgetSession` re-resolved every parked mailbox and pruned the ones that answered + // null. A momentarily null DB reference or a session mid-teardown made that EVERY worker, so + // the sibling's mail stayed durable but lost the edge that would have woken it. + const { delivery, send, goIdle, stopResolving, resumeResolving } = twoWorkerHarness() + delivery.deliverForHandle('dispatch:d1') + delivery.deliverForHandle('dispatch:d2') + await flush() + expect(send).not.toHaveBeenCalled() + + stopResolving() + delivery.forgetSession('session-1') + resumeResolving() + + goIdle() + delivery.onJournalActivity('session-2') + await flush() + expect(send).toHaveBeenCalledTimes(1) + expect(send.mock.calls[0]![0].sessionId).toBe('session-2') + }) + + it('still drops what the settled worker itself had parked', async () => { + const { delivery, send, goIdle, stopResolving } = twoWorkerHarness() + delivery.deliverForHandle('dispatch:d1') + await flush() + expect(send).not.toHaveBeenCalled() + + // Settlement forgets the identity, so the target no longer resolves — which is exactly why + // the recorded session id, not a re-resolution, has to be the test. + stopResolving() + delivery.forgetSession('session-1') + + goIdle() + delivery.onJournalActivity('session-1') + await flush() + expect(send).not.toHaveBeenCalled() + }) +}) diff --git a/src/main/runtime/orchestration/structured-mailbox-pointer-delivery.ts b/src/main/runtime/orchestration/structured-mailbox-pointer-delivery.ts new file mode 100644 index 00000000000..24794bf0850 --- /dev/null +++ b/src/main/runtime/orchestration/structured-mailbox-pointer-delivery.ts @@ -0,0 +1,267 @@ +/** + * The pointer-delivery lane for workers that ARE a structured agent session. + * + * The PTY lane types the nudge into a live pane and reads the idle edge off the terminal title. + * Neither exists here, so this is a sibling of `OrchestrationMailboxPointerDelivery` rather than a + * branch inside it: batch selection is literally shared (`selectOrchestrationPointerBatch`), and + * everything below it is different — the nudge is a session turn, the idle edge is the journal, + * and only an `accepted` dispatch may consume mail. + * + * Coordinators are in scope here, unlike the PTY lane's reasoning: a PTY coordinator blocks in + * `check --wait`, where a waiter preempts pointer delivery, but a structured coordinator is a chat + * session whose turn ends — so nothing else would ever wake it for its own `run:` mail. + */ + +import type { AgentJournalMessageItem } from '../../../shared/agent-session-journal-types' +import type { OrchestrationDb } from './db' +import { formatMessagePointer } from './formatter' +import { + selectOrchestrationPointerBatch, + type OrchestrationMessageWaiter +} from './mailbox-pointer-eligibility' +import { resolveStructuredPointerOperation } from './structured-pointer-operation-id' +import { + decideStructuredPointerDelivery, + decideStructuredSessionPointerDelivery, + retainReasonForDispatch, + retainWaitsForJournalEdge, + structuredDispatchDelivered, + type StructuredDispatchState, + type StructuredPointerRetainReason, + type StructuredSessionGateFacts +} from './structured-session-pointer-delivery' +import type { AgentSessionPtyWriteRefusal } from '../../../shared/agent-session-pty-write-admission' + +export type StructuredPointerTarget = { + sessionId: string + /** + * The dispatch whose mailbox this is, or null for direct peer mail addressed to the worker's own + * handle outside any dispatch. Nothing downstream needs a dispatch to deliver — it only scopes + * the operation-ledger budget — so a worker between dispatches is nudged, not dropped. + */ + dispatchId: string | null + /** Present only for an adopted pane, where a PTY write was refused in favour of this owner. */ + refusal?: AgentSessionPtyWriteRefusal +} + +type ParkedPointerDelivery = { + sessionId: string + reservedTypes: ReadonlySet | undefined +} + +export type StructuredPointerSendOutcome = + | { kind: 'sent'; state: StructuredDispatchState } + | { kind: 'unattached' } + +export type StructuredMailboxPointerHost = { + /** The idle gate, read off the session's full reduced timeline; `null` when it is not attached. */ + readGateFacts: (sessionId: string) => StructuredSessionGateFacts | null + send: (input: { + sessionId: string + dispatchId: string | null + operationId: string + payloadFingerprint: string + expectedRuntimeFence: number + body: AgentJournalMessageItem + }) => Promise + /** Current lease fence; `null` when no record backs the session any more. */ + currentFence: (sessionId: string) => number | null +} + +type StructuredPointerDeliveryDependencies = { + getDb: () => OrchestrationDb | null + getMessageWaiters: (mailboxHandle: string) => ReadonlySet | undefined + /** + * The session a mailbox must be nudged through, or null when a live PTY can take the bytes. + * + * Two shapes reach here. A NATIVE-BORN worker carries no refusal: it never had a PTY. An + * ADOPTED one does — its pane is bound to a session a native owner holds, so the PTY write is + * refused and the refusal is what proves the owner is settled enough to redirect to. + * + * The mailbox is a `dispatch:` address or the worker's own bearer handle; the second is how + * agents mail each other outside a dispatch, and no other lane can serve it. + */ + resolveStructuredTarget: (mailboxHandle: string) => StructuredPointerTarget | null + host: StructuredMailboxPointerHost + onRetain?: (input: { + mailboxHandle: string + sessionId: string + reason: StructuredPointerRetainReason + }) => void +} + +export class OrchestrationStructuredMailboxPointerDelivery< + TWaiter extends OrchestrationMessageWaiter +> { + private readonly inFlight = new Set() + /** + * Mailboxes whose retry must wait for the session's next journal edge, each remembering the + * session it is parked ON. + * + * Recorded rather than re-resolved: `resolveStructuredTarget` answers null whenever the runtime + * cannot look — a momentarily null DB reference, a session mid-teardown — and pruning on that + * absence dropped every OTHER worker's parked entry too, silently costing them their wake-up + * edge until the next explicit check. + */ + private readonly parkedUntilJournalEdge = new Map() + + constructor(private readonly deps: StructuredPointerDeliveryDependencies) {} + + deliverForHandle(mailboxHandle: string, reservedTypes?: ReadonlySet): boolean { + const target = this.deps.resolveStructuredTarget(mailboxHandle) + if (!target) { + return false + } + void this.deliver(mailboxHandle, target, reservedTypes).catch(() => { + // Durable mail stays available to an explicit check or the next settle edge. + }) + return true + } + + /** The session's journal moved — a turn settled, or a re-attach replayed it; retry what is + * parked on that edge. */ + onJournalActivity(sessionId: string): void { + for (const [mailboxHandle, parked] of Array.from(this.parkedUntilJournalEdge)) { + if (parked.sessionId !== sessionId) { + continue + } + this.parkedUntilJournalEdge.delete(mailboxHandle) + const target = this.deps.resolveStructuredTarget(mailboxHandle) + if (target?.sessionId !== sessionId) { + // The mailbox moved off this session (or cannot be resolved right now); its own edge or an + // explicit check is what retries it, not this session's journal. + continue + } + void this.deliver(mailboxHandle, target, parked.reservedTypes).catch(() => undefined) + } + } + + /** + * The worker settled; drop what IT had parked, and nothing else. + * + * The recorded session id is the whole test. Settlement forgets the worker's identity, so + * re-resolving the target here would answer null for exactly the entries this is meant to + * prune — and null for every sibling the runtime momentarily cannot resolve either. + */ + forgetSession(sessionId: string): void { + for (const [mailboxHandle, parked] of Array.from(this.parkedUntilJournalEdge)) { + if (parked.sessionId === sessionId) { + this.parkedUntilJournalEdge.delete(mailboxHandle) + } + } + } + + private async deliver( + mailboxHandle: string, + target: StructuredPointerTarget, + reservedTypes?: ReadonlySet + ): Promise { + const db = this.deps.getDb() + if (!db || this.inFlight.has(mailboxHandle)) { + return + } + // Don't re-nudge a mailbox whose consumer still holds an unacknowledged batch. The lookup is + // keyed on the exact handle being nudged, so a coordinator's own `run:` delivery is invisible + // to a worker's `dispatch:` gate and cannot suppress the nudges a coordinator sends its + // workers. Worth more here than in the PTY lane: a structured nudge costs a whole provider + // turn, not a line of text into a composer. + if (db.hasOutstandingMailboxDelivery?.(mailboxHandle)) { + return + } + const unread = selectOrchestrationPointerBatch({ + db, + mailboxHandle, + waiters: this.deps.getMessageWaiters(mailboxHandle), + reservedTypes + }) + if (unread.length === 0) { + return + } + this.inFlight.add(mailboxHandle) + try { + await this.attempt(db, mailboxHandle, target, unread, reservedTypes) + } finally { + this.inFlight.delete(mailboxHandle) + } + } + + private async attempt( + db: OrchestrationDb, + mailboxHandle: string, + target: StructuredPointerTarget, + unread: readonly { id: string; type: string; sequence: number }[], + reservedTypes: ReadonlySet | undefined + ): Promise { + const sessionId = target.sessionId + const session = this.deps.host.readGateFacts(sessionId) + // `target.refusal` is the snapshot the resolver already admitted, so this branch re-runs the + // owner test on frozen input and can only agree with it. What actually fences an owner that + // changed since resolution is `expectedRuntimeFence` below: a handoff bumps the lease fence, + // so the send is refused rather than landing in a lease on its way back to a TUI. The branch + // stays because the policy module is the one place that decides, and a later caller may pass + // an owner it did not pre-screen. + const decision = target.refusal + ? decideStructuredPointerDelivery({ session, refusal: target.refusal }) + : decideStructuredSessionPointerDelivery({ session }) + if (!decision.deliver) { + this.retain(mailboxHandle, sessionId, decision.retain, reservedTypes) + return + } + const fence = this.deps.host.currentFence(sessionId) + if (fence === null) { + this.retain(mailboxHandle, sessionId, 'session-not-attached', reservedTypes) + return + } + const body: AgentJournalMessageItem = { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: formatMessagePointer(unread.length, mailboxHandle).trim() }] + } + const staged = unread.map((message) => message.id) + const operation = resolveStructuredPointerOperation({ + db, + mailboxHandle, + sessionId, + body, + messageIds: staged + }) + const outcome = await this.deps.host.send({ + sessionId, + dispatchId: target.dispatchId, + operationId: operation.operationId, + payloadFingerprint: operation.payloadFingerprint, + expectedRuntimeFence: fence, + body + }) + if (outcome.kind === 'unattached') { + this.retain(mailboxHandle, sessionId, 'session-not-attached', reservedTypes) + return + } + if (!structuredDispatchDelivered(outcome.state)) { + this.retain( + mailboxHandle, + sessionId, + retainReasonForDispatch(outcome.state as Exclude), + reservedTypes + ) + return + } + db.markAsDelivered(staged) + // The nudge landed as its own turn, so the next settle edge is the natural retry point for + // anything that arrives while it runs. + db.deleteStructuredPointerOperation(mailboxHandle) + } + + /** No `markAsUndelivered` is owed: rows are marked delivered only after an accepted dispatch. */ + private retain( + mailboxHandle: string, + sessionId: string, + reason: StructuredPointerRetainReason, + reservedTypes: ReadonlySet | undefined + ): void { + this.deps.onRetain?.({ mailboxHandle, sessionId, reason }) + if (retainWaitsForJournalEdge(reason)) { + this.parkedUntilJournalEdge.set(mailboxHandle, { sessionId, reservedTypes }) + } + } +} diff --git a/src/main/runtime/orchestration/structured-mailbox-pointer-host.test.ts b/src/main/runtime/orchestration/structured-mailbox-pointer-host.test.ts new file mode 100644 index 00000000000..0bbdb74e037 --- /dev/null +++ b/src/main/runtime/orchestration/structured-mailbox-pointer-host.test.ts @@ -0,0 +1,161 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentJournalRenderItem } from '../../../shared/agent-session-journal-types' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { + createStructuredMailboxPointerHost, + structuredPointerCallerKey, + structuredSessionPointerCallerKey +} = await import('./structured-mailbox-pointer-host') + +function runningTurn(): AgentJournalRenderItem { + return { + itemId: 'lifecycle-1', + revision: 1, + body: { kind: 'status', text: 'working', turnLifecycle: { turnId: 'turn-1', state: 'running' } } + } as unknown as AgentJournalRenderItem +} + +function transcript(count: number): AgentJournalRenderItem[] { + return Array.from( + { length: count }, + (_unused, index) => + ({ + itemId: `tool-${index}`, + revision: 1, + body: { kind: 'tool-call', name: 'Bash', input: {}, state: 'completed' } + }) as unknown as AgentJournalRenderItem + ) +} + +describe('structured mailbox pointer host', () => { + beforeEach(() => { + hostRef.current = null + }) + + it('reads the gate facts from the FULL timeline, never a bounded tail', () => { + // The defect this pins: a running turn is announced by ONE lifecycle item, and settlement + // tombstones it rather than rewriting it. A long tool-calling turn pushes that item arbitrarily + // far from the tail, so any page-sized read reports a busy worker as idle — and the pointer is + // then delivered mid-turn, which Codex answers with `turn already running` and Claude settles + // `unknown` while the message is really queued. + const items = [runningTurn(), ...transcript(500)] + hostRef.current = { journalSnapshot: () => ({ items }) } + expect(createStructuredMailboxPointerHost().readGateFacts('s1')).toEqual({ + turnRunning: true, + awaitingHuman: false + }) + }) + + it('answers null rather than idle when the session cannot be read', () => { + // Null retains the pointer; `{turnRunning:false}` would deliver a nudge into a session this + // runtime cannot see at all. + expect(createStructuredMailboxPointerHost().readGateFacts('s1')).toBeNull() + hostRef.current = { + journalSnapshot: () => { + throw new Error('agent_session_ownership_unknown') + } + } + expect(createStructuredMailboxPointerHost().readGateFacts('s1')).toBeNull() + }) + + it('reports an unattached host rather than a rejection when nothing can be sent', async () => { + await expect( + createStructuredMailboxPointerHost().send({ + sessionId: 's1', + dispatchId: 'd1', + operationId: 'op1', + expectedRuntimeFence: 1, + payloadFingerprint: 'fp', + body: { kind: 'message', role: 'user', blocks: [] } + } as never) + ).resolves.toEqual({ kind: 'unattached' }) + }) + + it.each([ + ['accepted', 'accepted'], + ['rejected', 'rejected'], + // Neither is an acknowledgement, and only `accepted` may consume mail: both have to reach the + // caller as `unknown` so the pointer is retained for the next journal edge. + ['pending', 'unknown'], + ['unknown', 'unknown'] + ])('maps a %s submission to %s', async (dispatchState, expected) => { + const send = vi.fn( + async (_caller: { callerKey: string }, _payload: { retryUnknown?: boolean }) => ({ + ok: true, + value: { submission: { dispatchState } } + }) + ) + hostRef.current = { send } + await expect( + createStructuredMailboxPointerHost().send({ + sessionId: 's1', + dispatchId: 'd1', + operationId: 'op1', + expectedRuntimeFence: 1, + payloadFingerprint: 'fp', + body: { kind: 'message', role: 'user', blocks: [] } + } as never) + ).resolves.toEqual({ kind: 'sent', state: expected }) + // Per-dispatch, so one worker's nudges cannot exhaust the shared operation-ledger budget. + expect(send.mock.calls[0]![0]).toEqual({ callerKey: structuredPointerCallerKey('d1') }) + expect(send.mock.calls[0]![1]!.retryUnknown).toBe(true) + }) + + it('scopes direct peer mail to the session when there is no dispatch to scope to', async () => { + // Direct mail is addressed to the worker's own handle, so there may be no dispatch at all. + // The ledger is keyed on (callerKey, operationId): a key derived from the session keeps that + // nudge's own retry lane, and leaves the dispatch key byte-identical so nudges already in + // flight under it still replay rather than being re-minted as a second turn. + const send = vi.fn(async (_caller: { callerKey: string }) => ({ + ok: true, + value: { submission: { dispatchState: 'accepted' } } + })) + hostRef.current = { send } + await expect( + createStructuredMailboxPointerHost().send({ + sessionId: 's1', + dispatchId: null, + operationId: 'op1', + expectedRuntimeFence: 1, + payloadFingerprint: 'fp', + body: { kind: 'message', role: 'user', blocks: [] } + } as never) + ).resolves.toEqual({ kind: 'sent', state: 'accepted' }) + expect(send.mock.calls[0]![0]).toEqual({ + callerKey: structuredSessionPointerCallerKey('s1') + }) + expect(structuredSessionPointerCallerKey('s1')).not.toBe(structuredPointerCallerKey('s1')) + }) + + it('separates a not-attached refusal from a real one', async () => { + for (const [code, expected] of [ + ['agent_session_ownership_unknown', { kind: 'unattached' }], + ['agent_session_conflict', { kind: 'sent', state: 'rejected' }] + ] as const) { + hostRef.current = { send: async () => ({ ok: false, refusal: { code, message: 'no' } }) } + await expect( + createStructuredMailboxPointerHost().send({ + sessionId: 's1', + dispatchId: 'd1', + operationId: 'op1', + expectedRuntimeFence: 1, + payloadFingerprint: 'fp', + body: { kind: 'message', role: 'user', blocks: [] } + } as never) + ).resolves.toEqual(expected) + } + }) + + it('reads the runtime fence off the durable record', () => { + hostRef.current = { deps: { store: { getRecord: () => ({ lease: { runtimeFence: 9 } }) } } } + expect(createStructuredMailboxPointerHost().currentFence('s1')).toBe(9) + hostRef.current = { deps: { store: { getRecord: () => null } } } + expect(createStructuredMailboxPointerHost().currentFence('s1')).toBeNull() + }) +}) diff --git a/src/main/runtime/orchestration/structured-mailbox-pointer-host.ts b/src/main/runtime/orchestration/structured-mailbox-pointer-host.ts new file mode 100644 index 00000000000..4b80b8df5d7 --- /dev/null +++ b/src/main/runtime/orchestration/structured-mailbox-pointer-host.ts @@ -0,0 +1,107 @@ +/** + * The structured-session half of the structured pointer lane. + * + * Keeps every `getStructuredAgentSessionHost()` call in one place so the delivery policy above it + * stays pure and testable. Nothing here decides whether to deliver; it only performs the read and + * the send and reports what the host said. + */ + +import { AGENT_SESSION_NOT_ATTACHED } from '../../native-chat/agent-session-wire/structured-agent-session-mutation-admission' +import { getStructuredAgentSessionHost } from '../../native-chat/agent-session-wire/structured-agent-session-registry' +import type { StructuredMailboxPointerHost } from './structured-mailbox-pointer-delivery' +import { + structuredSessionGateFacts, + type StructuredSessionGateFacts +} from './structured-session-pointer-delivery' + +/** Per-dispatch so one worker's nudges cannot exhaust the shared runtime operation-ledger budget. */ +export function structuredPointerCallerKey(dispatchId: string): string { + return `trusted-local:orchestration:${dispatchId}` +} + +/** + * The same budget for direct peer mail, which is addressed to the worker's own handle and has no + * dispatch to scope to. + * + * A separate key rather than a reshaped one: the ledger is keyed on (callerKey, operationId), so + * changing the dispatch key's shape would orphan every nudge already in flight under the old one. + */ +export function structuredSessionPointerCallerKey(sessionId: string): string { + return `trusted-local:orchestration:session:${sessionId}` +} + +/** + * The idle gate for a structured session, read off its FULL reduced timeline. + * + * Never a bounded page. Settlement tombstones the running turn's lifecycle item rather than + * rewriting it to `completed`, so on any tail window an idle session and a busy one whose + * lifecycle item scrolled off look identical — and idle-with-history is the normal steady state of + * a working agent. Shared so the pointer lane and group addressing cannot disagree about it. + */ +export function readStructuredSessionGateFacts( + sessionId: string +): StructuredSessionGateFacts | null { + const host = getStructuredAgentSessionHost() + if (!host) { + return null + } + try { + return structuredSessionGateFacts(host.journalSnapshot(sessionId).items) + } catch (error) { + // Not attached is a retain reason, not a failure; anything else is still unreadable. + if ((error as Error)?.message !== AGENT_SESSION_NOT_ATTACHED.code) { + console.warn('[orchestration] structured journal unreadable', sessionId, error) + } + return null + } +} + +export function createStructuredMailboxPointerHost(): StructuredMailboxPointerHost { + return { + readGateFacts(sessionId) { + return readStructuredSessionGateFacts(sessionId) + }, + + currentFence(sessionId) { + return ( + getStructuredAgentSessionHost()?.deps.store.getRecord(sessionId)?.lease.runtimeFence ?? null + ) + }, + + async send(input) { + const host = getStructuredAgentSessionHost() + if (!host) { + return { kind: 'unattached' } + } + const result = await host.send( + { + callerKey: input.dispatchId + ? structuredPointerCallerKey(input.dispatchId) + : structuredSessionPointerCallerKey(input.sessionId) + }, + { + envelope: { + sessionId: input.sessionId, + clientOperationId: input.operationId, + expectedRuntimeFence: input.expectedRuntimeFence, + payloadFingerprint: input.payloadFingerprint + }, + body: input.body, + // The recorded unknown is the only thing that unlocks a redispatch of the same id. + retryUnknown: true + } + ) + if (!result.ok) { + return result.refusal.code === AGENT_SESSION_NOT_ATTACHED.code + ? { kind: 'unattached' } + : { kind: 'sent', state: 'rejected' } + } + // `pending` is not yet an acknowledgement; only `accepted` may consume mail. + const state = result.value.submission.dispatchState + return { + kind: 'sent', + state: state === 'accepted' ? 'accepted' : state === 'rejected' ? 'rejected' : 'unknown' + } + } + } +} diff --git a/src/main/runtime/orchestration/structured-pointer-operation-id.test.ts b/src/main/runtime/orchestration/structured-pointer-operation-id.test.ts new file mode 100644 index 00000000000..c1853be5a66 --- /dev/null +++ b/src/main/runtime/orchestration/structured-pointer-operation-id.test.ts @@ -0,0 +1,162 @@ +import { describe, expect, it } from 'vitest' +import { AGENT_SESSION_MAX_NEW_OPERATION_AGE_MS } from '../../../shared/agent-session-host-authority' +import type { AgentJournalMessageItem } from '../../../shared/agent-session-journal-types' +import { + mintAgentSessionOperationId, + resolveStructuredPointerOperation +} from './structured-pointer-operation-id' + +const OPERATION_ID_PATTERN = /^\d{13}-[0-9a-f]{32}$/ + +function body(text: string): AgentJournalMessageItem { + return { kind: 'message', role: 'user', blocks: [{ type: 'text', text }] } +} + +function fakeDb() { + const rows = new Map() + return { + rows, + getStructuredPointerOperation: (handle: string) => rows.get(handle), + putStructuredPointerOperation: (row: { mailbox_handle: string; operation_id: string }) => + rows.set(row.mailbox_handle, row) + } as never +} + +describe('structured pointer operation id', () => { + it('mints ids the host will admit', () => { + // Orchestration's own msg_ ids do not match and are refused before the first send. + expect(mintAgentSessionOperationId(Date.now())).toMatch(OPERATION_ID_PATTERN) + }) + + it('reuses one id for the same batch', () => { + const db = fakeDb() + const first = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m1', 'm2'], + now: 1_000 + }) + const second = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m1', 'm2'], + now: 2_000 + }) + expect(second.operationId).toBe(first.operationId) + expect(second.payloadFingerprint).toBe(first.payloadFingerprint) + }) + + it('re-mints when the batch grows', () => { + const db = fakeDb() + const first = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m1', 'm2'], + now: 1_000 + }) + const grown = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('3 messages'), + messageIds: ['m1', 'm2', 'm3'], + now: 1_500 + }) + expect(grown.operationId).not.toBe(first.operationId) + }) + + it('re-mints once the host would refuse the id as expired', () => { + const db = fakeDb() + const first = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m1', 'm2'], + now: 1_000 + }) + const aged = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m1', 'm2'], + now: 1_000 + AGENT_SESSION_MAX_NEW_OPERATION_AGE_MS + }) + expect(aged.operationId).not.toBe(first.operationId) + }) + + it('re-mints for a different batch of the same size', () => { + // The pointer body names only how many messages are waiting, so two unrelated same-size + // batches share a payload fingerprint. Reusing the live id across them makes the host replay + // its ledger answer — `accepted`, with no turn sent — and the lane then marks the NEW mail + // delivered. The worker is never told, and the mail is gone. + const db = fakeDb() + const first = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m1', 'm2'], + now: 1_000 + }) + const different = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m3', 'm4'], + now: 1_100 + }) + expect(different.operationId).not.toBe(first.operationId) + expect(different.payloadFingerprint).toBe(first.payloadFingerprint) + }) + + it('re-mints when a retained batch is reordered or partly consumed', () => { + const db = fakeDb() + const first = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m1', 'm2'], + now: 1_000 + }) + const shifted = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m2', 'm3'], + now: 1_100 + }) + expect(shifted.operationId).not.toBe(first.operationId) + }) + + it('re-mints when the mailbox moves to a different session', () => { + const db = fakeDb() + const first = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's1', + body: body('2 messages'), + messageIds: ['m1', 'm2'], + now: 1_000 + }) + const moved = resolveStructuredPointerOperation({ + db, + mailboxHandle: 'dispatch:d1', + sessionId: 's2', + body: body('2 messages'), + messageIds: ['m1', 'm2'], + now: 1_100 + }) + expect(moved.operationId).not.toBe(first.operationId) + }) +}) diff --git a/src/main/runtime/orchestration/structured-pointer-operation-id.ts b/src/main/runtime/orchestration/structured-pointer-operation-id.ts new file mode 100644 index 00000000000..6c1ecc7e032 --- /dev/null +++ b/src/main/runtime/orchestration/structured-pointer-operation-id.ts @@ -0,0 +1,77 @@ +/** + * The agent-session operation id one structured worker mailbox's pointer send runs under. + * + * Orchestration's own `msg_` ids do not match the host's `^\d{13}-[0-9a-f]{32}$` shape and are + * refused before the first send, so the id is minted here instead. It is durable and reused across + * retries, because the id IS the send's idempotency key: a fresh id for the same nudge would land + * as a second turn. It is re-minted only when the send is genuinely a different call — a different + * batch of mail, or a different session — or when the host would reject it as too old to admit. + * + * Reuse is keyed on the MESSAGE IDS in the batch, never on the pointer body: the body names only + * how many messages are waiting, so two unrelated same-size batches share a fingerprint. Reusing a + * live id across them makes the host answer from its operation ledger — `accepted`, with no turn + * sent — and this lane then marks the new mail delivered. That is silent mail loss. + */ + +import { createHash, randomBytes } from 'node:crypto' +import type { AgentJournalMessageItem } from '../../../shared/agent-session-journal-types' +import { computeAgentSessionPayloadFingerprint } from '../../../shared/agent-session-mutation-envelope' +import { AGENT_SESSION_MAX_NEW_OPERATION_AGE_MS } from '../../../shared/agent-session-host-authority' +import type { OrchestrationDb } from './db' + +export function mintAgentSessionOperationId(now: number): string { + return `${String(now).padStart(13, '0')}-${randomBytes(16).toString('hex')}` +} + +/** Batch identity, and the only thing reuse may be keyed on. */ +export function structuredPointerBatchFingerprint( + sessionId: string, + messageIds: readonly string[] +): string { + return createHash('sha256') + .update(JSON.stringify([sessionId, messageIds])) + .digest('base64url') +} + +export function structuredPointerPayloadFingerprint( + sessionId: string, + body: AgentJournalMessageItem +): string { + return computeAgentSessionPayloadFingerprint({ + method: 'agentSession.send', + sessionId, + fields: { body } + }) +} + +export function resolveStructuredPointerOperation(args: { + db: OrchestrationDb + mailboxHandle: string + sessionId: string + body: AgentJournalMessageItem + /** The rows this nudge stands for; batch identity, not the body, decides reuse. */ + messageIds: readonly string[] + now?: number +}): { operationId: string; payloadFingerprint: string } { + const now = args.now ?? Date.now() + const payloadFingerprint = structuredPointerPayloadFingerprint(args.sessionId, args.body) + const batchFingerprint = structuredPointerBatchFingerprint(args.sessionId, args.messageIds) + const stored = args.db.getStructuredPointerOperation(args.mailboxHandle) + if ( + stored && + stored.session_id === args.sessionId && + stored.batch_fingerprint === batchFingerprint && + now - stored.minted_at_ms < AGENT_SESSION_MAX_NEW_OPERATION_AGE_MS + ) { + return { operationId: stored.operation_id, payloadFingerprint } + } + const operationId = mintAgentSessionOperationId(now) + args.db.putStructuredPointerOperation({ + mailbox_handle: args.mailboxHandle, + session_id: args.sessionId, + operation_id: operationId, + batch_fingerprint: batchFingerprint, + minted_at_ms: now + }) + return { operationId, payloadFingerprint } +} diff --git a/src/main/runtime/orchestration/structured-session-pointer-delivery.test.ts b/src/main/runtime/orchestration/structured-session-pointer-delivery.test.ts new file mode 100644 index 00000000000..09e5010a104 --- /dev/null +++ b/src/main/runtime/orchestration/structured-session-pointer-delivery.test.ts @@ -0,0 +1,194 @@ +import { describe, expect, it } from 'vitest' +import type { AgentJournalRenderItem } from '../../../shared/agent-session-journal-types' +import type { AgentSessionPtyWriteRefusal } from '../../../shared/agent-session-pty-write-admission' +import { + decideStructuredPointerDelivery, + isSettledNativeOwner, + retainReasonForDispatch, + retainWaitsForJournalEdge, + structuredDispatchDelivered, + structuredSessionGateFacts +} from './structured-session-pointer-delivery' + +function refusal( + overrides: Partial = {} +): AgentSessionPtyWriteRefusal { + return { + code: 'agent_session_conflict', + sessionId: 'session-1', + ownerRuntimeKind: 'native', + handoffStage: null, + ownerPid: 4242, + runtimeFence: 7, + ...overrides + } +} + +function statusItem( + turnLifecycle: { turnId: string; state: 'running' } | undefined +): AgentJournalRenderItem { + return { + itemId: `item-${turnLifecycle?.turnId ?? 'plain'}`, + revision: 1, + body: { kind: 'status', text: 'working', ...(turnLifecycle ? { turnLifecycle } : {}) } + } as unknown as AgentJournalRenderItem +} + +/** A turn's worth of ordinary transcript: no lifecycle row, which is what a settled turn leaves. */ +function transcript(count: number): AgentJournalRenderItem[] { + return Array.from( + { length: count }, + (_unused, index) => + ({ + itemId: `tool-${index}`, + revision: 1, + body: { kind: 'tool-call', name: 'Bash', input: {}, state: 'completed' } + }) as unknown as AgentJournalRenderItem + ) +} + +function pendingApproval(): AgentJournalRenderItem { + return { + itemId: 'approval-1', + revision: 1, + body: { kind: 'approval', title: 'run it?', resolution: { state: 'pending' } } + } as unknown as AgentJournalRenderItem +} + +const IDLE = { turnRunning: false, awaitingHuman: false } + +describe('structured pointer owner admission', () => { + it('accepts only a settled native owner', () => { + expect(isSettledNativeOwner(refusal())).toBe(true) + }) + + it('refuses a tui owner', () => { + expect(isSettledNativeOwner(refusal({ ownerRuntimeKind: 'tui' }))).toBe(false) + }) + + it('refuses a native owner that is mid-handoff, so a to-tui takeover is not raced', () => { + expect(isSettledNativeOwner(refusal({ handoffStage: 'recovering' }))).toBe(false) + }) + + it('refuses a reconciling refusal even though it names a native owner', () => { + expect(isSettledNativeOwner(refusal({ code: 'execution_owner_reconciling' }))).toBe(false) + }) +}) + +describe('structured session gate facts', () => { + it('reads an empty journal as idle', () => { + expect(structuredSessionGateFacts([])).toEqual(IDLE) + }) + + it('reads a running turn as busy', () => { + expect( + structuredSessionGateFacts([statusItem({ turnId: 'turn-1', state: 'running' })]) + ).toEqual({ turnRunning: true, awaitingHuman: false }) + }) + + it('reads a tombstoned turn as idle, since settlement removes the running row', () => { + // A healthy completed turn leaves no turnLifecycle row behind at all. + expect(structuredSessionGateFacts([statusItem(undefined)])).toEqual(IDLE) + }) + + it('reads a worker that has finished a long turn as idle, however much history it has', () => { + // The steady state of a working agent: plenty of items, no lifecycle row anywhere. Answering + // this from a bounded tail page cannot distinguish it from a running turn whose lifecycle row + // was pushed off the end, which is why the facts come off the fully reduced timeline. + expect(structuredSessionGateFacts(transcript(120))).toEqual(IDLE) + }) + + it('sees a pending approval that scrolled out of any tail window', () => { + expect(structuredSessionGateFacts([pendingApproval(), ...transcript(120)])).toEqual({ + turnRunning: false, + awaitingHuman: true + }) + }) + + it('reports a prompt raised mid-turn as both busy and awaiting a human', () => { + expect( + structuredSessionGateFacts([ + statusItem({ turnId: 'turn-1', state: 'running' }), + pendingApproval() + ]) + ).toEqual({ turnRunning: true, awaitingHuman: true }) + }) +}) + +describe('decideStructuredPointerDelivery', () => { + it('delivers to a settled, attached, idle session', () => { + expect(decideStructuredPointerDelivery({ refusal: refusal(), session: IDLE })).toEqual({ + deliver: true + }) + }) + + it('retains when the session is not attached on this host', () => { + expect(decideStructuredPointerDelivery({ refusal: refusal(), session: null })).toEqual({ + deliver: false, + retain: 'session-not-attached' + }) + }) + + it('retains mid-turn rather than delegating the race to the provider', () => { + expect( + decideStructuredPointerDelivery({ + refusal: refusal(), + session: { turnRunning: true, awaitingHuman: false } + }) + ).toEqual({ deliver: false, retain: 'turn-unsettled' }) + }) + + it('names the human prompt ahead of the turn, so the retain reason is the actionable one', () => { + expect( + decideStructuredPointerDelivery({ + refusal: refusal(), + session: { turnRunning: true, awaitingHuman: true } + }) + ).toEqual({ deliver: false, retain: 'awaiting-human' }) + }) + + it('retains when the owner is not a settled native session', () => { + expect( + decideStructuredPointerDelivery({ + refusal: refusal({ handoffStage: 'preparing' }), + session: IDLE + }) + ).toEqual({ deliver: false, retain: 'owner-not-settled-native' }) + }) +}) + +describe('dispatch outcome classification', () => { + it('marks mail delivered only on an accepted dispatch', () => { + expect(structuredDispatchDelivered('accepted')).toBe(true) + expect(structuredDispatchDelivered('rejected')).toBe(false) + }) + + it('does not treat unknown as delivered, because a dead child settles unknown', () => { + expect(structuredDispatchDelivered('unknown')).toBe(false) + }) + + it('names the retain reason for each non-accepted dispatch', () => { + expect(retainReasonForDispatch('rejected')).toBe('dispatch-rejected') + expect(retainReasonForDispatch('unknown')).toBe('dispatch-unknown') + }) +}) + +describe('retry pacing', () => { + it('parks a nudge that may already be queued until the journal moves again', () => { + expect(retainWaitsForJournalEdge('dispatch-unknown')).toBe(true) + expect(retainWaitsForJournalEdge('turn-unsettled')).toBe(true) + expect(retainWaitsForJournalEdge('awaiting-human')).toBe(true) + }) + + it('parks a detached session, because the re-attach edge is the only thing that will notice', () => { + expect(retainWaitsForJournalEdge('session-not-attached')).toBe(true) + }) + + it('parks a rejected dispatch, because nothing else retries and no mail was consumed', () => { + expect(retainWaitsForJournalEdge('dispatch-rejected')).toBe(true) + }) + + it('allows a plain retry only for an owner the resolver would not have named', () => { + expect(retainWaitsForJournalEdge('owner-not-settled-native')).toBe(false) + }) +}) diff --git a/src/main/runtime/orchestration/structured-session-pointer-delivery.ts b/src/main/runtime/orchestration/structured-session-pointer-delivery.ts new file mode 100644 index 00000000000..272d7799947 --- /dev/null +++ b/src/main/runtime/orchestration/structured-session-pointer-delivery.ts @@ -0,0 +1,160 @@ +/** + * Delivery decisions for an orchestration mail pointer aimed at a host-owned + * structured ("native") agent session. + * + * A structured session has no PTY the pointer can be typed into, so the nudge + * travels as a session turn instead of as bytes. Everything here is pure: the + * caller supplies the refusal and the session's gate facts, and gets back a + * decision it can act on. Orchestration's database stays the source of truth — + * no decision here ever consumes mail, it only says whether the nudge may be + * attempted now. + */ + +import type { AgentJournalRenderItem } from '../../../shared/agent-session-journal-types' +import type { AgentSessionPtyWriteRefusal } from '../../../shared/agent-session-pty-write-admission' +import { + activeStructuredAgentSessionTurnId, + projectStructuredAgentSessionStatus +} from '../../../shared/structured-agent-session-projection' + +/** Every reason retains the pointer; none of them consume mail. */ +export type StructuredPointerRetainReason = + | 'owner-not-settled-native' + | 'session-not-attached' + | 'turn-unsettled' + | 'awaiting-human' + | 'dispatch-rejected' + | 'dispatch-unknown' + +export type StructuredPointerDecision = + | { deliver: true } + | { deliver: false; retain: StructuredPointerRetainReason } + +/** The dispatch states both provider adapters converge on. */ +export type StructuredDispatchState = 'accepted' | 'rejected' | 'unknown' + +/** + * A refusal names an owner this pointer may be redirected to only when that + * owner is native AND settled. A recovering or mid-handoff lease also reports + * `native`, but it may become a TUI again, so redirecting there races the + * takeover. + */ +export function isSettledNativeOwner(refusal: AgentSessionPtyWriteRefusal): boolean { + return ( + refusal.ownerRuntimeKind === 'native' && + refusal.code === 'agent_session_conflict' && + refusal.handoffStage === null + ) +} + +/** + * What the delivery gate needs to know about a session, read once per attempt. + * + * Deliberately two booleans rather than the journal: the caller reads the FULL reduced timeline + * (see `readGateFacts`), so nothing downstream can be tempted to re-derive them from a page. + */ +export type StructuredSessionGateFacts = { + turnRunning: boolean + /** A pending approval or question only a human can clear. */ + awaitingHuman: boolean +} + +/** + * Projects the gate facts off a session's live items. + * + * Reuses the projection the chat view already reads, so the delivery gate and the visible + * "working" state can never disagree. Both must be answered from the fully reduced timeline: a + * settled turn is TOMBSTONED rather than rewritten to `completed`, so on a bounded tail page an + * idle session and a running turn whose lifecycle item was pushed off the end look identical — + * and idle-with-history is the normal steady state of a working agent. + */ +export function structuredSessionGateFacts( + items: readonly AgentJournalRenderItem[] +): StructuredSessionGateFacts { + return { + turnRunning: activeStructuredAgentSessionTurnId(items) !== null, + awaitingHuman: projectStructuredAgentSessionStatus(items) === 'attention' + } +} + +/** + * Decide whether the nudge may be sent right now. + * + * Mid-turn delivery is refused for both providers rather than delegated to + * them: Codex answers a mid-turn `turn/start` with `turn already running`, and + * Claude accepts the frame but cannot acknowledge it inside the dispatch ack + * window, settling `unknown` while the message is really queued. Waiting for + * the turn to settle is the one contract that holds for both, and it preserves + * orchestration's existing idle-edge-only delivery policy. + */ +export function decideStructuredPointerDelivery(input: { + refusal: AgentSessionPtyWriteRefusal + /** Null when the session is not attached to this host. */ + session: StructuredSessionGateFacts | null +}): StructuredPointerDecision { + if (!isSettledNativeOwner(input.refusal)) { + return { deliver: false, retain: 'owner-not-settled-native' } + } + return decideStructuredSessionPointerDelivery(input) +} + +/** + * The same decision for a session that was BORN structured. + * + * There is no PTY write to be refused, so there is no refusal to read an owner off — the caller + * already knows the session is host-owned because it created it. Everything after that gate is + * identical, which is why the adopted-TUI path above delegates here rather than duplicating it. + */ +export function decideStructuredSessionPointerDelivery(input: { + session: StructuredSessionGateFacts | null +}): StructuredPointerDecision { + if (!input.session) { + return { deliver: false, retain: 'session-not-attached' } + } + // Checked before the turn gate: a pending prompt has no running turn, so the turn test alone + // reads it as idle, and sending there queues a nudge behind something only a human can clear. + if (input.session.awaitingHuman) { + return { deliver: false, retain: 'awaiting-human' } + } + if (input.session.turnRunning) { + return { deliver: false, retain: 'turn-unsettled' } + } + return { deliver: true } +} + +/** + * Only an accepted dispatch may mark mail delivered. + * + * `unknown` covers a dead provider child and a slow acknowledgement alike — the + * adapters cannot tell them apart — so it must retain. Treating it as delivered + * would drop mail whenever a child died mid-send. + */ +export function structuredDispatchDelivered(state: StructuredDispatchState): boolean { + return state === 'accepted' +} + +export function retainReasonForDispatch( + state: Exclude +): StructuredPointerRetainReason { + return state === 'rejected' ? 'dispatch-rejected' : 'dispatch-unknown' +} + +/** + * Whether a retained pointer should be parked for the session's next journal edge, or is cheap + * enough to re-attempt on any later trigger. + * + * `unknown` may mean the nudge is already sitting in the provider's input queue, so an immediate + * retry can stack duplicate nudges that each become a turn later. `session-not-attached` parks for + * the opposite reason: nothing else will ever notice the re-attach, and the dispatch preamble + * tells workers not to poll, so an unparked pointer leaves the worker idle on unread mail. + * `dispatch-rejected` parks for that same reason: a rejection consumes no mail and is usually a + * stale fence or a lease that has since moved, both of which the next journal edge re-reads. + * + * Only `owner-not-settled-native` is excluded, and it is unreachable in practice: the resolver + * refuses to name an unsettled owner, so the pointer falls through to the PTY lane before it can + * be retained here. Phrased as an exclusion so a reason added later parks by default — parking + * only adds a retry edge, while forgetting to park is how mail goes unnoticed. + */ +export function retainWaitsForJournalEdge(reason: StructuredPointerRetainReason): boolean { + return reason !== 'owner-not-settled-native' +} diff --git a/src/main/runtime/orchestration/structured-worker-direct-mailbox-target.test.ts b/src/main/runtime/orchestration/structured-worker-direct-mailbox-target.test.ts new file mode 100644 index 00000000000..e5b36cdf67f --- /dev/null +++ b/src/main/runtime/orchestration/structured-worker-direct-mailbox-target.test.ts @@ -0,0 +1,152 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentSessionRecord } from '../../../shared/agent-session-record' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { OrcaRuntimeWithGetPtyRecordForPaneKey } = + await import('../orca-runtime-get-pty-record-for-pane-key') +const { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} = await import('../structured-worker-identity') + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + +/** The real method through the real prototype chain; a re-declared copy would pin nothing. */ +class MailboxTargetProbe extends OrcaRuntimeWithGetPtyRecordForPaneKey { + probeResolveTarget(mailboxHandle: string): unknown { + return this.resolveStructuredMailboxTarget(mailboxHandle) + } +} + +function installRecord(lease: { runtimeKind: string; claimStatus: string }): void { + hostRef.current = { + deps: { + store: { + getRecord: (sessionId: string) => + ({ + sessionId, + location: { executionHostId: 'local', wslDistro: null }, + lease: { ...lease, runtimeFence: 1, deathEvidence: null } + }) as unknown as AgentSessionRecord + } + }, + hasSession: () => true + } +} + +function registerWorker(): string { + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + return handle +} + +function probe(activeDispatch: { id: string } | undefined, run?: { coordinator_handle: string }) { + const findActiveDispatchForAssignee = vi.fn(() => activeDispatch) + const getRun = vi.fn(() => run) + const instance = Object.assign(Object.create(MailboxTargetProbe.prototype), { + _orchestrationDb: { findActiveDispatchForAssignee, getRun }, + getLiveLeafForHandle: () => { + throw new Error('no leaf backs a native-born structured worker') + } + }) as MailboxTargetProbe + return { instance, findActiveDispatchForAssignee, getRun } +} + +describe('the mailbox target for direct peer mail to a structured worker', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + it('routes a bare worker handle through the active dispatch that worker holds', () => { + const handle = registerWorker() + installRecord({ runtimeKind: 'native', claimStatus: 'live' }) + const { instance, findActiveDispatchForAssignee } = probe({ id: 'd1' }) + expect(instance.probeResolveTarget(handle)).toEqual({ + sessionId: SESSION_ID, + dispatchId: 'd1' + }) + // The pane key is the remint-stable half of the lookup, exactly as the PTY path uses it. + expect(findActiveDispatchForAssignee).toHaveBeenCalledWith( + handle, + structuredWorkerIdentities.get(handle)!.paneKey + ) + }) + + it('still delivers to a worker that has no active dispatch', () => { + // The defect this pins: the send stored durably and reported success, and then NEITHER lane + // claimed the mailbox — the PTY lane refuses a structured handle outright and this resolver + // answered only `dispatch:` addresses. The worker never reacted and the peer waiting on a + // reply hung, with nothing logged. A dispatch says nothing about whether delivery is safe; + // the idle gate and the lease fence do, and both still run downstream. + const handle = registerWorker() + installRecord({ runtimeKind: 'native', claimStatus: 'live' }) + expect(probe(undefined).instance.probeResolveTarget(handle)).toEqual({ + sessionId: SESSION_ID, + dispatchId: null + }) + }) + + it('leaves a handle whose session this runtime no longer owns to the PTY lane', () => { + const handle = registerWorker() + installRecord({ runtimeKind: 'tui', claimStatus: 'live' }) + expect(probe({ id: 'd1' }).instance.probeResolveTarget(handle)).toBeNull() + installRecord({ runtimeKind: 'native', claimStatus: 'released' }) + expect(probe({ id: 'd1' }).instance.probeResolveTarget(handle)).toBeNull() + }) + + it('claims a PTY handle for neither lane', () => { + installRecord({ runtimeKind: 'native', claimStatus: 'live' }) + const { instance, findActiveDispatchForAssignee } = probe({ id: 'd1' }) + expect(instance.probeResolveTarget('term_abc')).toBeNull() + expect(findActiveDispatchForAssignee).not.toHaveBeenCalled() + }) +}) + +describe('the mailbox target for a Run whose coordinator is structured', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + it('owns the run mailbox, which neither lane used to claim', () => { + // The defect this pins: the PTY lane declines because the owner is structured, and this lane + // used to decline anything that was not `dispatch:`. Each half believed the other owned it, so + // a structured coordinator was never nudged for its own Run mail and nothing logged. A PTY + // coordinator is covered by blocking in `check --wait`; a chat session's turn just ends. + const handle = registerWorker() + installRecord({ runtimeKind: 'native', claimStatus: 'live' }) + expect( + probe(undefined, { coordinator_handle: handle }).instance.probeResolveTarget('run:run_1') + ).toEqual({ sessionId: SESSION_ID, dispatchId: null }) + }) + + it('leaves the run mailbox of a PTY coordinator to the PTY lane', () => { + installRecord({ runtimeKind: 'native', claimStatus: 'live' }) + expect( + probe(undefined, { coordinator_handle: 'term_coord' }).instance.probeResolveTarget( + 'run:run_1' + ) + ).toBeNull() + }) + + it('claims nothing for a run that does not resolve', () => { + installRecord({ runtimeKind: 'native', claimStatus: 'live' }) + expect(probe(undefined).instance.probeResolveTarget('run:run_1')).toBeNull() + }) +}) diff --git a/src/main/runtime/orchestration/structured-worker-group-addressing.test.ts b/src/main/runtime/orchestration/structured-worker-group-addressing.test.ts new file mode 100644 index 00000000000..ce9a9fbaf37 --- /dev/null +++ b/src/main/runtime/orchestration/structured-worker-group-addressing.test.ts @@ -0,0 +1,220 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentJournalRenderItem } from '../../../shared/agent-session-journal-types' +import type { AgentSessionRecord } from '../../../shared/agent-session-record' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { listAddressableStructuredWorkers, structuredWorkerAgentStatus } = + await import('./structured-worker-group-addressing') +const { resolveGroupAddress } = await import('./groups') +const { sendGroupMessage } = await import('../rpc/methods/orchestration/messaging/send-group') +const { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} = await import('../structured-worker-identity') + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + +function idleTurn(): AgentJournalRenderItem { + return { + itemId: 'lifecycle-1', + body: { kind: 'status', text: 'done', turnLifecycle: { turnId: 't1', state: 'completed' } } + } as unknown as AgentJournalRenderItem +} + +function runningTurn(): AgentJournalRenderItem { + return { + itemId: 'lifecycle-1', + body: { kind: 'status', text: 'working', turnLifecycle: { turnId: 't1', state: 'running' } } + } as unknown as AgentJournalRenderItem +} + +function transcript(count: number): AgentJournalRenderItem[] { + return Array.from( + { length: count }, + (_unused, index) => + ({ + itemId: `tool-${index}`, + body: { kind: 'tool-call', name: 'Bash', input: {}, state: 'completed' } + }) as unknown as AgentJournalRenderItem + ) +} + +function installHost(options: { + items?: AgentJournalRenderItem[] + lease?: { runtimeKind: string; claimStatus: string } + hasSession?: boolean +}): void { + const lease = options.lease ?? { runtimeKind: 'native', claimStatus: 'live' } + hostRef.current = { + deps: { + store: { + getRecord: (sessionId: string) => + ({ + sessionId, + provider: 'codex', + location: { executionHostId: 'local', wslDistro: null }, + lease: { ...lease, runtimeFence: 1, deathEvidence: null } + }) as unknown as AgentSessionRecord + } + }, + hasSession: () => options.hasSession ?? true, + journalSnapshot: () => ({ items: options.items ?? [idleTurn()] }) + } +} + +function registerWorker(worktreeId = 'wt_1'): string { + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'codex', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId, + hostScope: { kind: 'local', hostId: 'local' } + }) + return handle +} + +const PTY_TERMINAL = { handle: 'term_a', worktreeId: 'wt_1', agentIdentity: 'claude' as const } + +describe('group addressing and structured workers', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + it('enumerates a live structured worker as a candidate', () => { + const handle = registerWorker() + installHost({}) + expect(listAddressableStructuredWorkers()).toEqual([ + { handle, worktreeId: 'wt_1', agentIdentity: 'codex' } + ]) + }) + + it('leaves out a worker whose session is not proven live', () => { + // Addressing a settled worker would store mail no lane will ever deliver. + registerWorker() + installHost({ lease: { runtimeKind: 'native', claimStatus: 'live' }, hasSession: false }) + expect(listAddressableStructuredWorkers()).toEqual([]) + }) + + it('reaches a structured worker through @all', () => { + // The defect this pins: recipients came only from `listTerminals`, which enumerates leaves and + // PTYs, so a structured worker was excluded BEFORE per-recipient resolution — the warning + // machinery never ran and the sender got exit 0 with a receipt naming only who did resolve. + const handle = registerWorker() + installHost({}) + const recipients = [PTY_TERMINAL, ...listAddressableStructuredWorkers()] + expect(resolveGroupAddress('@all', 'term_sender', recipients, () => 'idle')).toContain(handle) + }) + + it('reaches a structured worker through @worktree: and @codex, but not @claude', () => { + const handle = registerWorker('wt_2') + installHost({}) + const recipients = [PTY_TERMINAL, ...listAddressableStructuredWorkers()] + expect(resolveGroupAddress('@worktree:wt_2', 'term_sender', recipients, () => 'idle')).toEqual([ + handle + ]) + expect(resolveGroupAddress('@codex', 'term_sender', recipients, () => 'idle')).toEqual([handle]) + expect(resolveGroupAddress('@claude', 'term_sender', recipients, () => 'idle')).toEqual([ + 'term_a' + ]) + }) + + it('reads @idle status off the FULL timeline, never a bounded tail', () => { + // The same trap that already cost this branch once: settlement tombstones the lifecycle item + // rather than rewriting it, so a long tool-calling turn pushes it arbitrarily far from the + // tail and any page-sized read reports a BUSY worker as idle — then `@idle` broadcasts into a + // running turn, which Codex refuses outright and Claude queues behind. + registerWorker() + installHost({ items: [runningTurn(), ...transcript(500)] }) + expect(structuredWorkerAgentStatus(SESSION_ID)).toBe('working') + }) + + it('answers idle only when no turn is running and no human is awaited', () => { + registerWorker() + installHost({ items: [idleTurn()] }) + expect(structuredWorkerAgentStatus(SESSION_ID)).toBe('idle') + installHost({ + items: [ + { + itemId: 'q1', + body: { + kind: 'question', + question: 'which?', + options: [], + resolution: { state: 'pending' } + } + } as unknown as AgentJournalRenderItem + ] + }) + expect(structuredWorkerAgentStatus(SESSION_ID)).toBe('attention') + }) + + it('answers null rather than idle when the session cannot be read', () => { + // Unknown must never read as idle, or `@idle` wakes a worker mid-turn. + hostRef.current = null + expect(structuredWorkerAgentStatus(SESSION_ID)).toBeNull() + }) +}) + +describe('sendGroupMessage actually composes structured workers in', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + /** + * Drives the real `sendGroupMessage`, not `resolveGroupAddress`. + * + * The suite above hand-composed `[PTY_TERMINAL, ...listAddressableStructuredWorkers()]` itself, + * so deleting the composition at the call site left it green — the exact regression the fix + * describes could come straight back. This test owns that seam. + */ + it('addresses a structured worker that only the call site can enumerate', async () => { + const handle = registerWorker() + installHost({}) + const inserted: { to: string }[] = [] + const db = { + getLegacyAdoptedRunMailboxOwner: () => null, + getCurrentRunForPane: () => undefined, + getActiveDispatchMailboxOwners: () => [], + getRunMailboxOwnerIdsForHandle: () => [], + insertMessages: (rows: { to: string }[]) => { + inserted.push(...rows) + return rows.map((row, index) => ({ id: `m${index}`, to_handle: row.to, type: 'status' })) + } + } + const runtime = { + // No PTY terminals at all: if the call site does not compose structured workers in, the + // group resolves empty and this throws instead of delivering. + listTerminals: async () => ({ terminals: [] }), + getAgentStatusForHandle: () => 'idle', + getLiveTerminalPaneKey: () => structuredWorkerIdentities.get(handle)!.paneKey, + notifyMessageArrived: () => {} + } + await sendGroupMessage({ + params: { subject: 's', body: 'b', type: 'status', priority: 'normal' }, + runtime: runtime as never, + db: db as never, + from: 'term_sender', + groupAddress: '@all', + senderPaneKey: undefined, + senderRunId: undefined, + explicitRunId: undefined, + legacyCoordinatorRunId: undefined, + revalidateLegacyCoordinator: undefined, + recordMutationReceipt: undefined, + withSendWarnings: (receipt) => receipt + } as never) + expect(inserted.map((row) => row.to)).toEqual([handle]) + }) +}) diff --git a/src/main/runtime/orchestration/structured-worker-group-addressing.ts b/src/main/runtime/orchestration/structured-worker-group-addressing.ts new file mode 100644 index 00000000000..8abad118de7 --- /dev/null +++ b/src/main/runtime/orchestration/structured-worker-group-addressing.ts @@ -0,0 +1,64 @@ +/** + * Structured workers as group-address recipients. + * + * `@all` and its siblings resolve recipients from `listTerminals`, which enumerates leaves and + * PTYs — so a structured worker was never a candidate. Worse, the exclusion happened BEFORE + * per-recipient resolution, so the `SendRecipientWarning` machinery never ran and the caller got + * exit 0 plus a receipt naming only the workers that did resolve. A broadcast "stop work" reached + * the PTY workers and silently missed the structured ones. + * + * Deliberately NOT solved by teaching `listTerminals` about structured sessions: that result is + * published to paired mobile and remote clients and to every consumer that assumes a summary has a + * `ptyId` or is writable, so it is its own change under + * `docs/reference/remote-wire-compatibility.md`. Group addressing needs three fields, and + * `RuntimeTerminalSummary` already satisfies them structurally — so the group resolver widens to + * the smaller shape instead, and nothing here has to invent a `worktreePath` or a `branch`. + */ + +import type { TuiAgent } from '../../../shared/tui-agent' +import { observeStructuredWorker, structuredWorkerAgent } from '../structured-worker-authority' +import { structuredWorkerIdentities } from '../structured-worker-identity' +import { readStructuredSessionGateFacts } from './structured-mailbox-pointer-host' + +/** The only facts group addressing reads off a recipient. */ +export type OrchestrationAddressableAgent = { + handle: string + worktreeId: string + /** Absent means "unknown", and `@claude`/`@codex` fail closed on it, exactly as for a pane. */ + agentIdentity?: TuiAgent +} + +/** + * Live structured workers of this runtime, as group-address candidates. + * + * Liveness-gated on the same observation the rest of the structured surface uses: a settled or + * handed-off worker is not a recipient, and addressing one would store mail no lane will deliver. + */ +export function listAddressableStructuredWorkers(): OrchestrationAddressableAgent[] { + return structuredWorkerIdentities + .list() + .filter((identity) => observeStructuredWorker(identity).status === 'live') + .map((identity) => ({ + handle: identity.handle, + worktreeId: identity.worktreeId, + agentIdentity: structuredWorkerAgent(identity) as TuiAgent + })) +} + +/** + * A structured worker's agent status, in the vocabulary `@idle` already matches on. + * + * Null when the session cannot be read: unknown must not read as idle, or a broadcast to `@idle` + * would wake a worker mid-turn — which Codex answers with `turn already running` and Claude queues + * behind the running turn. + */ +export function structuredWorkerAgentStatus(sessionId: string): string | null { + const facts = readStructuredSessionGateFacts(sessionId) + if (!facts) { + return null + } + if (facts.awaitingHuman) { + return 'attention' + } + return facts.turnRunning ? 'working' : 'idle' +} diff --git a/src/main/runtime/orchestration/structured-worker-journal-archive.test.ts b/src/main/runtime/orchestration/structured-worker-journal-archive.test.ts new file mode 100644 index 00000000000..3f529be726d --- /dev/null +++ b/src/main/runtime/orchestration/structured-worker-journal-archive.test.ts @@ -0,0 +1,78 @@ +import { describe, expect, it } from 'vitest' +import type { AgentJournalRenderItem } from '../../../shared/agent-session-journal-types' +import { buildStructuredJournalArchive } from './structured-worker-journal-archive' + +const STRUCTURED_ARCHIVE_MAX_BYTES = 262_144 + +/** A worker that actually did work: every turn is a full-width message, so the projected journal + * is several times the wire-size bound the forward transcript page uses. */ +function longJournal(count: number): AgentJournalRenderItem[] { + return Array.from( + { length: count }, + (_unused, index) => + ({ + itemId: `item-${index}`, + observedAt: index, + body: { + kind: 'message', + role: 'assistant', + blocks: Array.from({ length: 6 }, (_block, slot) => ({ + type: 'text', + text: `${index}:${slot}:${'x'.repeat(1_200)}` + })) + } + }) as unknown as AgentJournalRenderItem + ) +} + +function archive(items: AgentJournalRenderItem[], hasOlder = false) { + return buildStructuredJournalArchive({ + agent: 'claude', + processIncarnation: 'structured:session-1', + items, + hasOlder + }) +} + +describe('buildStructuredJournalArchive', () => { + it('keeps the worker final answer when the journal exceeds the bound', () => { + // The whole reason a released worker is read back. Bounding forward first kept the HEAD — the + // dispatch preamble and early exploration — and the newest-first cap then trimmed that head, + // so the answer was gone while the receipt said only the oldest messages had been dropped. + const items = longJournal(200) + const built = archive(items) + expect(built.limited).toBe(true) + expect(built.messages.at(-1)?.id).toBe('item-199') + expect(built.messages[0]?.id).not.toBe('item-0') + }) + + it('reports the end it actually dropped', () => { + const built = archive(longJournal(200)) + expect(built.warnings).toContain( + 'The oldest archived journal messages were dropped to fit the size bound.' + ) + expect(built.warnings).not.toContain('Transcript response was clipped to the wire-size limit.') + }) + + it('stays inside the durable bound', () => { + const built = archive(longJournal(200)) + expect(Buffer.byteLength(JSON.stringify(built.messages), 'utf8')).toBeLessThanOrEqual( + STRUCTURED_ARCHIVE_MAX_BYTES + ) + }) + + it('keeps a short journal whole and unflagged', () => { + const built = archive(longJournal(3)) + expect(built.limited).toBe(false) + expect(built.messages.map((message) => message.id)).toEqual(['item-0', 'item-1', 'item-2']) + expect(built.warnings).not.toContain( + 'The oldest archived journal messages were dropped to fit the size bound.' + ) + }) + + it('still reports omitted older items when the page itself was bounded', () => { + const built = archive(longJournal(2), true) + expect(built.limited).toBe(true) + expect(built.warnings).toContain('Older journal items were omitted from the bounded archive.') + }) +}) diff --git a/src/main/runtime/orchestration/structured-worker-journal-archive.ts b/src/main/runtime/orchestration/structured-worker-journal-archive.ts new file mode 100644 index 00000000000..4b196199d2d --- /dev/null +++ b/src/main/runtime/orchestration/structured-worker-journal-archive.ts @@ -0,0 +1,70 @@ +/** + * Freezing and re-reading a structured worker's journal. + * + * The terminal path archives a redacted PTY tail; there is no PTY here, so the durable evidence is + * the journal projected into the same message shape `worker-read --source transcript` already + * serves. It gets its own archive kind because its identity is a session, not a transcript file on + * disk, and because the read side must be able to say which of the three it is holding. + */ + +import type { AgentType, NativeChatMessage } from '../../../shared/native-chat-types' +import { projectStructuredItemsToNativeChat } from '../../../shared/structured-agent-session-projection' +import type { AgentJournalRenderItem } from '../../../shared/agent-session-journal-types' +import { boundWorkerTranscriptTail } from './worker-transcript-payload' + +// Same durable bound the terminal archive uses; a session journal can grow without limit. +const STRUCTURED_ARCHIVE_MAX_BYTES = 262_144 + +export type WorkerStructuredJournalArchive = { + version: 1 + agent: AgentType + processIncarnation: string + messages: NativeChatMessage[] + limited: boolean + warnings: string[] +} + +/** + * Project a journal page into messages and bound it NEWEST-first. + * + * One newest-first pass, never the forward wire bound first: that one keeps the HEAD, so a long + * worker's archive ended at its early exploration and dropped the answer it was released for — + * under a warning that said the OLDEST messages had gone. The same reasoning holds for any reader + * that wants a worker's RECENT output, which is why this is shared rather than inlined below. + * + * Redacts dispatch capabilities and clips oversized blocks, exactly as the transcript path does. + */ +export function boundStructuredJournalTail(items: readonly AgentJournalRenderItem[]): { + messages: NativeChatMessage[] + limited: boolean + warnings: string[] +} { + return boundWorkerTranscriptTail( + projectStructuredItemsToNativeChat(items), + STRUCTURED_ARCHIVE_MAX_BYTES + ) +} + +export function buildStructuredJournalArchive(input: { + agent: AgentType + processIncarnation: string + items: readonly AgentJournalRenderItem[] + hasOlder: boolean +}): WorkerStructuredJournalArchive { + const bounded = boundStructuredJournalTail(input.items) + const warnings = [...bounded.warnings] + if (input.hasOlder) { + warnings.push('Older journal items were omitted from the bounded archive.') + } + if (bounded.limited) { + warnings.push('The oldest archived journal messages were dropped to fit the size bound.') + } + return { + version: 1, + agent: input.agent, + processIncarnation: input.processIncarnation, + messages: bounded.messages, + limited: bounded.limited || input.hasOlder, + warnings + } +} diff --git a/src/main/runtime/orchestration/structured-worker-journal-page.ts b/src/main/runtime/orchestration/structured-worker-journal-page.ts new file mode 100644 index 00000000000..35676aaf52c --- /dev/null +++ b/src/main/runtime/orchestration/structured-worker-journal-page.ts @@ -0,0 +1,35 @@ +/** + * The one tail read every structured-journal reader shares. + * + * `worker-read`, the release archive and `terminal read` all want the same thing — the newest page + * of a session's reduced timeline, and `null` rather than a throw when the session is not attached. + * It lives here so none of them can drift onto a different page size or a different failure shape. + */ + +import type { AgentJournalRenderItem } from '../../../shared/agent-session-journal-types' +import { getStructuredAgentSessionHost } from '../../native-chat/agent-session-wire/structured-agent-session-registry' + +export const STRUCTURED_JOURNAL_PAGE_LIMIT = 200 + +export type StructuredJournalPage = { + items: readonly AgentJournalRenderItem[] + hasOlder: boolean +} + +/** The newest page of a session's journal, or null when this runtime cannot read it. */ +export function readStructuredJournalPage(sessionId: string): StructuredJournalPage | null { + const host = getStructuredAgentSessionHost() + if (!host) { + return null + } + try { + const result = host.history({ + sessionId, + direction: 'tail', + limit: STRUCTURED_JOURNAL_PAGE_LIMIT + }) + return { items: result.page.items, hasOlder: result.page.hasOlder } + } catch { + return null + } +} diff --git a/src/main/runtime/orchestration/worker-output-archive.ts b/src/main/runtime/orchestration/worker-output-archive.ts index 092fd03d34a..c1b93371bbe 100644 --- a/src/main/runtime/orchestration/worker-output-archive.ts +++ b/src/main/runtime/orchestration/worker-output-archive.ts @@ -3,6 +3,7 @@ import type { OrchestrationWorkerReadFallbackReason } from '../../../shared/orch import type { OrcaRuntimeService } from '../orca-runtime' import { OrchestrationError } from './orchestration-error' import type { + WorkerTerminalArchiveKind, WorkerTerminalArchiveRow, WorkerTerminalArchiveStatus } from './worker-terminal-ownership' @@ -13,6 +14,10 @@ import { import { readWorkerTranscript } from './worker-transcript-read' import { getSshFilesystemProvider } from '../../providers/ssh-filesystem-dispatch' import { isWslHookRelayConnectionId } from '../../../shared/wsl-hook-relay-contract' +import { captureStructuredWorkerArchive } from '../rpc/methods/orchestration-structured-worker-lifecycle' +import type { WorkerStructuredJournalArchive } from './structured-worker-journal-archive' +import { structuredWorkerAgent } from '../structured-worker-authority' +import type { StructuredWorkerIdentity } from '../structured-worker-identity' // Bound the durable copy of raw terminal output; the tail end is the evidence that matters. const TERMINAL_ARCHIVE_MAX_CHARS = 262_144 @@ -44,6 +49,11 @@ export type WorkerOutputArchiveCapture = content: WorkerTranscriptSnapshotArchive status: 'captured' } + | { + kind: 'structured_journal' + content: WorkerStructuredJournalArchive + status: 'captured' | 'empty' + } | { kind: 'terminal_tail'; content: WorkerTerminalTailArchive; status: 'captured' | 'empty' } export function summarizeWorkerOutputArchive(archive: WorkerTerminalArchiveRow): { @@ -53,6 +63,13 @@ export function summarizeWorkerOutputArchive(archive: WorkerTerminalArchiveRow): if (archive.kind === 'transcript_pin') { return { source: 'transcript', status: 'captured' } } + if (archive.kind === 'structured_journal') { + // A structured session's journal IS its transcript, so it reports as one. A third `source` + // would leak the structured/terminal split into a CLI surface this PR keeps deliberately + // uniform, and would widen a shape that already reaches paired clients. + const journal = JSON.parse(archive.content) as WorkerStructuredJournalArchive + return { source: 'transcript', status: journal.messages.length > 0 ? 'captured' : 'empty' } + } const content = JSON.parse(archive.content) as WorkerTerminalTailArchive const empty = content.lines.every((line) => line.trim() === '') && (content.draft?.trim() ?? '') === '' @@ -70,7 +87,20 @@ export async function captureWorkerOutputArchive(args: { dispatchId: string terminalHandle: string attachedAtMs: number + /** Present when the worker IS a structured session; its journal is the only output it has. */ + structuredWorker?: StructuredWorkerIdentity | null }): Promise { + if (args.structuredWorker) { + const content = captureStructuredWorkerArchive( + args.structuredWorker, + structuredWorkerAgent(args.structuredWorker) + ) + return { + kind: 'structured_journal', + status: content.messages.length > 0 ? 'captured' : 'empty', + content + } + } const session = args.runtime.getExactWorkerProviderSession(args.terminalHandle, args.attachedAtMs) let transcriptFallbackReason: OrchestrationWorkerReadFallbackReason = 'session_not_reported' if (session) { @@ -182,3 +212,10 @@ export function boundArchiveLines(lines: string[]): { lines: string[]; truncated keptReversed.reverse() return { lines: keptReversed, truncated: true } } + +/** Errors at compile time if a capture kind is ever added that the durable row cannot store. */ +type AssertAssignable = TValue +export type WorkerOutputArchiveCaptureKind = AssertAssignable< + WorkerOutputArchiveCapture['kind'], + WorkerTerminalArchiveKind +> diff --git a/src/main/runtime/orchestration/worker-terminal-ownership.ts b/src/main/runtime/orchestration/worker-terminal-ownership.ts index 096a9b7bf22..7e613050aa2 100644 --- a/src/main/runtime/orchestration/worker-terminal-ownership.ts +++ b/src/main/runtime/orchestration/worker-terminal-ownership.ts @@ -64,10 +64,19 @@ export type WorkerTerminalListState = export type WorkerDispatchListState = WorkerDispatchState | 'unsupervised' +/** + * The frozen output sources a released worker can be read back from. + * + * One name so widening it stays a single edit: the capture, the durable write, and the archived + * read all have to admit the same set, and a kind that reaches the row but not the read side is an + * archived worker that throws instead of answering. + */ +export type WorkerTerminalArchiveKind = 'transcript_pin' | 'terminal_tail' | 'structured_journal' + export type WorkerTerminalArchiveRow = { dispatch_id: string resource_id: string - kind: 'transcript_pin' | 'terminal_tail' + kind: WorkerTerminalArchiveKind content: string created_at: string } diff --git a/src/main/runtime/orchestration/worker-transcript-payload.ts b/src/main/runtime/orchestration/worker-transcript-payload.ts index d8b5e500503..9200f96cb14 100644 --- a/src/main/runtime/orchestration/worker-transcript-payload.ts +++ b/src/main/runtime/orchestration/worker-transcript-payload.ts @@ -64,6 +64,35 @@ export function boundWorkerTranscriptMessages( return { messages: bounded, limited: state.clipped, warnings: [...state.warnings] } } +/** + * The same per-message bounding, accumulated NEWEST-first. + * + * `boundWorkerTranscriptMessages` keeps the head, which is right for a forward page and wrong for + * an archive: the evidence anyone reads a released worker back for is its final answer, so the + * tail is what must survive the budget. + */ +export function boundWorkerTranscriptTail( + messages: readonly NativeChatMessage[], + maxBytes: number +): { messages: NativeChatMessage[]; limited: boolean; warnings: string[] } { + const state: TranscriptBoundState = { warnings: new Set(), clipped: false } + const keptReversed: NativeChatMessage[] = [] + let bytes = 2 + let limited = false + for (let index = messages.length - 1; index >= 0; index -= 1) { + const next = boundMessage(messages[index]!, undefined, state) + const serializedBytes = Buffer.byteLength(JSON.stringify(next), 'utf8') + 1 + if (keptReversed.length > 0 && bytes + serializedBytes > maxBytes) { + limited = true + break + } + keptReversed.push(next) + bytes += serializedBytes + } + keptReversed.reverse() + return { messages: keptReversed, limited, warnings: [...state.warnings] } +} + function boundMessage( message: NativeChatMessage, transcriptPath: string | undefined, diff --git a/src/main/runtime/push/desktop-push-service.test.ts b/src/main/runtime/push/desktop-push-service.test.ts new file mode 100644 index 00000000000..9177bcbc18f --- /dev/null +++ b/src/main/runtime/push/desktop-push-service.test.ts @@ -0,0 +1,294 @@ +import { mkdtempSync } from 'node:fs' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { describe, expect, it, vi } from 'vitest' +import type { MobileNotificationEvent } from '../runtime-mobile-notification-controller' +import { DeviceRegistry } from '../device-registry' +import { DesktopPushService } from './desktop-push-service' +import { PushRegisterThrottle } from './push-register-throttle' +import { PushUnregisterOutbox } from './push-unregister-outbox' +import { createPushHostKeypair } from './push-host-challenge-fixtures' + +const REGISTER_INPUT = { + platform: 'android' as const, + token: 'fcm-token', + filter: { sources: ['agent-task-complete'] as const, agentStates: ['finished'] as const } +} + +function createService( + options: { + registerFails?: boolean + deleteFails?: boolean + /** Runs before each delete resolves, so a suite can queue work mid-flush. */ + onDelete?: (registrationId: string) => void + now?: () => number + } = {} +): { + service: DesktopPushService + registry: DeviceRegistry + outbox: PushUnregisterOutbox + deviceId: string + deletes: string[] + send: ReturnType + dispatch: (event: MobileNotificationEvent) => void + retries: { run: () => void; delayMs: number }[] +} { + const userDataPath = mkdtempSync(join(tmpdir(), 'orca-push-service-')) + const registry = new DeviceRegistry(userDataPath) + const outbox = new PushUnregisterOutbox(userDataPath) + const device = registry.addDevice('phone', 'mobile') + const deletes: string[] = [] + let listener: ((event: MobileNotificationEvent) => void) | null = null + + const runtime = { + setMobilePushRegistrar: vi.fn(), + onNotificationDispatched: vi.fn((next: (event: MobileNotificationEvent) => void) => { + listener = next + return () => { + listener = null + } + }) + } + const runtimeRpc = { + getE2EEKeypair: () => createPushHostKeypair(), + getDeviceRegistry: () => registry, + getPushUnregisterOutbox: () => outbox, + setOnPushUnregisterQueued: vi.fn() + } + // A stub gateway keeps the suite on the service's own persistence decisions. + const client = { + registerDevice: vi.fn(async () => + options.registerFails + ? ({ ok: false, reason: 'unreachable' } as const) + : ({ ok: true, registrationId: 'reg-1' } as const) + ), + deleteDevice: vi.fn(async (registrationId: string) => { + deletes.push(registrationId) + options.onDelete?.(registrationId) + return options.deleteFails + ? { deleted: false, retryable: true } + : { deleted: true, retryable: false } + }), + send: vi.fn(async () => ({ ok: true, results: [] }) as const) + } + const retries: { run: () => void; delayMs: number }[] = [] + const service = DesktopPushService.create({ + runtime: runtime as never, + runtimeRpc: runtimeRpc as never, + gatewayUrl: 'https://push.onorca.dev', + client: client as never, + scheduleRetry: (run, delayMs) => { + retries.push({ run, delayMs }) + }, + ...(options.now ? { registerThrottle: new PushRegisterThrottle({ now: options.now }) } : {}) + })! + + service.start() + return { + service, + registry, + outbox, + deviceId: device.deviceId, + deletes, + send: client.send, + dispatch: (event) => listener?.(event), + retries + } +} + +describe('DesktopPushService', () => { + it('persists the registration the gateway hands back', async () => { + const harness = createService() + + expect( + await harness.service.register({ deviceId: harness.deviceId, ...REGISTER_INPUT }) + ).toEqual({ registered: true, registrationId: 'reg-1' }) + expect(harness.registry.getDevice(harness.deviceId)?.pushRegistration).toMatchObject({ + registrationId: 'reg-1', + platform: 'android', + filter: REGISTER_INPUT.filter + }) + }) + + it('persists nothing when the gateway is unreachable', async () => { + const harness = createService({ registerFails: true }) + + expect( + await harness.service.register({ deviceId: harness.deviceId, ...REGISTER_INPUT }) + ).toEqual({ registered: false, reason: 'gateway_unreachable' }) + expect(harness.registry.getDevice(harness.deviceId)?.pushRegistration).toBeUndefined() + }) + + it('refuses to register a device that is not a paired phone', async () => { + const harness = createService() + + expect(await harness.service.register({ deviceId: 'not-a-device', ...REGISTER_INPUT })).toEqual( + { + registered: false, + reason: 'not_mobile' + } + ) + }) + + it('clears the local registration and deletes at the gateway on unregister', async () => { + const harness = createService() + await harness.service.register({ deviceId: harness.deviceId, ...REGISTER_INPUT }) + + expect(await harness.service.unregister(harness.deviceId)).toEqual({ unregistered: true }) + await harness.service.flushUnregisterOutbox() + expect(harness.registry.getDevice(harness.deviceId)?.pushRegistration).toBeUndefined() + expect(harness.deletes).toEqual(['reg-1']) + expect(harness.outbox.pending()).toEqual([]) + }) + + it('keeps the delete queued when the gateway cannot be reached', async () => { + const harness = createService({ deleteFails: true }) + await harness.service.register({ deviceId: harness.deviceId, ...REGISTER_INPUT }) + + await harness.service.unregister(harness.deviceId) + + expect(harness.registry.getDevice(harness.deviceId)?.pushRegistration).toBeUndefined() + expect(harness.outbox.pending()).toEqual([ + expect.objectContaining({ registrationId: 'reg-1', deviceId: harness.deviceId }) + ]) + }) + + it('reports nothing to unregister for a device that never enabled push', async () => { + const harness = createService() + expect(await harness.service.unregister(harness.deviceId)).toEqual({ unregistered: false }) + }) + + it('drains a delete queued before this launch', async () => { + const harness = createService() + harness.outbox.enqueue({ registrationId: 'reg-stale', deviceId: 'device-gone' }) + + await harness.service.flushUnregisterOutbox() + + expect(harness.deletes).toEqual(['reg-stale']) + expect(harness.outbox.pending()).toEqual([]) + }) + + it('unregisters at the gateway when the device stopped being a phone mid-register', async () => { + const harness = createService() + vi.spyOn(harness.registry, 'setPushRegistration').mockReturnValue(false) + + expect( + await harness.service.register({ deviceId: harness.deviceId, ...REGISTER_INPUT }) + ).toEqual({ registered: false, reason: 'not_mobile' }) + // register() kicks the flush off without awaiting it; join the same run. + await harness.service.flushUnregisterOutbox() + expect(harness.deletes).toEqual(['reg-1']) + expect(harness.outbox.pending()).toEqual([]) + }) + + it('unregisters at the gateway when the registration cannot be written', async () => { + const harness = createService({ deleteFails: true }) + vi.spyOn(harness.registry, 'setPushRegistration').mockImplementation(() => { + throw new Error('disk full') + }) + const warn = vi.spyOn(console, 'warn').mockImplementation(() => {}) + + expect( + await harness.service.register({ deviceId: harness.deviceId, ...REGISTER_INPUT }) + ).toEqual({ registered: false, reason: 'registration_storage_failed' }) + // The gateway kept the token, so the delete stays queued until it lands. + expect(harness.outbox.pending()).toEqual([ + expect.objectContaining({ registrationId: 'reg-1', deviceId: harness.deviceId }) + ]) + warn.mockRestore() + }) + + it('drains a delete queued while a flush is already running', async () => { + let queued = false + const harness = createService({ + onDelete: () => { + if (queued) { + return + } + queued = true + harness.outbox.enqueue({ registrationId: 'reg-late', deviceId: 'device-late' }) + // Mirrors unregister(): the trigger arrives while the flush is mid-await. + void harness.service.flushUnregisterOutbox() + } + }) + harness.outbox.enqueue({ registrationId: 'reg-first', deviceId: 'device-first' }) + + await harness.service.flushUnregisterOutbox() + + expect(harness.deletes).toEqual(['reg-first', 'reg-late']) + expect(harness.outbox.pending()).toEqual([]) + }) + + it('retries a failed drain on a capped backoff instead of waiting for a relaunch', async () => { + const harness = createService({ deleteFails: true }) + harness.outbox.enqueue({ registrationId: 'reg-stuck', deviceId: 'device-1' }) + + await harness.service.flushUnregisterOutbox() + expect(harness.retries.map((entry) => entry.delayMs)).toEqual([30_000]) + + harness.retries[0]?.run() + await new Promise((resolve) => setImmediate(resolve)) + expect(harness.deletes).toEqual(['reg-stuck', 'reg-stuck']) + expect(harness.retries.map((entry) => entry.delayMs)).toEqual([30_000, 60_000]) + expect(harness.outbox.pending()).toHaveLength(1) + }) + + it('stops re-arming the retry once the service is stopped', async () => { + const harness = createService({ deleteFails: true }) + harness.outbox.enqueue({ registrationId: 'reg-stuck', deviceId: 'device-1' }) + await harness.service.flushUnregisterOutbox() + + harness.service.stop() + harness.retries[0]?.run() + await new Promise((resolve) => setImmediate(resolve)) + + expect(harness.retries).toHaveLength(1) + }) + + it('throttles a device that registers in a loop and lets it back in a minute later', async () => { + let clock = 1_700_000_000_000 + const harness = createService({ now: () => clock }) + const input = { deviceId: harness.deviceId, ...REGISTER_INPUT } + + for (let index = 0; index < 10; index++) { + expect(await harness.service.register(input)).toEqual({ + registered: true, + registrationId: 'reg-1' + }) + } + expect(await harness.service.register(input)).toEqual({ + registered: false, + reason: 'throttled' + }) + // The registration it already made stands; only the new write is refused. + expect(harness.registry.getDevice(harness.deviceId)?.pushRegistration?.registrationId).toBe( + 'reg-1' + ) + + clock += 60_000 + expect(await harness.service.register(input)).toEqual({ + registered: true, + registrationId: 'reg-1' + }) + }) + + it('pushes a dispatched notification through the subscribed dispatcher', async () => { + const harness = createService() + await harness.service.register({ deviceId: harness.deviceId, ...REGISTER_INPUT }) + + harness.dispatch({ + type: 'notification', + source: 'agent-task-complete', + title: 'feat/x - Claude finished', + body: 'Done.', + notificationSeq: 3, + notificationEpoch: 'epoch-1', + agentState: 'done' + }) + await new Promise((resolve) => setImmediate(resolve)) + + expect(harness.send).toHaveBeenCalledWith( + expect.objectContaining({ registrationIds: ['reg-1'] }) + ) + }) +}) diff --git a/src/main/runtime/push/desktop-push-service.ts b/src/main/runtime/push/desktop-push-service.ts new file mode 100644 index 00000000000..a459798625a --- /dev/null +++ b/src/main/runtime/push/desktop-push-service.ts @@ -0,0 +1,267 @@ +// Why: owns the desktop half of background push — the gateway session, the +// registration each paired phone asked for, and the durable delete queue. Built +// alongside DesktopRelayService but deliberately not gated on cloud sign-in: the +// gateway authenticates with the host keypair, so accountless hosts push too. +import type { + MobilePushRegisterInput, + MobilePushRegisterResult +} from '../../../shared/mobile-push-contract' +import { runKeyedSerializedOperation } from '../../cli/keyed-promise-queue' +import type { DeviceRegistry } from '../device-registry' +import type { OrcaRuntimeService } from '../orca-runtime' +import type { OrcaRuntimeRpcServer } from '../runtime-rpc' +import { PushDispatcher } from './push-dispatcher' +import { PushGatewayClient } from './push-gateway-client' +import { PushRegisterThrottle } from './push-register-throttle' +import type { PushUnregisterOutbox } from './push-unregister-outbox' + +const OUTBOX_RETRY_BASE_MS = 30_000 +const OUTBOX_RETRY_MAX_MS = 10 * 60_000 + +type RegisterStorageFailure = 'not_mobile' | 'registration_storage_failed' + +type DesktopPushServiceOptions = { + runtime: OrcaRuntimeService + runtimeRpc: OrcaRuntimeRpcServer + gatewayUrl: string + /** Test seam: lets a suite drive the service without a live gateway. */ + client?: PushGatewayClient + /** Test seam: lets a suite drive the outbox backoff without real timers. */ + scheduleRetry?: (run: () => void, delayMs: number) => void + /** Test seam: lets a suite drive the per-device register bucket on its own clock. */ + registerThrottle?: PushRegisterThrottle +} + +export class DesktopPushService { + private readonly runtime: OrcaRuntimeService + private readonly runtimeRpc: OrcaRuntimeRpcServer + private readonly registry: DeviceRegistry + private readonly outbox: PushUnregisterOutbox + private readonly client: PushGatewayClient + private readonly dispatcher: PushDispatcher + private readonly registerThrottle: PushRegisterThrottle + private readonly scheduleRetry: (run: () => void, delayMs: number) => void + private unsubscribe: (() => void) | null = null + private flushLoop: Promise | null = null + private flushRequested = false + private retryArmed = false + private retryDelayMs = OUTBOX_RETRY_BASE_MS + private stopped = false + private readonly deviceOperations = new Map>() + + private constructor( + options: DesktopPushServiceOptions, + registry: DeviceRegistry, + client: PushGatewayClient + ) { + this.runtime = options.runtime + this.runtimeRpc = options.runtimeRpc + this.registry = registry + this.client = client + this.outbox = options.runtimeRpc.getPushUnregisterOutbox() + this.dispatcher = new PushDispatcher({ client, registry }) + this.registerThrottle = options.registerThrottle ?? new PushRegisterThrottle() + this.scheduleRetry = + options.scheduleRetry ?? + ((run, delayMs) => { + // Why: a queued gateway delete must never hold the app open at quit. + setTimeout(run, delayMs).unref?.() + }) + } + + /** Returns null when the mobile runtime never came up, so there is nothing to push for. */ + static create(options: DesktopPushServiceOptions): DesktopPushService | null { + const keypair = options.runtimeRpc.getE2EEKeypair() + const registry = options.runtimeRpc.getDeviceRegistry() + if (!keypair || !registry) { + return null + } + const client = + options.client ?? new PushGatewayClient({ gatewayUrl: options.gatewayUrl, keypair }) + return new DesktopPushService(options, registry, client) + } + + start(): void { + this.stopped = false + this.dispatcher.start() + this.runtime.setMobilePushRegistrar(this) + this.unsubscribe = this.runtime.onNotificationDispatched((event) => { + this.dispatcher.enqueue(event) + }) + // Unpairing queues a delete without going through this service; drain on that too. + this.runtimeRpc.setOnPushUnregisterQueued(() => { + void this.flushUnregisterOutbox() + }) + // Deletes queued while the gateway was unreachable — including across restarts. + void this.flushUnregisterOutbox() + } + + stop(): void { + this.stopped = true + this.dispatcher.stop() + this.unsubscribe?.() + this.unsubscribe = null + this.runtimeRpc.setOnPushUnregisterQueued(null) + this.runtime.setMobilePushRegistrar(null) + } + + async register(input: MobilePushRegisterInput): Promise { + if (this.registry.getDevice(input.deviceId)?.scope !== 'mobile') { + return { registered: false, reason: 'not_mobile' } + } + // Unregister needs no bucket: with nothing registered it is a lookup, and + // with something registered it can only run once per successful register. + if (!this.registerThrottle.allow(input.deviceId)) { + return { registered: false, reason: 'throttled' } + } + return runKeyedSerializedOperation(this.deviceOperations, input.deviceId, () => + this.registerAfterCleanup(input) + ) + } + + private async registerAfterCleanup( + input: MobilePushRegisterInput + ): Promise { + // A stable gateway ID must not inherit a delete from an earlier registration. + for (const item of this.outbox.pending().filter((entry) => entry.deviceId === input.deviceId)) { + if (!(await this.deleteQueued(item.reqId, item.registrationId))) { + this.scheduleFlushRetry() + return { registered: false, reason: 'gateway_unreachable' } + } + } + if (this.registry.getDevice(input.deviceId)?.scope !== 'mobile' || this.stopped) { + return { registered: false, reason: 'not_mobile' } + } + const result = await this.client.registerDevice(input) + if (!result.ok) { + return { + registered: false, + reason: result.reason === 'unreachable' ? 'gateway_unreachable' : 'gateway_rejected' + } + } + const failure = this.storeRegistration(input, result.registrationId) + if (failure) { + // Why: the gateway now holds a token this host will never push to. Queue its + // delete instead of leaking it until the phone happens to register again. + this.outbox.enqueue({ registrationId: result.registrationId, deviceId: input.deviceId }) + } + void this.flushUnregisterOutbox() + return failure + ? { registered: false, reason: failure } + : { registered: true, registrationId: result.registrationId } + } + + async unregister(deviceId: string): Promise<{ unregistered: boolean }> { + return runKeyedSerializedOperation(this.deviceOperations, deviceId, async () => + this.unregisterCurrent(deviceId) + ) + } + + private unregisterCurrent(deviceId: string): { unregistered: boolean } { + const registrationId = this.registry.getDevice(deviceId)?.pushRegistration?.registrationId + if (!registrationId) { + return { unregistered: false } + } + // Persist cleanup before forgetting its ID; neither write waits on the gateway. + this.outbox.enqueue({ registrationId, deviceId }) + this.registry.setPushRegistration(deviceId, null) + void this.flushUnregisterOutbox() + return { unregistered: true } + } + + /** Joining an in-flight drain still waits for the item this call queued. */ + async flushUnregisterOutbox(): Promise { + this.flushRequested = true + this.flushLoop ??= this.runFlushLoop().finally(() => { + this.flushLoop = null + }) + await this.flushLoop + } + + private async runFlushLoop(): Promise { + while (this.flushRequested && !this.stopped) { + // Cleared before the pass, so a delete queued mid-drain earns another one. + this.flushRequested = false + if (await this.drainPending()) { + this.scheduleFlushRetry() + } else { + this.retryDelayMs = OUTBOX_RETRY_BASE_MS + } + } + } + + /** Returns the refusal reason when a gateway-accepted registration cannot be stored. */ + private storeRegistration( + input: MobilePushRegisterInput, + registrationId: string + ): RegisterStorageFailure | null { + try { + const stored = this.registry.setPushRegistration(input.deviceId, { + registrationId, + platform: input.platform, + filter: input.filter, + registeredAt: Date.now() + }) + // False means the device was removed or left mobile scope while the gateway + // call was in flight. + return stored ? null : 'not_mobile' + } catch (error) { + console.warn('[push] Failed to persist a push registration:', error) + return 'registration_storage_failed' + } + } + + /** Returns true when the pass left behind an item the gateway may still accept. */ + private async drainPending(): Promise { + const attempted = new Set() + let retryable = false + for (;;) { + // Re-read per item: a snapshot taken at loop entry misses anything queued + // while an await was in flight, and the outbox swaps arrays on every write. + const item = this.outbox.pending().find((candidate) => !attempted.has(candidate.reqId)) + if (!item) { + return retryable + } + attempted.add(item.reqId) + try { + const deleted = await runKeyedSerializedOperation( + this.deviceOperations, + item.deviceId, + () => this.deleteQueued(item.reqId, item.registrationId) + ) + if (!deleted) { + retryable = true + } + } catch (error) { + // One bad delete must not strand the rest of the queue. + console.warn('[push] Failed to drain the push unregister outbox:', error) + retryable = true + } + } + } + + private async deleteQueued(reqId: string, registrationId: string): Promise { + if (!this.outbox.pending().some((item) => item.reqId === reqId)) { + return true + } + const result = await this.client.deleteDevice(registrationId) + if (!result.deleted) { + return false + } + this.outbox.remove(reqId) + return true + } + + private scheduleFlushRetry(): void { + if (this.retryArmed || this.stopped) { + return + } + this.retryArmed = true + const delayMs = this.retryDelayMs + this.retryDelayMs = Math.min(delayMs * 2, OUTBOX_RETRY_MAX_MS) + this.scheduleRetry(() => { + this.retryArmed = false + void this.flushUnregisterOutbox() + }, delayMs) + } +} diff --git a/src/main/runtime/push/push-agent-state.test.ts b/src/main/runtime/push/push-agent-state.test.ts new file mode 100644 index 00000000000..e56d39ffb01 --- /dev/null +++ b/src/main/runtime/push/push-agent-state.test.ts @@ -0,0 +1,21 @@ +import { describe, expect, it } from 'vitest' +import { mapPushAgentState } from './push-dispatcher' + +describe('mapPushAgentState', () => { + it.each([ + ['blocked', 'needs-input'], + ['waiting', 'needs-input'], + ['done', 'finished'], + [undefined, 'finished'] + ] as const)('maps agent-task-complete %s to %s', (agentState, expected) => { + expect(mapPushAgentState('agent-task-complete', agentState)).toBe(expected) + }) + + it('suppresses a still-working agent', () => { + expect(mapPushAgentState('agent-task-complete', 'working')).toBeUndefined() + }) + + it('leaves non-agent sources without a state', () => { + expect(mapPushAgentState('terminal-bell', undefined)).toBeNull() + }) +}) diff --git a/src/main/runtime/push/push-cleanup-auth-expiry.test.ts b/src/main/runtime/push/push-cleanup-auth-expiry.test.ts new file mode 100644 index 00000000000..b746a03a002 --- /dev/null +++ b/src/main/runtime/push/push-cleanup-auth-expiry.test.ts @@ -0,0 +1,41 @@ +import { createHash } from 'node:crypto' +import { expect, it } from 'vitest' +import { PushGatewayClient } from './push-gateway-client' +import { buildPushChallengeFixture, createPushHostKeypair } from './push-host-challenge-fixtures' + +it('retains a delete when its session proof expires before the DELETE is attempted', async () => { + const keypair = createPushHostKeypair() + const hostFingerprint = createHash('sha256') + .update(keypair.publicKey) + .digest('base64url') + .slice(0, 16) + let now = 1_770_000_000_000 + let deletes = 0 + const client = new PushGatewayClient({ + gatewayUrl: 'https://push.example.test', + keypair, + now: () => now, + fetch: (async (url, init) => { + if (String(url).endsWith('/challenge')) { + const fixture = buildPushChallengeFixture({ + hostKeypair: keypair, + hostFingerprint, + gatewayOrigin: 'https://push.example.test', + issuedAt: now, + challengeId: 'challenge-1' + }) + now += 11_000 + return Response.json(fixture.challenge) + } + if (String(url).endsWith('/session')) { + return Response.json({ error: 'invalid_proof' }, { status: 401 }) + } + if (init?.method === 'DELETE') { + deletes++ + } + return new Response(null, { status: 204 }) + }) as typeof fetch + }) + expect(await client.deleteDevice('registration-1')).toEqual({ deleted: false, retryable: true }) + expect(deletes).toBe(0) +}) diff --git a/src/main/runtime/push/push-device-registration-persistence.test.ts b/src/main/runtime/push/push-device-registration-persistence.test.ts new file mode 100644 index 00000000000..43a7dc5266a --- /dev/null +++ b/src/main/runtime/push/push-device-registration-persistence.test.ts @@ -0,0 +1,106 @@ +import { mkdtempSync, readFileSync, writeFileSync } from 'node:fs' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { describe, expect, it } from 'vitest' +import { DeviceRegistry } from '../device-registry' +import { DEVICE_REGISTRY_FILENAME } from '../mobile-pairing-files' +import type { MobilePushRegistration } from '../../../shared/mobile-push-contract' + +const REGISTRATION: MobilePushRegistration = { + registrationId: 'reg-1', + platform: 'ios', + filter: { sources: ['agent-task-complete'], agentStates: ['needs-input', 'finished'] }, + registeredAt: 1_770_000_000_000 +} + +function userDataDir(): string { + return mkdtempSync(join(tmpdir(), 'orca-push-registry-')) +} + +function rewriteRegistry(dir: string, mutate: (devices: Record[]) => void): void { + const path = join(dir, DEVICE_REGISTRY_FILENAME) + const devices: Record[] = JSON.parse(readFileSync(path, 'utf-8')) + mutate(devices) + writeFileSync(path, JSON.stringify(devices)) +} + +describe('DeviceRegistry push registrations', () => { + it('persists a registration across a restart', () => { + const dir = userDataDir() + const device = new DeviceRegistry(dir).addDevice('phone', 'mobile') + expect(new DeviceRegistry(dir).setPushRegistration(device.deviceId, REGISTRATION)).toBe(true) + + expect(new DeviceRegistry(dir).getDevice(device.deviceId)?.pushRegistration).toEqual( + REGISTRATION + ) + }) + + it('clears a registration when the gateway reports the token dead', () => { + const dir = userDataDir() + const registry = new DeviceRegistry(dir) + const device = registry.addDevice('phone', 'mobile') + registry.setPushRegistration(device.deviceId, REGISTRATION) + + expect(registry.setPushRegistration(device.deviceId, null)).toBe(true) + expect(new DeviceRegistry(dir).getDevice(device.deviceId)?.pushRegistration).toBeUndefined() + }) + + it('refuses to register a runtime-scoped device', () => { + const dir = userDataDir() + const registry = new DeviceRegistry(dir) + const cli = registry.addDevice('cli', 'runtime') + + expect(registry.setPushRegistration(cli.deviceId, REGISTRATION)).toBe(false) + }) + + it('loads a registry written before push existed', () => { + const dir = userDataDir() + const device = new DeviceRegistry(dir).addDevice('phone', 'mobile') + rewriteRegistry(dir, (devices) => { + for (const entry of devices) { + delete entry.pushRegistration + } + }) + + const reloaded = new DeviceRegistry(dir) + expect(reloaded.listDevices()).toHaveLength(1) + expect(reloaded.getDevice(device.deviceId)?.pushRegistration).toBeUndefined() + }) + + it.each([ + ['a malformed registration', { registrationId: 'reg-1' }], + ['an unknown platform', { ...REGISTRATION, platform: 'windows-phone' }], + ['a missing filter', { ...REGISTRATION, filter: undefined }], + ['a non-object', 'nonsense'] + ])('keeps the device but drops %s', (_name, pushRegistration) => { + const dir = userDataDir() + const device = new DeviceRegistry(dir).addDevice('phone', 'mobile') + rewriteRegistry(dir, (devices) => { + for (const entry of devices) { + entry.pushRegistration = pushRegistration + } + }) + + const reloaded = new DeviceRegistry(dir) + expect(reloaded.listDevices()).toHaveLength(1) + expect(reloaded.getDevice(device.deviceId)?.pushRegistration).toBeUndefined() + }) + + it('drops only the unknown members of a stored filter', () => { + const dir = userDataDir() + const device = new DeviceRegistry(dir).addDevice('phone', 'mobile') + rewriteRegistry(dir, (devices) => { + for (const entry of devices) { + entry.pushRegistration = { + ...REGISTRATION, + filter: { sources: ['agent-task-complete', 'smoke-signal'], agentStates: ['finished'] } + } + } + }) + + expect(new DeviceRegistry(dir).getDevice(device.deviceId)?.pushRegistration?.filter).toEqual({ + sources: ['agent-task-complete'], + agentStates: ['finished'] + }) + }) +}) diff --git a/src/main/runtime/push/push-dispatcher.test-fixture.ts b/src/main/runtime/push/push-dispatcher.test-fixture.ts new file mode 100644 index 00000000000..9137ed8ea9f --- /dev/null +++ b/src/main/runtime/push/push-dispatcher.test-fixture.ts @@ -0,0 +1,94 @@ +import { vi } from 'vitest' +import type { MobilePushFilter, MobilePushRegistration } from '../../../shared/mobile-push-contract' +import type { MobileNotificationEvent } from '../runtime-mobile-notification-controller' +import type { PushGatewayClient, PushSendResult } from './push-gateway-client' +import { PushDispatcher, type PushDispatcherRegistry } from './push-dispatcher' + +const ALL_SOURCES: MobilePushFilter = { + sources: ['agent-task-complete', 'terminal-bell', 'plugin'], + agentStates: ['needs-input', 'finished'] +} + +export function registration( + overrides: Partial = {} +): MobilePushRegistration { + return { + registrationId: 'reg-1', + platform: 'ios', + filter: ALL_SOURCES, + registeredAt: 1, + ...overrides + } +} + +export type SendCall = Parameters[0] + +export function createHarness(options: { + devices: { deviceId: string; pushRegistration?: MobilePushRegistration }[] + results?: PushSendResult[] + sendImpl?: () => Promise +}): { + dispatcher: PushDispatcher + sends: SendCall[] + cleared: (string | null)[] + runRetry: () => void +} { + const sends: SendCall[] = [] + const cleared: (string | null)[] = [] + let retry: (() => void) | null = null + const client = { + send: vi.fn(async (input: SendCall) => { + sends.push(input) + if (options.sendImpl) { + return await options.sendImpl() + } + return { + ok: true as const, + results: + options.results ?? + input.registrationIds.map((registrationId) => ({ + registrationId, + status: 'queued' as const + })) + } + }) + } as unknown as PushGatewayClient + const registry: PushDispatcherRegistry = { + listDevices: () => options.devices, + setPushRegistration: (deviceId, value) => { + cleared.push(value === null ? deviceId : null) + return true + } + } + return { + dispatcher: new PushDispatcher({ + client, + registry, + scheduleRetry: (run) => { + retry = run + } + }), + sends, + cleared, + runRetry: () => retry?.() + } +} + +export function notification( + overrides: Partial = {} +): MobileNotificationEvent { + return { + type: 'notification', + source: 'agent-task-complete', + title: 'feat/x - Claude finished', + body: 'All done.', + worktreeId: 'repo::wt1', + notificationId: 'agent:one', + notificationSeq: 7, + notificationEpoch: 'epoch-1', + agentState: 'done', + ...overrides + } as MobileNotificationEvent +} + +export const flush = (): Promise => new Promise((resolve) => setImmediate(resolve)) diff --git a/src/main/runtime/push/push-dispatcher.test.ts b/src/main/runtime/push/push-dispatcher.test.ts new file mode 100644 index 00000000000..221383a34b1 --- /dev/null +++ b/src/main/runtime/push/push-dispatcher.test.ts @@ -0,0 +1,229 @@ +import { describe, expect, it, vi } from 'vitest' +import type { PushGatewayClient } from './push-gateway-client' +import { PushDispatcher } from './push-dispatcher' +import { + createHarness, + flush, + notification, + registration, + type SendCall +} from './push-dispatcher.test-fixture' + +describe('PushDispatcher', () => { + it('batches every matching registration into one send', async () => { + const harness = createHarness({ + devices: [ + { deviceId: 'a', pushRegistration: registration({ registrationId: 'reg-a' }) }, + { deviceId: 'b', pushRegistration: registration({ registrationId: 'reg-b' }) }, + { deviceId: 'c' } + ] + }) + + harness.dispatcher.enqueue(notification()) + await flush() + + expect(harness.sends).toHaveLength(1) + expect(harness.sends[0]?.registrationIds).toEqual(['reg-a', 'reg-b']) + expect(harness.sends[0]?.notification).toMatchObject({ + source: 'agent-task-complete', + agentState: 'finished', + notificationSeq: 7, + notificationEpoch: 'epoch-1', + worktreeId: 'repo::wt1' + }) + }) + + it('fans out past the per-request cap instead of starving the extra devices', async () => { + const devices = Array.from({ length: 25 }, (_, index) => ({ + deviceId: `device-${index}`, + pushRegistration: registration({ registrationId: `reg-${index}` }) + })) + const harness = createHarness({ devices }) + + harness.dispatcher.enqueue(notification()) + await flush() + + expect(harness.sends).toHaveLength(2) + expect(harness.sends[0]?.registrationIds).toHaveLength(20) + expect(harness.sends[1]?.registrationIds).toEqual([ + 'reg-20', + 'reg-21', + 'reg-22', + 'reg-23', + 'reg-24' + ]) + }) + + it('drops a dead registration reported by a later chunk', async () => { + const devices = Array.from({ length: 25 }, (_, index) => ({ + deviceId: `device-${index}`, + pushRegistration: registration({ registrationId: `reg-${index}` }) + })) + const harness = createHarness({ + devices, + results: [{ registrationId: 'reg-24', status: 'dead' }] + }) + + harness.dispatcher.enqueue(notification()) + await flush() + + expect(harness.cleared).toEqual(['device-24']) + }) + + it('never pushes a dismissal', async () => { + const harness = createHarness({ + devices: [{ deviceId: 'a', pushRegistration: registration() }] + }) + + harness.dispatcher.enqueue({ + type: 'dismiss', + notificationId: 'agent:one', + notificationSeq: 8, + notificationEpoch: 'epoch-1' + }) + await flush() + + expect(harness.sends).toHaveLength(0) + }) + + it('stays silent while the agent is still working', async () => { + const harness = createHarness({ + devices: [{ deviceId: 'a', pushRegistration: registration() }] + }) + + harness.dispatcher.enqueue(notification({ agentState: 'working' })) + await flush() + + expect(harness.sends).toHaveLength(0) + }) + + it('applies each device filter independently', async () => { + const harness = createHarness({ + devices: [ + { + deviceId: 'needs-input-only', + pushRegistration: registration({ + registrationId: 'reg-needs', + filter: { sources: ['agent-task-complete'], agentStates: ['needs-input'] } + }) + }, + { + deviceId: 'bells-only', + pushRegistration: registration({ + registrationId: 'reg-bell', + filter: { sources: ['terminal-bell'], agentStates: ['needs-input', 'finished'] } + }) + }, + { deviceId: 'everything', pushRegistration: registration({ registrationId: 'reg-all' }) } + ] + }) + + harness.dispatcher.enqueue(notification({ agentState: 'blocked' })) + await flush() + + expect(harness.sends[0]?.registrationIds).toEqual(['reg-needs', 'reg-all']) + }) + + it('pushes a bell to a device that filtered agent states out', async () => { + const harness = createHarness({ + devices: [ + { + deviceId: 'a', + pushRegistration: registration({ + filter: { sources: ['terminal-bell'], agentStates: [] } + }) + } + ] + }) + + harness.dispatcher.enqueue( + notification({ source: 'terminal-bell', agentState: undefined, title: 'Bell in x' }) + ) + await flush() + + expect(harness.sends[0]?.notification.agentState).toBeNull() + }) + + it('drops a registration the gateway reports dead', async () => { + const harness = createHarness({ + devices: [ + { deviceId: 'a', pushRegistration: registration({ registrationId: 'reg-a' }) }, + { deviceId: 'b', pushRegistration: registration({ registrationId: 'reg-b' }) } + ], + results: [ + { registrationId: 'reg-a', status: 'dead' }, + { registrationId: 'reg-b', status: 'queued' } + ] + }) + + harness.dispatcher.enqueue(notification()) + await flush() + + expect(harness.cleared).toEqual(['a']) + }) + + it('retries once when the gateway is unreachable', async () => { + const sends: SendCall[] = [] + const client = { + send: vi.fn(async (input: SendCall) => { + sends.push(input) + return { ok: false as const, reason: 'unreachable' as const } + }) + } as unknown as PushGatewayClient + const scheduled: (() => void)[] = [] + const devices = [{ deviceId: 'a', pushRegistration: registration() }] + const dispatcher = new PushDispatcher({ + client, + registry: { + listDevices: () => devices, + setPushRegistration: () => true + }, + scheduleRetry: (run, delayMs) => { + expect(delayMs).toBe(2_000) + scheduled.push(run) + } + }) + + dispatcher.enqueue(notification()) + await flush() + expect(sends).toHaveLength(1) + expect(scheduled).toHaveLength(1) + + scheduled[0]?.() + await flush() + expect(sends).toHaveLength(2) + // The second attempt is the last one; a further retry is never scheduled. + expect(scheduled).toHaveLength(1) + }) + + it('never throws into the caller when the client rejects', async () => { + const harness = createHarness({ + devices: [{ deviceId: 'a', pushRegistration: registration() }], + sendImpl: async () => { + throw new Error('boom') + } + }) + const warn = vi.spyOn(console, 'warn').mockImplementation(() => {}) + + expect(() => harness.dispatcher.enqueue(notification())).not.toThrow() + await flush() + expect(warn).toHaveBeenCalled() + warn.mockRestore() + }) + + it('never throws when the registry itself fails', async () => { + const dispatcher = new PushDispatcher({ + client: { send: vi.fn() } as unknown as PushGatewayClient, + registry: { + listDevices: () => { + throw new Error('registry unavailable') + }, + setPushRegistration: () => true + } + }) + const warn = vi.spyOn(console, 'warn').mockImplementation(() => {}) + + expect(() => dispatcher.enqueue(notification())).not.toThrow() + warn.mockRestore() + }) +}) diff --git a/src/main/runtime/push/push-dispatcher.ts b/src/main/runtime/push/push-dispatcher.ts new file mode 100644 index 00000000000..1a53113f20c --- /dev/null +++ b/src/main/runtime/push/push-dispatcher.ts @@ -0,0 +1,222 @@ +import { reserveNotificationCooldown } from '../../../shared/notification-burst-cooldown' +// Why: the out-of-band leg of the mobile notification fan-out. Every event that +// already went to connected sockets is offered to the push gateway so a phone +// with Orca closed still hears about it. Fire-and-forget by construction: the +// socket fan-out must never wait on, or fail because of, a push. +import type { MobilePushRegistration } from '../../../shared/mobile-push-contract' +import { PushOutcomeCounters } from './push-outcome-counters' +import { MOBILE_PUSH_SOURCES } from '../../../shared/mobile-push-contract' +import type { MobileNotificationEvent } from '../runtime-mobile-notification-controller' +import type { PushGatewayClient, PushSendNotification } from './push-gateway-client' + +const PUSH_RETRY_DELAY_MS = 2_000 +// The gateway rejects a whole request above this, so a host with more paired +// phones fans out across several sends rather than starving the extras. +const MAX_REGISTRATIONS_PER_SEND = 20 +const PUSH_TITLE_MAX_LENGTH = 80 +const PUSH_BODY_MAX_LENGTH = 180 + +export type PushDispatcherRegistry = { + listDevices(): readonly { deviceId: string; pushRegistration?: MobilePushRegistration }[] + setPushRegistration(deviceId: string, registration: MobilePushRegistration | null): boolean +} + +type PushDispatcherOptions = { + client: PushGatewayClient + registry: PushDispatcherRegistry + /** Test seam: lets a suite drive the single retry without real time. */ + scheduleRetry?: (run: () => void, delayMs: number) => void +} + +type PushTarget = { deviceId: string; registrationId: string; registration: MobilePushRegistration } + +function clip(value: string, maxLength: number): string { + const normalized = value.replace(/\s+/g, ' ').trim() + return normalized.length <= maxLength ? normalized : `${normalized.slice(0, maxLength - 1)}…` +} + +export { mapPushAgentState } from '../../../shared/mobile-notification-policy' +import { + allowsMobileNotification, + mapPushAgentState +} from '../../../shared/mobile-notification-policy' + +export class PushDispatcher { + private readonly recentNotifications = new Map() + private readonly outcomes = new PushOutcomeCounters() + private stopped = false + private readonly client: PushGatewayClient + private readonly registry: PushDispatcherRegistry + private readonly scheduleRetry: (run: () => void, delayMs: number) => void + + constructor(options: PushDispatcherOptions) { + this.client = options.client + this.registry = options.registry + this.scheduleRetry = + options.scheduleRetry ?? + ((run, delayMs) => { + // Why: a pending push retry must never hold the app open at quit. + setTimeout(run, delayMs).unref?.() + }) + } + + start(): void { + this.stopped = false + } + + stop(): void { + this.stopped = true + this.outcomes.flush() + } + + enqueue(event: MobileNotificationEvent): void { + if (this.stopped) { + return + } + try { + const plan = this.planSend(event) + if (!plan) { + return + } + for (const sound of [true, false]) { + const targets = plan.targets.filter( + (target) => (target.registration.filter.sound !== false) === sound + ) + for (let start = 0; start < targets.length; start += MAX_REGISTRATIONS_PER_SEND) { + void this.deliver( + targets.slice(start, start + MAX_REGISTRATIONS_PER_SEND), + { ...plan.notification, ...(!sound ? { sound: false } : {}) }, + 0 + ) + } + } + } catch (error) { + console.warn('[push] Failed to prepare a push notification:', error) + } + } + + private planSend( + event: MobileNotificationEvent + ): { targets: PushTarget[]; notification: PushSendNotification } | null { + // Dismissals are a socket-only concern; the phone clears its own banner. + if (event.type !== 'notification') { + return null + } + const source = MOBILE_PUSH_SOURCES.find((candidate) => candidate === event.source) + if (!source || event.notificationSeq === undefined || event.notificationEpoch === undefined) { + return null + } + const agentState = mapPushAgentState(source, event.agentState) + if (agentState === undefined) { + return null + } + const targets = this.registry.listDevices().flatMap((device) => { + const registration = device.pushRegistration + if (!registration || !allowsMobileNotification(registration.filter, event)) { + return [] + } + if ( + event.emittedAt !== undefined && + !reserveNotificationCooldown( + this.recentNotifications, + JSON.stringify([device.deviceId, event.worktreeId ?? 'global']), + event.emittedAt + ) + ) { + return [] + } + return [ + { deviceId: device.deviceId, registrationId: registration.registrationId, registration } + ] + }) + if (targets.length === 0) { + return null + } + return { + targets, + notification: { + ...(event.notificationId ? { notificationId: event.notificationId } : {}), + notificationSeq: event.notificationSeq, + notificationEpoch: event.notificationEpoch, + source, + agentState, + title: clip(event.title, PUSH_TITLE_MAX_LENGTH), + body: clip(event.body, PUSH_BODY_MAX_LENGTH), + ...(event.worktreeId ? { worktreeId: event.worktreeId } : {}) + } + } + } + + private async deliver( + targets: readonly PushTarget[], + notification: PushSendNotification, + attempt: number + ): Promise { + if (this.stopped) { + return + } + const currentTargets = targets.filter((target) => + this.registry + .listDevices() + .some( + (device) => + device.deviceId === target.deviceId && device.pushRegistration === target.registration + ) + ) + if (!currentTargets.length) { + return + } + try { + const result = await this.client.send({ + registrationIds: currentTargets.map((target) => target.registrationId), + notification + }) + if (this.stopped) { + return + } + if (result.ok) { + for (const entry of result.results) { + if (entry.status === 'error' || entry.status === 'rate_limited') { + this.outcomes.record(entry.status) + } + } + this.dropDeadRegistrations(targets, result.results) + return + } + this.outcomes.record(result.reason) + // Only a transport-level miss is worth repeating; a gateway that refused + // this payload will refuse the identical retry. + if (attempt === 0 && result.reason === 'unreachable') { + this.scheduleRetry(() => { + void this.deliver(targets, notification, attempt + 1) + }, PUSH_RETRY_DELAY_MS) + } + } catch (error) { + console.warn('[push] Push send failed:', error) + } + } + + private dropDeadRegistrations( + targets: readonly PushTarget[], + results: readonly { registrationId: string; status: string }[] + ): void { + for (const result of results) { + if (result.status !== 'dead') { + continue + } + const target = targets.find((entry) => entry.registrationId === result.registrationId) + if ( + !target || + this.registry.listDevices().find((device) => device.deviceId === target.deviceId) + ?.pushRegistration !== target.registration + ) { + continue + } + try { + this.registry.setPushRegistration(target.deviceId, null) + } catch (error) { + console.warn('[push] Failed to drop a dead push registration:', error) + } + } + } +} diff --git a/src/main/runtime/push/push-gateway-client.test.ts b/src/main/runtime/push/push-gateway-client.test.ts new file mode 100644 index 00000000000..5f86b10c7e4 --- /dev/null +++ b/src/main/runtime/push/push-gateway-client.test.ts @@ -0,0 +1,260 @@ +import { describe, expect, it, vi } from 'vitest' +import { createHash } from 'node:crypto' +import { buildPushChallengeFixture, createPushHostKeypair } from './push-host-challenge-fixtures' +import { PushGatewayClient } from './push-gateway-client' + +const GATEWAY_URL = 'https://push.onorca.dev' +const NOW = 1_770_000_000_000 + +type Recorded = { + url: string + method: string + authorization: string | null + body: unknown + redirect: RequestRedirect | undefined +} + +function fingerprintOf(publicKey: Uint8Array): string { + return createHash('sha256').update(publicKey).digest('base64url').slice(0, 16) +} + +function jsonResponse(status: number, body: unknown): Response { + return new Response(JSON.stringify(body), { + status, + headers: { 'content-type': 'application/json' } + }) +} + +function createFakeGateway( + options: { sessionTtlMs?: number; devicesStatus?: number; rejectBearer?: boolean } = {} +): { + client: PushGatewayClient + calls: Recorded[] + expireSession: () => void + now: { value: number } +} { + const hostKeypair = createPushHostKeypair() + const hostFingerprint = fingerprintOf(hostKeypair.publicKey) + const now = { value: NOW } + const calls: Recorded[] = [] + const liveTokens = new Set() + const knownRegistrations = new Set() + let issued = 0 + let pendingProof: string | null = null + + const fetchImpl = (async (input: string, init?: RequestInit): Promise => { + const url = String(input) + const headers = new Headers(init?.headers) + const body: unknown = init?.body ? JSON.parse(String(init.body)) : undefined + calls.push({ + url, + method: init?.method ?? 'GET', + authorization: headers.get('authorization'), + body, + redirect: init?.redirect + }) + if (url.endsWith('/v1/host/challenge')) { + const built = buildPushChallengeFixture({ + hostKeypair, + gatewayOrigin: GATEWAY_URL, + hostFingerprint, + issuedAt: now.value, + challengeId: `challenge-${++issued}` + }) + pendingProof = built.proof + return jsonResponse(200, built.challenge) + } + if (url.endsWith('/v1/host/session')) { + const params = body as { proofB64: string } + if (params.proofB64 !== pendingProof) { + return jsonResponse(401, { error: 'bad_proof' }) + } + const sessionToken = `session-${issued}` + liveTokens.add(sessionToken) + return jsonResponse(200, { + sessionToken, + expiresAt: now.value + (options.sessionTtlMs ?? 24 * 60 * 60_000), + hostFingerprint + }) + } + const bearer = headers.get('authorization')?.replace('Bearer ', '') ?? '' + if (options.rejectBearer || !liveTokens.has(bearer)) { + return jsonResponse(401, { error: 'session_expired' }) + } + if (url.endsWith('/v1/devices')) { + if (options.devicesStatus) { + return jsonResponse(options.devicesStatus, { error: 'nope' }) + } + knownRegistrations.add('reg-1') + return jsonResponse(200, { registrationId: 'reg-1' }) + } + if (url.endsWith('/v1/send')) { + return jsonResponse(200, { results: [{ registrationId: 'reg-1', status: 'queued' }] }) + } + // Why explicit: a catch-all 204 would report every delete as accepted and + // leave the 404 branch of deleteDevice untested. + const deleted = /\/v1\/devices\/([^/]+)$/.exec(url) + if (deleted && init?.method === 'DELETE') { + const registrationId = decodeURIComponent(deleted[1] ?? '') + return new Response(null, { status: knownRegistrations.has(registrationId) ? 204 : 404 }) + } + throw new Error(`unexpected request: ${init?.method ?? 'GET'} ${url}`) + }) as unknown as typeof globalThis.fetch + + return { + client: new PushGatewayClient({ + gatewayUrl: GATEWAY_URL, + keypair: hostKeypair, + fetch: fetchImpl, + now: () => now.value + }), + calls, + expireSession: () => liveTokens.clear(), + now + } +} + +const REGISTER_INPUT = { + deviceId: 'device-1', + platform: 'ios' as const, + token: 'a'.repeat(64), + apnsEnvironment: 'sandbox' as const, + filter: { sources: ['agent-task-complete'] as const, agentStates: ['finished'] as const } +} + +describe('PushGatewayClient', () => { + it('runs the challenge handshake once and reuses the cached session', async () => { + const gateway = createFakeGateway() + + expect(await gateway.client.registerDevice(REGISTER_INPUT)).toEqual({ + ok: true, + registrationId: 'reg-1' + }) + expect( + await gateway.client.send({ + registrationIds: ['reg-1'], + notification: { + notificationSeq: 1, + notificationEpoch: 'epoch-1', + source: 'agent-task-complete', + agentState: 'finished', + title: 'Done', + body: 'Body' + } + }) + ).toEqual({ ok: true, results: [{ registrationId: 'reg-1', status: 'queued' }] }) + + const handshakes = gateway.calls.filter((call) => call.url.includes('/v1/host/')) + expect(handshakes).toHaveLength(2) + expect(gateway.calls.at(-1)?.authorization).toBe('Bearer session-1') + }) + + it('re-authenticates once when the gateway rejects the cached session', async () => { + const gateway = createFakeGateway() + await gateway.client.registerDevice(REGISTER_INPUT) + gateway.expireSession() + + expect(await gateway.client.registerDevice(REGISTER_INPUT)).toEqual({ + ok: true, + registrationId: 'reg-1' + }) + expect(gateway.calls.filter((call) => call.url.endsWith('/v1/host/challenge'))).toHaveLength(2) + expect(gateway.calls.at(-1)?.authorization).toBe('Bearer session-2') + }) + + it('re-authenticates before a session that is about to expire', async () => { + const gateway = createFakeGateway({ sessionTtlMs: 90_000 }) + await gateway.client.registerDevice(REGISTER_INPUT) + gateway.now.value += 60_000 + + await gateway.client.registerDevice(REGISTER_INPUT) + expect(gateway.calls.filter((call) => call.url.endsWith('/v1/host/challenge'))).toHaveLength(2) + }) + + it('shares one handshake across concurrent calls', async () => { + const gateway = createFakeGateway() + await Promise.all([ + gateway.client.registerDevice(REGISTER_INPUT), + gateway.client.registerDevice(REGISTER_INPUT) + ]) + expect(gateway.calls.filter((call) => call.url.endsWith('/v1/host/challenge'))).toHaveLength(1) + }) + + it('reports an unreachable gateway instead of throwing', async () => { + const keypair = createPushHostKeypair() + const client = new PushGatewayClient({ + gatewayUrl: GATEWAY_URL, + keypair, + fetch: vi.fn(async () => { + throw new Error('network down') + }) as unknown as typeof globalThis.fetch, + now: () => NOW + }) + expect(await client.registerDevice(REGISTER_INPUT)).toEqual({ + ok: false, + reason: 'unreachable' + }) + }) + + it('reports a refused registration as rejected', async () => { + const gateway = createFakeGateway({ devicesStatus: 400 }) + expect(await gateway.client.registerDevice(REGISTER_INPUT)).toEqual({ + ok: false, + reason: 'rejected' + }) + }) + + it('never follows a redirect, on the handshake or on an authorized call', async () => { + const gateway = createFakeGateway() + + await gateway.client.registerDevice(REGISTER_INPUT) + await gateway.client.deleteDevice('reg-1') + + // A 307 would replay the host proof, then the phone's token, to whatever + // origin the redirect named. + expect(gateway.calls.length).toBeGreaterThanOrEqual(4) + expect(gateway.calls.every((call) => call.redirect === 'error')).toBe(true) + }) + + it('reports a gateway 5xx as unreachable so the caller can retry', async () => { + const gateway = createFakeGateway({ devicesStatus: 503 }) + expect(await gateway.client.registerDevice(REGISTER_INPUT)).toEqual({ + ok: false, + reason: 'unreachable' + }) + }) + + it('treats a delete the gateway accepted as done', async () => { + const gateway = createFakeGateway() + await gateway.client.registerDevice(REGISTER_INPUT) + + expect(await gateway.client.deleteDevice('reg-1')).toEqual({ deleted: true, retryable: false }) + expect(gateway.calls.at(-1)).toMatchObject({ method: 'DELETE' }) + }) + + it('treats a delete of an unknown registration as done', async () => { + const gateway = createFakeGateway() + + expect(await gateway.client.deleteDevice('reg-gone')).toEqual({ + deleted: true, + retryable: false + }) + }) + + it('reports a 401 that survives the forced re-auth as unreachable', async () => { + const gateway = createFakeGateway({ rejectBearer: true }) + + expect(await gateway.client.registerDevice(REGISTER_INPUT)).toEqual({ + ok: false, + reason: 'unreachable' + }) + // Exactly one forced re-auth, not a handshake loop. + expect(gateway.calls.filter((call) => call.url.endsWith('/v1/host/challenge'))).toHaveLength(2) + }) + + it('keeps an unreachable-classified 401 retryable for a queued delete', async () => { + const gateway = createFakeGateway({ rejectBearer: true }) + + expect(await gateway.client.deleteDevice('reg-1')).toEqual({ deleted: false, retryable: true }) + }) +}) diff --git a/src/main/runtime/push/push-gateway-client.ts b/src/main/runtime/push/push-gateway-client.ts new file mode 100644 index 00000000000..e1097f3dc77 --- /dev/null +++ b/src/main/runtime/push/push-gateway-client.ts @@ -0,0 +1,177 @@ +// Why: talks to the Orca push gateway (docs/reference/mobile-push-contract.md). +// Every method returns a result instead of throwing — push is best-effort and +// must never break the socket fan-out it rides along with. +import { z } from 'zod' +import { cancelUnreadResponseBody } from '../../lib/unread-response-body' +import type { E2EEKeypair } from '../e2ee-keypair' +import type { + MobilePushAgentState, + MobilePushApnsEnvironment, + MobilePushFilter, + MobilePushPlatform, + MobilePushSource +} from '../../../shared/mobile-push-contract' +import { + PUSH_REQUEST_DEADLINE_MS, + readPushGatewayJson, + type PushGatewayFailure, + type PushGatewayResponse, + type PushGatewayResult +} from './push-gateway-response' +import { PushGatewaySession } from './push-gateway-session' + +export type { PushGatewayFailure, PushGatewayResult } + +const RegisterResponseSchema = z.object({ registrationId: z.string().min(1).max(512) }) + +const SendResponseSchema = z.object({ + results: z + .array( + z.object({ + registrationId: z.string().min(1).max(512), + status: z.enum(['queued', 'dead', 'rate_limited', 'error']) + }) + ) + .max(64) +}) + +export type PushSendResult = z.infer['results'][number] + +export type PushSendNotification = { + sound?: boolean + notificationId?: string + notificationSeq: number + notificationEpoch: string + source: MobilePushSource + agentState: MobilePushAgentState | null + title: string + body: string + worktreeId?: string +} + +type PushGatewayClientOptions = { + gatewayUrl: string + keypair: E2EEKeypair + fetch?: typeof globalThis.fetch + now?: () => number +} + +type AuthorizedResponse = { ok: true; response: Response; token: string } | PushGatewayFailure + +export class PushGatewayClient { + private readonly origin: string + private readonly fetchImpl: typeof globalThis.fetch + private readonly session: PushGatewaySession + readonly hostFingerprint: string + + constructor(options: PushGatewayClientOptions) { + this.origin = new URL(options.gatewayUrl).origin + this.fetchImpl = options.fetch ?? globalThis.fetch + this.session = new PushGatewaySession({ + origin: this.origin, + keypair: options.keypair, + fetchImpl: this.fetchImpl, + now: options.now ?? Date.now + }) + this.hostFingerprint = this.session.hostFingerprint + } + + async registerDevice(input: { + deviceId: string + platform: MobilePushPlatform + token: string + apnsEnvironment?: MobilePushApnsEnvironment + filter: MobilePushFilter + }): Promise> { + const response = await this.authorized('/v1/devices', { + method: 'POST', + body: { + v: 1, + deviceId: input.deviceId, + platform: input.platform, + token: input.token, + ...(input.apnsEnvironment ? { apnsEnvironment: input.apnsEnvironment } : {}), + filter: { sources: [...input.filter.sources], agentStates: [...input.filter.agentStates] } + } + }) + const parsed = await readPushGatewayJson(response, RegisterResponseSchema) + return parsed.ok ? { ok: true, registrationId: parsed.value.registrationId } : parsed + } + + /** `retryable` tells the outbox whether to keep the delete queued. */ + async deleteDevice(registrationId: string): Promise<{ deleted: boolean; retryable: boolean }> { + const response = await this.authorized(`/v1/devices/${encodeURIComponent(registrationId)}`, { + method: 'DELETE' + }) + if (!response.ok) { + return { deleted: false, retryable: true } + } + await cancelUnreadResponseBody(response.response) + // A gateway that no longer knows the registration is as deleted as it gets. + const gone = response.response.ok || response.response.status === 404 + return { deleted: gone, retryable: !gone } + } + + async send(input: { + registrationIds: readonly string[] + notification: PushSendNotification + }): Promise> { + const response = await this.authorized('/v1/send', { + method: 'POST', + body: { + v: 1, + registrationIds: [...input.registrationIds], + notification: input.notification + } + }) + const parsed = await readPushGatewayJson(response, SendResponseSchema) + return parsed.ok ? { ok: true, results: parsed.value.results } : parsed + } + + private async authorized( + path: string, + init: { method: string; body?: unknown } + ): Promise { + const first = await this.sendAuthorized(path, init, null) + if (!first.ok || first.response.status !== 401) { + return first + } + // A 401 means that one session died server-side; one forced re-auth, then stop. + await cancelUnreadResponseBody(first.response) + const retried = await this.sendAuthorized(path, init, first.token) + if (retried.ok && retried.response.status === 401) { + await cancelUnreadResponseBody(retried.response) + // A 401 that survives a freshly minted session is the gateway being unusable + // right now, not this request being wrong: register should report it as + // unreachable, and send should still spend its one retry. + return { ok: false, reason: 'unreachable' } + } + return retried + } + + private async sendAuthorized( + path: string, + init: { method: string; body?: unknown }, + staleToken: string | null + ): Promise { + const outcome = await this.session.ensure(staleToken) + if (!outcome.ok) { + return outcome + } + try { + const response = await this.fetchImpl(`${this.origin}${path}`, { + method: init.method, + headers: { + authorization: `Bearer ${outcome.session.token}`, + ...(init.body === undefined ? {} : { 'content-type': 'application/json' }) + }, + redirect: 'error', + signal: AbortSignal.timeout(PUSH_REQUEST_DEADLINE_MS), + ...(init.body === undefined ? {} : { body: JSON.stringify(init.body) }) + }) + return { ok: true, response, token: outcome.session.token } + } catch { + return { ok: false, reason: 'unreachable' } + } + } +} diff --git a/src/main/runtime/push/push-gateway-response.ts b/src/main/runtime/push/push-gateway-response.ts new file mode 100644 index 00000000000..12a901b2943 --- /dev/null +++ b/src/main/runtime/push/push-gateway-response.ts @@ -0,0 +1,61 @@ +// Why: the authorized request path and the handshake that authorizes it must +// classify a gateway response identically — otherwise the same 503 means "retry" +// on one leg and "give up" on the other, and register/send disagree about why. +import type { z } from 'zod' +import { cancelUnreadResponseBody } from '../../lib/unread-response-body' + +export const PUSH_REQUEST_DEADLINE_MS = 15_000 + +export type PushGatewayFailure = { ok: false; reason: 'unreachable' | 'rejected' } +export type PushGatewayResult = ({ ok: true } & T) | PushGatewayFailure +export type PushGatewayResponse = { ok: true; response: Response } | PushGatewayFailure + +/** Unauthenticated POST; the handshake legs run before any session exists. */ +export async function postPushGatewayJson( + fetchImpl: typeof globalThis.fetch, + url: string, + body: unknown +): Promise { + try { + const response = await fetchImpl(url, { + method: 'POST', + headers: { 'content-type': 'application/json' }, + // A 307 would replay the proof, and later the phone's token, to whatever + // origin the redirect named. + redirect: 'error', + signal: AbortSignal.timeout(PUSH_REQUEST_DEADLINE_MS), + body: JSON.stringify(body) + }) + return { ok: true, response } + } catch { + return { ok: false, reason: 'unreachable' } + } +} + +export async function readPushGatewayJson( + result: PushGatewayResponse, + schema: TSchema +): Promise<{ ok: true; value: z.infer } | PushGatewayFailure> { + if (!result.ok) { + return result + } + const { response } = result + if (!response.ok) { + await cancelUnreadResponseBody(response) + // 5xx and 429 are worth another attempt later; anything else is the gateway + // refusing this request as written. + return { + ok: false, + reason: response.status >= 500 || response.status === 429 ? 'unreachable' : 'rejected' + } + } + let payload: unknown + try { + payload = await response.json() + } catch { + await cancelUnreadResponseBody(response) + return { ok: false, reason: 'unreachable' } + } + const parsed = schema.safeParse(payload) + return parsed.success ? { ok: true, value: parsed.data } : { ok: false, reason: 'rejected' } +} diff --git a/src/main/runtime/push/push-gateway-session.test.ts b/src/main/runtime/push/push-gateway-session.test.ts new file mode 100644 index 00000000000..8527430365a --- /dev/null +++ b/src/main/runtime/push/push-gateway-session.test.ts @@ -0,0 +1,169 @@ +import { createHash } from 'node:crypto' +import { describe, expect, it, vi } from 'vitest' +import { buildPushChallengeFixture, createPushHostKeypair } from './push-host-challenge-fixtures' +import { PushGatewaySession, type PushSessionOutcome } from './push-gateway-session' + +const GATEWAY_ORIGIN = 'https://push.onorca.dev' +const NOW = 1_770_000_000_000 + +function jsonResponse(status: number, body: unknown): Response { + return new Response(JSON.stringify(body), { + status, + headers: { 'content-type': 'application/json' } + }) +} + +function tokenOf(outcome: PushSessionOutcome): string | null { + return outcome.ok ? outcome.session.token : null +} + +function createSessionHarness( + options: { sessionStatus?: number; challengeStatus?: number; wrongFingerprint?: boolean } = {} +): { + session: PushGatewaySession + challenges: () => number + requests: () => number + now: { value: number } +} { + const hostKeypair = createPushHostKeypair() + const hostFingerprint = createHash('sha256') + .update(hostKeypair.publicKey) + .digest('base64url') + .slice(0, 16) + const now = { value: NOW } + let issued = 0 + let requests = 0 + let pendingProof: string | null = null + + const fetchImpl = (async (input: string, init?: RequestInit): Promise => { + const url = String(input) + requests += 1 + if (url.endsWith('/v1/host/challenge')) { + if (options.challengeStatus) { + return jsonResponse(options.challengeStatus, { error: 'rate_limited' }) + } + const built = buildPushChallengeFixture({ + hostKeypair, + gatewayOrigin: GATEWAY_ORIGIN, + hostFingerprint, + issuedAt: now.value, + challengeId: `challenge-${++issued}` + }) + pendingProof = built.proof + return jsonResponse(200, built.challenge) + } + if (options.sessionStatus) { + return jsonResponse(options.sessionStatus, { error: 'nope' }) + } + const body = init?.body ? (JSON.parse(String(init.body)) as { proofB64: string }) : null + if (body?.proofB64 !== pendingProof) { + return jsonResponse(401, { error: 'bad_proof' }) + } + return jsonResponse(200, { + sessionToken: `session-${issued}`, + expiresAt: now.value + 24 * 60 * 60_000, + hostFingerprint: options.wrongFingerprint ? 'someone-else' : hostFingerprint + }) + }) as unknown as typeof globalThis.fetch + + return { + session: new PushGatewaySession({ + origin: GATEWAY_ORIGIN, + keypair: hostKeypair, + fetchImpl, + now: () => now.value + }), + challenges: () => issued, + requests: () => requests, + now + } +} + +describe('PushGatewaySession', () => { + it('reuses the cached session until it nears expiry', async () => { + const harness = createSessionHarness() + + expect(tokenOf(await harness.session.ensure(null))).toBe('session-1') + expect(tokenOf(await harness.session.ensure(null))).toBe('session-1') + expect(harness.challenges()).toBe(1) + }) + + it('drops only the exact session that received the 401', async () => { + const harness = createSessionHarness() + expect(tokenOf(await harness.session.ensure(null))).toBe('session-1') + + // A request that 401ed on session-1 forces a fresh handshake. + expect(tokenOf(await harness.session.ensure('session-1'))).toBe('session-2') + // A second request whose 401 also named session-1 must keep the new token. + expect(tokenOf(await harness.session.ensure('session-1'))).toBe('session-2') + expect(harness.challenges()).toBe(2) + }) + + it('reports a refused handshake as rejected rather than unreachable', async () => { + const harness = createSessionHarness({ sessionStatus: 403 }) + + expect(await harness.session.ensure(null)).toEqual({ ok: false, reason: 'rejected' }) + }) + + it('reports a session minted for another host as rejected', async () => { + const harness = createSessionHarness({ wrongFingerprint: true }) + + expect(await harness.session.ensure(null)).toEqual({ ok: false, reason: 'rejected' }) + }) + + it('caches a refusal briefly instead of re-handshaking on every call', async () => { + const harness = createSessionHarness({ sessionStatus: 403 }) + + await harness.session.ensure(null) + await harness.session.ensure(null) + expect(harness.challenges()).toBe(1) + + harness.now.value += 30_000 + await harness.session.ensure(null) + expect(harness.challenges()).toBe(2) + }) + + it('never caches a transport failure, which may clear on the next try', async () => { + const fetchImpl = vi.fn(async () => { + throw new Error('network down') + }) as unknown as typeof globalThis.fetch + const session = new PushGatewaySession({ + origin: GATEWAY_ORIGIN, + keypair: createPushHostKeypair(), + fetchImpl, + now: () => NOW + }) + + expect(await session.ensure(null)).toEqual({ ok: false, reason: 'unreachable' }) + expect(await session.ensure(null)).toEqual({ ok: false, reason: 'unreachable' }) + expect(fetchImpl).toHaveBeenCalledTimes(2) + }) + + it('reports a rate-limited challenge as unreachable and backs off', async () => { + const harness = createSessionHarness({ challengeStatus: 429 }) + + expect(await harness.session.ensure(null)).toEqual({ ok: false, reason: 'unreachable' }) + expect(await harness.session.ensure(null)).toEqual({ ok: false, reason: 'unreachable' }) + expect(harness.requests()).toBe(1) + + harness.now.value += 60_000 + await harness.session.ensure(null) + expect(harness.requests()).toBe(2) + }) + + it('reports a rate-limited session mint as unreachable, not refused', async () => { + const harness = createSessionHarness({ sessionStatus: 429 }) + + expect(await harness.session.ensure(null)).toEqual({ ok: false, reason: 'unreachable' }) + // Cached for a minute, so the next dispatch does not spend more of the bucket. + expect(await harness.session.ensure(null)).toEqual({ ok: false, reason: 'unreachable' }) + expect(harness.challenges()).toBe(1) + }) + + it('shares one handshake across concurrent callers', async () => { + const harness = createSessionHarness() + + await Promise.all([harness.session.ensure(null), harness.session.ensure(null)]) + expect(harness.challenges()).toBe(1) + }) +}) diff --git a/src/main/runtime/push/push-gateway-session.ts b/src/main/runtime/push/push-gateway-session.ts new file mode 100644 index 00000000000..dd50b813f1d --- /dev/null +++ b/src/main/runtime/push/push-gateway-session.ts @@ -0,0 +1,157 @@ +// Why: the challenge/proof handshake every push request rides on, split out of +// push-gateway-client.ts so the session cache and its refusal cache stay readable +// next to the request methods rather than buried under them. +import { z } from 'zod' +import { cancelUnreadResponseBody } from '../../lib/unread-response-body' +import type { E2EEKeypair } from '../e2ee-keypair' +import { deriveRelayHostId } from '../relay/relay-http-client' +import { answerPushHostChallenge } from './push-host-proof' +import { + postPushGatewayJson, + readPushGatewayJson, + type PushGatewayFailure +} from './push-gateway-response' + +// Re-auth a little early so a send never spends its one retry on a token that +// expired between the check and the request. +const SESSION_RENEWAL_MARGIN_MS = 60_000 +// Why: a gateway that refuses this host's proof refuses the identical next one, +// so without this every dispatch pays two full handshake round trips to relearn it. +const HANDSHAKE_REFUSAL_TTL_MS = 30_000 +// Why: the handshake routes sit behind a per-IP bucket. Backing off keeps this +// host from spending the whole bucket on challenges it will never get to use. +const HANDSHAKE_RATE_LIMIT_TTL_MS = 60_000 + +const ChallengeResponseSchema = z + .object({ + challengeId: z.string().min(1).max(512), + gatewayEphemeralPublicKeyB64: z.string().min(1).max(128), + nonceB64: z.string().min(1).max(128), + ciphertextB64: z + .string() + .min(1) + .max(8 * 1024), + expiresAt: z.number().int().positive().max(Number.MAX_SAFE_INTEGER) + }) + .strict() + +const SessionResponseSchema = z + .object({ + sessionToken: z.string().min(1).max(1024), + expiresAt: z.number().int().positive().max(Number.MAX_SAFE_INTEGER), + hostFingerprint: z.string().min(1).max(64) + }) + .strict() + +export type PushSession = { token: string; expiresAt: number } +export type PushSessionOutcome = { ok: true; session: PushSession } | PushGatewayFailure + +type PushGatewaySessionOptions = { + origin: string + keypair: E2EEKeypair + fetchImpl: typeof globalThis.fetch + now: () => number +} + +export class PushGatewaySession { + private readonly origin: string + private readonly keypair: E2EEKeypair + private readonly fetchImpl: typeof globalThis.fetch + private readonly now: () => number + readonly hostFingerprint: string + private session: PushSession | null = null + private pending: Promise | null = null + private negative: { until: number; reason: PushGatewayFailure['reason'] } | null = null + + constructor(options: PushGatewaySessionOptions) { + this.origin = options.origin + this.keypair = options.keypair + this.fetchImpl = options.fetchImpl + this.now = options.now + this.hostFingerprint = deriveRelayHostId(options.keypair.publicKey) + } + + /** + * `staleToken` is the token that just received a 401. Only that exact session is + * dropped: a concurrent request may already have installed a good one, and + * clearing unconditionally would throw it away and re-handshake for nothing. + */ + async ensure(staleToken: string | null): Promise { + if (staleToken !== null && this.session?.token === staleToken) { + this.session = null + } + const cached = this.session + if (cached && cached.expiresAt - SESSION_RENEWAL_MARGIN_MS > this.now()) { + return { ok: true, session: cached } + } + if (this.negative && this.negative.until > this.now()) { + return { ok: false, reason: this.negative.reason } + } + // Concurrent sends must not each burn a challenge; share one handshake. + this.pending ??= this.open().finally(() => { + this.pending = null + }) + return await this.pending + } + + private async open(): Promise { + const challenge = await this.handshakePost( + '/v1/host/challenge', + { v: 1, hostPublicKeyB64: this.keypair.publicKeyB64 }, + ChallengeResponseSchema + ) + if (!challenge.ok) { + return this.remember(challenge) + } + const proofB64 = answerPushHostChallenge(challenge.value, { + gatewayOrigin: this.origin, + hostFingerprint: this.hostFingerprint, + hostPublicKey: this.keypair.publicKey, + hostSecretKey: this.keypair.secretKey, + now: this.now + }) + if (!proofB64) { + // A challenge this host cannot answer is a refusal, not a dropped packet. + return this.remember({ ok: false, reason: 'rejected' }) + } + const parsed = await this.handshakePost( + '/v1/host/session', + { v: 1, challengeId: challenge.value.challengeId, proofB64 }, + SessionResponseSchema + ) + if (!parsed.ok) { + return this.remember(parsed) + } + if (parsed.value.hostFingerprint !== this.hostFingerprint) { + // The gateway answered for some other host; that token is never usable here. + return this.remember({ ok: false, reason: 'rejected' }) + } + this.session = { token: parsed.value.sessionToken, expiresAt: parsed.value.expiresAt } + this.negative = null + return { ok: true, session: this.session } + } + + private async handshakePost( + path: string, + body: unknown, + schema: TSchema + ): Promise<{ ok: true; value: z.infer } | PushGatewayFailure> { + const response = await postPushGatewayJson(this.fetchImpl, `${this.origin}${path}`, body) + if (response.ok && response.response.status === 429) { + await cancelUnreadResponseBody(response.response) + // Rate limiting refuses the moment, not this host: back off, stay retryable + // so register reports gateway_unreachable and send keeps its one retry. + this.negative = { until: this.now() + HANDSHAKE_RATE_LIMIT_TTL_MS, reason: 'unreachable' } + return { ok: false, reason: 'unreachable' } + } + return await readPushGatewayJson(response, schema) + } + + /** Caches refusals only: a transport failure may clear on the very next try. */ + private remember(failure: PushGatewayFailure): PushGatewayFailure { + if (failure.reason === 'rejected') { + this.negative = { until: this.now() + HANDSHAKE_REFUSAL_TTL_MS, reason: 'rejected' } + } + return failure + } +} diff --git a/src/main/runtime/push/push-host-challenge-fixtures.ts b/src/main/runtime/push/push-host-challenge-fixtures.ts new file mode 100644 index 00000000000..e48dec33c7a --- /dev/null +++ b/src/main/runtime/push/push-host-challenge-fixtures.ts @@ -0,0 +1,136 @@ +// Test fixtures: builds the sealed challenge the push gateway would issue, so the +// proof answerer and the gateway client can both be exercised against a real box. +import { createHmac, randomBytes } from 'node:crypto' +import nacl from 'tweetnacl' +import type { E2EEKeypair } from '../e2ee-keypair' +import type { PushHostChallenge, PushHostProofContext } from './push-host-proof' + +const encoder = new TextEncoder() +export const PUSH_PROOF_DOMAIN = 'orca-push-host-proof/v1' +export const PUSH_CHALLENGE_DOMAIN = 'orca-push-host-challenge/v1' + +function concat(parts: readonly Uint8Array[]): Uint8Array { + const output = new Uint8Array(parts.reduce((total, part) => total + part.byteLength, 0)) + let offset = 0 + for (const part of parts) { + output.set(part, offset) + offset += part.byteLength + } + return output +} + +function uint32(value: number): Uint8Array { + const bytes = new Uint8Array(4) + new DataView(bytes.buffer).setUint32(0, value, false) + return bytes +} + +function uint64(value: number): Uint8Array { + const bytes = new Uint8Array(8) + new DataView(bytes.buffer).setBigUint64(0, BigInt(value), false) + return bytes +} + +function field(name: string, value: Uint8Array): Uint8Array { + const encodedName = encoder.encode(name) + return concat([uint32(encodedName.byteLength), encodedName, uint32(value.byteLength), value]) +} + +export function text(value: string): Uint8Array { + return encoder.encode(value) +} + +export type PushTranscriptInput = { + gatewayOrigin: string + gatewayKey: Uint8Array + nonce: Uint8Array + challengeId: string + issuedAt: number + expiresAt: number + hostFingerprint: string + hostKey: Uint8Array +} + +export function buildPushTranscript(input: PushTranscriptInput): Uint8Array { + return concat([ + field('protocol', text(PUSH_PROOF_DOMAIN)), + field('version', new Uint8Array([1])), + field('gatewayOrigin', text(input.gatewayOrigin)), + field('gatewayEphemeralPublicKey', input.gatewayKey), + field('challengeNonce', input.nonce), + field('challengeId', text(input.challengeId)), + field('issuedAt', uint64(input.issuedAt)), + field('expiresAt', uint64(input.expiresAt)), + field('hostFingerprint', text(input.hostFingerprint)), + field('hostPublicKey', input.hostKey) + ]) +} + +export function pushAckProof(secret: Uint8Array, transcript: Uint8Array): string { + return createHmac('sha256', secret) + .update(text(`${PUSH_PROOF_DOMAIN}\0ack\0`)) + .update(transcript) + .digest('base64') +} + +export function createPushHostKeypair(): E2EEKeypair { + const keys = nacl.box.keyPair() + return { + publicKey: keys.publicKey, + secretKey: keys.secretKey, + publicKeyB64: Buffer.from(keys.publicKey).toString('base64') + } +} + +/** Seals a challenge for `hostPublicKey`; overrides let a suite corrupt one field at a time. */ +export function buildPushChallengeFixture(input: { + hostKeypair: E2EEKeypair + gatewayOrigin: string + hostFingerprint: string + issuedAt: number + challengeId?: string + transcript?: Partial + challenge?: Partial +}): { challenge: PushHostChallenge; context: Omit; proof: string } { + const gatewayKeys = nacl.box.keyPair() + const nonce = randomBytes(24) + const secret = randomBytes(32) + const expiresAt = input.issuedAt + 10_000 + const challengeId = input.challengeId ?? 'challenge-1' + const transcript = buildPushTranscript({ + gatewayOrigin: input.gatewayOrigin, + gatewayKey: gatewayKeys.publicKey, + nonce, + challengeId, + issuedAt: input.issuedAt, + expiresAt, + hostFingerprint: input.hostFingerprint, + hostKey: input.hostKeypair.publicKey, + ...input.transcript + }) + const plaintext = concat([ + text(`${PUSH_CHALLENGE_DOMAIN}\0`), + uint32(transcript.byteLength), + transcript, + secret + ]) + return { + challenge: { + challengeId, + gatewayEphemeralPublicKeyB64: Buffer.from(gatewayKeys.publicKey).toString('base64'), + nonceB64: nonce.toString('base64'), + ciphertextB64: Buffer.from( + nacl.box(plaintext, nonce, input.hostKeypair.publicKey, gatewayKeys.secretKey) + ).toString('base64'), + expiresAt, + ...input.challenge + }, + context: { + gatewayOrigin: input.gatewayOrigin, + hostFingerprint: input.hostFingerprint, + hostPublicKey: input.hostKeypair.publicKey, + hostSecretKey: input.hostKeypair.secretKey + }, + proof: pushAckProof(secret, transcript) + } +} diff --git a/src/main/runtime/push/push-host-proof-vector.test.ts b/src/main/runtime/push/push-host-proof-vector.test.ts new file mode 100644 index 00000000000..6a012d9cd05 --- /dev/null +++ b/src/main/runtime/push/push-host-proof-vector.test.ts @@ -0,0 +1,30 @@ +import { describe, expect, it } from 'vitest' +import { createHmac } from 'node:crypto' +import vector from '../../../../cloud/packages/push-contract/src/push-host-proof-vector.json' +import { answerPushHostChallenge } from './push-host-proof' + +// Why: the gateway builds the challenge and this file answers it, in two +// workspaces that cannot import each other in CI. Both replay one checked-in +// vector; a transcript field drift on either side fails here and in the +// gateway's copy of this test. +describe('push host proof vector', () => { + it('answers the checked-in gateway challenge with the expected proof', () => { + const secret = Buffer.from(vector.challengeSecretB64, 'base64') + const transcript = Buffer.from(vector.transcriptB64, 'base64') + const expected = createHmac('sha256', secret) + .update(Buffer.from('orca-push-host-proof/v1\0ack\0')) + .update(transcript) + .digest('base64') + const reasons: string[] = [] + const proof = answerPushHostChallenge(vector.challenge, { + gatewayOrigin: vector.gatewayOrigin, + hostFingerprint: vector.hostFingerprint, + hostPublicKey: Buffer.from(vector.hostPublicKeyB64, 'base64'), + hostSecretKey: Buffer.from(vector.hostSecretKeyB64, 'base64'), + now: () => vector.issuedAt + 1_000, + onInvalid: (reason) => reasons.push(reason) + }) + expect(reasons).toEqual([]) + expect(proof).toBe(expected) + }) +}) diff --git a/src/main/runtime/push/push-host-proof.test.ts b/src/main/runtime/push/push-host-proof.test.ts new file mode 100644 index 00000000000..7ec59f3a1b4 --- /dev/null +++ b/src/main/runtime/push/push-host-proof.test.ts @@ -0,0 +1,106 @@ +import { describe, expect, it } from 'vitest' +import nacl from 'tweetnacl' +import { + buildPushChallengeFixture, + createPushHostKeypair, + type PushTranscriptInput +} from './push-host-challenge-fixtures' +import { answerPushHostChallenge, type PushHostProofContext } from './push-host-proof' + +const GATEWAY_ORIGIN = 'https://push.onorca.dev' +const HOST_FINGERPRINT = 'abcdef0123456789' +const ISSUED_AT = 1_770_000_000_000 + +function fixture( + overrides: { + transcript?: Partial + challenge?: Partial[0]> + context?: Partial + } = {} +): { + challenge: Parameters[0] + context: PushHostProofContext + proof: string +} { + const built = buildPushChallengeFixture({ + hostKeypair: createPushHostKeypair(), + gatewayOrigin: GATEWAY_ORIGIN, + hostFingerprint: HOST_FINGERPRINT, + issuedAt: ISSUED_AT, + transcript: overrides.transcript, + challenge: overrides.challenge + }) + return { + challenge: built.challenge, + context: { ...built.context, now: () => ISSUED_AT + 1_000, ...overrides.context }, + proof: built.proof + } +} + +describe('answerPushHostChallenge', () => { + it('answers a well-formed challenge with the ack HMAC', () => { + const { challenge, context, proof } = fixture() + expect(answerPushHostChallenge(challenge, context)).toBe(proof) + }) + + it('tolerates clock skew inside the 30s allowance', () => { + const { challenge, context, proof } = fixture({ context: { now: () => ISSUED_AT - 20_000 } }) + expect(answerPushHostChallenge(challenge, context)).toBe(proof) + }) + + it('refuses a challenge whose secret was sealed to another host', () => { + const { challenge, context } = fixture() + expect( + answerPushHostChallenge(challenge, { + ...context, + hostSecretKey: nacl.box.keyPair().secretKey + }) + ).toBeNull() + }) + + it.each([ + ['gatewayOrigin', { gatewayOrigin: 'https://push.evil.example' }], + ['hostFingerprint', { hostFingerprint: 'ffffffffffffffff' }], + ['challengeId', { challengeId: 'challenge-other' }], + ['issuedAt', { issuedAt: ISSUED_AT + 120_000 }] + ] as const)('refuses a transcript whose %s does not match the challenge', (_name, transcript) => { + const invalid: string[] = [] + const { challenge, context } = fixture({ + transcript, + context: { onInvalid: (reason) => invalid.push(reason) } + }) + expect(answerPushHostChallenge(challenge, context)).toBeNull() + expect(invalid.join(',')).toContain('transcript') + }) + + it('refuses a transcript that swaps in a different gateway ephemeral key', () => { + const { challenge, context } = fixture({ + transcript: { gatewayKey: nacl.box.keyPair().publicKey } + }) + expect(answerPushHostChallenge(challenge, context)).toBeNull() + }) + + it('refuses an expired challenge beyond the skew allowance', () => { + const { challenge, context } = fixture({ + context: { now: () => ISSUED_AT + 10_000 + 30_001 } + }) + expect(answerPushHostChallenge(challenge, context)).toBeNull() + }) + + it('refuses a challenge whose declared expiry disagrees with the transcript', () => { + const { challenge, context } = fixture() + expect( + answerPushHostChallenge({ ...challenge, expiresAt: challenge.expiresAt + 1 }, context) + ).toBeNull() + }) + + it('refuses a non-canonical base64 ephemeral key without opening the box', () => { + const { challenge, context } = fixture() + expect( + answerPushHostChallenge( + { ...challenge, gatewayEphemeralPublicKeyB64: 'not base64!' }, + context + ) + ).toBeNull() + }) +}) diff --git a/src/main/runtime/push/push-host-proof.ts b/src/main/runtime/push/push-host-proof.ts new file mode 100644 index 00000000000..a48eaade01f --- /dev/null +++ b/src/main/runtime/push/push-host-proof.ts @@ -0,0 +1,113 @@ +// Why: the push gateway authenticates this host the same way the relay does — +// a sealed box the host can only open with its X25519 E2EE secret key — but with +// its own domain strings and a transcript that names the host by fingerprint +// instead of by account. See docs/reference/mobile-push-contract.md. +import { + encodeText, + equalBytes, + hostChallengeAckProof, + openHostChallengeEnvelope, + parseHostChallengeTranscript, + readTranscriptUint64 +} from '../host-challenge-envelope' + +const PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN = 'orca-push-host-proof/v1' +const PUSH_HOST_CHALLENGE_PLAINTEXT_DOMAIN = 'orca-push-host-challenge/v1' +const PUSH_HOST_PROOF_CLOCK_SKEW_MS = 30_000 +const MAX_PUSH_HOST_PROOF_CHALLENGE_WINDOW_MS = 10_000 +const PUSH_HOST_PROOF_TRANSCRIPT_FIELD_COUNT = 10 + +export type PushHostChallenge = { + challengeId: string + gatewayEphemeralPublicKeyB64: string + nonceB64: string + ciphertextB64: string + expiresAt: number +} + +export type PushHostProofContext = { + gatewayOrigin: string + hostFingerprint: string + hostPublicKey: Uint8Array + hostSecretKey: Uint8Array + now?: () => number + /** Reports the failing check by name only; never receives field values. */ + onInvalid?: (reason: string) => void +} + +function validateTranscript( + transcript: Uint8Array, + challenge: PushHostChallenge, + context: PushHostProofContext, + gatewayKey: Uint8Array, + nonce: Uint8Array +): boolean { + const fields = parseHostChallengeTranscript(transcript) + if (!fields || fields.size !== PUSH_HOST_PROOF_TRANSCRIPT_FIELD_COUNT) { + context.onInvalid?.('transcript-structure') + return false + } + const now = (context.now ?? Date.now)() + const issuedAt = readTranscriptUint64(fields.get('issuedAt')) + const expiresAt = readTranscriptUint64(fields.get('expiresAt')) + const checks: [string, boolean][] = [ + ['issuedAt-readable', issuedAt !== null], + ['issuedAt-not-future', issuedAt === null || issuedAt - PUSH_HOST_PROOF_CLOCK_SKEW_MS <= now], + ['not-expired', now - PUSH_HOST_PROOF_CLOCK_SKEW_MS <= challenge.expiresAt], + ['issuedAt-before-expiry', issuedAt === null || issuedAt <= challenge.expiresAt], + [ + 'window', + issuedAt === null || challenge.expiresAt - issuedAt <= MAX_PUSH_HOST_PROOF_CHALLENGE_WINDOW_MS + ], + ['expiry-consistent', expiresAt === challenge.expiresAt], + ['protocol', equalBytes(fields.get('protocol'), encodeText(PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN))], + ['version', equalBytes(fields.get('version'), new Uint8Array([1]))], + ['gatewayOrigin', equalBytes(fields.get('gatewayOrigin'), encodeText(context.gatewayOrigin))], + ['gatewayEphemeralPublicKey', equalBytes(fields.get('gatewayEphemeralPublicKey'), gatewayKey)], + ['challengeNonce', equalBytes(fields.get('challengeNonce'), nonce)], + ['challengeId', equalBytes(fields.get('challengeId'), encodeText(challenge.challengeId))], + [ + 'hostFingerprint', + equalBytes(fields.get('hostFingerprint'), encodeText(context.hostFingerprint)) + ], + ['hostPublicKey', equalBytes(fields.get('hostPublicKey'), context.hostPublicKey)] + ] + const failed = checks.filter(([, ok]) => !ok).map(([name]) => name) + if (failed.length > 0) { + context.onInvalid?.(`transcript:${failed.join('+')}`) + return false + } + return true +} + +/** Returns the base64 HMAC proof for a valid challenge, or null for anything else. */ +export function answerPushHostChallenge( + challenge: PushHostChallenge, + context: PushHostProofContext +): string | null { + const envelope = openHostChallengeEnvelope({ + peerEphemeralPublicKeyB64: challenge.gatewayEphemeralPublicKeyB64, + nonceB64: challenge.nonceB64, + ciphertextB64: challenge.ciphertextB64, + hostSecretKey: context.hostSecretKey, + plaintextDomain: PUSH_HOST_CHALLENGE_PLAINTEXT_DOMAIN, + onInvalid: context.onInvalid + }) + if ( + !envelope || + !validateTranscript( + envelope.transcript, + challenge, + context, + envelope.peerEphemeralPublicKey, + envelope.nonce + ) + ) { + return null + } + return hostChallengeAckProof({ + secret: envelope.secret, + transcript: envelope.transcript, + proofDomain: PUSH_HOST_PROOF_TRANSCRIPT_DOMAIN + }) +} diff --git a/src/main/runtime/push/push-outcome-counters.test.ts b/src/main/runtime/push/push-outcome-counters.test.ts new file mode 100644 index 00000000000..67ccc475cfc --- /dev/null +++ b/src/main/runtime/push/push-outcome-counters.test.ts @@ -0,0 +1,25 @@ +import { expect, it, vi } from 'vitest' +import { PushOutcomeCounters } from './push-outcome-counters' +it('limits failure logs while retaining category counts', () => { + let now = 0 + const log = vi.spyOn(console, 'warn').mockImplementation(() => {}) + try { + const counters = new PushOutcomeCounters(() => now) + counters.record('rejected') + counters.record('error') + counters.record('error') + expect(log).toHaveBeenCalledTimes(1) + now += 60_000 + counters.record('rate_limited') + expect(JSON.parse(String(log.mock.calls[1]![0]))).toEqual({ + event: 'orca_desktop_push_failures', + error: 2, + rate_limited: 1 + }) + counters.record('unreachable') + counters.flush() + expect(log).toHaveBeenCalledTimes(3) + } finally { + log.mockRestore() + } +}) diff --git a/src/main/runtime/push/push-outcome-counters.ts b/src/main/runtime/push/push-outcome-counters.ts new file mode 100644 index 00000000000..6b2507e5a18 --- /dev/null +++ b/src/main/runtime/push/push-outcome-counters.ts @@ -0,0 +1,27 @@ +type PushOutcome = 'error' | 'rate_limited' | 'rejected' | 'unreachable' + +export class PushOutcomeCounters { + private readonly counts = new Map() + private nextLogAt = 0 + + constructor(private readonly now: () => number = Date.now) {} + + record(outcome: PushOutcome): void { + this.counts.set(outcome, (this.counts.get(outcome) ?? 0) + 1) + if (this.now() < this.nextLogAt) { + return + } + this.nextLogAt = this.now() + 60_000 + this.flush() + } + + flush(): void { + if (!this.counts.size) { + return + } + console.warn( + JSON.stringify({ event: 'orca_desktop_push_failures', ...Object.fromEntries(this.counts) }) + ) + this.counts.clear() + } +} diff --git a/src/main/runtime/push/push-preferences.test.ts b/src/main/runtime/push/push-preferences.test.ts new file mode 100644 index 00000000000..8ab84fbea65 --- /dev/null +++ b/src/main/runtime/push/push-preferences.test.ts @@ -0,0 +1,87 @@ +import { expect, it } from 'vitest' +import { createHarness, notification, registration, flush } from './push-dispatcher.test-fixture' + +it('routes a desktop-disabled bell only to a phone that independently permits bells', async () => { + const filter = registration().filter + const harness = createHarness({ + devices: [ + { + deviceId: 'mirror', + pushRegistration: registration({ + registrationId: 'mirror', + filter: { ...filter, followDesktop: true } + }) + }, + { + deviceId: 'override', + pushRegistration: registration({ + registrationId: 'override', + filter: { ...filter, followDesktop: false, sound: false } + }) + }, + { + deviceId: 'no-bells', + pushRegistration: registration({ + registrationId: 'no-bells', + filter: { ...filter, followDesktop: false, sources: ['agent-task-complete'] } + }) + } + ] + }) + harness.dispatcher.enqueue(notification({ source: 'terminal-bell', desktopAllowed: false })) + await flush() + expect(harness.sends).toHaveLength(1) + expect(harness.sends[0]).toMatchObject({ + registrationIds: ['override'], + notification: { sound: false } + }) +}) + +it('keeps sound preferences separate when several phones receive the same event', async () => { + const harness = createHarness({ + devices: [ + { deviceId: 'loud', pushRegistration: registration({ registrationId: 'loud' }) }, + { + deviceId: 'quiet', + pushRegistration: registration({ + registrationId: 'quiet', + filter: { ...registration().filter, sound: false } + }) + } + ] + }) + harness.dispatcher.enqueue(notification()) + await flush() + expect(harness.sends).toHaveLength(2) + expect(harness.sends[0]).toMatchObject({ registrationIds: ['loud'] }) + expect(harness.sends[0].notification.sound).toBeUndefined() + expect(harness.sends[1]).toMatchObject({ + registrationIds: ['quiet'], + notification: { sound: false } + }) +}) + +it('applies burst suppression after each phone filters event types', async () => { + const harness = createHarness({ + devices: [ + { + deviceId: 'all', + pushRegistration: registration({ + registrationId: 'all', + filter: { ...registration().filter, followDesktop: false } + }) + }, + { + deviceId: 'no-bells', + pushRegistration: registration({ + registrationId: 'no-bells', + filter: { ...registration().filter, sources: ['agent-task-complete'] } + }) + } + ] + }) + harness.dispatcher.enqueue(notification({ source: 'terminal-bell', emittedAt: 10000 })) + harness.dispatcher.enqueue(notification({ emittedAt: 10250 })) + await flush() + expect(harness.sends.map((send) => send.registrationIds)).toEqual([['all'], ['no-bells']]) +}) diff --git a/src/main/runtime/push/push-register-throttle.ts b/src/main/runtime/push/push-register-throttle.ts new file mode 100644 index 00000000000..7cc31bbb11d --- /dev/null +++ b/src/main/runtime/push/push-register-throttle.ts @@ -0,0 +1,45 @@ +// Why: notifications.registerPush costs a gateway write and a synchronous +// registry write on the main thread, and a paired phone may call it as often +// as it likes. A phone legitimately registers on switch-on, on each host +// connect, and on a token change, so a small per-device bucket bounds a loop +// without getting in the way of any of those. +const DEFAULT_CAPACITY = 10 +const DEFAULT_WINDOW_MS = 60_000 + +type Bucket = { tokens: number; updatedAt: number } + +export type PushRegisterThrottleOptions = { + capacity?: number + windowMs?: number + now?: () => number +} + +export class PushRegisterThrottle { + private readonly buckets = new Map() + private readonly capacity: number + private readonly windowMs: number + private readonly now: () => number + + constructor(options: PushRegisterThrottleOptions = {}) { + this.capacity = options.capacity ?? DEFAULT_CAPACITY + this.windowMs = options.windowMs ?? DEFAULT_WINDOW_MS + this.now = options.now ?? Date.now + } + + allow(deviceId: string): boolean { + const now = this.now() + const bucket = this.buckets.get(deviceId) + const refilled = bucket + ? Math.min( + this.capacity, + bucket.tokens + Math.max(0, ((now - bucket.updatedAt) * this.capacity) / this.windowMs) + ) + : this.capacity + if (refilled < 1) { + this.buckets.set(deviceId, { tokens: refilled, updatedAt: now }) + return false + } + this.buckets.set(deviceId, { tokens: refilled - 1, updatedAt: now }) + return true + } +} diff --git a/src/main/runtime/push/push-registration-races.test.ts b/src/main/runtime/push/push-registration-races.test.ts new file mode 100644 index 00000000000..afdba983a58 --- /dev/null +++ b/src/main/runtime/push/push-registration-races.test.ts @@ -0,0 +1,160 @@ +import { mkdtempSync, rmSync } from 'node:fs' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { afterEach, expect, it, vi } from 'vitest' +import { DeviceRegistry } from '../device-registry' +import { DesktopPushService } from './desktop-push-service' +import { PushUnregisterOutbox } from './push-unregister-outbox' +import { createPushHostKeypair } from './push-host-challenge-fixtures' +import { PushDispatcher } from './push-dispatcher' + +const paths: string[] = [] +afterEach(() => { + for (const path of paths.splice(0)) { + rmSync(path, { recursive: true, force: true }) + } +}) +const input = { + platform: 'android' as const, + token: 'synthetic', + filter: { sources: ['plugin'] as const, agentStates: [] } +} +const tick = () => new Promise((resolve) => setImmediate(resolve)) + +function harness() { + const path = mkdtempSync(join(tmpdir(), 'push-races-')) + paths.push(path) + const registry = new DeviceRegistry(path) + const deviceId = registry.addDevice('phone', 'mobile').deviceId + const outbox = new PushUnregisterOutbox(path) + let live = false + let reachable = true + const client = { + registerDevice: vi.fn(async () => { + live = true + return { ok: true, registrationId: 'stable-id' } + }), + deleteDevice: vi.fn(async () => { + if (!reachable) { + return { deleted: false, retryable: true } + } + live = false + return { deleted: true, retryable: false } + }), + send: vi.fn() + } + const service = DesktopPushService.create({ + gatewayUrl: 'https://push.example.test', + client: client as never, + scheduleRetry: () => {}, + runtime: { + setMobilePushRegistrar: () => {}, + onNotificationDispatched: () => () => {} + } as never, + runtimeRpc: { + getE2EEKeypair: createPushHostKeypair, + getDeviceRegistry: () => registry, + getPushUnregisterOutbox: () => outbox, + setOnPushUnregisterQueued: () => {} + } as never + })! + service.start() + return { + registry, + deviceId, + outbox, + client, + service, + live: () => live, + reachable: (value: boolean) => { + reachable = value + } + } +} + +it('deletes obsolete gateway state before reporting successful re-enable', async () => { + const h = harness() + await h.service.register({ ...input, deviceId: h.deviceId }) + h.reachable(false) + await h.service.unregister(h.deviceId) + await h.service.flushUnregisterOutbox() + expect(h.outbox.pending()).toHaveLength(1) + expect(await h.service.register({ ...input, deviceId: h.deviceId })).toMatchObject({ + registered: false + }) + h.reachable(true) + expect(await h.service.register({ ...input, deviceId: h.deviceId })).toMatchObject({ + registered: true + }) + await h.service.flushUnregisterOutbox() + expect(h.live()).toBe(true) + expect(h.outbox.pending()).toEqual([]) +}) + +it('waits for an already-running delete before re-registering', async () => { + const h = harness() + await h.service.register({ ...input, deviceId: h.deviceId }) + let release!: () => void + const normalDelete = h.client.deleteDevice.getMockImplementation()! + h.client.deleteDevice.mockImplementationOnce(async () => { + await new Promise((resolve) => { + release = resolve + }) + return normalDelete() + }) + await h.service.unregister(h.deviceId) + await tick() + const registration = h.service.register({ ...input, deviceId: h.deviceId }) + await tick() + expect(h.client.registerDevice).toHaveBeenCalledTimes(1) + release() + await registration + await h.service.flushUnregisterOutbox() + expect(h.live()).toBe(true) +}) + +it('orders unregister after a register already in flight', async () => { + const h = harness() + let release!: () => void + const normalRegister = h.client.registerDevice.getMockImplementation()! + h.client.registerDevice.mockImplementationOnce(async () => { + await new Promise((resolve) => { + release = resolve + }) + return normalRegister() + }) + const registered = h.service.register({ ...input, deviceId: h.deviceId }) + await tick() + const unregistered = h.service.unregister(h.deviceId) + release() + await Promise.all([registered, unregistered]) + await h.service.flushUnregisterOutbox() + expect(h.registry.getDevice(h.deviceId)?.pushRegistration).toBeUndefined() + expect(h.live()).toBe(false) +}) + +it('does not clear a replacement with the same ID and timestamp after a stale dead response', async () => { + const h = harness() + await h.service.register({ ...input, deviceId: h.deviceId }) + let finish!: (value: unknown) => void + h.client.send.mockImplementation( + () => + new Promise((resolve) => { + finish = resolve + }) + ) + const dispatcher = new PushDispatcher({ registry: h.registry, client: h.client as never }) + dispatcher.enqueue({ + type: 'notification', + source: 'plugin', + title: 'test', + body: '', + notificationEpoch: 'epoch', + notificationSeq: 1 + }) + const original = h.registry.getDevice(h.deviceId)!.pushRegistration! + h.registry.setPushRegistration(h.deviceId, { ...original }) + finish({ ok: true, results: [{ registrationId: 'stable-id', status: 'dead' }] }) + await tick() + expect(h.registry.getDevice(h.deviceId)?.pushRegistration).toEqual(original) +}) diff --git a/src/main/runtime/push/push-registration-rpc.test.ts b/src/main/runtime/push/push-registration-rpc.test.ts new file mode 100644 index 00000000000..cf7ba46b83c --- /dev/null +++ b/src/main/runtime/push/push-registration-rpc.test.ts @@ -0,0 +1,157 @@ +import { mkdtempSync } from 'node:fs' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { describe, expect, it, vi } from 'vitest' +import type { RpcContext, RpcMethod } from '../rpc/core' +import { NOTIFICATION_METHODS } from '../rpc/methods/notifications' +import { DeviceRegistry } from '../device-registry' +import { OrcaRuntimeRpcServer } from '../runtime-rpc' +import { OrcaRuntimeService } from '../orca-runtime' + +function method(name: string): RpcMethod { + const found = NOTIFICATION_METHODS.find((candidate) => candidate.name === name) + if (!found || 'stream' in found) { + throw new Error(`${name} is not a one-shot RPC method`) + } + return found +} + +const REGISTER_PARAMS = { + platform: 'ios', + token: 'a'.repeat(64), + apnsEnvironment: 'sandbox', + filter: { sources: ['agent-task-complete'], agentStates: ['finished'] } +} + +function contextFor(overrides: Partial): RpcContext { + return { + runtime: { + registerMobilePushDevice: vi.fn(async () => ({ + registered: true, + registrationId: 'reg-1' + })), + unregisterMobilePushDevice: vi.fn(async () => ({ unregistered: true })) + }, + ...overrides + } as unknown as RpcContext +} + +describe('notifications.registerPush', () => { + it('registers under the authenticated paired device id', async () => { + const registerPush = method('notifications.registerPush') + const ctx = contextFor({ clientKind: 'mobile', pairedDeviceId: 'device-1' }) + + const result = await registerPush.handler(registerPush.params!.parse(REGISTER_PARAMS), ctx) + + expect(result).toEqual({ registered: true, registrationId: 'reg-1' }) + expect(ctx.runtime.registerMobilePushDevice).toHaveBeenCalledWith({ + deviceId: 'device-1', + platform: 'ios', + token: REGISTER_PARAMS.token, + apnsEnvironment: 'sandbox', + filter: REGISTER_PARAMS.filter + }) + }) + + it.each([ + ['a runtime-scoped caller', { clientKind: 'runtime' as const, pairedDeviceId: 'device-1' }], + ['an in-process caller', {}], + ['a mobile caller with no paired device', { clientKind: 'mobile' as const }] + ])('refuses %s', async (_name, overrides) => { + const registerPush = method('notifications.registerPush') + const ctx = contextFor(overrides) + + expect(await registerPush.handler(registerPush.params!.parse(REGISTER_PARAMS), ctx)).toEqual({ + registered: false, + reason: 'not_mobile' + }) + expect(ctx.runtime.registerMobilePushDevice).not.toHaveBeenCalled() + }) + + it('requires an APNs environment for an iOS token', () => { + const registerPush = method('notifications.registerPush') + expect( + registerPush.params!.safeParse({ ...REGISTER_PARAMS, apnsEnvironment: undefined }).success + ).toBe(false) + expect( + registerPush.params!.safeParse({ + ...REGISTER_PARAMS, + platform: 'android', + apnsEnvironment: undefined + }).success + ).toBe(true) + }) + + it('rejects a caller-supplied device id instead of dropping it', () => { + const registerPush = method('notifications.registerPush') + expect( + registerPush.params!.safeParse({ ...REGISTER_PARAMS, deviceId: 'device-9' }).success + ).toBe(false) + }) + + it('rejects a source the contract does not define', () => { + const registerPush = method('notifications.registerPush') + expect( + registerPush.params!.safeParse({ + ...REGISTER_PARAMS, + filter: { sources: ['smoke-signal'], agentStates: [] } + }).success + ).toBe(false) + }) +}) + +describe('notifications.unregisterPush', () => { + it('unregisters the authenticated paired device', async () => { + const unregisterPush = method('notifications.unregisterPush') + const ctx = contextFor({ clientKind: 'mobile', pairedDeviceId: 'device-1' }) + + expect(await unregisterPush.handler(undefined, ctx)).toEqual({ unregistered: true }) + expect(ctx.runtime.unregisterMobilePushDevice).toHaveBeenCalledWith('device-1') + }) + + it('refuses a non-mobile caller', async () => { + const unregisterPush = method('notifications.unregisterPush') + const ctx = contextFor({ clientKind: 'runtime', pairedDeviceId: 'device-1' }) + + expect(await unregisterPush.handler(undefined, ctx)).toEqual({ unregistered: false }) + expect(ctx.runtime.unregisterMobilePushDevice).not.toHaveBeenCalled() + }) +}) + +describe('revokeMobileDevice', () => { + it('queues the gateway delete before the device row disappears', async () => { + const userDataPath = mkdtempSync(join(tmpdir(), 'orca-push-revoke-')) + const server = new OrcaRuntimeRpcServer({ + runtime: new OrcaRuntimeService(), + userDataPath, + enableWebSocket: false + }) + server['deviceRegistry'] = new DeviceRegistry(userDataPath) + const device = server['deviceRegistry']!.addDevice('phone', 'mobile') + server['deviceRegistry']!.setPushRegistration(device.deviceId, { + registrationId: 'reg-1', + platform: 'android', + filter: { sources: ['agent-task-complete'], agentStates: ['finished'] }, + registeredAt: 1 + }) + + expect(await server.revokeMobileDevice(device.deviceId)).toBe(true) + expect(server.getPushUnregisterOutbox().pending()).toEqual([ + expect.objectContaining({ registrationId: 'reg-1', deviceId: device.deviceId }) + ]) + }) + + it('queues nothing for a device that never enabled push', async () => { + const userDataPath = mkdtempSync(join(tmpdir(), 'orca-push-revoke-')) + const server = new OrcaRuntimeRpcServer({ + runtime: new OrcaRuntimeService(), + userDataPath, + enableWebSocket: false + }) + server['deviceRegistry'] = new DeviceRegistry(userDataPath) + const device = server['deviceRegistry']!.addDevice('phone', 'mobile') + + expect(await server.revokeMobileDevice(device.deviceId)).toBe(true) + expect(server.getPushUnregisterOutbox().pending()).toEqual([]) + }) +}) diff --git a/src/main/runtime/push/push-unregister-outbox.test.ts b/src/main/runtime/push/push-unregister-outbox.test.ts new file mode 100644 index 00000000000..f0ca35fa144 --- /dev/null +++ b/src/main/runtime/push/push-unregister-outbox.test.ts @@ -0,0 +1,64 @@ +import { mkdtempSync, readFileSync, writeFileSync } from 'node:fs' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { describe, expect, it } from 'vitest' +import { PushUnregisterOutbox } from './push-unregister-outbox' + +const OUTBOX_FILENAME = 'mobile-push-unregister-outbox.json' + +function userDataDir(): string { + return mkdtempSync(join(tmpdir(), 'orca-push-outbox-')) +} + +describe('PushUnregisterOutbox', () => { + it('survives a restart with the queued delete intact', () => { + const dir = userDataDir() + const first = new PushUnregisterOutbox(dir) + const item = first.enqueue({ registrationId: 'reg-1', deviceId: 'device-1' }) + + const reopened = new PushUnregisterOutbox(dir) + expect(reopened.pending()).toEqual([item]) + }) + + it('coalesces repeat enqueues of the same registration', () => { + const dir = userDataDir() + const outbox = new PushUnregisterOutbox(dir) + const first = outbox.enqueue({ registrationId: 'reg-1', deviceId: 'device-1' }) + const second = outbox.enqueue({ registrationId: 'reg-1', deviceId: 'device-1' }) + + expect(second.reqId).toBe(first.reqId) + expect(outbox.pending()).toHaveLength(1) + }) + + it('keeps a removal durable across a restart', () => { + const dir = userDataDir() + const outbox = new PushUnregisterOutbox(dir) + const kept = outbox.enqueue({ registrationId: 'reg-keep', deviceId: 'device-1' }) + const dropped = outbox.enqueue({ registrationId: 'reg-drop', deviceId: 'device-2' }) + outbox.remove(dropped.reqId) + + expect(new PushUnregisterOutbox(dir).pending()).toEqual([kept]) + }) + + it('drops malformed rows instead of failing the whole load', () => { + const dir = userDataDir() + const valid = new PushUnregisterOutbox(dir).enqueue({ + registrationId: 'reg-1', + deviceId: 'device-1' + }) + const path = join(dir, OUTBOX_FILENAME) + const stored: unknown[] = JSON.parse(readFileSync(path, 'utf-8')) + writeFileSync( + path, + JSON.stringify([...stored, { reqId: 'broken' }, null, 'nope', { registrationId: '' }]) + ) + + expect(new PushUnregisterOutbox(dir).pending()).toEqual([valid]) + }) + + it('starts empty when the file is not JSON at all', () => { + const dir = userDataDir() + writeFileSync(join(dir, OUTBOX_FILENAME), 'not json') + expect(new PushUnregisterOutbox(dir).pending()).toEqual([]) + }) +}) diff --git a/src/main/runtime/push/push-unregister-outbox.ts b/src/main/runtime/push/push-unregister-outbox.ts new file mode 100644 index 00000000000..a5b4bd1d989 --- /dev/null +++ b/src/main/runtime/push/push-unregister-outbox.ts @@ -0,0 +1,83 @@ +// Why: a phone that turns background notifications off, or gets unpaired, must +// have its token deleted at the gateway even if the gateway is unreachable right +// then. Modelled on relay-revoke-outbox.ts: durable, hardened, drained on start. +import { randomUUID } from 'node:crypto' +import { existsSync, readFileSync } from 'node:fs' +import { join } from 'node:path' +import { hardenExistingSecureFile, writeSecureJsonFile } from '../../../shared/secure-file' + +export type PushUnregisterOutboxItem = { + reqId: string + registrationId: string + deviceId: string + createdAt: number +} + +const OUTBOX_FILENAME = 'mobile-push-unregister-outbox.json' + +function isItem(value: unknown): value is PushUnregisterOutboxItem { + if (!value || typeof value !== 'object') { + return false + } + const item = value as Partial + return ( + typeof item.reqId === 'string' && + typeof item.registrationId === 'string' && + item.registrationId.length > 0 && + typeof item.deviceId === 'string' && + typeof item.createdAt === 'number' && + Number.isFinite(item.createdAt) + ) +} + +export class PushUnregisterOutbox { + private readonly path: string + private items: PushUnregisterOutboxItem[] + + constructor(userDataPath: string) { + this.path = join(userDataPath, OUTBOX_FILENAME) + this.items = this.load() + } + + enqueue(entry: { registrationId: string; deviceId: string }): PushUnregisterOutboxItem { + const existing = this.items.find((item) => item.registrationId === entry.registrationId) + if (existing) { + return existing + } + const item = { ...entry, reqId: randomUUID(), createdAt: Date.now() } + const next = [...this.items, item] + this.save(next) + this.items = next + return item + } + + pending(): readonly PushUnregisterOutboxItem[] { + return this.items + } + + remove(reqId: string): void { + const next = this.items.filter((item) => item.reqId !== reqId) + if (next.length === this.items.length) { + return + } + this.save(next) + this.items = next + } + + private load(): PushUnregisterOutboxItem[] { + if (!existsSync(this.path)) { + return [] + } + try { + hardenExistingSecureFile(this.path) + const parsed: unknown = JSON.parse(readFileSync(this.path, 'utf-8')) + return Array.isArray(parsed) ? parsed.filter(isItem) : [] + } catch { + return [] + } + } + + private save(items: readonly PushUnregisterOutboxItem[]): void { + writeSecureJsonFile(this.path, items) + } +} diff --git a/src/main/runtime/relay/relay-host-proof.ts b/src/main/runtime/relay/relay-host-proof.ts index 59c028b1ab1..a169540b5ee 100644 --- a/src/main/runtime/relay/relay-host-proof.ts +++ b/src/main/runtime/relay/relay-host-proof.ts @@ -1,13 +1,18 @@ -import { createHmac, timingSafeEqual } from 'node:crypto' -import nacl from 'tweetnacl' +import { + encodeText, + encodeUint64, + equalBytes, + hostChallengeAckProof, + openHostChallengeEnvelope, + parseHostChallengeTranscript, + readTranscriptUint64 +} from '../host-challenge-envelope' const HOST_PROOF_TRANSCRIPT_DOMAIN = 'orca-relay-host-proof/v1' const HOST_CHALLENGE_PLAINTEXT_DOMAIN = 'orca-relay-host-challenge/v1' // Covers routine NTP drift without extending the signed challenge window. const RELAY_HOST_PROOF_CLOCK_SKEW_MS = 30_000 const MAX_HOST_PROOF_CHALLENGE_WINDOW_MS = 10_000 -const textEncoder = new TextEncoder() -const textDecoder = new TextDecoder() export type RelayHostChallenge = { challengeId: string @@ -33,61 +38,6 @@ export type RelayHostProofContext = { onInvalid?: (reason: string) => void } -function decodeCanonicalBase64(value: string, expectedBytes: number): Uint8Array | null { - if (!/^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/.test(value)) { - return null - } - const decoded = Buffer.from(value, 'base64') - return decoded.byteLength === expectedBytes && decoded.toString('base64') === value - ? decoded - : null -} - -function uint64(value: number): Uint8Array { - const bytes = new Uint8Array(8) - new DataView(bytes.buffer).setBigUint64(0, BigInt(value), false) - return bytes -} - -function equal(left: Uint8Array | undefined, right: Uint8Array): boolean { - return Boolean(left && left.byteLength === right.byteLength && timingSafeEqual(left, right)) -} - -function parseTranscript(transcript: Uint8Array): Map | null { - const fields = new Map() - const view = new DataView(transcript.buffer, transcript.byteOffset, transcript.byteLength) - let offset = 0 - try { - while (offset < transcript.byteLength) { - const nameLength = view.getUint32(offset, false) - offset += 4 - const name = textDecoder.decode(transcript.slice(offset, offset + nameLength)) - offset += nameLength - const valueLength = view.getUint32(offset, false) - offset += 4 - if (fields.has(name) || offset + valueLength > transcript.byteLength) { - return null - } - fields.set(name, transcript.slice(offset, offset + valueLength)) - offset += valueLength - } - } catch { - return null - } - return offset === transcript.byteLength ? fields : null -} - -function readUint64(value: Uint8Array | undefined): number | null { - if (!value || value.byteLength !== 8) { - return null - } - const parsed = new DataView(value.buffer, value.byteOffset, value.byteLength).getBigUint64( - 0, - false - ) - return parsed <= BigInt(Number.MAX_SAFE_INTEGER) ? Number(parsed) : null -} - function validateTranscript( transcript: Uint8Array, challenge: RelayHostChallenge, @@ -95,17 +45,19 @@ function validateTranscript( relayKey: Uint8Array, nonce: Uint8Array ): boolean { - const fields = parseTranscript(transcript) + const fields = parseHostChallengeTranscript(transcript) if (!fields || fields.size !== 16) { context.onInvalid?.('transcript-structure') return false } const now = (context.now ?? Date.now)() - const issuedAt = readUint64(fields.get('issuedAt')) - const expiresAt = readUint64(fields.get('expiresAt')) + const issuedAt = readTranscriptUint64(fields.get('issuedAt')) + const expiresAt = readTranscriptUint64(fields.get('expiresAt')) const previousGeneration = fields.get('previousGeneration') const expectedPrevious = - context.previousGeneration === undefined ? new Uint8Array() : uint64(context.previousGeneration) + context.previousGeneration === undefined + ? new Uint8Array() + : encodeUint64(context.previousGeneration) // Main's 30s skew bounds with named-check reporting kept from the incident // instrumentation; deltas are relative offsets only, never absolute values. const checks: [string, boolean][] = [ @@ -124,25 +76,28 @@ function validateTranscript( issuedAt === null || challenge.expiresAt - issuedAt <= MAX_HOST_PROOF_CHALLENGE_WINDOW_MS ], ['expiry-consistent', expiresAt === challenge.expiresAt], - ['protocol', equal(fields.get('protocol'), textEncoder.encode(HOST_PROOF_TRANSCRIPT_DOMAIN))], - ['version', equal(fields.get('version'), new Uint8Array([1]))], - ['relayOrigin', equal(fields.get('relayOrigin'), textEncoder.encode(context.relayOrigin))], - ['relayEphemeralPublicKey', equal(fields.get('relayEphemeralPublicKey'), relayKey)], - ['challengeNonce', equal(fields.get('challengeNonce'), nonce)], - ['challengeId', equal(fields.get('challengeId'), textEncoder.encode(challenge.challengeId))], - ['userId', equal(fields.get('userId'), textEncoder.encode(context.userId))], - ['profileId', equal(fields.get('profileId'), textEncoder.encode(context.profileId))], + ['protocol', equalBytes(fields.get('protocol'), encodeText(HOST_PROOF_TRANSCRIPT_DOMAIN))], + ['version', equalBytes(fields.get('version'), new Uint8Array([1]))], + ['relayOrigin', equalBytes(fields.get('relayOrigin'), encodeText(context.relayOrigin))], + ['relayEphemeralPublicKey', equalBytes(fields.get('relayEphemeralPublicKey'), relayKey)], + ['challengeNonce', equalBytes(fields.get('challengeNonce'), nonce)], + ['challengeId', equalBytes(fields.get('challengeId'), encodeText(challenge.challengeId))], + ['userId', equalBytes(fields.get('userId'), encodeText(context.userId))], + ['profileId', equalBytes(fields.get('profileId'), encodeText(context.profileId))], [ 'organizationId', - equal(fields.get('organizationId'), textEncoder.encode(context.organizationId)) + equalBytes(fields.get('organizationId'), encodeText(context.organizationId)) ], - ['relayHostId', equal(fields.get('relayHostId'), textEncoder.encode(context.relayHostId))], - ['hostPublicKey', equal(fields.get('hostPublicKey'), context.hostPublicKey)], - ['assignmentEpoch', equal(fields.get('assignmentEpoch'), uint64(context.assignmentEpoch))], - ['previousGeneration', equal(previousGeneration, expectedPrevious)], + ['relayHostId', equalBytes(fields.get('relayHostId'), encodeText(context.relayHostId))], + ['hostPublicKey', equalBytes(fields.get('hostPublicKey'), context.hostPublicKey)], + [ + 'assignmentEpoch', + equalBytes(fields.get('assignmentEpoch'), encodeUint64(context.assignmentEpoch)) + ], + ['previousGeneration', equalBytes(previousGeneration, expectedPrevious)], [ 'resumeRequested', - equal(fields.get('resumeRequested'), new Uint8Array([context.resumeRequested ? 1 : 0])) + equalBytes(fields.get('resumeRequested'), new Uint8Array([context.resumeRequested ? 1 : 0])) ] ] const failed = checks.filter(([, ok]) => !ok).map(([name]) => name) @@ -157,41 +112,29 @@ export function answerRelayHostChallenge( challenge: RelayHostChallenge, context: RelayHostProofContext ): string | null { - const relayKey = decodeCanonicalBase64(challenge.relayEphemeralPublicKeyB64, 32) - const nonce = decodeCanonicalBase64(challenge.nonceB64, 24) - const ciphertext = Buffer.from(challenge.ciphertextB64, 'base64') - if (!relayKey || !nonce || ciphertext.toString('base64') !== challenge.ciphertextB64) { - return null - } - const plaintext = nacl.box.open(ciphertext, nonce, relayKey, context.hostSecretKey) - if (!plaintext) { - context.onInvalid?.('challenge-box-open') - return null - } - const domain = textEncoder.encode(`${HOST_CHALLENGE_PLAINTEXT_DOMAIN}\0`) + const envelope = openHostChallengeEnvelope({ + peerEphemeralPublicKeyB64: challenge.relayEphemeralPublicKeyB64, + nonceB64: challenge.nonceB64, + ciphertextB64: challenge.ciphertextB64, + hostSecretKey: context.hostSecretKey, + plaintextDomain: HOST_CHALLENGE_PLAINTEXT_DOMAIN, + onInvalid: context.onInvalid + }) if ( - !equal(plaintext.slice(0, domain.byteLength), domain) || - plaintext.byteLength < domain.byteLength + 36 + !envelope || + !validateTranscript( + envelope.transcript, + challenge, + context, + envelope.peerEphemeralPublicKey, + envelope.nonce + ) ) { return null } - const transcriptLength = new DataView( - plaintext.buffer, - plaintext.byteOffset + domain.byteLength, - 4 - ).getUint32(0, false) - const transcriptStart = domain.byteLength + 4 - const secretStart = transcriptStart + transcriptLength - if (secretStart + 32 !== plaintext.byteLength) { - return null - } - const transcript = plaintext.slice(transcriptStart, secretStart) - if (!validateTranscript(transcript, challenge, context, relayKey, nonce)) { - return null - } - const secret = plaintext.slice(secretStart) - return createHmac('sha256', secret) - .update(textEncoder.encode(`${HOST_PROOF_TRANSCRIPT_DOMAIN}\0ack\0`)) - .update(transcript) - .digest('base64') + return hostChallengeAckProof({ + secret: envelope.secret, + transcript: envelope.transcript, + proofDomain: HOST_PROOF_TRANSCRIPT_DOMAIN + }) } diff --git a/src/main/runtime/renderer-browser-session-reconciliation.test.ts b/src/main/runtime/renderer-browser-session-reconciliation.test.ts new file mode 100644 index 00000000000..7dc929c231d --- /dev/null +++ b/src/main/runtime/renderer-browser-session-reconciliation.test.ts @@ -0,0 +1,154 @@ +import { expect, it, vi } from 'vitest' +import type { + RuntimeMobileSessionBrowserTab, + RuntimeMobileSessionTabsSnapshot +} from '../../shared/runtime-types' +import { OrcaRuntimeWithCloseStructuredAgentSessionTab } from './orca-runtime-close-structured-agent-session-tab' +import { OrcaRuntimeWithReconcileHeadlessMobileSessionBrowserTabs } from './orca-runtime-reconcile-headless-mobile-session-browser-tabs' + +const rendererPage: RuntimeMobileSessionBrowserTab = { + type: 'browser', + id: 'renderer-tab', + browserWorkspaceId: 'renderer-workspace', + browserPageId: 'renderer-page', + title: 'Server page', + url: 'https://example.com/server', + loading: false, + canGoBack: false, + canGoForward: false, + isActive: false +} +const clientPage: RuntimeMobileSessionBrowserTab = { + ...rendererPage, + id: 'client', + browserWorkspaceId: 'client', + browserPageId: 'client', + placement: { + kind: 'client', + browserHostClientId: 'host', + browserHostGeneration: 1, + pageHostGeneration: 1 + } +} +const snapshot: RuntimeMobileSessionTabsSnapshot = { + worktree: 'wt', + publicationEpoch: 'renderer:1', + snapshotVersion: 1, + activeGroupId: 'group', + activeTabId: 'renderer-tab', + activeTabType: 'browser', + tabs: [rendererPage], + tabGroups: [{ id: 'group', activeTabId: 'renderer-tab', tabOrder: ['renderer-tab'] }] +} + +/** Drives the reconcile against a stub host and returns the published snapshot, if any. */ +function reconcile( + host: { + live?: RuntimeMobileSessionBrowserTab[] + attached?: boolean + offscreen?: boolean + }, + existing: RuntimeMobileSessionTabsSnapshot = snapshot +): RuntimeMobileSessionTabsSnapshot | undefined { + const storeMobileSessionSnapshot = vi.fn() + const runtime = OrcaRuntimeWithReconcileHeadlessMobileSessionBrowserTabs.prototype as unknown as { + reconcileHeadlessMobileSessionBrowserTabs( + worktreeId: string, + existing: RuntimeMobileSessionTabsSnapshot + ): void + } + runtime.reconcileHeadlessMobileSessionBrowserTabs.call( + { + buildHeadlessMobileSessionBrowserTabs: () => host.live ?? [], + getAvailableAuthoritativeWindow: () => (host.attached === false ? null : {}), + offscreenBrowserBackend: host.offscreen === true ? {} : null, + storeMobileSessionSnapshot + }, + 'wt', + existing + ) + return storeMobileSessionSnapshot.mock.calls[0]?.[1] +} + +it('keeps renderer-owned browser pages when refreshing client-hosted pages on an attached desktop', () => { + const published = reconcile({}) ?? snapshot + + expect(published.tabs).toContainEqual(rendererPage) + expect(published.tabGroups?.[0].tabOrder).toContain('renderer-tab') +}) + +it.each([false, true])('retires absent offscreen pages when attached=%s', (attached) => { + expect(reconcile({ attached, offscreen: true })?.tabs).toEqual([]) +}) + +it('removes retired client pages and publishes live ones while retaining renderer rows and group order', () => { + const livePage = { ...clientPage, id: 'live', browserWorkspaceId: 'live', browserPageId: 'live' } + + const published = reconcile( + { live: [livePage] }, + { + ...snapshot, + tabs: [rendererPage, clientPage], + tabGroups: [ + { id: 'group', activeTabId: 'renderer-tab', tabOrder: ['renderer-tab', 'client'] } + ] + } + ) + + expect(published?.tabs).toEqual([rendererPage, livePage]) + expect(published?.tabGroups?.[0].tabOrder).toEqual(['renderer-tab', 'live']) + expect(published?.activeTabId).toBe('renderer-tab') + expect(published?.publicationEpoch).toBe(snapshot.publicationEpoch) + expect(published?.snapshotVersion).toBe(snapshot.snapshotVersion + 1) +}) + +it('never publishes a row twice when the live build reclaims a renderer-owned id', () => { + const reclaimed = { + ...clientPage, + id: rendererPage.id, + browserPageId: rendererPage.browserPageId + } + + const published = reconcile({ live: [reclaimed] }) + + expect(published?.tabs).toEqual([reclaimed]) + expect(published?.tabGroups?.[0].tabOrder).toEqual([rendererPage.id]) +}) + +it('does not republish when a client row merely sits before a renderer row', () => { + const interleaved = { + ...snapshot, + tabs: [clientPage, rendererPage], + tabGroups: [{ id: 'group', activeTabId: 'renderer-tab', tabOrder: ['client', 'renderer-tab'] }] + } + + expect(reconcile({ live: [clientPage] }, interleaved)).toBeUndefined() +}) + +it('keeps the renderer publication epoch when selecting a client-hosted browser tab', () => { + const storeMobileSessionSnapshot = vi.fn() + const runtime = OrcaRuntimeWithCloseStructuredAgentSessionTab.prototype as unknown as { + markHeadlessBrowserSessionTabActive( + worktreeId: string, + browserPageId: string, + options: { focusesHost: boolean } + ): void + } + + runtime.markHeadlessBrowserSessionTabActive.call( + { + offscreenBrowserBackend: {}, + hydrateHeadlessMobileSessionTabsFromWorkspaceSession: () => undefined, + mobileSessionTabsByWorktree: new Map([['wt', snapshot]]), + storeMobileSessionSnapshot, + emitMobileSessionTabsSnapshot: vi.fn() + }, + 'wt', + 'renderer-page', + { focusesHost: false } + ) + + expect(storeMobileSessionSnapshot.mock.calls[0]?.[1].publicationEpoch).toBe( + snapshot.publicationEpoch + ) +}) diff --git a/src/main/runtime/rpc/errors.ts b/src/main/runtime/rpc/errors.ts index bbf918f4f55..ef4b0b3024d 100644 --- a/src/main/runtime/rpc/errors.ts +++ b/src/main/runtime/rpc/errors.ts @@ -90,6 +90,9 @@ const STRUCTURED_RUNTIME_PASSTHROUGH_CODES: ReadonlySet = new Set([ 'dispatch_not_found', 'dispatch_run_mismatch', 'terminal_not_found', + // A handle that names a live agent session with no terminal. Distinct from + // `terminal_handle_stale`, which claims the handle went dead — nothing went stale here. + 'terminal_unsupported_for_agent_session', 'recipient_ambiguous', 'recipient_run_mismatch', 'dispatch_inactive', diff --git a/src/main/runtime/rpc/methods/browser-client-host-attach-adoption.test.ts b/src/main/runtime/rpc/methods/browser-client-host-attach-adoption.test.ts index 51fa6304d2d..0d6bf3b80ef 100644 --- a/src/main/runtime/rpc/methods/browser-client-host-attach-adoption.test.ts +++ b/src/main/runtime/rpc/methods/browser-client-host-attach-adoption.test.ts @@ -183,6 +183,52 @@ describe('browser.clientHost.attach adoption', () => { await rig.dispatch }) + it('does not recover a page created after the attach inventory was captured', async () => { + let releaseAdoption!: (value: BrowserExecutionHostKeyResolution) => void + const route = new Promise((resolve) => { + releaseAdoption = resolve + }) + const resolveExecutionHostKey = vi.fn(() => route) + const rig = attachHost([orphanedPage()], { resolveExecutionHostKey }) + const settleAttach = async (): Promise => { + rig.cleanups.get(`browser-client-host:${HOST_CLIENT_ID}`)?.() + await rig.dispatch + } + try { + await vi.waitFor(() => expect(resolveExecutionHostKey).toHaveBeenCalled()) + const authority = getBrowserHostLeaseRegistry(rig.hostRuntime) + const pages = getRuntimeBrowserPageRegistry(rig.hostRuntime) + const placement = authority.placeClientPage('page-created-after-attach', HOST_CLIENT_ID) + if (placement.kind !== 'client') { + throw new Error('expected client placement') + } + pages.publishClientPage({ + browserPageId: 'page-created-after-attach', + workspaceId: WORKSPACE_ID, + browserProfileId: 'default', + executionHostKey: EXECUTION_HOST_KEY, + placement, + pairedDeviceId: 'device-a', + url: 'https://remote.internal/new', + loading: false, + active: true + }) + releaseAdoption({ status: 'resolved', executionHostKey: EXECUTION_HOST_KEY }) + await vi.waitFor(() => expect(rig.markClientHostedPagesReconciled).toHaveBeenCalled()) + // Attach only settles once recovery has returned, so this is the barrier the assertions need: + // draining microtasks would let a regression slip through as a not-yet-issued command. + await settleAttach() + + expect(authority.getPlacement('page-created-after-attach')).toEqual(placement) + expect(pages.getPage('page-created-after-attach')).toMatchObject({ placement, active: true }) + expect( + rig.commands().filter((event) => event.browserPageId === 'page-created-after-attach') + ).toEqual([]) + } finally { + await settleAttach() + } + }) + it('does not re-enter recovery for a page it just adopted', async () => { const warn = vi.spyOn(console, 'warn').mockImplementation(() => {}) const rig = attachHost([orphanedPage({ browserPageId: 'page-d' })], { diff --git a/src/main/runtime/rpc/methods/browser-client-host.ts b/src/main/runtime/rpc/methods/browser-client-host.ts index 5126a870d24..525fd4fde96 100644 --- a/src/main/runtime/rpc/methods/browser-client-host.ts +++ b/src/main/runtime/rpc/methods/browser-client-host.ts @@ -39,6 +39,12 @@ export const BROWSER_CLIENT_HOST_METHODS: RpcAnyMethod[] = [ } const registry = getBrowserHostLeaseRegistry(runtime) + // Attach inventory cannot describe pages created or replaced after readiness is published. + const pagePlacementsAtAttach = new Map( + getRuntimeBrowserPageRegistry(runtime) + .listPages() + .map((page) => [page.browserPageId, page.placement]) + ) const handle = registry.attach({ browserHostClientId: params.browserHostClientId, connectionId, @@ -127,6 +133,7 @@ export const BROWSER_CLIENT_HOST_METHODS: RpcAnyMethod[] = [ lease: handle.lease, authority: registry, pages: getRuntimeBrowserPageRegistry(runtime), + pagePlacementsAtAttach, notifyWorkspace: (workspaceId) => runtime.notifyMobileSessionTabsChanged(workspaceId), releaseUnrecoverablePage: (page) => releaseRuntimeBrowserClientPageRecord(runtime, page.browserPageId, page.placement), diff --git a/src/main/runtime/rpc/methods/client-native-chat-settings.test.ts b/src/main/runtime/rpc/methods/client-native-chat-settings.test.ts new file mode 100644 index 00000000000..e26ed4acae3 --- /dev/null +++ b/src/main/runtime/rpc/methods/client-native-chat-settings.test.ts @@ -0,0 +1,78 @@ +import { describe, expect, it, vi } from 'vitest' +import type { OrcaRuntimeService } from '../../orca-runtime' +import type { RpcRequest } from '../core' +import { RpcDispatcher } from '../dispatcher' +import { CLIENT_UI_METHODS } from './client-ui' + +const request = (params: unknown): RpcRequest => ({ + id: 'req-1', + authToken: 'tok', + method: 'settings.mutateNativeChatSessionOptions', + params +}) + +describe('native-chat settings RPC', () => { + it('routes option deltas to the runtime-owned atomic update', async () => { + const updateClientNativeChatSessionOptions = vi.fn() + const runtime = { + getRuntimeId: () => 'test-runtime', + updateClientNativeChatSessionOptions + } as unknown as OrcaRuntimeService + const dispatcher = new RpcDispatcher({ runtime, methods: CLIENT_UI_METHODS }) + const mutation = { + type: 'apply-picks' as const, + agent: 'codex' as const, + picks: [ + { modelId: 'gpt-fast', optionId: 'model' as const, value: 'gpt-fast' }, + { modelId: 'gpt-fast', optionId: 'effort' as const, value: 'low' } + ] + } + + const response = await dispatcher.dispatch(request(mutation)) + + expect(updateClientNativeChatSessionOptions).toHaveBeenCalledExactlyOnceWith(mutation) + expect(response).toMatchObject({ ok: true, result: { ok: true } }) + }) + + it('rejects malformed option deltas', async () => { + const updateClientNativeChatSessionOptions = vi.fn() + const runtime = { + getRuntimeId: () => 'test-runtime', + updateClientNativeChatSessionOptions + } as unknown as OrcaRuntimeService + const dispatcher = new RpcDispatcher({ runtime, methods: CLIENT_UI_METHODS }) + + for (const mutation of [ + { type: 'apply-picks', agent: 'codex', picks: [] }, + { + type: 'apply-picks', + agent: 'opencode', + picks: [{ modelId: 'model', optionId: 'model', value: 'model' }] + }, + { + type: 'apply-picks', + agent: 'codex', + picks: [{ modelId: 'model', optionId: 'arbitrary', value: 'value' }] + }, + { + type: 'apply-picks', + agent: 'codex', + picks: [{ modelId: 'model', optionId: 'effort', value: true }] + }, + { + type: 'apply-picks', + agent: 'cursor', + picks: [{ modelId: 'model', optionId: 'fastMode', value: 'true' }] + }, + { + type: 'clear-model-if-missing', + agent: 'grok', + availableModelIds: [] + } + ]) { + const response = await dispatcher.dispatch(request(mutation)) + expect(response).toMatchObject({ ok: false, error: { code: 'invalid_argument' } }) + } + expect(updateClientNativeChatSessionOptions).not.toHaveBeenCalled() + }) +}) diff --git a/src/main/runtime/rpc/methods/client-settings-schemas.ts b/src/main/runtime/rpc/methods/client-settings-schemas.ts index 7244879c350..e389ed9d12b 100644 --- a/src/main/runtime/rpc/methods/client-settings-schemas.ts +++ b/src/main/runtime/rpc/methods/client-settings-schemas.ts @@ -18,6 +18,45 @@ export const PRBotAuthorOverrideUpdate = z .object({ author: z.string(), isBot: z.boolean() }) .strict() +const NativeChatSessionOptionPickBase = { + modelId: z.string().trim().min(1).max(512), + adoptModelAsLaunchDefault: z.boolean().optional() +} + +const NativeChatSessionOptionPick = z.union([ + z + .object({ + ...NativeChatSessionOptionPickBase, + optionId: z.enum(['model', 'effort']), + value: z.string().trim().min(1).max(512) + }) + .strict(), + z + .object({ + ...NativeChatSessionOptionPickBase, + optionId: z.enum(['fastMode', 'thinking']), + value: z.boolean() + }) + .strict() +]) + +export const NativeChatSessionOptionsMutation = z.discriminatedUnion('type', [ + z + .object({ + type: z.literal('apply-picks'), + agent: z.enum(['claude', 'codex', 'gemini', 'cursor', 'grok']), + picks: z.array(NativeChatSessionOptionPick).min(1).max(8) + }) + .strict(), + z + .object({ + type: z.literal('clear-model-if-missing'), + agent: z.enum(['claude', 'codex', 'gemini', 'cursor', 'grok']), + availableModelIds: z.array(z.string().trim().min(1).max(512)).min(1).max(256) + }) + .strict() +]) + const GitHubProjectRef = z .object({ owner: z.string(), @@ -72,6 +111,7 @@ export const SettingsUpdate = z compactWorktreeCards: z.boolean().optional(), minimaxGroupId: z.string().optional(), minimaxUsageModels: z.string().optional(), + minimaxEndpoint: z.enum(['overseas', 'cn']).optional(), githubProjects: GitHubProjectSettings.optional(), prBotAuthorOverrides: z .unknown() diff --git a/src/main/runtime/rpc/methods/client-ui.test.ts b/src/main/runtime/rpc/methods/client-ui.test.ts index 34596835294..39048611155 100644 --- a/src/main/runtime/rpc/methods/client-ui.test.ts +++ b/src/main/runtime/rpc/methods/client-ui.test.ts @@ -35,6 +35,7 @@ describe('client UI RPC methods', () => { compactWorktreeCards: true, minimaxGroupId: 'group-42', minimaxUsageModels: 'general,abab6.5', + minimaxEndpoint: 'cn', githubProjects: { pinned: [ { @@ -133,6 +134,7 @@ describe('client UI RPC methods', () => { compactWorktreeCards: true, minimaxGroupId: 'group-42', minimaxUsageModels: 'general,abab6.5', + minimaxEndpoint: 'cn', defaultRepoSelection: settings.defaultRepoSelection, defaultLinearTeamSelection: ['team-1', 'team-2'], githubProjects: settings.githubProjects @@ -157,6 +159,7 @@ describe('client UI RPC methods', () => { compactWorktreeCards: true, minimaxGroupId: 'group-42', minimaxUsageModels: 'general,abab6.5', + minimaxEndpoint: 'cn', defaultRepoSelection: settings.defaultRepoSelection, defaultLinearTeamSelection: ['team-1', 'team-2'], githubProjects: settings.githubProjects diff --git a/src/main/runtime/rpc/methods/client-ui.ts b/src/main/runtime/rpc/methods/client-ui.ts index 4161064be01..ffd964b6be6 100644 --- a/src/main/runtime/rpc/methods/client-ui.ts +++ b/src/main/runtime/rpc/methods/client-ui.ts @@ -1,7 +1,11 @@ import { omitPairingLocalUiFields } from '../../../../shared/pairing-local-ui-fields' import type { PersistedUIState } from '../../../../shared/persisted-ui-state-types' import { defineMethod, type RpcMethod } from '../core' -import { PRBotAuthorOverrideUpdate, SettingsUpdate } from './client-settings-schemas' +import { + NativeChatSessionOptionsMutation, + PRBotAuthorOverrideUpdate, + SettingsUpdate +} from './client-settings-schemas' import { FeatureInteractionIdParam, UiUpdate } from './client-ui-schemas' // Type-only side effect: keeps the schema/PersistedUIState parity assertions in // the typecheck graph so drift fails the build instead of a paired client. @@ -44,6 +48,14 @@ export const CLIENT_UI_METHODS: RpcMethod[] = [ settings: runtime.updateClientPRBotAuthorOverride(params) }) }), + defineMethod({ + name: 'settings.mutateNativeChatSessionOptions', + params: NativeChatSessionOptionsMutation, + handler: (params, { runtime }) => { + runtime.updateClientNativeChatSessionOptions(params) + return { ok: true as const } + } + }), defineMethod({ name: 'ui.get', params: null, diff --git a/src/main/runtime/rpc/methods/notification-preferences.test.ts b/src/main/runtime/rpc/methods/notification-preferences.test.ts new file mode 100644 index 00000000000..9ff372f5314 --- /dev/null +++ b/src/main/runtime/rpc/methods/notification-preferences.test.ts @@ -0,0 +1,79 @@ +import { expect, it } from 'vitest' +import { NOTIFICATION_METHODS } from './notifications' +import { RuntimeMobileNotificationController } from '../../runtime-mobile-notification-controller' +import type { RpcContext, RpcStreamingMethod, RpcMethod } from '../core' + +it('keeps desktop-disabled events out of legacy live and replay streams', async () => { + const controller = new RuntimeMobileNotificationController() + const cleanups: (() => void)[] = [] + const runtime = { + onNotificationDispatched: controller.onDispatched.bind(controller), + getMobileNotificationEpoch: controller.getEpoch.bind(controller), + getMissedNotificationsSince: controller.getMissedSince.bind(controller), + registerSubscriptionCleanup: (_id: string, cleanup: () => void) => cleanups.push(cleanup) + } + const ctx = { runtime } as unknown as RpcContext + const subscribe = NOTIFICATION_METHODS.find( + (method) => method.name === 'notifications.subscribe' + ) as RpcStreamingMethod + const replay = NOTIFICATION_METHODS.find( + (method) => method.name === 'notifications.getMissedSince' + ) as RpcMethod + const legacy: unknown[] = [] + const current: unknown[] = [] + const pending = [ + subscribe.handler({}, ctx, (event) => legacy.push(event)), + subscribe.handler({ includeDesktopSuppressed: true }, ctx, (event) => current.push(event)) + ] + controller.dispatch({ + type: 'notification', + source: 'terminal-bell', + title: 'bell', + body: '', + desktopAllowed: false + }) + controller.dispatch({ + type: 'notification', + source: 'agent-task-complete', + title: 'done', + body: '' + }) + expect(legacy).toHaveLength(2) + expect(current).toHaveLength(3) + expect(legacy[1]).toMatchObject({ title: 'done' }) + expect(current[1]).toMatchObject({ desktopAllowed: false }) + expect(await replay.handler({ lastSeenSeq: 0 }, ctx)).toMatchObject({ + notifications: [{ title: 'done' }] + }) + const result = (await replay.handler( + { lastSeenSeq: 0, includeDesktopSuppressed: true }, + ctx + )) as { notifications: unknown[] } + expect(result.notifications).toHaveLength(2) + cleanups.forEach((cleanup) => cleanup()) + await Promise.all(pending) +}) + +it('preserves legacy workspace cooldown while letting current phones filter before cooldown', async () => { + const { createNotificationStreamFilter } = await import('./notification-stream-policy') + const events = [ + { + type: 'notification' as const, + source: 'terminal-bell' as const, + title: '', + body: '', + worktreeId: 'folder', + emittedAt: 10000 + }, + { + type: 'notification' as const, + source: 'agent-task-complete' as const, + title: '', + body: '', + worktreeId: 'folder', + emittedAt: 10250 + } + ] + expect(events.filter(createNotificationStreamFilter())).toEqual([events[0]]) + expect(events.filter(createNotificationStreamFilter(true))).toEqual(events) +}) diff --git a/src/main/runtime/rpc/methods/notification-stream-policy.ts b/src/main/runtime/rpc/methods/notification-stream-policy.ts new file mode 100644 index 00000000000..2210545ab3a --- /dev/null +++ b/src/main/runtime/rpc/methods/notification-stream-policy.ts @@ -0,0 +1,19 @@ +import { reserveNotificationCooldown } from '../../../../shared/notification-burst-cooldown' +import type { MobileNotificationEvent } from '../../runtime-mobile-notification-controller' + +export function createNotificationStreamFilter(includeDesktopSuppressed = false) { + const recent = new Map() + return (event: MobileNotificationEvent): boolean => { + if (includeDesktopSuppressed || event.type !== 'notification') { + return true + } + if (event.desktopAllowed === false) { + return false + } + // Old phones rely on the host for workspace-wide burst suppression. + return ( + event.emittedAt === undefined || + reserveNotificationCooldown(recent, event.worktreeId ?? 'global', event.emittedAt) + ) + } +} diff --git a/src/main/runtime/rpc/methods/notifications.ts b/src/main/runtime/rpc/methods/notifications.ts index 80c6af7caec..10a48f2b49f 100644 --- a/src/main/runtime/rpc/methods/notifications.ts +++ b/src/main/runtime/rpc/methods/notifications.ts @@ -1,4 +1,11 @@ import { z } from 'zod' +import { createNotificationStreamFilter } from './notification-stream-policy' +import { + MOBILE_PUSH_AGENT_STATES, + MOBILE_PUSH_APNS_ENVIRONMENTS, + MOBILE_PUSH_PLATFORMS, + MOBILE_PUSH_SOURCES +} from '../../../../shared/mobile-push-contract' import { defineStreamingMethod, defineMethod, type RpcAnyMethod } from '../core' // Why: monotonically increasing per-process counter eliminates the @@ -26,9 +33,36 @@ const NotificationUnsubscribeParams = z.object({ // client that predates the field keeps the seq-only cut. const NotificationGetMissedSinceParams = z.object({ lastSeenSeq: z.number().int().min(0, 'lastSeenSeq must be a non-negative integer'), - epoch: z.string().optional() + epoch: z.string().optional(), + includeDesktopSuppressed: z.boolean().optional() }) +// Why: the phone owns which alerts are worth waking it for; the host stores the +// filter per device and applies it before it ever calls the gateway. Native push +// tokens are long (FCM registration strings), so the bound is generous. +const NotificationPushFilterParams = z.object({ + followDesktop: z.boolean().optional(), + sound: z.boolean().optional(), + sources: z.array(z.enum(MOBILE_PUSH_SOURCES)).max(MOBILE_PUSH_SOURCES.length), + agentStates: z.array(z.enum(MOBILE_PUSH_AGENT_STATES)).max(MOBILE_PUSH_AGENT_STATES.length) +}) + +const NotificationRegisterPushParams = z + .object({ + platform: z.enum(MOBILE_PUSH_PLATFORMS), + token: z.string().min(1).max(4096), + apnsEnvironment: z.enum(MOBILE_PUSH_APNS_ENVIRONMENTS).optional(), + filter: NotificationPushFilterParams + }) + // Why strict: the device identity is added by the handler, so a caller-supplied + // `deviceId` must be an error, not a key silently dropped. + .strict() + // Why: an APNs token is only routable against the environment it was minted in, + // so a missing environment must fail loudly rather than default to production. + .refine((params) => params.platform !== 'ios' || params.apnsEnvironment !== undefined, { + message: 'apnsEnvironment is required for ios' + }) + // Why: notifications.subscribe streams desktop notification events to mobile // clients over WebSocket. The mobile client shows a local push notification // for each event. This avoids requiring Firebase/APNs — the existing @@ -36,11 +70,14 @@ const NotificationGetMissedSinceParams = z.object({ export const NOTIFICATION_METHODS: readonly RpcAnyMethod[] = [ defineStreamingMethod({ name: 'notifications.subscribe', - params: null, - handler: async (_params, { runtime, connectionId }, emit) => { + params: z.object({ includeDesktopSuppressed: z.boolean().optional() }).optional(), + handler: async (params, { runtime, connectionId }, emit) => { + const shouldEmit = createNotificationStreamFilter(params?.includeDesktopSuppressed) await new Promise((resolve) => { const unsubscribe = runtime.onNotificationDispatched((event) => { - emit(event) + if (shouldEmit(event)) { + emit(event) + } }) // Why: scope by per-ws connectionId + per-process counter so @@ -79,7 +116,38 @@ export const NOTIFICATION_METHODS: readonly RpcAnyMethod[] = [ // client missed while its socket was reaped. handler: async (params, { runtime }) => { const missed = runtime.getMissedNotificationsSince(params.lastSeenSeq, params.epoch) - return { notifications: missed, epoch: runtime.getMobileNotificationEpoch() } + return { + notifications: missed.filter( + createNotificationStreamFilter(params.includeDesktopSuppressed) + ), + epoch: runtime.getMobileNotificationEpoch() + } + } + }), + defineMethod({ + name: 'notifications.registerPush', + params: NotificationRegisterPushParams, + // Why: the registration is keyed by the revocable paired device identity, never + // by anything the caller can assert, so an in-process or CLI caller has no device + // to register and is refused outright. + handler: async (params, { runtime, clientKind, pairedDeviceId }) => { + if (clientKind !== 'mobile' || !pairedDeviceId) { + return { registered: false, reason: 'not_mobile' } + } + // The paired identity is spread last so no parameter can ever override it. + return await runtime.registerMobilePushDevice({ ...params, deviceId: pairedDeviceId }) + } + }), + defineMethod({ + name: 'notifications.unregisterPush', + params: null, + // Deleting the gateway token is durable (outbox), so an offline gateway still + // reports success to the phone that asked to stop being pushed to. + handler: async (_params, { runtime, clientKind, pairedDeviceId }) => { + if (clientKind !== 'mobile' || !pairedDeviceId) { + return { unregistered: false } + } + return await runtime.unregisterMobilePushDevice(pairedDeviceId) } }) ] diff --git a/src/main/runtime/rpc/methods/orchestration-caller-workspace.ts b/src/main/runtime/rpc/methods/orchestration-caller-workspace.ts new file mode 100644 index 00000000000..556938eef23 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-caller-workspace.ts @@ -0,0 +1,30 @@ +/** + * The workspace a dispatching pane sits in, for either kind of coordinator. + * + * `showTerminal` resolves a live PTY or a live renderer leaf, and a worker that IS a structured + * agent session has neither — so routing every coordinator through it would have made "can dispatch + * sub-workers" a property of how the coordinator itself was started. The worker mode is a runtime + * implementation detail: an agent is taught the same verbs and reads the same receipts either way, + * so the one fact `worker-start` actually needs from `--from` is resolved from the same authority + * the pane-key and process-incarnation getters already use. + * + * Deliberately NOT a `showTerminal` branch: that returns a `RuntimeTerminalShow` with a ptyId, a + * leaf id and a pane runtime id, and synthesising those for a session with no PTY would hand every + * caller of a public terminal verb something that looks writable and is not. + */ + +import type { OrcaRuntimeService } from '../../orca-runtime' +import { isStructuredWorkerHandle } from '../../structured-worker-identity' + +export async function resolveDispatchCallerWorktreeId( + runtime: Pick, + callerHandle: string +): Promise { + if (isStructuredWorkerHandle(callerHandle)) { + const worktreeId = runtime.getOrchestrationDispatchAuthority?.(callerHandle)?.worktreeId ?? null + if (worktreeId) { + return worktreeId + } + } + return (await runtime.showTerminal(callerHandle)).worktreeId +} diff --git a/src/main/runtime/rpc/methods/orchestration-structured-worker-abandon.test.ts b/src/main/runtime/rpc/methods/orchestration-structured-worker-abandon.test.ts new file mode 100644 index 00000000000..983bd20ad95 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-structured-worker-abandon.test.ts @@ -0,0 +1,54 @@ +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { createOrchestrationRpcHarness } from './orchestration/rpc-test-harness' +import type { OrchestrationRpcState } from './orchestration/rpc-test-harness' + +const released: string[] = [] + +vi.mock('./orchestration-structured-worker-session', () => ({ + releaseStructuredWorkerSession: (dispatchId: string) => released.push(dispatchId), + createStructuredWorkerSession: vi.fn(), + sendStructuredWorkerPreamble: vi.fn(), + structuredWorkerHoldId: (dispatchId: string) => `orchestration:dispatch:${dispatchId}` +})) + +const harness = createOrchestrationRpcHarness() + +describe('workerAbandon settles the structured hold', () => { + let state: OrchestrationRpcState + + beforeEach(() => { + released.length = 0 + state = harness.setup() + }) + + afterEach(() => { + harness.cleanup() + vi.restoreAllMocks() + }) + + async function startedDispatch(): Promise { + const task = state.db.createTask({ spec: 'do it' }) + const started = state.db.createStartingWorkerDispatch({ + creator: { kind: 'system' }, + maxDepth: Number.MAX_SAFE_INTEGER, + taskId: task.id, + startOptions: {} + }) + return started.dispatch.id + } + + it('releases the hold when the dispatch actually settles', async () => { + const dispatchId = await startedDispatch() + // Without this, the resume-capable hold outlives settlement: the provider child can never be + // evicted and host crash recovery keeps respawning an abandoned worker. + await harness.call('orchestration.workerAbandon', { dispatch: dispatchId }, state.ctx) + expect(released).toEqual([dispatchId]) + }) + + it('does not release twice when the dispatch was already settled', async () => { + const dispatchId = await startedDispatch() + await harness.call('orchestration.workerAbandon', { dispatch: dispatchId }, state.ctx) + await harness.call('orchestration.workerAbandon', { dispatch: dispatchId }, state.ctx) + expect(released).toEqual([dispatchId]) + }) +}) diff --git a/src/main/runtime/rpc/methods/orchestration-structured-worker-lifecycle.test.ts b/src/main/runtime/rpc/methods/orchestration-structured-worker-lifecycle.test.ts new file mode 100644 index 00000000000..984dad9996f --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-structured-worker-lifecycle.test.ts @@ -0,0 +1,423 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentJournalRenderItem } from '../../../../shared/agent-session-journal-types' +import type { StructuredWorkerIdentity } from '../../structured-worker-identity' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../../../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) +vi.mock('./orchestration-structured-worker-session', () => ({ + releaseStructuredWorkerSession: vi.fn() +})) + +const { + captureStructuredWorkerArchive, + observeStructuredWorker, + readArchivedStructuredJournal, + readStructuredWorkerJournal, + stopStructuredWorker +} = await import('./orchestration-structured-worker-lifecycle') +const { readArchivedWorkerOutput } = await import('./orchestration/worker/worker-archive-read') + +const IDENTITY: StructuredWorkerIdentity = { + handle: 'structworker_1', + sessionId: 'session-1', + agent: 'claude', + paneKey: 'structured-agent-session-session-1:11111111-1111-4111-a111-111111111111', + processIncarnation: 'structured:session-1', + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } +} + +const ITEMS: AgentJournalRenderItem[] = [ + { + itemId: 'i1', + observedAt: 1, + body: { kind: 'message', role: 'assistant', blocks: [{ type: 'text', text: 'hello' }] } + } as unknown as AgentJournalRenderItem +] + +function installHost(options: { + items?: AgentJournalRenderItem[] + hasSession?: boolean + claimStatus?: string + runtimeKind?: string + deathEvidence?: unknown + record?: unknown + close?: () => Promise + setSessionTabVisibility?: () => Promise + historyThrows?: boolean +}) { + const record = + options.record === undefined + ? { + location: { executionHostId: 'local', wslDistro: null }, + lease: { + runtimeKind: options.runtimeKind ?? 'native', + claimStatus: options.claimStatus ?? 'live', + deathEvidence: options.deathEvidence ?? null, + runtimeFence: 3 + } + } + : options.record + let closed = false + hostRef.current = { + deps: { store: { getRecord: () => record } }, + hasSession: () => (closed ? false : (options.hasSession ?? true)), + setSessionTabVisibility: options.setSessionTabVisibility ?? (async () => {}), + close: + options.close ?? + (async () => { + closed = true + }), + history: () => { + if (options.historyThrows) { + throw new Error('agent_session_not_attached') + } + return { ok: true, page: { items: options.items ?? ITEMS, hasOlder: false } } + } + } +} + +describe('structured worker observation', () => { + beforeEach(() => { + hostRef.current = null + }) + + it('is unverifiable, never exited, when the host is not installed', () => { + // Not being able to look is not a death certificate. + expect(observeStructuredWorker(IDENTITY)).toEqual({ + status: 'unverifiable', + reason: expect.stringContaining('not installed') + }) + }) + + it('is live when the host holds the session under a live native lease', () => { + installHost({}) + expect(observeStructuredWorker(IDENTITY).status).toBe('live') + }) + + it('is exited only on a released lease with death evidence', () => { + installHost({ + claimStatus: 'released', + deathEvidence: { kind: 'exit-observed', detail: 'x', observedAt: 1 } + }) + expect(observeStructuredWorker(IDENTITY).status).toBe('exited') + }) + + it('is unverifiable when the lease moved to a terminal owner', () => { + installHost({ runtimeKind: 'tui' }) + expect(observeStructuredWorker(IDENTITY).status).toBe('unverifiable') + }) +}) + +describe('structured worker stop', () => { + beforeEach(() => { + hostRef.current = null + }) + + it('settles only when the session is proven gone after the close', async () => { + installHost({ + claimStatus: 'released', + deathEvidence: { kind: 'exit-observed', detail: 'closed', observedAt: 1 } + }) + await expect(stopStructuredWorker(IDENTITY, 'd1')).resolves.toEqual({ + stopped: true, + closeAttempted: true + }) + }) + + it.each([ + { hasSession: false }, + { runtimeKind: 'tui' }, + { claimStatus: 'released' }, + { record: null } + ])('retains without positive exit evidence: %j', async (options) => { + installHost({ ...options, close: async () => {} }) + const retireStructuredAgentSessionTabFromSnapshot = vi.fn() + const result = await stopStructuredWorker(IDENTITY, 'd1', { + forgetStructuredSessionMail: vi.fn(), + retireStructuredAgentSessionTabFromSnapshot + }) + expect(result).toMatchObject({ stopped: false, closeAttempted: true }) + expect(retireStructuredAgentSessionTabFromSnapshot).not.toHaveBeenCalled() + }) + + it('retains when the close throws, and admits the close was issued', async () => { + installHost({ + close: async () => { + throw new Error('close is queued for retry') + } + }) + const result = await stopStructuredWorker(IDENTITY, 'd1') + expect(result.stopped).toBe(false) + expect(result.closeAttempted).toBe(true) + expect(result.reason).toContain('retry') + }) + + it('retains when the session is still attached after the close', async () => { + installHost({ hasSession: true, close: async () => {} }) + const result = await stopStructuredWorker(IDENTITY, 'd1') + expect(result.stopped).toBe(false) + }) + + it('claims no close when the tab-visibility step threw before one was issued', async () => { + // `closeAttempted` is what the receipt turns into `processAction: 'closed_agent_terminal'`. + // Reporting it here would claim a close for a child that is still running. + const close = vi.fn(async () => {}) + installHost({ + close, + setSessionTabVisibility: async () => { + throw new Error('the durable tab index is wedged') + } + }) + const result = await stopStructuredWorker(IDENTITY, 'd1') + expect(result.stopped).toBe(false) + expect(result.closeAttempted).toBe(false) + expect(close).not.toHaveBeenCalled() + }) + + it('retains when the host is not installed, and claims no close', async () => { + // `closed_agent_terminal` on a runtime that never reached a host is the receipt claiming an + // action it did not take. + const result = await stopStructuredWorker(IDENTITY, 'd1') + expect(result.stopped).toBe(false) + expect(result.closeAttempted).toBe(false) + }) +}) + +describe('structured worker output', () => { + beforeEach(() => { + hostRef.current = null + }) + + it('round-trips the journal through the archive and back out of a released read', () => { + installHost({}) + const live = readStructuredWorkerJournal({ + identity: IDENTITY, + dispatchId: 'd1', + workerState: 'ready', + liveness: 'live', + agent: 'claude' + }) + expect(live.source).toBe('transcript') + const archive = captureStructuredWorkerArchive(IDENTITY, 'claude') + hostRef.current = null + const archived = readArchivedStructuredJournal({ + dispatchId: 'd1', + workerState: 'succeeded', + resourceId: 'res_1', + createdAt: '2026-09-05 00:00:00', + releaseState: 'released', + archive + }) + expect(archived.source).toBe('transcript') + expect(archived.archived).toBe(true) + expect(archived.transcript?.messages).toHaveLength(1) + expect(archived.transcript?.messages[0]?.blocks[0]).toMatchObject({ text: 'hello' }) + // The frozen source has its own identity, so a live cursor cannot be replayed against it. + expect(archived.sourceIdentity).not.toBe(live.sourceIdentity) + }) + + it('redacts dispatch capabilities from the archived journal', () => { + installHost({ + items: [ + { + itemId: 'i1', + observedAt: 1, + body: { + kind: 'message', + role: 'assistant', + blocks: [{ type: 'text', text: `token dcap_${'a'.repeat(30)} here` }] + } + } as unknown as AgentJournalRenderItem + ] + }) + const archive = captureStructuredWorkerArchive(IDENTITY, 'claude') + expect(JSON.stringify(archive)).not.toContain('dcap_aaa') + expect(JSON.stringify(archive)).toContain('[dispatch capability redacted]') + }) + + it('refuses to read a session the host no longer holds', () => { + expect(() => + readStructuredWorkerJournal({ + identity: IDENTITY, + dispatchId: 'd1', + workerState: 'ready', + liveness: 'live', + agent: 'claude' + }) + ).toThrow(/not attached/) + }) + + it('reports an unverifiable worker as unknown, never as running', () => { + // The `could not look, therefore it is alive` inversion. After a restart the runtime observes + // `unverifiable` — no attached provider child in this generation — while the journal is still + // readable, and a coordinator reading `running` waits on a worker that may already be gone. + installHost({}) + const read = readStructuredWorkerJournal({ + identity: IDENTITY, + dispatchId: 'd1', + workerState: 'ready', + liveness: 'unverifiable', + agent: 'claude' + }) + expect(read.status.terminal).toBe('unknown') + expect(read.status.liveness).toBe('unverifiable') + }) + + it('carries each proven verdict through unchanged', () => { + installHost({}) + const live = readStructuredWorkerJournal({ + identity: IDENTITY, + dispatchId: 'd1', + workerState: 'ready', + liveness: 'live', + agent: 'claude' + }) + expect(live.status).toMatchObject({ terminal: 'running', liveness: 'live' }) + const exited = readStructuredWorkerJournal({ + identity: IDENTITY, + dispatchId: 'd1', + workerState: 'succeeded', + liveness: 'exited', + agent: 'claude' + }) + expect(exited.status).toMatchObject({ terminal: 'exited', liveness: 'exited' }) + }) + + it('states that a settled release is exited', () => { + installHost({}) + const archive = captureStructuredWorkerArchive(IDENTITY, 'claude') + const archived = readArchivedStructuredJournal({ + dispatchId: 'd1', + workerState: 'succeeded', + resourceId: 'res_1', + createdAt: '2026-09-05 00:00:00', + releaseState: 'released', + archive + }) + expect(archived.status).toMatchObject({ terminal: 'exited', liveness: 'exited' }) + }) + + it('never calls an unproven release exited', () => { + // The archive is frozen BEFORE the close. `release_unknown` is the state that records a close + // that did NOT land, and a coordinator reading `exited` there starts a replacement worker over + // the same worktree while the original provider child may still be attached. + installHost({}) + const archive = captureStructuredWorkerArchive(IDENTITY, 'claude') + for (const releaseState of ['unknown', 'releasing'] as const) { + const archived = readArchivedStructuredJournal({ + dispatchId: 'd1', + workerState: 'stop_unknown', + resourceId: 'res_1', + createdAt: '2026-09-05 00:00:00', + releaseState, + archive + }) + expect(archived.status).toMatchObject({ terminal: 'unknown', liveness: 'unverifiable' }) + } + }) + + it('carries the resource release state through the archived read', async () => { + // The wiring, not just the mapping: `worker-read` reaches the archive through + // `readArchivedWorkerOutput`, and the resource row it already holds is the only thing that + // knows whether the close landed. + installHost({}) + const archive = captureStructuredWorkerArchive(IDENTITY, 'claude') + const db = { + getWorkerTerminalArchive: () => ({ + dispatch_id: 'd1', + resource_id: 'res_1', + kind: 'structured_journal', + content: JSON.stringify(archive), + created_at: '2026-09-05 00:00:00' + }) + } + const read = async (releaseState: string) => + readArchivedWorkerOutput({ + db: db as never, + dispatchId: 'd1', + workerState: 'stop_unknown', + resource: { + id: 'res_1', + terminal_handle: IDENTITY.handle, + release_state: releaseState + } as never + }) + expect((await read('unknown')).status).toMatchObject({ + terminal: 'unknown', + liveness: 'unverifiable' + }) + expect((await read('released')).status).toMatchObject({ + terminal: 'exited', + liveness: 'exited' + }) + }) + + it('refuses a cursor once the tail window has slid past it', () => { + // The cursor is an index into the bounded tail, and `sourceIdentity` was constant for the + // worker's life, so a coordinator paging a growing journal resumed at the newest items and + // skipped the middle without a word. + installHost({ items: ITEMS }) + const first = readStructuredWorkerJournal({ + identity: IDENTITY, + dispatchId: 'd1', + workerState: 'ready', + liveness: 'live', + agent: 'claude' + }) + installHost({ + items: [ + { + itemId: 'i2', + observedAt: 2, + body: { kind: 'message', role: 'assistant', blocks: [{ type: 'text', text: 'later' }] } + } as unknown as AgentJournalRenderItem + ] + }) + expect(() => + readStructuredWorkerJournal({ + identity: IDENTITY, + dispatchId: 'd1', + workerState: 'ready', + liveness: 'live', + agent: 'claude', + cursor: first.cursor + }) + ).toThrow(/source changed/i) + }) +}) + +describe('archiving a structured worker whose journal cannot be read', () => { + beforeEach(() => { + hostRef.current = null + }) + + it('settles with an empty, warned archive once the session is PROVEN gone', () => { + // Closing the worker's chat tab is a routine user action: it evicts the child and detaches the + // journal permanently. Throwing archive_failed there wedged release on evidence that could + // never arrive, leaving worker-abandon as the only way out. + installHost({ + historyThrows: true, + claimStatus: 'released', + deathEvidence: { kind: 'exit-observed', detail: 'surface released', observedAt: 1 } + }) + const archive = captureStructuredWorkerArchive(IDENTITY, 'claude') + expect(archive.messages).toEqual([]) + expect(archive.processIncarnation).toBe(IDENTITY.processIncarnation) + expect(archive.warnings).toContain( + 'The structured session was already closed, so its journal could not be preserved.' + ) + }) + + it('still retains when the journal is unreadable but nothing proves the child is gone', () => { + installHost({ historyThrows: true }) + expect(() => captureStructuredWorkerArchive(IDENTITY, 'claude')).toThrow(/retained/) + }) + + it('still retains when there is no host to look with', () => { + expect(() => captureStructuredWorkerArchive(IDENTITY, 'claude')).toThrow(/retained/) + }) +}) diff --git a/src/main/runtime/rpc/methods/orchestration-structured-worker-lifecycle.ts b/src/main/runtime/rpc/methods/orchestration-structured-worker-lifecycle.ts new file mode 100644 index 00000000000..13d3ce1dbce --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-structured-worker-lifecycle.ts @@ -0,0 +1,345 @@ +/** + * The lifecycle verbs for a worker that IS a structured agent session. + * + * Observation follows the SSH execution-boundary vocabulary — `live` / `unverifiable` / `exited` — + * because losing contact with a host generation is not a death certificate. In particular a + * runtime that has not installed the structured host cannot see a session's child at all, and that + * is `unverifiable`, never `exited`. + */ + +import type { AgentType, NativeChatMessage } from '../../../../shared/native-chat-types' +import type { OrchestrationWorkerReadTranscriptResult } from '../../../../shared/orchestration-worker-output' +import type { OrcaRuntimeService } from '../../orca-runtime' +import type { OrchestrationDb } from '../../orchestration/db' +import { OrchestrationError } from '../../orchestration/orchestration-error' +import { + buildStructuredJournalArchive, + type WorkerStructuredJournalArchive +} from '../../orchestration/structured-worker-journal-archive' +import { + readStructuredJournalPage, + type StructuredJournalPage +} from '../../orchestration/structured-worker-journal-page' +import { + createWorkerOutputSourceIdentity, + decodeWorkerOutputCursor, + encodeWorkerOutputCursor +} from '../../orchestration/worker-output-cursor' +import { + boundWorkerTranscriptMessages, + clampWorkerTranscriptLimit +} from '../../orchestration/worker-transcript-payload' +import { + projectStructuredItemToNativeChat, + projectStructuredItemsToNativeChat +} from '../../../../shared/structured-agent-session-projection' +import { + observeStructuredWorker, + resolveStructuredWorkerIdentity, + structuredWorkerAgent, + structuredWorkerTerminalState, + type StructuredWorkerObservation +} from '../../structured-worker-authority' +import type { StructuredWorkerIdentity } from '../../structured-worker-identity' +import type { WorkerTerminalReleaseState } from '../../orchestration/worker-terminal-ownership' +import { releaseStructuredWorkerSession } from './orchestration-structured-worker-session' +import { closeStructuredAgentSessionChild } from '../../structured-agent-session-close' + +export { observeStructuredWorker, type StructuredWorkerObservation } + +/** The structured worker behind a dispatch, or null when a PTY worker owns it. */ +export function resolveStructuredWorkerForDispatch( + db: OrchestrationDb, + dispatchId: string +): StructuredWorkerIdentity | null { + const handle = + db.getWorkerDispatch(dispatchId)?.agent_terminal_handle ?? + db.getDispatchContextById(dispatchId)?.assignee_handle + return handle ? resolveStructuredWorkerIdentity(handle, db) : null +} + +export type StructuredWorkerStopOutcome = { + stopped: boolean + /** Whether a close was actually issued; the receipt's `processAction` may claim nothing more. */ + closeAttempted: boolean + reason?: string +} + +/** + * Stopping a structured worker. + * + * `host.close` returns void and keeps a failed close indexed for retry, so the only settlement + * evidence is the observation AFTER it: a session the host no longer holds and whose lease is no + * longer live is proven gone. Anything else is retained rather than settled. + */ +export async function stopStructuredWorker( + identity: StructuredWorkerIdentity, + dispatchId: string, + runtime?: Pick< + OrcaRuntimeService, + 'forgetStructuredSessionMail' | 'retireStructuredAgentSessionTabFromSnapshot' + > +): Promise { + return closeStructuredAgentSessionChild(identity.sessionId, { + ...(runtime ? { runtime } : {}), + // Between the close and the proof, never after: an unsettled close returns early, and a + // surviving hold keeps the provider child un-evictable for the life of the app. + afterClose: () => releaseStructuredWorkerSession(dispatchId, runtime) + }) +} + +/** The structured half of `worker-read`, or null when a PTY worker owns the dispatch. */ +export function readStructuredWorkerOutput(args: { + db: OrchestrationDb + dispatchId: string + workerState: string + /** What the caller's observation actually proved; never inferred from being able to read. */ + liveness: StructuredWorkerObservation['status'] + source?: 'auto' | 'transcript' | 'terminal' + cursor?: string | number + limit?: number +}): OrchestrationWorkerReadTranscriptResult | null { + const identity = resolveStructuredWorkerForDispatch(args.db, args.dispatchId) + if (!identity) { + return null + } + if (args.source === 'terminal') { + throw new OrchestrationError( + 'archive_unavailable', + // Mode-neutral on purpose: a coordinator is never told which kind of worker it started, so + // a refusal must not be the thing that discloses it. `auto` and `transcript` both work here. + `Worker Dispatch ${args.dispatchId} has no terminal output; read it with --source auto or --source transcript.` + ) + } + return readStructuredWorkerJournal({ + identity, + dispatchId: args.dispatchId, + workerState: args.workerState, + liveness: args.liveness, + agent: structuredWorkerAgent(identity), + ...(args.cursor === undefined ? {} : { cursor: args.cursor }), + ...(args.limit === undefined ? {} : { limit: args.limit }) + }) +} + +/** Journal page in the shape `worker-read --source transcript` already serves. */ +export function readStructuredWorkerJournal(args: { + identity: StructuredWorkerIdentity + dispatchId: string + workerState: string + liveness: StructuredWorkerObservation['status'] + agent: AgentType + cursor?: string | number + limit?: number +}): OrchestrationWorkerReadTranscriptResult { + const page = readStructuredJournalPage(args.identity.sessionId) + if (!page) { + throw new OrchestrationError( + 'transcript_required', + `The transcript for Dispatch ${args.dispatchId} could not be read; its session is not attached.` + ) + } + const bounded = boundWorkerTranscriptMessages(projectStructuredItemsToNativeChat(page.items)) + // Identity of the PREFIX the caller already holds — see `structuredJournalPrefixIdentity`. + const identityAt = (position: number): string => + structuredJournalPrefixIdentity({ identity: args.identity, page, position }) + const cursor = decodeWorkerOutputCursor(args.cursor, args.dispatchId) + if ( + cursor && + (cursor.source !== 'transcript' || cursor.sourceIdentity !== identityAt(cursor.position)) + ) { + throw new OrchestrationError( + 'source_changed', + 'The worker output source changed. Start a fresh worker-read without the old cursor.' + ) + } + return pageMessages({ + messages: bounded.messages, + warnings: [ + ...bounded.warnings, + ...(page.hasOlder ? ['Older journal items were omitted from this page.'] : []) + ], + limited: bounded.limited || page.hasOlder, + dispatchId: args.dispatchId, + workerState: args.workerState, + agent: args.agent, + identityAt, + start: cursor?.position ?? 0, + limit: args.limit, + archived: false, + liveness: args.liveness + }) +} + +/** + * The cursor's `source_changed` anchor: the window's oldest item, plus every item whose projected + * message sits BELOW `position`, by id AND revision. + * + * The journal is a reduced, MUTABLE timeline, so a message index over it is not self-validating and + * the old oldest-item-only fingerprint could not see the normal case. A `running` tool item gains + * its `[tool result]` at its original sequence once later items exist, the delta coalescer revises a + * message in place, settlement can rewrite an item smaller, and a pending approval projects to null + * until it resolves and then appears in the MIDDLE of the array. Under a stable oldest item that + * fingerprint stayed valid through all of it: a caller could be handed `hel`, resume past it and + * never receive the revision to `hello world` (omission), or have a resolved approval insert ahead + * of its saved index and re-read what it already had (duplication) — both returning ok. + * + * Scoped to the prefix rather than the whole page ON PURPOSE. Fingerprinting every item would flip + * the identity every 60ms with the coalescer window during an active turn, making the cursor + * unusable exactly while the worker is working — a useless verb in place of a silent bug. Tail + * growth the caller has not read yet cannot invalidate; a change to what it already holds does. + * Position-dependence is safe because `p` rides in the same opaque payload as the identity. + * + * The oldest item stays in the anchor as the window-slide detector: a slide shifts every index. + */ +function structuredJournalPrefixIdentity(args: { + identity: StructuredWorkerIdentity + page: StructuredJournalPage + position: number +}): string { + // Items that project to a message, in message order. `projectStructuredItemsToNativeChat` keeps + // order and drops the rest, and `boundWorkerTranscriptMessages` returns a PREFIX of that, so + // message index i is item i here for every index a cursor can name. + const projected = args.page.items.filter( + (item) => projectStructuredItemToNativeChat(item) !== null + ) + return createWorkerOutputSourceIdentity([ + 'structured-journal', + args.identity.processIncarnation, + args.identity.paneKey, + args.page.items[0]?.itemId ?? '', + ...projected.slice(0, args.position).flatMap((item) => [item.itemId, String(item.revision)]) + ]) +} + +/** Freezes the journal before the session is closed, so a released worker is still readable. */ +export function captureStructuredWorkerArchive( + identity: StructuredWorkerIdentity, + agent: AgentType +): WorkerStructuredJournalArchive { + const page = readStructuredJournalPage(identity.sessionId) + if (page) { + return buildStructuredJournalArchive({ + agent, + processIncarnation: identity.processIncarnation, + items: page.items, + hasOlder: page.hasOlder + }) + } + // An unreadable journal is `archive_failed`, and release retains the worker so the evidence can + // still be preserved later — the same contract the PTY path keeps. It holds only while the + // evidence might still arrive. A session PROVEN gone detaches its journal for good, and closing + // the worker's chat tab is a routine user action that does exactly that, so throwing there wedges + // release on evidence that can never come and leaves `worker-abandon` as the only exit. + // + // `exited` is the only verdict that qualifies: it needs a released lease WITH death evidence. + // `unverifiable` — no host installed, a lease handed to a TUI owner — means we could not look, + // and retaining is still right. + if (observeStructuredWorker(identity).status !== 'exited') { + throw new OrchestrationError( + 'archive_failed', + 'Output could not be preserved for this structured worker; the session was retained.' + ) + } + const empty = buildStructuredJournalArchive({ + agent, + processIncarnation: identity.processIncarnation, + items: [], + hasOlder: false + }) + return { + ...empty, + warnings: [ + ...empty.warnings, + 'The structured session was already closed, so its journal could not be preserved.' + ] + } +} + +export function readArchivedStructuredJournal(args: { + dispatchId: string + workerState: string + resourceId: string + createdAt: string + /** Only a SETTLED release proves the session is gone; `releasing` and `unknown` never do. */ + releaseState: WorkerTerminalReleaseState + archive: WorkerStructuredJournalArchive + cursor?: string | number + limit?: number +}): OrchestrationWorkerReadTranscriptResult { + const sourceIdentity = createWorkerOutputSourceIdentity([ + 'released-structured-journal', + args.resourceId, + args.archive.processIncarnation, + args.createdAt + ]) + const cursor = decodeWorkerOutputCursor(args.cursor, args.dispatchId) + if (cursor && (cursor.source !== 'transcript' || cursor.sourceIdentity !== sourceIdentity)) { + throw new OrchestrationError( + 'source_changed', + 'The worker output source changed. Start a fresh worker-read without the old cursor.' + ) + } + return pageMessages({ + messages: args.archive.messages, + warnings: args.archive.warnings, + limited: args.archive.limited, + dispatchId: args.dispatchId, + workerState: args.workerState, + agent: args.archive.agent, + // Constant on purpose: the archive is FROZEN before the close, so no item can be revised under + // a caller and there is no prefix to fingerprint. + identityAt: () => sourceIdentity, + start: cursor?.position ?? 0, + limit: args.limit, + archived: true, + // The archive is frozen BEFORE the close, so it proves nothing about the child. Only a + // settled release row proves the close landed; `releasing` and `unknown` are the states + // that exist to say it did not, and answering `exited` from one of them is the death + // certificate `docs/reference/ssh-execution-boundary.md` forbids. + liveness: args.releaseState === 'released' ? 'exited' : 'unverifiable' + }) +} + +function pageMessages(input: { + messages: readonly NativeChatMessage[] + warnings: string[] + limited: boolean + dispatchId: string + workerState: string + agent: AgentType + /** Identity of the prefix below a position; the returned cursor is stamped with its own end. */ + identityAt: (position: number) => string + start: number + limit: number | undefined + archived: boolean + liveness: StructuredWorkerObservation['status'] +}): OrchestrationWorkerReadTranscriptResult { + const start = Math.min(input.start, input.messages.length) + const end = Math.min(start + clampWorkerTranscriptLimit(input.limit), input.messages.length) + // Stamped with the identity of everything up to `end`, which is exactly what the next read + // recomputes and compares — so a later in-place revision below it is caught. + const sourceIdentity = input.identityAt(end) + const nextCursor = encodeWorkerOutputCursor(input.dispatchId, 'transcript', sourceIdentity, end) + return { + dispatchId: input.dispatchId, + source: 'transcript', + sourceIdentity, + provider: input.agent, + transcript: { + messages: input.messages.slice(start, end), + nextCursor, + limited: input.limited || end < input.messages.length, + returnedMessageCount: end - start + }, + cursor: nextCursor, + status: { + worker: input.workerState, + terminal: structuredWorkerTerminalState(input.liveness), + liveness: input.liveness + }, + fallbackReason: null, + warnings: input.warnings, + ...(input.archived ? { archived: true } : {}) + } +} diff --git a/src/main/runtime/rpc/methods/orchestration-structured-worker-redrive.test.ts b/src/main/runtime/rpc/methods/orchestration-structured-worker-redrive.test.ts new file mode 100644 index 00000000000..34db73f88ef --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-structured-worker-redrive.test.ts @@ -0,0 +1,173 @@ +/** + * The redrive edge is coalesced, and coalescing is not allowed to change what gets delivered. + * + * Every journal batch is a redrive candidate, because a settled turn is tombstoned rather than + * rewritten. Once mail is parked on a session, each candidate re-resolves the dispatch, queries + * unread mail and reads the host's gate facts — so a turn that streams tool calls paid the full + * gate per batch, only to re-park because the turn was still running. + */ + +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { setStructuredAgentSessionHost } from '../../../native-chat/agent-session-wire/structured-agent-session-registry' +import { OrcaRuntimeService } from '../../orca-runtime' +import { OrchestrationDb } from '../../orchestration/db' +import { structuredWorkerIdentities } from '../../structured-worker-identity' +import { createStructuredWorkerSession } from './orchestration-structured-worker-session' + +vi.mock('./structured-agent-session-create', () => ({ + createStructuredAgentSessionForWorktree: async (args: { envelope: { sessionId: string } }) => ({ + ok: true, + value: { sessionId: args.envelope.sessionId } + }) +})) + +type JournalEmit = (event: { type: string }) => void + +/** Captures the redrive subscription so the test can drive journal batches by hand. */ +function installHost(): { emit: (type: string) => void; unsubscribed: () => boolean } { + let emitter: JournalEmit | null = null + let disposed = false + setStructuredAgentSessionHost({ + hasSession: () => true, + hold: async () => {}, + release: () => {}, + subscribe: (subscription: { emit: JournalEmit }) => { + emitter = subscription.emit + return () => { + disposed = true + } + }, + deps: { + store: { + getRecord: () => ({ + provider: 'claude', + location: { executionHostId: 'local', wslDistro: null }, + lease: { + runtimeKind: 'native', + claimStatus: 'live', + deathEvidence: null, + runtimeFence: 1 + } + }) + } + } + } as never) + return { + emit: (type: string) => emitter?.({ type }), + unsubscribed: () => disposed + } +} + +describe('the structured redrive edge', () => { + let db: OrchestrationDb + let runtime: OrcaRuntimeService + + beforeEach(() => { + vi.useFakeTimers() + structuredWorkerIdentities.clear() + db = new OrchestrationDb(':memory:') + runtime = new OrcaRuntimeService() + runtime.setOrchestrationDb(db) + vi.spyOn(runtime, 'ensureStructuredAgentSessionHost').mockResolvedValue(undefined as never) + }) + + afterEach(() => { + vi.useRealTimers() + db.close() + setStructuredAgentSessionHost(null) + structuredWorkerIdentities.clear() + vi.restoreAllMocks() + }) + + async function startWorker(onJournalActivity: (sessionId: string) => void) { + return createStructuredWorkerSession({ + runtime, + worktreeId: 'repo::wt', + agent: 'claude', + dispatchId: 'd_redrive', + onJournalActivity + }) + } + + it('collapses a burst of mid-turn batches into one gate evaluation', async () => { + const host = installHost() + const onJournalActivity = vi.fn() + await startWorker(onJournalActivity) + + for (let batch = 0; batch < 25; batch += 1) { + host.emit('batch') + vi.advanceTimersByTime(10) + } + + // Still inside the quiet window: nothing has fired for 25 batches. + expect(onJournalActivity).not.toHaveBeenCalled() + vi.advanceTimersByTime(300) + expect(onJournalActivity).toHaveBeenCalledTimes(1) + }) + + it('delivers the settle edge once the journal goes quiet', async () => { + const host = installHost() + const onJournalActivity = vi.fn() + const { identity } = await startWorker(onJournalActivity) + + host.emit('batch') + vi.advanceTimersByTime(300) + + expect(onJournalActivity).toHaveBeenCalledTimes(1) + expect(onJournalActivity).toHaveBeenCalledWith(identity.sessionId) + }) + + it('still re-evaluates a turn that never goes quiet', async () => { + const host = installHost() + const onJournalActivity = vi.fn() + await startWorker(onJournalActivity) + + // Sustained churn inside the quiet window would starve a plain trailing edge forever. + for (let batch = 0; batch < 60; batch += 1) { + host.emit('batch') + vi.advanceTimersByTime(100) + } + + expect(onJournalActivity.mock.calls.length).toBeGreaterThan(0) + // ...but nowhere near one per batch. + expect(onJournalActivity.mock.calls.length).toBeLessThan(10) + }) + + it('treats a re-attach reset as the same coalesced edge', async () => { + const host = installHost() + const onJournalActivity = vi.fn() + await startWorker(onJournalActivity) + + host.emit('reset') + host.emit('batch') + vi.advanceTimersByTime(300) + + expect(onJournalActivity).toHaveBeenCalledTimes(1) + }) + + it('drops a pending redrive when the worker settles, rather than nudging a released session', async () => { + const host = installHost() + const onJournalActivity = vi.fn() + await startWorker(onJournalActivity) + + host.emit('batch') + const { releaseStructuredWorkerSession } = + await import('./orchestration-structured-worker-session') + releaseStructuredWorkerSession('d_redrive', runtime) + vi.advanceTimersByTime(5_000) + + expect(onJournalActivity).not.toHaveBeenCalled() + expect(host.unsubscribed()).toBe(true) + }) + + it('ignores journal events that are not a batch or a reset', async () => { + const host = installHost() + const onJournalActivity = vi.fn() + await startWorker(onJournalActivity) + + host.emit('snapshot') + vi.advanceTimersByTime(5_000) + + expect(onJournalActivity).not.toHaveBeenCalled() + }) +}) diff --git a/src/main/runtime/rpc/methods/orchestration-structured-worker-session.test.ts b/src/main/runtime/rpc/methods/orchestration-structured-worker-session.test.ts new file mode 100644 index 00000000000..c9187302484 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-structured-worker-session.test.ts @@ -0,0 +1,265 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' + +const hostRef: { current: unknown } = { current: null } +const createSpy = vi.fn() + +vi.mock('../../../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) +vi.mock('./structured-agent-session-create', () => ({ + createStructuredAgentSessionForWorktree: (...args: unknown[]) => createSpy(...args) +})) + +const { + createStructuredWorkerSession, + releaseStructuredWorkerSession, + sendStructuredWorkerPreamble, + structuredWorkerHoldId +} = await import('./orchestration-structured-worker-session') +const { isUnknownWorkerStartOutcome } = await import('./orchestration/worker/worker-topology') +const { structuredWorkerIdentities } = await import('../../structured-worker-identity') +const { structuredWorkerChildIdentityEnv } = + await import('../../structured-worker-child-identity-env') + +function installHost() { + const hold = vi.fn(async () => {}) + const release = vi.fn() + const dispose = vi.fn() + hostRef.current = { + setSessionTabVisibility: async () => {}, + close: async () => {}, + deps: { + store: { + getRecord: () => ({ + location: { executionHostId: 'local', wslDistro: null }, + lease: { runtimeFence: 2, runtimeKind: 'native', claimStatus: 'live' } + }) + } + }, + hold, + release, + subscribe: () => dispose + } + return { hold, release, dispose } +} + +describe('structured worker session hold', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + createSpy.mockReset() + createSpy.mockImplementation(async (args: { envelope: { sessionId: string } }) => ({ + ok: true, + value: { sessionId: args.envelope.sessionId } + })) + }) + + it('takes a resume-capable hold at start and releases it only on settlement', async () => { + const { hold, release, dispose } = installHost() + const created = await createStructuredWorkerSession({ + runtime: { ensureStructuredAgentSessionHost: async () => {} } as never, + worktreeId: 'wt_1', + agent: 'claude', + dispatchId: 'd1', + onJournalActivity: () => {} + }) + // Without the hold, the release clock evicts the provider child 15s after a user closes the + // worker's chat tab, killing an idle worker mid-dispatch. + expect(hold).toHaveBeenCalledWith(created.identity.sessionId, structuredWorkerHoldId('d1')) + expect(release).not.toHaveBeenCalled() + + releaseStructuredWorkerSession('d1') + expect(release).toHaveBeenCalledWith(created.identity.sessionId, structuredWorkerHoldId('d1')) + expect(dispose).toHaveBeenCalledTimes(1) + expect(structuredWorkerIdentities.get(created.identity.handle)).toBeNull() + // A second settlement is a no-op rather than a second release of the same holder. + releaseStructuredWorkerSession('d1') + expect(release).toHaveBeenCalledTimes(1) + }) + + it('registers the identity BEFORE the session is created, so the child gets the handle', async () => { + installHost() + let envAtSpawn: Record | undefined + createSpy.mockImplementation(async (args: { envelope: { sessionId: string } }) => { + // `attach` is what spawns the provider child, and the child's env is read from the registry + // at spawn time. Registering afterwards ships a worker with no ORCA_TERMINAL_HANDLE. + envAtSpawn = structuredWorkerChildIdentityEnv(args.envelope.sessionId, {}) + return { ok: true, value: { sessionId: args.envelope.sessionId } } + }) + const created = await createStructuredWorkerSession({ + runtime: { ensureStructuredAgentSessionHost: async () => {} } as never, + worktreeId: 'wt_1', + agent: 'claude', + dispatchId: 'd_spawn', + onJournalActivity: () => {} + }) + expect(envAtSpawn?.ORCA_TERMINAL_HANDLE).toBe(created.identity.handle) + expect(envAtSpawn?.ORCA_CLI_COMMAND).toBe('orca') + expect(envAtSpawn?.ORCA_PANE_KEY).toBeUndefined() + releaseStructuredWorkerSession('d_spawn') + }) + + it('forgets the identity and discards the session when the start fails', async () => { + const { hold } = installHost() + hold.mockRejectedValueOnce(new Error('hold refused')) + const closed: string[] = [] + ;(hostRef.current as { close: (id: string) => Promise }).close = async (id) => { + closed.push(id) + } + await expect( + createStructuredWorkerSession({ + runtime: { ensureStructuredAgentSessionHost: async () => {} } as never, + worktreeId: 'wt_1', + agent: 'claude', + dispatchId: 'd_fail', + onJournalActivity: () => {} + }) + ).rejects.toThrow('hold refused') + // Neither a live provider child nor a registry entry may outlive the failed start. + expect(closed).toHaveLength(1) + expect(structuredWorkerIdentities.getBySessionId(closed[0]!)).toBeNull() + }) + + it('discards the session when the create settled UNKNOWN after attach', async () => { + installHost() + const closed: string[] = [] + ;(hostRef.current as { close: (id: string) => Promise }).close = async (id) => { + closed.push(id) + } + // `commit` answers this after `attach` SUCCEEDED and only the tab publish failed, so the + // provider child is live. Reading it as "refused, nothing created" strands that child with no + // hold and no binding, and nothing else in the runtime ever retires it. + createSpy.mockImplementation(async () => ({ + ok: false, + refusal: { + code: 'agent_session_operation_unknown', + message: 'The chat may have been created, but its tab could not be confirmed.' + } + })) + await expect( + createStructuredWorkerSession({ + runtime: { ensureStructuredAgentSessionHost: async () => {} } as never, + worktreeId: 'wt_1', + agent: 'claude', + dispatchId: 'd_unknown', + onJournalActivity: () => {} + }) + ).rejects.toThrow(/was refused/) + expect(closed).toHaveLength(1) + expect(structuredWorkerIdentities.getBySessionId(closed[0]!)).toBeNull() + }) + + it('does not close anything when the create refusal proves nothing was created', async () => { + installHost() + const closed: string[] = [] + ;(hostRef.current as { close: (id: string) => Promise }).close = async (id) => { + closed.push(id) + } + createSpy.mockImplementation(async () => ({ + ok: false, + refusal: { + code: 'structured_agent_session_unsupported', + message: 'Orca cannot open a structured agent chat for this workspace.' + } + })) + await expect( + createStructuredWorkerSession({ + runtime: { ensureStructuredAgentSessionHost: async () => {} } as never, + worktreeId: 'wt_1', + agent: 'claude', + dispatchId: 'd_definitive', + onJournalActivity: () => {} + }) + ).rejects.toThrow(/was refused/) + expect(closed).toEqual([]) + }) + + it('registers a random handle bound to the created session', async () => { + installHost() + const created = await createStructuredWorkerSession({ + runtime: { ensureStructuredAgentSessionHost: async () => {} } as never, + worktreeId: 'wt_1', + agent: 'codex', + dispatchId: 'd2', + onJournalActivity: () => {} + }) + expect(created.identity.handle.startsWith('structworker_')).toBe(true) + expect(created.identity.processIncarnation).toBe(`structured:${created.identity.sessionId}`) + expect(structuredWorkerIdentities.getBySessionId(created.identity.sessionId)?.agent).toBe( + 'codex' + ) + releaseStructuredWorkerSession('d2') + }) + + it('does not activate the worker session, so a dispatch cannot steal the surface', async () => { + installHost() + await createStructuredWorkerSession({ + runtime: { ensureStructuredAgentSessionHost: async () => {} } as never, + worktreeId: 'wt_1', + agent: 'claude', + dispatchId: 'd3', + onJournalActivity: () => {} + }) + expect(createSpy.mock.calls[0]![0].activate).toBe(false) + releaseStructuredWorkerSession('d3') + }) + + it('refuses a session pinned to a non-local execution host', async () => { + installHost() + ;(hostRef.current as { deps: { store: { getRecord: () => unknown } } }).deps.store.getRecord = + () => ({ + location: { executionHostId: 'ssh-1', wslDistro: null }, + lease: { runtimeFence: 2, runtimeKind: 'native', claimStatus: 'live' } + }) + await expect( + createStructuredWorkerSession({ + runtime: { ensureStructuredAgentSessionHost: async () => {} } as never, + worktreeId: 'wt_1', + agent: 'claude', + dispatchId: 'd4', + onJournalActivity: () => {} + }) + ).rejects.toThrow(/local execution host/) + }) +}) + +describe('structured worker dispatch preamble', () => { + function hostWithSubmission(submission: Record) { + return { + deps: { store: { getRecord: () => ({ lease: { runtimeFence: 7 } }) } }, + send: async () => ({ ok: true, value: { clientMessageId: 'c1', submission } }) + } as never + } + + const send = (host: never) => + sendStructuredWorkerPreamble({ host, sessionId: 's1', dispatchId: 'd1', preamble: 'spec' }) + + it('reports the preamble delivered only on an accepted submission', async () => { + await expect( + send(hostWithSubmission({ dispatchState: 'accepted', reason: null })) + ).resolves.toBeUndefined() + }) + + it('never claims delivery for a submission the provider never acknowledged', async () => { + // `dispatchSafely` turns ANY thrown adapter call — provider child dead, transport dropped — + // into `unknown`, and `performSend` still returns ok. Reporting that as `dispatch_input: + // accepted` marks the worker ready with no task, and the coordinator blocks in + // `check --wait --types worker_done` until it times out. + for (const dispatchState of ['unknown', 'pending'] as const) { + const error = await send( + hostWithSubmission({ dispatchState, reason: 'provider child exited' }) + ).catch((thrown: unknown) => thrown) + expect((error as { code?: string }).code).toBe('operation_unknown') + // The wiring, not just the throw: this is the code that makes the start receipt + // `outcome_unknown` with the worker-show / worker-abandon recovery commands. + expect(isUnknownWorkerStartOutcome(error, 'dispatch_input')).toBe(true) + } + }) + + it('keeps a rejected preamble a proven failure rather than an unknown one', async () => { + const error = await send( + hostWithSubmission({ dispatchState: 'rejected', reason: 'fence moved' }) + ).catch((thrown: unknown) => thrown) + expect((error as Error).message).toMatch(/rejected: fence moved/) + expect(isUnknownWorkerStartOutcome(error, 'dispatch_input')).toBe(false) + }) +}) diff --git a/src/main/runtime/rpc/methods/orchestration-structured-worker-session.ts b/src/main/runtime/rpc/methods/orchestration-structured-worker-session.ts new file mode 100644 index 00000000000..9ed0c05ad1c --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-structured-worker-session.ts @@ -0,0 +1,326 @@ +/** + * Starting, holding and retiring a worker that IS a structured agent session. + * + * Three things make this different from the PTY worker path, and all three live here: + * + * - The session is created directly as structured, so readiness is the attach returning ok. There + * is no boot-to-idle gap to wait on and no `tui-idle` edge to read. + * - A structured session's provider child is evicted 15s after its last HOLDER leaves, and holds + * come only from bound surfaces. A dispatched worker parked on mail is exactly that state, so + * the dispatch takes its own resume-capable hold and keeps it until the worker settles. + * - The dispatch preamble is a turn, not keystrokes. + */ + +import { randomUUID } from 'node:crypto' +import { isDefinitiveAgentSessionCreateRefusal } from '../../../../shared/agent-session-definitive-refusal' +import type { AgentJournalMessageItem } from '../../../../shared/agent-session-journal-types' +import type { StructuredAgentSessionHost } from '../../../native-chat/agent-session-wire/structured-agent-session-host' +import { getStructuredAgentSessionHost } from '../../../native-chat/agent-session-wire/structured-agent-session-registry' +import type { OrcaRuntimeService } from '../../orca-runtime' +import { OrchestrationError } from '../../orchestration/orchestration-error' +import { + mintAgentSessionOperationId, + structuredPointerPayloadFingerprint +} from '../../orchestration/structured-pointer-operation-id' +import { structuredPointerCallerKey } from '../../orchestration/structured-mailbox-pointer-host' +import { retireSettledStructuredWorkerTab } from '../../structured-agent-session-tab-retirement' +import { + mintStructuredWorkerHandle, + structuredWorkerHostScope, + structuredWorkerIdentities, + mintStructuredWorkerPaneKey, + structuredWorkerProcessIncarnation, + type StructuredWorkerIdentity +} from '../../structured-worker-identity' +import { createKeyedTrailingEdgeCoalescer } from '../../keyed-trailing-edge-coalescer' +import { createStructuredAgentSessionForWorktree } from './structured-agent-session-create' + +type StructuredWorkerBinding = { + sessionId: string + handle: string + holderId: string + disposeSubscription: () => void +} + +const bindingsByDispatchId = new Map() + +export function structuredWorkerHoldId(dispatchId: string): string { + return `orchestration:dispatch:${dispatchId}` +} + +/** + * Drops the dispatch's hold, its redrive subscription and its parked mail; the release clock takes + * it from here. + * + * EVERY settlement has to reach this — stop, release AND abandon. A surviving hold does not just + * leak: it keeps the provider child un-evictable for the life of the app, and makes host crash + * recovery respawn a child for a worker that was settled long ago. + */ +export function releaseStructuredWorkerSession( + dispatchId: string, + runtime?: Pick +): void { + const binding = bindingsByDispatchId.get(dispatchId) + if (!binding) { + return + } + bindingsByDispatchId.delete(dispatchId) + binding.disposeSubscription() + structuredWorkerIdentities.forget(binding.handle) + runtime?.forgetStructuredSessionMail?.(binding.sessionId) + try { + getStructuredAgentSessionHost()?.release(binding.sessionId, binding.holderId) + } catch (error) { + console.warn('[orchestration] structured worker hold release failed', dispatchId, error) + } +} + +export async function createStructuredWorkerSession(args: { + runtime: OrcaRuntimeService + worktreeId: string + agent: 'claude' | 'codex' + dispatchId: string + /** Retried whenever the session's journal moves, which is the structured idle edge. */ + onJournalActivity: (sessionId: string) => void +}): Promise<{ identity: StructuredWorkerIdentity; host: StructuredAgentSessionHost }> { + const sessionId = randomUUID() + // Registered BEFORE the session is created, because `attach` is what spawns the provider child + // and the child's environment is read from this registry at spawn time. Registering afterwards + // ships a worker with no ORCA_TERMINAL_HANDLE, whose bare `orca orchestration check` then + // resolves to whatever single leaf sits in the worktree — by default the COORDINATOR's pane. + // + // The scope is provisionally local; the record's own location is asserted local below, and a + // session that resolves anywhere else never reaches a hold. + const identity = structuredWorkerIdentities.register({ + handle: mintStructuredWorkerHandle(), + sessionId, + agent: args.agent, + paneKey: mintStructuredWorkerPaneKey(sessionId), + processIncarnation: structuredWorkerProcessIncarnation(sessionId), + worktreeId: args.worktreeId, + hostScope: { kind: 'local', hostId: 'local' } + }) + let created: Awaited> | undefined + try { + created = await createStructuredAgentSessionForWorktree({ + runtime: args.runtime, + ensureHost: async () => { + await args.runtime.ensureStructuredAgentSessionHost() + return requireInstalledHost() + }, + caller: { callerKey: structuredPointerCallerKey(args.dispatchId) }, + envelope: { + sessionId, + clientOperationId: mintAgentSessionOperationId(Date.now()), + expectedRuntimeFence: null, + // Empty on purpose: `prepare` overwrites this with the host's own attach fingerprint, and + // the create-intent conflict check it would otherwise feed guards the RPC boundary against + // a replayed operation id — there is no such boundary on this in-process call. + payloadFingerprint: '' + }, + worktree: `id:${args.worktreeId}`, + agent: args.agent, + // Dispatching a worker is background work; it must not pull the surface away from the user. + activate: false + }) + if (!created.ok) { + throw new OrchestrationError( + 'agent_unconfigured', + `The structured ${args.agent} session for this worker was refused: ${created.refusal.message}` + ) + } + const host = requireInstalledHost() + const record = host.deps.store.getRecord(sessionId) + if (!record || !structuredWorkerHostScope(record.location)) { + throw new OrchestrationError( + 'agent_unconfigured', + 'A structured worker must run on the local execution host outside WSL.' + ) + } + const holderId = structuredWorkerHoldId(args.dispatchId) + await host.hold(sessionId, holderId) + const disposeSubscription = subscribeForRedrive(host, sessionId, args.onJournalActivity) + bindingsByDispatchId.set(args.dispatchId, { + sessionId, + handle: identity.handle, + holderId, + disposeSubscription + }) + return { identity, host } + } catch (error) { + // A start that fails after the session exists would otherwise strand a live provider child + // that no dispatch owns and that nothing else in the runtime will ever retire. + structuredWorkerIdentities.forget(identity.handle) + if (structuredCreateMayHaveCommitted(created)) { + await discardStructuredWorkerSession(sessionId, args.runtime) + } + throw error + } +} + +/** + * Whether a create may have attached a session, which is the question cleanup has to ask. + * + * `ok` is not the test. `commit` answers `agent_session_operation_unknown` when `attach` SUCCEEDED + * and only the tab publish failed, and a throw out of the commit half is past `attach` too — the + * pre-commit half never throws, it refuses. Both leave a live provider child that took no hold and + * has no binding, so nothing else in the runtime will ever retire it. Only a DEFINITIVE refusal + * proves there is nothing to discard; everything else gets the best-effort close. + */ +function structuredCreateMayHaveCommitted( + created: Awaited> | undefined +): boolean { + return !created || created.ok || !isDefinitiveAgentSessionCreateRefusal(created.refusal.code) +} + +/** + * Best-effort teardown of a session created by a worker start that then failed. + * + * Stops the provider child, drops the DURABLE tab reference so nothing restores the chat after a + * restart, and — only once the close came back without throwing — retires the background tab this + * start published from the live snapshot. All three are no-ops for a session that was never + * attached, which is why a non-definitive refusal can reach here unconditionally. A close that + * threw leaves the tab alone: the child may still be running, and the tab is the way to reach it. + * + * Exported because a start can also fail AFTER `createStructuredWorkerSession` returned — on the + * authority gate, or on the preamble turn — and that is the fourth settlement path. Dropping only + * the hold there left one dead "Claude Chat"/"Codex Chat" tab per failed start, durably restored + * on every subsequent app launch. + */ +export async function discardStructuredWorkerSession( + sessionId: string, + runtime: Pick +): Promise { + const host = getStructuredAgentSessionHost() + if (!host) { + return + } + try { + await host.setSessionTabVisibility?.(sessionId, false) + await host.close(sessionId) + } catch (error) { + console.warn( + '[orchestration] failed to discard a half-started structured worker', + sessionId, + error + ) + return + } + retireSettledStructuredWorkerTab(sessionId, runtime) +} + +/** Delivers the dispatch preamble as the worker's first turn. */ +export async function sendStructuredWorkerPreamble(args: { + host: StructuredAgentSessionHost + sessionId: string + dispatchId: string + preamble: string +}): Promise { + const body: AgentJournalMessageItem = { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: args.preamble }] + } + const fence = args.host.deps.store.getRecord(args.sessionId)?.lease.runtimeFence + if (fence === undefined) { + throw new Error('The structured worker session has no durable record to dispatch into.') + } + const result = await args.host.send( + { callerKey: structuredPointerCallerKey(args.dispatchId) }, + { + envelope: { + sessionId: args.sessionId, + clientOperationId: mintAgentSessionOperationId(Date.now()), + expectedRuntimeFence: fence, + payloadFingerprint: structuredPointerPayloadFingerprint(args.sessionId, body) + }, + body, + retryUnknown: true + } + ) + if (!result.ok) { + throw new Error(`The dispatch preamble was refused: ${result.refusal.message}`) + } + const submission = result.value.submission + if (submission.dispatchState === 'accepted') { + return + } + if (submission.dispatchState === 'rejected') { + throw new Error(`The dispatch preamble was rejected: ${submission.reason ?? 'no reason given'}`) + } + // Only `accepted` is an acknowledgement — the same rule the mail lane already applies. A thrown + // adapter call settles as `unknown`, which is indistinguishable from a lost reply, so the start + // may claim neither delivery nor failure: `operation_unknown` is what turns this into the + // `outcome_unknown` receipt whose nextCommands send the coordinator to look. + throw new OrchestrationError( + 'operation_unknown', + `The dispatch preamble was submitted but not acknowledged (${submission.dispatchState}): ${submission.reason ?? 'no reason given'}.` + ) +} + +function requireInstalledHost(): StructuredAgentSessionHost { + const host = getStructuredAgentSessionHost() + if (!host) { + throw new OrchestrationError( + 'agent_unconfigured', + 'Structured agent sessions are unavailable on this runtime.' + ) + } + return host +} + +/** + * Quiet window before a coalesced redrive runs. A settled turn stops emitting, so this is how long + * after the last batch the nudge lands — short enough to read as immediate, long enough that a + * streaming turn collapses into a handful of evaluations instead of one per batch. + */ +const REDRIVE_FLUSH_MS = 300 + +/** A turn that streams without pause still gets re-evaluated this often. */ +const REDRIVE_MAX_WAIT_MS = 2_000 + +/** + * Any journal movement is the redrive edge, coalesced. + * + * A settled turn is TOMBSTONED rather than rewritten, so watching for a completed lifecycle row + * would miss the common case — every batch has to be a candidate. Running the gate on each one is + * not free once mail IS parked on the session: the edge re-resolves the dispatch, queries unread + * mail and reads the host's gate facts, only to re-park because the turn is still running. A + * streaming turn paid that per batch. + * + * Coalescing costs nothing in delivery terms. The pointer body names only HOW MANY messages are + * waiting, so the edge is inherently batch-shaped, and this is not the path fresh mail takes to an + * idle worker — that is `deliverForHandle`, called when the message is enqueued and untouched + * here. This is only the retry for mail already parked because the worker was busy. + */ +function subscribeForRedrive( + host: StructuredAgentSessionHost, + sessionId: string, + onJournalActivity: (sessionId: string) => void +): () => void { + const coalescer = createKeyedTrailingEdgeCoalescer(onJournalActivity, { + flushMs: REDRIVE_FLUSH_MS, + maxWaitMs: REDRIVE_MAX_WAIT_MS + }) + try { + const unsubscribe = host.subscribe({ + id: `orchestration:redrive:${sessionId}`, + sessionId, + emit: (event) => { + if (event.type === 'batch' || event.type === 'reset') { + coalescer.schedule(sessionId) + } + } + }) + // Disposal drops the pending timer rather than flushing it: every settlement reaches here, and + // a redrive that fires after the hold is gone would nudge a session no dispatch owns. + return () => { + coalescer.dispose() + unsubscribe() + } + } catch (error) { + console.warn('[orchestration] structured worker redrive subscription failed', sessionId, error) + coalescer.dispose() + return () => {} + } +} diff --git a/src/main/runtime/rpc/methods/orchestration-structured-worker-start-failure.test.ts b/src/main/runtime/rpc/methods/orchestration-structured-worker-start-failure.test.ts new file mode 100644 index 00000000000..1f29dfb4904 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-structured-worker-start-failure.test.ts @@ -0,0 +1,151 @@ +/** + * A worker start that fails AFTER its structured session exists is the fourth settlement path. + * + * The create publishes a "Claude Chat"/"Codex Chat" tab and writes it into the durable restore + * index before the start can fail on the authority gate or on the preamble turn. Dropping only the + * dispatch hold there left one dead tab per failed start, re-published on every app launch and + * re-attaching a session no dispatch owns. + */ + +import { beforeEach, describe, expect, it, vi } from 'vitest' +import { setStructuredAgentSessionHost } from '../../../native-chat/agent-session-wire/structured-agent-session-registry' +import type { OrcaRuntimeService } from '../../orca-runtime' +import type { OrchestrationDb } from '../../orchestration/db' +import { structuredWorkerIdentities } from '../../structured-worker-identity' + +vi.mock('./structured-agent-session-create', () => ({ + createStructuredAgentSessionForWorktree: async (args: { envelope: { sessionId: string } }) => ({ + ok: true, + value: { sessionId: args.envelope.sessionId } + }) +})) +vi.mock('./orchestration-structured-worker-session', async (importOriginal) => ({ + ...(await importOriginal>()), + // The realistic post-create failure: the session is live, the preamble turn is not acknowledged. + sendStructuredWorkerPreamble: async () => { + throw new Error('The dispatch preamble was rejected: no capacity') + } +})) +vi.mock('./orchestration/worker/worker-start-validation', () => ({ + prepareLocalWorkerStart: () => ({ + agent: 'claude', + launch: { receipt: { requested: null, effective: null }, preferences: undefined } + }) +})) +vi.mock('./orchestration/worker/worker-setup-gate', () => ({ + persistGatedSetupSpawnFailure: () => false, + persistWorkerReadinessStage: () => {}, + persistWorkerSetupWaitOutcome: () => {} +})) +vi.mock('./orchestration/worker/worker-start-receipt', () => ({ + failWorkerStartWithReceipt: (args: { failedStage: string }) => ({ + state: 'failed', + stage: args.failedStage + }) +})) +vi.mock('./orchestration/runs/dispatch-creator', () => ({ + resolveDispatchCreator: () => ({ kind: 'terminal', handle: 'term_c' }) +})) +vi.mock('../../orchestration/preamble', () => ({ buildDispatchPreamble: () => 'preamble' })) + +const { startLocalWorker } = await import('./orchestration/worker/local-worker-start') + +const WORKTREE = 'wt_1' + +function installHost() { + const closed: string[] = [] + const visibility: [string, boolean][] = [] + setStructuredAgentSessionHost({ + setSessionTabVisibility: async (sessionId: string, visible: boolean) => { + visibility.push([sessionId, visible]) + }, + close: async (sessionId: string) => { + closed.push(sessionId) + }, + hasSession: () => true, + hold: async () => {}, + release: () => {}, + subscribe: () => () => {}, + deps: { + store: { + getRecord: () => ({ + location: { executionHostId: 'local', wslDistro: null }, + lease: { + runtimeKind: 'native', + claimStatus: 'live', + deathEvidence: null, + runtimeFence: 1 + } + }) + } + } + } as never) + return { closed, visibility } +} + +function fakes() { + const retireStructuredAgentSessionTabFromSnapshot = vi.fn(() => true) + const runtime = { + showTerminal: async () => ({ worktreeId: WORKTREE }), + showManagedTerminalWorkspace: async () => ({ id: WORKTREE }), + getNestedWorkerMaxDepth: () => 3, + getRuntimeId: () => 'epoch-1', + ensureStructuredAgentSessionHost: async () => {}, + getTerminalOrchestrationCliCommand: () => 'orca', + getStructuredAgentSessionCreateSupport: async () => ({ supported: true }), + getOrchestrationDispatchAuthority: () => ({ + paneKey: 'pane', + processIncarnation: 'structured:x', + hostScope: { kind: 'local', hostId: 'local' } + }), + forgetStructuredSessionMail: vi.fn(), + validateOrchestrationAgentLauncher: vi.fn(), + getTerminalProcessIncarnation: vi.fn(() => 'inc_1'), + getTerminalPaneKey: vi.fn(() => 'pane_1'), + retireStructuredAgentSessionTabFromSnapshot + } as unknown as OrcaRuntimeService + const db = { + createStartingWorkerDispatch: () => ({ + dispatch: { id: 'd_fail', depth: 0 }, + task: { id: 't1', spec: 'do the thing' } + }), + recordWorkerStage: () => {}, + prepareStartingWorkerAuthority: () => 'capability' + } as unknown as OrchestrationDb + return { runtime, db, retireStructuredAgentSessionTabFromSnapshot } +} + +beforeEach(() => { + structuredWorkerIdentities.clear() +}) + +describe('a structured worker-start that fails after the session exists', () => { + it('closes the session and retires the tab it published', async () => { + const host = installHost() + const { runtime, db, retireStructuredAgentSessionTabFromSnapshot } = fakes() + + const receipt = await startLocalWorker({ + params: { from: 'term_c', timeoutMs: 1_000, agent: 'claude' } as never, + mode: { + mode: 'structured', + preferred: 'structured', + reason: 'user_default', + detail: 'structured by default' + } as const, + runtime, + db, + run: { id: 'run_1' } as never, + existingTask: { id: 't1', spec: 'do the thing' } as never, + coordinatorPane: null, + orchestrationMutation: undefined + }) + + expect(receipt).toMatchObject({ state: 'failed', stage: 'dispatch_input' }) + expect(host.closed).toHaveLength(1) + const sessionId = host.closed[0] as string + // Durable restore index first, then the live snapshot; without both, the dead tab comes back + // on the next launch. + expect(host.visibility).toContainEqual([sessionId, false]) + expect(retireStructuredAgentSessionTabFromSnapshot).toHaveBeenCalledWith(sessionId) + }) +}) diff --git a/src/main/runtime/rpc/methods/orchestration-worker-mode-opacity.test.ts b/src/main/runtime/rpc/methods/orchestration-worker-mode-opacity.test.ts new file mode 100644 index 00000000000..aa8c7c58574 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-worker-mode-opacity.test.ts @@ -0,0 +1,281 @@ +/** + * The worker mode is a runtime implementation detail, not part of the orchestration contract. + * + * Two properties are pinned here, because both were false at some point in this lane: + * + * - a worker is TAUGHT the same thing whichever mode it runs in, byte for byte once the handle and + * dispatch id are normalised. The sub-dispatch section used to be withheld from a structured + * worker, which is a two-tier capability model dressed as a preamble tweak; + * - a structured worker can actually BE a coordinator. `worker-start` used to resolve `--from` + * through `showTerminal`, which needs a PTY, so the capability the preamble withheld was in fact + * missing rather than merely unadvertised. + */ + +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { setStructuredAgentSessionHost } from '../../../native-chat/agent-session-wire/structured-agent-session-registry' +import { OrcaRuntimeService } from '../../orca-runtime' +import { OrchestrationDb } from '../../orchestration/db' +import { + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} from '../../structured-worker-identity' +import { ORCHESTRATION_METHODS } from './orchestration' +import { readStructuredWorkerOutput } from './orchestration-structured-worker-lifecycle' +import { inspectWorkerTerminal } from './orchestration/worker/worker-observation' + +const WORKTREE = 'repo::wt' +const STRUCTURED_HANDLE = 'structworker_worker' +const TERMINAL_HANDLE = 'term_worker' + +const structuredPreambles: string[] = [] + +vi.mock('./orchestration/worker/worker-topology', async (importOriginal) => ({ + ...(await importOriginal>()), + createStructuredWorkerSessionForWorktree: async (args: { effects: unknown[] }) => { + args.effects.push({ kind: 'terminal', role: 'agent', action: 'created' }) + return { identity: { handle: STRUCTURED_HANDLE, sessionId: 'sess_worker' }, host: {} } + }, + createExistingWorktreeWorkerTerminal: async () => ({ handle: TERMINAL_HANDLE }) +})) +vi.mock('./orchestration-structured-worker-session', async (importOriginal) => ({ + ...(await importOriginal>()), + sendStructuredWorkerPreamble: async (args: { preamble: string }) => { + structuredPreambles.push(args.preamble) + }, + releaseStructuredWorkerSession: () => {}, + discardStructuredWorkerSession: async () => {} +})) + +const STRUCTURED_DEFAULT = { + experimentalNativeChat: true, + openAgentTabsInChatByDefault: true, + experimentalStructuredNativeChat: true, + agentCmdOverrides: {}, + agentDefaultArgs: {}, + agentDefaultEnv: {} +} +const TERMINAL_DEFAULT = { ...STRUCTURED_DEFAULT, experimentalStructuredNativeChat: false } + +/** A coordinator that IS a structured session: registry identity plus a live durable record. */ +function installStructuredCoordinator(handle: string, sessionId: string): string { + const paneKey = mintStructuredWorkerPaneKey(sessionId) + structuredWorkerIdentities.register({ + handle, + sessionId, + agent: 'claude', + paneKey, + processIncarnation: structuredWorkerProcessIncarnation(sessionId), + worktreeId: WORKTREE, + hostScope: { kind: 'local', hostId: 'local' } + }) + setStructuredAgentSessionHost({ + hasSession: () => true, + deps: { + store: { + getRecord: () => ({ + provider: 'claude', + location: { executionHostId: 'local', wslDistro: null }, + lease: { + runtimeKind: 'native', + claimStatus: 'live', + deathEvidence: null, + runtimeFence: 1 + } + }) + } + } + } as never) + return paneKey +} + +/** Strips the ids that legitimately differ per dispatch, leaving what the agent is taught. */ +function normalizePreamble(preamble: string, handle: string, dispatchId: string): string { + return preamble + .split(handle) + .join('') + .split(dispatchId) + .join('') + .replace(/dcap_[\w-]+/g, '') + .replace(/task_[0-9a-f]+/g, '') +} + +describe('a worker cannot tell which mode it is running in', () => { + const coordinatorPaneKey = 'tab_coord:aaaaaaaa-aaaa-4aaa-8aaa-aaaaaaaaaaaa' + let db: OrchestrationDb + let runtime: OrcaRuntimeService + + beforeEach(() => { + structuredPreambles.length = 0 + structuredWorkerIdentities.clear() + db = new OrchestrationDb(':memory:') + runtime = new OrcaRuntimeService() + runtime.setOrchestrationDb(db) + // Deferred to the real getters for a structured handle, because resolving one through the + // registry is exactly what is under test; stubbed only for the PTY handles that have no runtime. + const realPaneKey = runtime.getTerminalPaneKey.bind(runtime) + const realIncarnation = runtime.getTerminalProcessIncarnation.bind(runtime) + vi.spyOn(runtime, 'getTerminalPaneKey').mockImplementation((handle) => + handle === 'term_coord' ? coordinatorPaneKey : (realPaneKey(handle) ?? `tab_worker:${handle}`) + ) + vi.spyOn(runtime, 'getTerminalProcessIncarnation').mockImplementation( + (handle) => realIncarnation(handle) ?? 'runtime_test:worker:1' + ) + vi.spyOn(runtime, 'validateOrchestrationAgentLauncher').mockImplementation(() => {}) + // Above the default of 1, so the sub-dispatch section is on the table for both modes; at the + // default a depth-1 worker is refused nesting whatever mode it runs in. + vi.spyOn(runtime, 'getNestedWorkerMaxDepth').mockReturnValue(3) + vi.spyOn(runtime, 'showManagedTerminalWorkspace').mockResolvedValue({ + id: WORKTREE, + repoId: 'repo' + } as never) + vi.spyOn(runtime, 'getStructuredAgentSessionCreateSupport').mockResolvedValue({ + supported: true + }) + vi.spyOn(runtime, 'waitForTerminal').mockResolvedValue({ + handle: TERMINAL_HANDLE, + condition: 'tui-idle', + satisfied: true, + status: 'running', + exitCode: null + }) + vi.spyOn(runtime, 'getTerminalOrchestrationCliCommand').mockReturnValue('orca') + vi.spyOn(runtime, 'sendTerminalAgentPrompt').mockResolvedValue({ + handle: TERMINAL_HANDLE, + accepted: true, + bytesWritten: 1 + }) + }) + + afterEach(() => { + db.close() + setStructuredAgentSessionHost(null) + structuredWorkerIdentities.clear() + vi.restoreAllMocks() + }) + + async function startWorker(args: { + settings: Record + from: string + coordinatorPaneKey: string + }) { + vi.spyOn(runtime, 'getClientSettings').mockReturnValue(args.settings as never) + const runId = db.createRun({ + objective: 'mode opacity', + coordinatorHandle: args.from, + coordinatorPaneKey: args.coordinatorPaneKey + }).id + const task = db.createTask({ spec: 'do the thing', runId }) + const method = ORCHESTRATION_METHODS.find( + (candidate) => candidate.name === 'orchestration.workerStart' + )! + const result = (await method.handler( + method.params!.parse({ + task: task.id, + from: args.from, + worktree: 'current', + agent: 'claude' + }), + { runtime } + )) as { state: string; dispatchId: string; mode: { mode: string } } + return result + } + + it('teaches byte-identical instructions in both modes', async () => { + vi.spyOn(runtime, 'showTerminal').mockResolvedValue({ + handle: 'term_coord', + worktreeId: WORKTREE, + status: 'running' + } as never) + + const structured = await startWorker({ + settings: STRUCTURED_DEFAULT, + from: 'term_coord', + coordinatorPaneKey + }) + const terminal = await startWorker({ + settings: TERMINAL_DEFAULT, + from: 'term_coord', + coordinatorPaneKey + }) + + expect(structured.mode.mode).toBe('structured') + expect(terminal.mode.mode).toBe('terminal') + const structuredPreamble = structuredPreambles[0] as string + const terminalPreamble = vi.mocked(runtime.sendTerminalAgentPrompt).mock.calls[0]?.[1] as string + expect(normalizePreamble(structuredPreamble, STRUCTURED_HANDLE, structured.dispatchId)).toBe( + normalizePreamble(terminalPreamble, TERMINAL_HANDLE, terminal.dispatchId) + ) + // The section the structured lane used to withhold, asserted by name so the equality above + // cannot pass by both preambles losing it. + expect(structuredPreamble).toContain('=== SUB-DISPATCH ===') + }) + + it('lets a structured worker dispatch a sub-worker like any other coordinator', async () => { + const paneKey = installStructuredCoordinator('structworker_coord', 'sess_coord') + // Proves the resolution is not falling through to a PTY: showTerminal cannot answer here. + const showTerminal = vi + .spyOn(runtime, 'showTerminal') + .mockRejectedValue(new Error('no_active_terminal')) + + const result = await startWorker({ + settings: TERMINAL_DEFAULT, + from: 'structworker_coord', + coordinatorPaneKey: paneKey + }) + + expect(result).toMatchObject({ state: 'ready' }) + expect(showTerminal).not.toHaveBeenCalled() + expect(vi.mocked(runtime.sendTerminalAgentPrompt).mock.calls[0]?.[1]).toContain( + '=== SUB-DISPATCH ===' + ) + }) + + it('refuses an unavailable output source without disclosing the mode', async () => { + installStructuredCoordinator(STRUCTURED_HANDLE, 'sess_worker') + vi.spyOn(runtime, 'showTerminal').mockResolvedValue({ + handle: 'term_coord', + worktreeId: WORKTREE, + status: 'running' + } as never) + const started = await startWorker({ + settings: STRUCTURED_DEFAULT, + from: 'term_coord', + coordinatorPaneKey + }) + + const read = () => + readStructuredWorkerOutput({ + db, + dispatchId: started.dispatchId, + workerState: 'ready', + liveness: 'live', + source: 'terminal' + }) + + expect(read).toThrow(/has no terminal output/) + // The refusal names a source that works instead of naming the worker's kind. + expect(read).toThrow(/--source auto or --source transcript/) + expect(read).not.toThrow(/structured/i) + }) + + it('never claims a structured worker was checked for a human-answerable prompt', async () => { + installStructuredCoordinator(STRUCTURED_HANDLE, 'sess_worker') + vi.spyOn(runtime, 'showTerminal').mockResolvedValue({ + handle: 'term_coord', + worktreeId: WORKTREE, + status: 'running' + } as never) + const started = await startWorker({ + settings: STRUCTURED_DEFAULT, + from: 'term_coord', + coordinatorPaneKey + }) + + const observation = await inspectWorkerTerminal(runtime, db, started.dispatchId) + + // Absent, not null: null is the contract's "looked and found none", and a journal question is + // invisible to every prompt scan, so null would be a false negative a coordinator acts on. + expect('agentWait' in observation).toBe(false) + }) +}) diff --git a/src/main/runtime/rpc/methods/orchestration-worker-start-mode-selection.test.ts b/src/main/runtime/rpc/methods/orchestration-worker-start-mode-selection.test.ts new file mode 100644 index 00000000000..e87f8182037 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-worker-start-mode-selection.test.ts @@ -0,0 +1,198 @@ +/** + * End of the seam: `orchestration.workerStart` reads the user's own setting and starts the worker + * that setting describes. No flag reaches this decision, and no combination refuses the start. + */ + +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { OrcaRuntimeService } from '../../orca-runtime' +import { OrchestrationDb } from '../../orchestration/db' +import { ORCHESTRATION_METHODS } from './orchestration' + +const STRUCTURED_HANDLE = 'structworker_abc' +const TERMINAL_HANDLE = 'term_worker' + +const createStructuredWorkerSessionForWorktree = vi.fn( + async (args: { effects: { kind: string }[] }) => { + args.effects.push({ kind: 'terminal' }) + return { identity: { handle: STRUCTURED_HANDLE, sessionId: 'sess_1' }, host: {} } + } +) +const createExistingWorktreeWorkerTerminal = vi.fn(async () => ({ handle: TERMINAL_HANDLE })) + +vi.mock('./orchestration/worker/worker-topology', async (importOriginal) => ({ + ...(await importOriginal>()), + createStructuredWorkerSessionForWorktree: (args: never) => + createStructuredWorkerSessionForWorktree(args), + createExistingWorktreeWorkerTerminal: () => createExistingWorktreeWorkerTerminal() +})) +vi.mock('./orchestration/federation/federated-worker-start', () => ({ + startFederatedWorker: async () => ({ state: 'ready', dispatchId: 'ctx_remote' }) +})) +vi.mock('./orchestration-structured-worker-session', async (importOriginal) => ({ + ...(await importOriginal>()), + sendStructuredWorkerPreamble: async () => {}, + releaseStructuredWorkerSession: () => {}, + discardStructuredWorkerSession: async () => {} +})) + +const STRUCTURED_DEFAULT = { + experimentalNativeChat: true, + openAgentTabsInChatByDefault: true, + experimentalStructuredNativeChat: true, + agentCmdOverrides: {}, + agentDefaultArgs: {}, + agentDefaultEnv: {} +} + +describe('worker-start honours the settings default', () => { + const coordinatorPaneKey = 'tab_coord:aaaaaaaa-aaaa-4aaa-8aaa-aaaaaaaaaaaa' + let db: OrchestrationDb + let runtime: OrcaRuntimeService + let runId: string + + beforeEach(() => { + createStructuredWorkerSessionForWorktree.mockClear() + createExistingWorktreeWorkerTerminal.mockClear() + db = new OrchestrationDb(':memory:') + runtime = new OrcaRuntimeService() + runtime.setOrchestrationDb(db) + runId = db.createRun({ + objective: 'Settings-driven worker mode', + coordinatorHandle: 'term_coord', + coordinatorPaneKey + }).id + vi.spyOn(runtime, 'getTerminalPaneKey').mockImplementation((handle) => + handle === 'term_coord' ? coordinatorPaneKey : `tab_worker:${handle}` + ) + vi.spyOn(runtime, 'getTerminalProcessIncarnation').mockReturnValue('runtime_test:worker:1') + vi.spyOn(runtime, 'validateOrchestrationAgentLauncher').mockImplementation(() => {}) + vi.spyOn(runtime, 'showTerminal').mockResolvedValue({ + handle: 'term_coord', + worktreeId: 'repo::wt', + status: 'running' + } as never) + vi.spyOn(runtime, 'showManagedTerminalWorkspace').mockResolvedValue({ + id: 'repo::wt', + repoId: 'repo' + } as never) + vi.spyOn(runtime, 'getStructuredAgentSessionCreateSupport').mockResolvedValue({ + supported: true + }) + vi.spyOn(runtime, 'waitForTerminal').mockResolvedValue({ + handle: TERMINAL_HANDLE, + condition: 'tui-idle', + satisfied: true, + status: 'running', + exitCode: null + }) + vi.spyOn(runtime, 'getTerminalOrchestrationCliCommand').mockReturnValue('orca') + vi.spyOn(runtime, 'sendTerminalAgentPrompt').mockResolvedValue({ + handle: TERMINAL_HANDLE, + accepted: true, + bytesWritten: 1 + }) + }) + + afterEach(() => { + db.close() + vi.restoreAllMocks() + }) + + async function startWorker( + settings: Record | null, + overrides: Record = {} + ) { + vi.spyOn(runtime, 'getClientSettings').mockImplementation(() => { + if (!settings) { + throw new Error('runtime_unavailable') + } + return settings as never + }) + const task = db.createTask({ spec: 'settings-driven task', runId }) + const method = ORCHESTRATION_METHODS.find( + (candidate) => candidate.name === 'orchestration.workerStart' + )! + const params = method.params!.parse({ + task: task.id, + from: 'term_coord', + worktree: 'current', + agent: 'claude', + ...overrides + }) + return (await method.handler(params, { runtime })) as { + state: string + mode: { mode: string; preferred: string; reason: string; detail: string } + } + } + + it('starts a structured chat worker when structured native chat is the default', async () => { + const result = await startWorker(STRUCTURED_DEFAULT) + + expect(result).toMatchObject({ + state: 'ready', + mode: { mode: 'structured', preferred: 'structured', reason: 'user_default' } + }) + expect(createStructuredWorkerSessionForWorktree).toHaveBeenCalledTimes(1) + expect(createExistingWorktreeWorkerTerminal).not.toHaveBeenCalled() + }) + + it('starts a terminal agent worker when it is not', async () => { + const result = await startWorker({ + ...STRUCTURED_DEFAULT, + experimentalStructuredNativeChat: false + }) + + expect(result).toMatchObject({ + state: 'ready', + mode: { mode: 'terminal', preferred: 'terminal', reason: 'user_default' } + }) + expect(createExistingWorktreeWorkerTerminal).toHaveBeenCalledTimes(1) + expect(createStructuredWorkerSessionForWorktree).not.toHaveBeenCalled() + }) + + it('starts a terminal worker rather than failing when the host refuses a structured session', async () => { + vi.mocked(runtime.getStructuredAgentSessionCreateSupport).mockResolvedValue({ + supported: false, + reason: 'wsl' + }) + + const result = await startWorker(STRUCTURED_DEFAULT) + + expect(result).toMatchObject({ + state: 'ready', + mode: { mode: 'terminal', preferred: 'structured', reason: 'wsl_execution_runtime' } + }) + expect(createExistingWorktreeWorkerTerminal).toHaveBeenCalledTimes(1) + }) + + it('still starts a worker when the runtime has no settings to read', async () => { + const result = await startWorker(null) + + expect(result).toMatchObject({ state: 'ready', mode: { mode: 'terminal' } }) + expect(createExistingWorktreeWorkerTerminal).toHaveBeenCalledTimes(1) + }) + + it('falls back instead of refusing a launch preference the structured default cannot apply', async () => { + const result = await startWorker(STRUCTURED_DEFAULT, { model: 'opus', effort: 'high' }) + + expect(result).toMatchObject({ + state: 'ready', + mode: { mode: 'terminal', preferred: 'structured', reason: 'launch_preferences' } + }) + expect(createExistingWorktreeWorkerTerminal).toHaveBeenCalledTimes(1) + expect(createStructuredWorkerSessionForWorktree).not.toHaveBeenCalled() + }) + + it('tells a remote dispatch why its structured default did not apply', async () => { + const result = await startWorker(STRUCTURED_DEFAULT, { + on: 'server-1', + worktree: 'repo::remote' + }) + + expect(result).toMatchObject({ + state: 'ready', + mode: { mode: 'terminal', preferred: 'structured', reason: 'remote_execution_host' } + }) + expect(createStructuredWorkerSessionForWorktree).not.toHaveBeenCalled() + }) +}) diff --git a/src/main/runtime/rpc/methods/orchestration-worker-start-mode.test.ts b/src/main/runtime/rpc/methods/orchestration-worker-start-mode.test.ts new file mode 100644 index 00000000000..7eafe9c86d4 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-worker-start-mode.test.ts @@ -0,0 +1,128 @@ +/** + * The worker mode is the user's own setting, not a flag, and the fallback is never silent. + * + * Every case here is one a coordinator can hit on a routine `worker-start`. Before this became + * settings-driven each of them was a REFUSAL, which was right for an explicit `--structured` and + * wrong for a preference: a dispatch that cannot be a structured session must still start. + */ + +import { describe, expect, it } from 'vitest' +import { + decideWorkerStartMode, + downgradeWorkerStartModeForHost, + type WorkerStartModeReceipt +} from './orchestration-worker-start-mode' + +const STRUCTURED_DEFAULT = { + experimentalNativeChat: true, + openAgentTabsInChatByDefault: true, + experimentalStructuredNativeChat: true +} + +function decide( + overrides: { + params?: Parameters[0]['params'] + settings?: Parameters[0]['settings'] + platform?: NodeJS.Platform + } = {} +): WorkerStartModeReceipt { + return decideWorkerStartMode({ + params: { agent: 'claude', ...overrides.params }, + settings: overrides.settings === undefined ? STRUCTURED_DEFAULT : overrides.settings, + platform: overrides.platform ?? 'darwin' + }) +} + +describe('worker start mode from the user default', () => { + it.each(['claude', 'codex'] as const)('starts a local %s worker structured', (agent) => { + expect(decide({ params: { agent } })).toMatchObject({ + mode: 'structured', + preferred: 'structured', + reason: 'user_default' + }) + }) + + it.each([ + ['native chat off', { ...STRUCTURED_DEFAULT, experimentalNativeChat: false }], + ['chat-by-default off', { ...STRUCTURED_DEFAULT, openAgentTabsInChatByDefault: false }], + ['structured off', { ...STRUCTURED_DEFAULT, experimentalStructuredNativeChat: false }], + ['no settings at all', null] + ])('starts a terminal worker when %s', (_name, settings) => { + expect(decide({ settings })).toMatchObject({ + mode: 'terminal', + preferred: 'terminal', + reason: 'user_default' + }) + }) + + it('says which mode ran even when the default was honoured', () => { + expect(decide().detail).toContain('structured chat session') + expect(decide({ settings: null }).detail).toContain('terminal agent') + }) +}) + +describe('a structured default this dispatch cannot honour', () => { + it.each([ + ['a remote --on', { on: 'server-1' }, 'remote_execution_host'], + ['an existing --terminal', { terminal: 'term_1' }, 'reused_terminal'], + ['a new-child worktree', { worktree: 'new-child' }, 'worktree_creation'], + ['a new-top-level worktree', { worktree: 'new-top-level' }, 'worktree_creation'], + ['--model', { model: 'opus' }, 'launch_preferences'], + ['--effort', { effort: 'high' }, 'launch_preferences'], + ['a non-structured agent', { agent: 'cursor' }, 'agent_without_structured_session'], + ['no agent at all', { agent: undefined }, 'agent_without_structured_session'] + ])('falls back to a terminal worker for %s', (_name, params, reason) => { + const receipt = decide({ params: { agent: 'claude', ...params } }) + expect(receipt).toMatchObject({ mode: 'terminal', preferred: 'structured', reason }) + // Never a silent fallback: the receipt states the default AND why it did not apply. + expect(receipt.detail).toContain('Your default is a structured chat session') + }) + + it('keeps the current worktree structured, which is the ordinary dispatch', () => { + expect(decide({ params: { agent: 'codex', worktree: 'current' } }).mode).toBe('structured') + }) + + it('falls back rather than dropping a custom TUI launch the session cannot apply', () => { + expect( + decide({ + settings: { ...STRUCTURED_DEFAULT, agentCmdOverrides: { claude: 'claude-wrapper' } } + }) + ).toMatchObject({ mode: 'terminal', reason: 'tui_launch_customization' }) + }) + + it('keeps Codex terminal-backed on Windows and leaves Claude to the host', () => { + expect(decide({ params: { agent: 'codex' }, platform: 'win32' })).toMatchObject({ + mode: 'terminal', + reason: 'codex_on_windows' + }) + expect(decide({ params: { agent: 'claude' }, platform: 'win32' }).mode).toBe('structured') + }) +}) + +describe('the executing host settles what the client cannot', () => { + it.each([ + ['wsl', 'wsl_execution_runtime'], + ['remote', 'remote_execution_host'], + ['agent', 'structured_unsupported_on_host'] + ] as const)('downgrades on a %s refusal', (reason, expected) => { + expect(downgradeWorkerStartModeForHost(decide(), { supported: false, reason })).toMatchObject({ + mode: 'terminal', + preferred: 'structured', + reason: expected + }) + }) + + it('downgrades on a refusal that names no reason', () => { + expect(downgradeWorkerStartModeForHost(decide(), { supported: false }).reason).toBe( + 'structured_unsupported_on_host' + ) + }) + + it('leaves a supported structured start and an already-terminal receipt alone', () => { + expect(downgradeWorkerStartModeForHost(decide(), { supported: true }).mode).toBe('structured') + const terminal = decide({ settings: null }) + expect(downgradeWorkerStartModeForHost(terminal, { supported: false, reason: 'wsl' })).toBe( + terminal + ) + }) +}) diff --git a/src/main/runtime/rpc/methods/orchestration-worker-start-mode.ts b/src/main/runtime/rpc/methods/orchestration-worker-start-mode.ts new file mode 100644 index 00000000000..7c02c2a688f --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration-worker-start-mode.ts @@ -0,0 +1,237 @@ +/** + * Which kind of worker `orchestration.workerStart` starts, decided from the user's own settings. + * + * There is no `--structured` flag: if the user's default is that a new agent tab opens as a + * structured native chat, an orchestration worker is one too. That default is a preference, not a + * demand, so a dispatch it cannot apply to falls back to an ordinary PTY terminal worker and the + * receipt says which mode ran and why — a routine `worker-start` must never fail because the user + * happens to have a chat preference on. + * + * The settings default and the per-launch feasibility both come from + * `shared/structured-native-chat-launch-route`, the same module the renderer's + * `resolveAgentLaunchRoute` uses; only the placement options that exist solely on this command are + * decided here. + */ + +import type { GlobalSettings } from '../../../../shared/global-settings-types' +import { RUNTIME_CAPABILITIES } from '../../../../shared/protocol-version' +import { + prefersStructuredNativeChatByDefault, + resolveStructuredNativeChatSupport, + type NativeChatDefaultSettings, + type StructuredNativeChatBlocker +} from '../../../../shared/structured-native-chat-launch-route' +import type { TuiAgent } from '../../../../shared/tui-agent' +import { hasExplicitTuiLaunchCustomization } from '../../../../shared/tui-agent-launch-customization' +import type { OrcaRuntimeService } from '../../orca-runtime' + +export type WorkerStartMode = 'structured' | 'terminal' + +export type WorkerStartModeReason = + | 'user_default' + | 'remote_execution_host' + | 'reused_terminal' + | 'worktree_creation' + | 'launch_preferences' + | 'agent_without_structured_session' + | 'tui_launch_customization' + | 'structured_sessions_unavailable' + | 'wsl_execution_runtime' + | 'codex_on_windows' + | 'structured_unsupported_on_host' + +export type WorkerStartModeReceipt = { + /** The mode the worker actually started in. */ + mode: WorkerStartMode + /** The user's settings default for a new agent tab. */ + preferred: WorkerStartMode + reason: WorkerStartModeReason + /** One sentence, always present, so a fallback is never silent. */ + detail: string +} + +type WorkerStartModeSettings = Partial< + NativeChatDefaultSettings & + Pick +> + +type WorkerStartModePlacement = { + agent?: string + on?: string + terminal?: string + worktree?: string + model?: string + effort?: string +} + +const DOWNGRADE_DETAIL: Record, string> = { + remote_execution_host: '--on runs the worker on a remote execution host', + reused_terminal: '--terminal reuses a running terminal agent', + worktree_creation: 'a new worktree is created with its agent terminal', + launch_preferences: '--model and --effort apply only to a terminal agent', + agent_without_structured_session: 'this agent has no structured session', + tui_launch_customization: + 'this agent has a custom launch command, arguments or environment that only a terminal applies', + structured_sessions_unavailable: 'this runtime does not support structured agent sessions', + wsl_execution_runtime: 'this workspace runs under WSL', + codex_on_windows: 'Codex has no structured session on Windows', + structured_unsupported_on_host: 'the execution host cannot create one here' +} + +const BLOCKER_REASON: Record< + StructuredNativeChatBlocker, + Exclude +> = { + 'agent-without-structured-session': 'agent_without_structured_session', + 'draft-prompt': 'structured_unsupported_on_host', + 'floating-workspace': 'structured_unsupported_on_host', + 'tui-launch-customization': 'tui_launch_customization', + 'remote-execution-host': 'remote_execution_host', + 'codex-on-windows': 'codex_on_windows', + 'project-runtime': 'wsl_execution_runtime', + 'runtime-capability': 'structured_sessions_unavailable' +} + +/** The host's own create-support verdict (`agentSession.createSupport`) in this vocabulary. */ +const HOST_SUPPORT_REASON: Record< + 'agent' | 'remote' | 'wsl', + Exclude +> = { + agent: 'structured_unsupported_on_host', + remote: 'remote_execution_host', + wsl: 'wsl_execution_runtime' +} + +export function decideWorkerStartMode(args: { + params: WorkerStartModePlacement + settings: WorkerStartModeSettings | null | undefined + platform: NodeJS.Platform +}): WorkerStartModeReceipt { + const { params, settings } = args + if (!prefersStructuredNativeChatByDefault(settings)) { + return { + mode: 'terminal', + preferred: 'terminal', + reason: 'user_default', + detail: 'Started a terminal agent worker, the default for new agent tabs in your settings.' + } + } + const placementReason = resolvePlacementReason(params) + if (placementReason) { + return downgraded(placementReason) + } + const agent = params.agent as TuiAgent + const support = resolveStructuredNativeChatSupport({ + agent, + // Set only by --on, which the placement check above already turned into a fallback. + executionHostId: 'local', + platform: args.platform, + hostCapabilities: RUNTIME_CAPABILITIES, + // Orchestration resolves a managed worktree or folder workspace; a floating terminal is never + // a worker placement. WSL is left to the executing host's own create-support probe, which + // reads the resolved workspace rather than guessing from a client-side project runtime. + requiresTuiLaunchCustomization: hasExplicitTuiLaunchCustomization(settings, agent) + }) + if (!support.supported) { + return downgraded(BLOCKER_REASON[support.blocker]) + } + return { + mode: 'structured', + preferred: 'structured', + reason: 'user_default', + detail: + 'Started a structured chat session worker, the default for new agent tabs in your settings.' + } +} + +/** + * Second half of the decision, once the worktree is resolved: the host that will run the worker + * answers whether it can create a structured session there at all. Asked before anything is + * created, so a refusal becomes a terminal worker rather than a failed start. + */ +export async function resolveWorkerStartModeOnHost( + runtime: Pick, + mode: WorkerStartModeReceipt, + worktreeId: string | undefined, + agent: TuiAgent | undefined +): Promise { + if (mode.mode !== 'structured' || !worktreeId) { + return mode + } + return downgradeWorkerStartModeForHost( + mode, + await readStructuredCreateSupport(runtime, worktreeId, agent) + ) +} + +/** A host that cannot answer has not proved it can create one, so the worker stays a PTY agent. */ +async function readStructuredCreateSupport( + runtime: Pick, + worktreeId: string, + agent: TuiAgent | undefined +): Promise<{ supported: boolean; reason?: 'agent' | 'remote' | 'wsl' }> { + if (agent !== 'claude' && agent !== 'codex') { + return { supported: false, reason: 'agent' } + } + try { + return await runtime.getStructuredAgentSessionCreateSupport(`id:${worktreeId}`, agent) + } catch { + return { supported: false } + } +} + +/** + * Applies the executing host's `agentSession.createSupport` answer, which is the authority on WSL, + * remoteness and the Windows process-start-time gate for the resolved workspace. + */ +export function downgradeWorkerStartModeForHost( + receipt: WorkerStartModeReceipt, + support: { supported: boolean; reason?: 'agent' | 'remote' | 'wsl' } +): WorkerStartModeReceipt { + if (receipt.mode !== 'structured' || support.supported) { + return receipt + } + return downgraded( + support.reason ? HOST_SUPPORT_REASON[support.reason] : 'structured_unsupported_on_host' + ) +} + +function resolvePlacementReason( + params: WorkerStartModePlacement +): Exclude | null { + if (params.on) { + return 'remote_execution_host' + } + if (params.terminal) { + return 'reused_terminal' + } + if (params.worktree === 'new-child' || params.worktree === 'new-top-level') { + return 'worktree_creation' + } + if (params.model || params.effort) { + return 'launch_preferences' + } + return null +} + +function downgraded( + reason: Exclude +): WorkerStartModeReceipt { + return { + mode: 'terminal', + preferred: 'structured', + reason, + detail: `Your default is a structured chat session, but ${DOWNGRADE_DETAIL[reason]}; started a terminal agent worker instead.` + } +} + +/** The store can be missing on a runtime that never opened one; that reads as no preference. */ +export function readWorkerStartModeSettings( + runtime: Pick +): WorkerStartModeSettings | null { + try { + return runtime.getClientSettings() + } catch { + return null + } +} diff --git a/src/main/runtime/rpc/methods/orchestration/cli-runtime-boundary.test.ts b/src/main/runtime/rpc/methods/orchestration/cli-runtime-boundary.test.ts index 419a1a5af55..e0c5d5dd8c9 100644 --- a/src/main/runtime/rpc/methods/orchestration/cli-runtime-boundary.test.ts +++ b/src/main/runtime/rpc/methods/orchestration/cli-runtime-boundary.test.ts @@ -43,8 +43,10 @@ describe('orchestration CLI/runtime boundary', () => { isRemote: false, /** Preserves terminal-handle validation while routing other calls through runtime RPC. */ async call(method: string, params?: unknown): Promise<{ result: T }> { - if (method === 'terminal.show') { - return { result: { terminal: { handle: objectParams(params).terminal } } as T } + if (method === 'terminal.resolveIdentity') { + return { + result: { identity: { handle: objectParams(params).terminal, live: true } } as T + } } return { result: (await callRpc(method, objectParams(params))) as T } } diff --git a/src/main/runtime/rpc/methods/orchestration/messaging/send-group.ts b/src/main/runtime/rpc/methods/orchestration/messaging/send-group.ts index d58e5f8afda..6681ce3236a 100644 --- a/src/main/runtime/rpc/methods/orchestration/messaging/send-group.ts +++ b/src/main/runtime/rpc/methods/orchestration/messaging/send-group.ts @@ -3,6 +3,7 @@ import type { OrcaRuntimeService } from '../../../../orca-runtime' import { OrchestrationError } from '../../../../orchestration/orchestration-error' import { resolveGroupAddress } from '../../../../orchestration/groups' import { resolveBareOrchestrationRecipient } from './recipient-routing' +import { listAddressableStructuredWorkers } from '../../../../orchestration/structured-worker-group-addressing' import { legacyWorkerDeliveryContract } from '../routing' import { exposeMessages } from './mailbox-message-receipt' import { recordReceiptBeforeNudge } from './mutation-replay-nudge' @@ -44,7 +45,11 @@ export async function sendGroupMessage(args: { const { terminals } = await runtime.listTerminals(undefined, undefined, { includeVisualLayouts: false }) - const handles = resolveGroupAddress(groupAddress, from, terminals, (handle: string) => + // Structured workers are on no PTY surface, so `listTerminals` cannot see them and a broadcast + // silently missed every one. Composed here rather than inside `listTerminals`, whose result is + // published to paired clients and to consumers that assume a summary is writable. + const recipients = [...terminals, ...listAddressableStructuredWorkers()] + const handles = resolveGroupAddress(groupAddress, from, recipients, (handle: string) => runtime.getAgentStatusForHandle(handle) ) if (handles.length === 0) { diff --git a/src/main/runtime/rpc/methods/orchestration/worker/deliver-worker-dispatch-preamble.ts b/src/main/runtime/rpc/methods/orchestration/worker/deliver-worker-dispatch-preamble.ts new file mode 100644 index 00000000000..3da57f9c530 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration/worker/deliver-worker-dispatch-preamble.ts @@ -0,0 +1,60 @@ +import type { RuntimeTerminalSend } from '../../../../../../shared/runtime-terminal-contracts' +import type { OrcaRuntimeService } from '../../../../orca-runtime' +import { buildDispatchPreamble } from '../../../../orchestration/preamble' +import { sendStructuredWorkerPreamble } from '../../orchestration-structured-worker-session' +import type { createStructuredWorkerSessionForWorktree } from './worker-topology' + +type StructuredSession = Awaited> | null + +/** + * Hands a started worker the dispatch preamble, over whichever transport it has. + * + * The preamble itself is identical for both: a worker is taught the same verbs whichever mode it + * runs in, and only the delivery differs — a PTY write returns a queued/accepted receipt, while a + * structured turn either is acknowledged or throws. + */ +export async function deliverWorkerDispatchPreamble(args: { + runtime: OrcaRuntimeService + structuredSession: StructuredSession + terminalHandle: string + dispatchId: string + dispatchDepth: number + taskId: string + taskSpec: string + coordinatorHandle: string + dispatchCapability: string + devMode: boolean | undefined + requestId: string +}): Promise { + const { runtime, structuredSession, terminalHandle } = args + const preamble = buildDispatchPreamble({ + // Depth only. A worker is taught the same verbs whichever mode it runs in, so this must not + // become a second gate: resolving the caller's worktree is what lets a structured worker + // dispatch sub-workers exactly like a PTY one. + canDispatchSubWorkers: args.dispatchDepth < runtime.getNestedWorkerMaxDepth(), + taskId: args.taskId, + dispatchId: args.dispatchId, + taskSpec: args.taskSpec, + coordinatorHandle: args.coordinatorHandle, + workerHandle: terminalHandle, + dispatchCapability: args.dispatchCapability, + devMode: args.devMode, + cliCommand: runtime.getTerminalOrchestrationCliCommand(terminalHandle) + }) + if (structuredSession) { + await sendStructuredWorkerPreamble({ + host: structuredSession.host, + sessionId: structuredSession.identity.sessionId, + dispatchId: args.dispatchId, + preamble + }) + return undefined + } + return ( + await runtime.sendTerminalAgentPrompt(terminalHandle, preamble, { + acceptQueued: true, + observationTimeoutMs: 0, + requestId: args.requestId + }) + ).prompt +} diff --git a/src/main/runtime/rpc/methods/orchestration/worker/explicit-worker-terminal-validation.ts b/src/main/runtime/rpc/methods/orchestration/worker/explicit-worker-terminal-validation.ts new file mode 100644 index 00000000000..6e20cc40a29 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration/worker/explicit-worker-terminal-validation.ts @@ -0,0 +1,49 @@ +import type { OrcaRuntimeService } from '../../../../orca-runtime' +import { OrchestrationError } from '../../../../orchestration/orchestration-error' +import { isStructuredWorkerHandle } from '../../../../structured-worker-identity' + +/** + * Admits a caller-supplied `--terminal` as this dispatch's worker pane. + * + * Three refusals, all of which must happen before anything is created: a coordinator adopted as its + * own worker answers its own dispatch preamble forever, a pane in another worktree is not this + * dispatch's to take, and a pane with no agent cannot read a preamble at all. + */ +export async function assertExplicitWorkerTerminalUsable(args: { + runtime: OrcaRuntimeService + terminal: string + from: string + coordinatorPane: string | null + resolvedWorktreeId: string | undefined +}): Promise { + const { runtime, terminal, from, coordinatorPane, resolvedWorktreeId } = args + const explicitTerminal = await runtime.showTerminal(terminal) + const targetPane = runtime.getTerminalPaneKey(terminal) + const callerPane = coordinatorPane ?? runtime.getTerminalPaneKey(from) + // A structured coordinator has no terminal to show, so its own identity is the raw handle plus + // the pane key; showing `from` unconditionally would throw for exactly those callers. + const coordinatorHandle = isStructuredWorkerHandle(from) + ? from + : (await runtime.showTerminal(from)).handle + if ( + explicitTerminal.handle === coordinatorHandle || + (targetPane !== null && targetPane === callerPane) + ) { + throw new OrchestrationError( + 'terminal_is_coordinator', + `Terminal ${terminal} is this coordinator's own terminal. Pass --terminal for a different agent pane, or omit it so worker-start creates one.` + ) + } + if (explicitTerminal.worktreeId !== resolvedWorktreeId) { + throw new OrchestrationError( + 'terminal_worktree_mismatch', + `Terminal ${terminal} does not belong to worktree ${resolvedWorktreeId}.` + ) + } + if (!(await runtime.isTerminalRunningAgent(terminal))) { + throw new OrchestrationError( + 'agent_unconfigured', + `Terminal ${terminal} is not running a recognized agent.` + ) + } +} diff --git a/src/main/runtime/rpc/methods/orchestration/worker/failed-start-residual-terminal.test.ts b/src/main/runtime/rpc/methods/orchestration/worker/failed-start-residual-terminal.test.ts index 6a30dcd2c4d..bdf5daad565 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/failed-start-residual-terminal.test.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/failed-start-residual-terminal.test.ts @@ -147,6 +147,12 @@ describe('failed worker-start receipt for a residual terminal', () => { }) return failWorkerStartWithReceipt({ db: d, + mode: { + mode: 'terminal', + preferred: 'terminal', + reason: 'user_default', + detail: 'terminal by default' + } as const, runId: 'run_residual', taskId: task.id, dispatchId: started.dispatch.id, diff --git a/src/main/runtime/rpc/methods/orchestration/worker/failed-worker-start-teardown.ts b/src/main/runtime/rpc/methods/orchestration/worker/failed-worker-start-teardown.ts new file mode 100644 index 00000000000..32827377b54 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration/worker/failed-worker-start-teardown.ts @@ -0,0 +1,42 @@ +import type { OrcaRuntimeService } from '../../../../orca-runtime' +import { + discardStructuredWorkerSession, + releaseStructuredWorkerSession +} from '../../orchestration-structured-worker-session' +import { resolveResidualAgentTerminal } from './failed-start-residual-terminal' +import type { createStructuredWorkerSessionForWorktree } from './worker-topology' +import type { FailedStartTerminalAdoption } from '../../../../orchestration/db/worker-terminal/failed-start-terminal-adoption' + +/** + * Undoes what a start created before it failed, and reports what `worker-release` still owns. + * + * A start that never reached ready leaves no settlement to release the hold later, and its session + * was already published as a chat tab — without the discard, a failed start strands a dead chat tab + * that the durable restore index republishes on every app launch. Both halves are best-effort by + * construction, so neither can replace the real error. + */ +export async function tearDownFailedWorkerStart(args: { + runtime: OrcaRuntimeService + structuredSession: Awaited> | null + dispatchId: string + effects: unknown[] + terminalHandle: string | undefined + worktreeId: string | null +}): Promise { + const { runtime, structuredSession } = args + // A structured session is torn down outright here, so it must never also be adopted as a residual + // terminal for `worker-release` to close a second time. + const residualAgentTerminal = structuredSession + ? undefined + : resolveResidualAgentTerminal({ + runtime, + effects: args.effects as never, + terminalHandle: args.terminalHandle, + worktreeId: args.worktreeId + }) + releaseStructuredWorkerSession(args.dispatchId, runtime) + if (structuredSession) { + await discardStructuredWorkerSession(structuredSession.identity.sessionId, runtime) + } + return residualAgentTerminal +} diff --git a/src/main/runtime/rpc/methods/orchestration/worker/local-worker-start.ts b/src/main/runtime/rpc/methods/orchestration/worker/local-worker-start.ts index eb1ce43817d..48b14f9a84e 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/local-worker-start.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/local-worker-start.ts @@ -1,10 +1,13 @@ import type { TuiAgent } from '../../../../../../shared/tui-agent' import type { OrcaRuntimeService } from '../../../../orca-runtime' import type { OrchestrationDb } from '../../../../orchestration/db' -import { OrchestrationError } from '../../../../orchestration/orchestration-error' -import { buildDispatchPreamble } from '../../../../orchestration/preamble' import type { RunRow, TaskRow } from '../../../../orchestration/types' import { resolveDispatchCreator } from '../runs/dispatch-creator' +import { resolveDispatchCallerWorktreeId } from '../../orchestration-caller-workspace' +import { + resolveWorkerStartModeOnHost, + type WorkerStartModeReceipt +} from '../../orchestration-worker-start-mode' import { assertOrchestrationWorktreeCreationSupported } from './folder-worktree-placement' import type { WorkerStartInput } from './worker-start-schema' import { @@ -13,10 +16,13 @@ import { persistWorkerSetupWaitOutcome } from './worker-setup-gate' import { failWorkerStartWithReceipt } from './worker-start-receipt' -import { resolveResidualAgentTerminal } from './failed-start-residual-terminal' import { parseTaskDeps } from './task-deps-argument' +import { assertExplicitWorkerTerminalUsable } from './explicit-worker-terminal-validation' +import { deliverWorkerDispatchPreamble } from './deliver-worker-dispatch-preamble' +import { tearDownFailedWorkerStart } from './failed-worker-start-teardown' import { createExistingWorktreeWorkerTerminal, + createStructuredWorkerSessionForWorktree, createWorkerWorktree, monitorWorkerSetup, requireWorkerAuthority, @@ -40,15 +46,17 @@ export async function startLocalWorker(args: { coordinatorPane: string | null existingTask?: TaskRow orchestrationMutation?: WorkerStartMutation + /** Settings-driven; the executing host still gets to refuse below. */ + mode: WorkerStartModeReceipt }): Promise { const { params, runtime, db, run, coordinatorPane, existingTask, orchestrationMutation } = args const requestedWorktree = params.worktree ?? 'current' const createsWorktree = requestedWorktree === 'new-child' || requestedWorktree === 'new-top-level' const { agent, launch } = prepareLocalWorkerStart({ params, createsWorktree, runtime }) - const coordinatorTerminal = await runtime.showTerminal(params.from) + const coordinatorWorktreeId = await resolveDispatchCallerWorktreeId(runtime, params.from) const creationWorktree = createsWorktree - ? await runtime.showManagedWorktree(`id:${coordinatorTerminal.worktreeId}`) + ? await runtime.showManagedWorktree(`id:${coordinatorWorktreeId}`) : undefined if (creationWorktree) { await assertOrchestrationWorktreeCreationSupported({ @@ -60,38 +68,22 @@ export async function startLocalWorker(args: { let resolvedWorktree = creationWorktree ? undefined : requestedWorktree === 'current' - ? await runtime.showManagedTerminalWorkspace(`id:${coordinatorTerminal.worktreeId}`) + ? await runtime.showManagedTerminalWorkspace(`id:${coordinatorWorktreeId}`) : await runtime.showManagedTerminalWorkspace(requestedWorktree) if (params.terminal) { - const explicitTerminal = await runtime.showTerminal(params.terminal) - const targetPane = runtime.getTerminalPaneKey(params.terminal) - const callerPane = coordinatorPane ?? runtime.getTerminalPaneKey(params.from) - if ( - explicitTerminal.handle === coordinatorTerminal.handle || - (targetPane !== null && targetPane === callerPane) - ) { - // A coordinator adopted as its own worker answers its own dispatch preamble forever. - throw new OrchestrationError( - 'terminal_is_coordinator', - `Terminal ${params.terminal} is this coordinator's own terminal. Pass --terminal for a different agent pane, or omit it so worker-start creates one.` - ) - } - if (explicitTerminal.worktreeId !== resolvedWorktree?.id) { - throw new OrchestrationError( - 'terminal_worktree_mismatch', - `Terminal ${params.terminal} does not belong to worktree ${resolvedWorktree?.id}.` - ) - } - if (!(await runtime.isTerminalRunningAgent(params.terminal))) { - throw new OrchestrationError( - 'agent_unconfigured', - `Terminal ${params.terminal} is not running a recognized agent.` - ) - } + await assertExplicitWorkerTerminalUsable({ + runtime, + terminal: params.terminal, + from: params.from, + coordinatorPane, + resolvedWorktreeId: resolvedWorktree?.id + }) } + const mode = await resolveWorkerStartModeOnHost(runtime, args.mode, resolvedWorktree?.id, agent) const startOptions = { worktree: requestedWorktree, + mode, resolvedWorktreeId: resolvedWorktree?.id ?? null, name: params.name ?? null, repo: params.repo ?? creationWorktree?.repoId ?? null, @@ -135,6 +127,9 @@ export async function startLocalWorker(args: { ) } let terminalHandle = params.terminal + let structuredSession: Awaited< + ReturnType + > | null = null let terminalRevealWarning: string | undefined let failedStage = 'terminal_create' let setupReceipt: WorkerSetupReceipt = { @@ -162,6 +157,21 @@ export async function startLocalWorker(args: { resolvedWorktree = created.worktree terminalHandle = created.terminalHandle setupReceipt = created.setupReceipt + } else if (!terminalHandle && mode.mode === 'structured') { + db.recordWorkerStage({ + dispatchId: started.dispatch.id, + stage: 'terminal_creating', + worktreeId: resolvedWorktree!.id, + effects + }) + structuredSession = await createStructuredWorkerSessionForWorktree({ + runtime, + worktreeId: resolvedWorktree!.id, + agent: agent as TuiAgent, + dispatchId: started.dispatch.id, + effects + }) + terminalHandle = structuredSession.identity.handle } else if (!terminalHandle) { db.recordWorkerStage({ dispatchId: started.dispatch.id, @@ -200,20 +210,24 @@ export async function startLocalWorker(args: { persistWorkerReadinessStage(setupStage) failedStage = 'agent_readiness' - const wait = await runtime.waitForTerminal(terminalHandle, { - condition: 'tui-idle', - timeoutMs: params.timeoutMs ?? 60_000 - }) - persistWorkerSetupWaitOutcome({ ...setupStage, wait }) - if (!wait.satisfied) { - if (setupReceipt.state === 'failed') { - failedStage = 'setup_wait' + // A structured session is ready the moment its attach returns ok: there is no boot-to-idle + // gap and no terminal title to read an idle edge from. + if (!structuredSession) { + const wait = await runtime.waitForTerminal(terminalHandle, { + condition: 'tui-idle', + timeoutMs: params.timeoutMs ?? 60_000 + }) + persistWorkerSetupWaitOutcome({ ...setupStage, wait }) + if (!wait.satisfied) { + if (setupReceipt.state === 'failed') { + failedStage = 'setup_wait' + } + throw new Error( + wait.blockedReason + ? `Agent startup blocked: ${wait.blockedReason}` + : `Agent did not become ready (${wait.status}).` + ) } - throw new Error( - wait.blockedReason - ? `Agent startup blocked: ${wait.blockedReason}` - : `Agent did not become ready (${wait.status}).` - ) } const terminalAuthority = requireWorkerAuthority(runtime, terminalHandle) const capability = db.prepareStartingWorkerAuthority({ @@ -227,19 +241,17 @@ export async function startLocalWorker(args: { }) failedStage = 'dispatch_input' - const preamble = buildDispatchPreamble({ - taskId: task.id, + const promptDelivery = await deliverWorkerDispatchPreamble({ + runtime, + structuredSession, + terminalHandle, dispatchId: started.dispatch.id, + dispatchDepth: started.dispatch.depth, + taskId: task.id, taskSpec: task.spec, coordinatorHandle: params.from, - workerHandle: terminalHandle, dispatchCapability: capability, devMode: params.devMode, - cliCommand: runtime.getTerminalOrchestrationCliCommand(terminalHandle) - }) - const prompt = await runtime.sendTerminalAgentPrompt(terminalHandle, preamble, { - acceptQueued: true, - observationTimeoutMs: 0, requestId: orchestrationMutation?.requestId ?? started.dispatch.id }) effects.push({ @@ -265,15 +277,18 @@ export async function startLocalWorker(args: { stage: worker.stage, setup: setupReceipt, launch: launch.receipt, + mode, timeoutMs: params.timeoutMs ?? 60_000, effects, - ...(prompt.prompt ? { prompt: prompt.prompt } : {}), + ...(promptDelivery ? { prompt: promptDelivery } : {}), residualResources: [], ...(terminalRevealWarning ? { warning: terminalRevealWarning } : {}) } } catch (error) { - const residualAgentTerminal = resolveResidualAgentTerminal({ + const residualAgentTerminal = await tearDownFailedWorkerStart({ runtime, + structuredSession, + dispatchId: started.dispatch.id, effects, terminalHandle, worktreeId: resolvedWorktree?.id ?? null @@ -287,6 +302,7 @@ export async function startLocalWorker(args: { error, setup: setupReceipt, launch: launch.receipt, + mode, ...(residualAgentTerminal ? { residualAgentTerminal } : {}) }) } diff --git a/src/main/runtime/rpc/methods/orchestration/worker/structured-worker-release-stop.ts b/src/main/runtime/rpc/methods/orchestration/worker/structured-worker-release-stop.ts new file mode 100644 index 00000000000..4a32147d1a6 --- /dev/null +++ b/src/main/runtime/rpc/methods/orchestration/worker/structured-worker-release-stop.ts @@ -0,0 +1,50 @@ +import type { OrcaRuntimeService } from '../../../../orca-runtime' +import type { OrchestrationDb } from '../../../../orchestration/db' +import type { WorkerTerminalResourceRow } from '../../../../orchestration/worker-terminal-ownership' +import { stopStructuredWorker } from '../../orchestration-structured-worker-lifecycle' +import type { StructuredWorkerIdentity } from '../../../../structured-worker-identity' +import { archiveSummary } from './worker-terminal-resource-presentation' +import type { WorkerReleaseReceipt } from './worker-release-completion' + +/** + * The close half of a release for a worker that IS a structured session. + * + * Separate from the PTY close for the same reason the delivery lane is: there is no terminal to + * close and no exit to observe, so the host's own settlement is the only proof available. Only a + * proven close may settle; an unproven one reports `release_unknown` and stays retryable under the + * same request id. + */ +export async function stopStructuredWorkerForRelease(args: { + structured: StructuredWorkerIdentity + dispatchId: string + resource: WorkerTerminalResourceRow + runtime: OrcaRuntimeService + db: OrchestrationDb + archiveSource: string | null + archiveStatus: string | null +}): Promise { + const { structured, dispatchId, resource, runtime, db } = args + const stop = await stopStructuredWorker(structured, dispatchId, runtime) + if (!stop.stopped) { + const unknown = db.markWorkerTerminalReleaseUnknown( + resource.id, + stop.reason ?? 'The structured session close was not proven.' + ) + return { + dispatchId, + state: 'release_unknown', + processAction: stop.closeAttempted ? 'closed_agent_terminal' : 'none', + archive: { source: args.archiveSource, status: args.archiveStatus }, + lastError: unknown.release_error ?? stop.reason, + recovery: `Inspect with: orca orchestration worker-show --dispatch ${dispatchId} --json — then repeat worker-release with the same --retry-request.` + } + } + const settled = db.settleWorkerTerminalRelease(resource.id) + runtime.notifyMessageArrived(`dispatch:${dispatchId}`, 'status') + return { + dispatchId, + state: 'released', + processAction: 'closed_agent_terminal', + archive: archiveSummary(settled) + } +} diff --git a/src/main/runtime/rpc/methods/orchestration/worker/worker-archive-read.ts b/src/main/runtime/rpc/methods/orchestration/worker/worker-archive-read.ts index 4f2a4f7e2a1..8996f8e40a4 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/worker-archive-read.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/worker-archive-read.ts @@ -19,6 +19,8 @@ import { decodeWorkerOutputCursor, encodeWorkerOutputCursor } from '../../../../orchestration/worker-output-cursor' +import type { WorkerStructuredJournalArchive } from '../../../../orchestration/structured-worker-journal-archive' +import { readArchivedStructuredJournal } from '../../orchestration-structured-worker-lifecycle' const ARCHIVED_TERMINAL_PAGE_LINES = 2_000 @@ -43,11 +45,29 @@ export async function readArchivedWorkerOutput(args: { `Dispatch ${args.dispatchId} was released without a preserved output archive.` ) } + if (archive.kind === 'structured_journal') { + if (args.source === 'terminal') { + throw new OrchestrationError( + 'archive_unavailable', + `Dispatch ${args.dispatchId} preserved transcript output only; terminal output was released.` + ) + } + return readArchivedStructuredJournal({ + dispatchId: args.dispatchId, + workerState: args.workerState, + resourceId: args.resource.id, + createdAt: archive.created_at, + releaseState: args.resource.release_state, + archive: JSON.parse(archive.content) as WorkerStructuredJournalArchive, + ...(args.cursor === undefined ? {} : { cursor: args.cursor }), + ...(args.limit === undefined ? {} : { limit: args.limit }) + }) + } if (archive.kind === 'transcript_pin') { if (args.source === 'terminal') { throw new OrchestrationError( 'archive_unavailable', - `Dispatch ${args.dispatchId} preserved structured transcript output only; terminal output was released.` + `Dispatch ${args.dispatchId} preserved transcript output only; terminal output was released.` ) } return readFrozenTranscript( diff --git a/src/main/runtime/rpc/methods/orchestration/worker/worker-control.ts b/src/main/runtime/rpc/methods/orchestration/worker/worker-control.ts index 3ba64a29918..cf08ce31f69 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/worker-control.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/worker-control.ts @@ -14,6 +14,8 @@ import { showContextOnlyWorker } from './worker-observation' import { readArchivedWorkerOutput } from './worker-archive-read' +import { readStructuredWorkerOutput } from '../../orchestration-structured-worker-lifecycle' +import { releaseStructuredWorkerSession } from '../../orchestration-structured-worker-session' import { readExactWorkerOutput } from './worker-output' import { exposeWorkerTerminalResource } from './worker-release-completion' import { readFederatedWorkerOutput } from '../federation/federated-worker-read' @@ -146,6 +148,23 @@ export const ORCHESTRATION_WORKER_CONTROL_METHODS: RpcMethod[] = [ `Worker Dispatch ${params.dispatch} no longer resolves to its exact process.` ) } + const structured = readStructuredWorkerOutput({ + db, + dispatchId: params.dispatch, + workerState: worker?.state ?? 'unsupervised', + // Reused, never re-derived: being able to read the journal proves the host is installed, + // not that the provider child is alive. + liveness: + observation.status === 'live' || observation.status === 'exited' + ? observation.status + : 'unverifiable', + source: params.source, + cursor: params.cursor, + limit: params.limit + }) + if (structured) { + return structured + } const output = await readExactWorkerOutput({ runtime, dispatchId: params.dispatch, @@ -186,6 +205,10 @@ export const ORCHESTRATION_WORKER_CONTROL_METHODS: RpcMethod[] = [ const abandoned = runtime.getOrchestrationDb().abandonWorkerDispatch(params.dispatch) if (abandoned.disposition === 'context_only') { if (!abandoned.alreadySettled) { + // Abandon settles the Dispatch, so it owes the same hold release stop and release do. + // A surviving hold pins the provider child for the life of the app and makes host crash + // recovery respawn a worker nobody is waiting on. + releaseStructuredWorkerSession(params.dispatch, runtime) runtime.notifyMessageArrived(`dispatch:${params.dispatch}`, 'status') } return { @@ -200,6 +223,7 @@ export const ORCHESTRATION_WORKER_CONTROL_METHODS: RpcMethod[] = [ } const worker = abandoned.worker if (abandoned.disposition === 'abandoned') { + releaseStructuredWorkerSession(params.dispatch, runtime) runtime.notifyMessageArrived(`dispatch:${params.dispatch}`, 'status') } return { diff --git a/src/main/runtime/rpc/methods/orchestration/worker/worker-observation.ts b/src/main/runtime/rpc/methods/orchestration/worker/worker-observation.ts index 610195b4249..83b3d020f4c 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/worker-observation.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/worker-observation.ts @@ -5,6 +5,10 @@ import { OrchestrationError } from '../../../../orchestration/orchestration-erro import { parseWorkerTerminalHostScope } from '../../../../orchestration/worker-terminal-process-liveness' import type { OrchestrationFleetWorker } from '../../../../../../shared/orchestration-fleet-projection' import { projectWorkerFleet } from './worker-list-projection' +import { + observeStructuredWorker, + resolveStructuredWorkerForDispatch +} from '../../orchestration-structured-worker-lifecycle' import type { DispatchContextRow, FederatedDispatchRow, @@ -30,6 +34,28 @@ export async function inspectWorkerTerminal( if (!terminalHandle) { return { terminal: null, exact: false, status: 'unattached' } } + const structured = resolveStructuredWorkerForDispatch(db, dispatchId) + if (structured) { + // Exactness is the recorded pane and lineage, which the runtime getters answer from the + // structured registry; there is no terminal to show. + // + // `agentWait` is deliberately ABSENT rather than null. Null is the contract's "Orca looked and + // found no wait", and nothing here looks: a structured worker parks on a journal question item, + // which no terminal prompt scan can see. Reporting null would tell a coordinator the worker is + // not waiting, which is the one thing the field's own documentation forbids inferring. + const exact = db.isDispatchProcessCurrent({ + dispatchId, + paneKey: structured.paneKey, + processIncarnation: structured.processIncarnation + }) + const observation = observeStructuredWorker(structured) + return { + terminal: null, + exact, + status: exact ? observation.status : 'identity_changed', + ...(exact && observation.reason ? { reason: observation.reason } : {}) + } + } const terminal = await runtime.showTerminal(terminalHandle).catch(() => null) if (!terminal) { return { terminal: null, exact: false, status: 'missing' } diff --git a/src/main/runtime/rpc/methods/orchestration/worker/worker-release-completion.ts b/src/main/runtime/rpc/methods/orchestration/worker/worker-release-completion.ts index a5fcee1e921..b686cdfe6cd 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/worker-release-completion.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/worker-release-completion.ts @@ -1,5 +1,6 @@ import type { OrchestrationDb } from '../../../../orchestration/db' import type { + WorkerTerminalArchiveKind, WorkerTerminalArchiveStatus, WorkerTerminalResourceRow, WorkerTerminalRetainedReason @@ -15,6 +16,9 @@ import { orchestrationTimestampToMs } from './worker-output' import { archiveSummary } from './worker-terminal-resource-presentation' import { classifyWorkerTerminalCloseError } from './worker-release-close-error' import { workerTerminalLeaseIsCurrent } from './worker-terminal-release-lease' +import { resolveStructuredWorkerForDispatch } from '../../orchestration-structured-worker-lifecycle' +import { stopStructuredWorkerForRelease } from './structured-worker-release-stop' +import { isStructuredWorkerHandle } from '../../../../structured-worker-identity' export { archiveSummary, @@ -88,6 +92,23 @@ async function completeWorkerTerminalReleaseOnce( args: WorkerTerminalReleaseArgs ): Promise { const { runtime, db, dispatchId, resource } = args + if (isStructuredWorkerHandle(resource.terminal_handle)) { + // Observation and archive capture both read the structured host, and after a restart nothing + // has installed it yet — the startup recovery reconciler runs exactly this path. Installing it + // here is what lets the release see the session instead of reporting it unreadable. + // + // NOT yet handled, and deliberately follow-up: rebinding a restarted runtime to a structured + // worker's hold and redrive subscription. Until that exists, a worker that survives a restart + // keeps no hold, so its child is evictable and its parked mail waits for the next arrival + // rather than a settle edge. + await runtime.ensureStructuredAgentSessionHost().catch((error: unknown) => { + console.warn( + '[orchestration] structured host install failed before release', + dispatchId, + error + ) + }) + } const worker = db.getWorkerDispatch(dispatchId) if (!worker || worker.agent_terminal_handle !== resource.terminal_handle) { const retained = db.revertWorkerTerminalReleaseToRetained(resource.id, 'identity_unproven') @@ -176,16 +197,18 @@ async function completeWorkerTerminalReleaseOnce( const archive = db.getWorkerTerminalArchive(dispatchId) let archiveSource = resource.archive_source as 'transcript' | 'terminal' | null let archiveStatus: WorkerTerminalArchiveStatus | null = resource.archive_status - let capturedArchive: { kind: 'transcript_pin' | 'terminal_tail'; content: string } | undefined + let capturedArchive: { kind: WorkerTerminalArchiveKind; content: string } | undefined + const structured = resolveStructuredWorkerForDispatch(db, dispatchId) if (!archive) { const captured = await captureWorkerOutputArchive({ runtime, dispatchId, terminalHandle: resource.terminal_handle, - attachedAtMs: orchestrationTimestampToMs(worker.created_at) + attachedAtMs: orchestrationTimestampToMs(worker.created_at), + structuredWorker: structured }) capturedArchive = { kind: captured.kind, content: JSON.stringify(captured.content) } - archiveSource = captured.kind === 'transcript_pin' ? 'transcript' : 'terminal' + archiveSource = captured.kind === 'terminal_tail' ? 'terminal' : 'transcript' archiveStatus = captured.status } else { const stored = summarizeWorkerOutputArchive(archive) @@ -220,6 +243,17 @@ async function completeWorkerTerminalReleaseOnce( } try { + if (structured) { + return await stopStructuredWorkerForRelease({ + structured, + dispatchId, + resource, + runtime, + db, + archiveSource, + archiveStatus + }) + } const close = await runtime.closeTerminal(resource.terminal_handle) if (!close.ptyKilled) { const reason = describeUnconfirmedAgentStop(close) diff --git a/src/main/runtime/rpc/methods/orchestration/worker/worker-release.ts b/src/main/runtime/rpc/methods/orchestration/worker/worker-release.ts index 46aa8a62175..2a24a3efd0e 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/worker-release.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/worker-release.ts @@ -1,7 +1,6 @@ import { z } from 'zod' import { OrchestrationError } from '../../../../orchestration/orchestration-error' import { defineMethod, type RpcMethod } from '../../../core' -import { requiredString } from '../../../schemas' import { releaseFederatedWorker } from '../federation/federated-worker-release' import { ORCHESTRATION_WORKER_LIST_METHOD } from './worker-list-method' import { resolvePinnedFederatedServer } from './worker-observation' @@ -134,11 +133,21 @@ export const ORCHESTRATION_WORKER_RELEASE_METHODS: RpcMethod[] = [ ORCHESTRATION_WORKER_LIST_METHOD, defineMethod({ name: 'orchestration.workerTerminalUserInput', - params: z.object({ paneKey: requiredString('Missing paneKey') }), + // `sessionId` addresses a worker that IS a structured agent session. Its pane key is a random + // identity credential that never leaves main, so the caller names the session and the owning + // runtime resolves it — a renderer echoing the pane key back would make it learnable. + params: z + .object({ paneKey: z.string().min(1).optional(), sessionId: z.string().min(1).optional() }) + .refine((value) => Boolean(value.paneKey ?? value.sessionId), 'Missing paneKey or sessionId'), // Real user keystrokes durably relinquish orchestration ownership on the owning runtime, so // restarts, SSH drops, remote viewing, and renderer remounts cannot erase the takeover. handler: (params, { runtime }) => { - const changed = runtime.getOrchestrationDb().markWorkerTerminalUserOwned(params.paneKey) + // A structured worker reports by session id; it has no pane of its own to name. + const paneKey = + params.paneKey ?? runtime.getStructuredWorkerPaneKeyForSession(params.sessionId!) + const changed = paneKey + ? runtime.getOrchestrationDb().markWorkerTerminalUserOwned(paneKey) + : 0 if (changed > 0) { // Only a real takeover retires the resource; ordinary panes report here too and must not // pay for a plan read on every keystroke window. diff --git a/src/main/runtime/rpc/methods/orchestration/worker/worker-start-receipt.ts b/src/main/runtime/rpc/methods/orchestration/worker/worker-start-receipt.ts index 08b1aeab735..9fd98dd9db3 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/worker-start-receipt.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/worker-start-receipt.ts @@ -2,6 +2,7 @@ import type { OrchestrationDb } from '../../../../orchestration/db' import { isAgentPromptStalledError } from '../../../../agent-prompt-submission-verification' import { isUnknownWorkerStartOutcome, type WorkerSetupReceipt } from './worker-topology' import type { OrchestrationWorkerLaunchReceipt } from './worker-launch-preferences' +import type { WorkerStartModeReceipt } from '../../orchestration-worker-start-mode' import { isAgentSessionPtyWriteRefusedError } from '../../../../../../shared/agent-session-pty-write-admission' import type { FailedStartTerminalAdoption } from '../../../../orchestration/db/worker-terminal/failed-start-terminal-adoption' import { structuredChatPtyWriteRefusalCopy } from '../../../../../../shared/agent-session-pty-write-refusal-copy' @@ -15,6 +16,7 @@ export function failWorkerStartWithReceipt(args: { error: unknown setup: WorkerSetupReceipt launch: OrchestrationWorkerLaunchReceipt + mode: WorkerStartModeReceipt /** The terminal this start created and never handed to an owner. */ residualAgentTerminal?: FailedStartTerminalAdoption }): unknown { @@ -49,6 +51,7 @@ export function failWorkerStartWithReceipt(args: { lastError: reason, setup: args.setup, launch: args.launch, + mode: args.mode, effects: JSON.parse(worker.effects) as unknown[], residualResources: JSON.parse(worker.residual_resources) as unknown[], ...(agentSessionRefusal ? { agentSessionRefusal } : {}), diff --git a/src/main/runtime/rpc/methods/orchestration/worker/worker-stop.ts b/src/main/runtime/rpc/methods/orchestration/worker/worker-stop.ts index b0cc88c51c1..605d8c52d4a 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/worker-stop.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/worker-stop.ts @@ -7,6 +7,11 @@ import { ORCHESTRATION_WORKER_STOP_VERDICT_RUNTIME_CAPABILITY } from '../../../. import type { RuntimeStatus } from '../../../../../../shared/runtime-types' import type { OrcaRuntimeService } from '../../../../orca-runtime' import { inspectWorkerTerminal, resolvePinnedFederatedServer } from './worker-observation' +import { + resolveStructuredWorkerForDispatch, + stopStructuredWorker +} from '../../orchestration-structured-worker-lifecycle' +import { isStructuredWorkerHandle } from '../../../../structured-worker-identity' const WorkerDispatchParams = z.object({ dispatch: requiredString('Missing --dispatch') }) @@ -126,6 +131,18 @@ export const ORCHESTRATION_WORKER_STOP_METHODS: RpcMethod[] = [ 'unknown' ) } + if (isStructuredWorkerHandle(handle)) { + // The same install release performs, for the same reason: after a restart nothing has + // installed the structured host, and both the observation below and the close read it. + // Without this a restarted worker answers `unknown` forever and can never be stopped. + await runtime.ensureStructuredAgentSessionHost().catch((error: unknown) => { + console.warn( + '[orchestration] structured host install failed before stop', + handle, + error + ) + }) + } const observation = await inspectWorkerTerminal(runtime, db, params.dispatch) // The host exit can settle this stop while terminal inspection is awaiting inventory. if (db.getWorkerDispatch(params.dispatch)?.state === 'stopped') { @@ -164,6 +181,28 @@ export const ORCHESTRATION_WORKER_STOP_METHODS: RpcMethod[] = [ 'none' ) } + const structured = resolveStructuredWorkerForDispatch(db, params.dispatch) + if (structured) { + const stop = await stopStructuredWorker(structured, params.dispatch, runtime) + if (!stop.stopped) { + // Close is retried by the host; only a proven exit may settle the dispatch. And when no + // close was issued at all — no host in this runtime generation — the receipt says so + // rather than crediting this runtime with a terminal it never touched. + return unknownReceipt( + params.dispatch, + db.markWorkerStopUnknown(params.dispatch, stop.reason ?? 'The close was not proven.'), + stop.closeAttempted ? 'closed_agent_terminal' : 'none' + ) + } + const stopped = db.settleWorkerStop(params.dispatch) + runtime.notifyMessageArrived(`dispatch:${params.dispatch}`, 'status') + return { + dispatchId: params.dispatch, + state: stopped.state, + alreadySettled: false, + processAction: 'closed_agent_terminal' + } + } const closed = await runtime .closeTerminal(handle) .then((close) => ({ close }) as const) diff --git a/src/main/runtime/rpc/methods/orchestration/worker/worker-terminal-release-lease.ts b/src/main/runtime/rpc/methods/orchestration/worker/worker-terminal-release-lease.ts index d0d5dd0a40d..50a97c9de4b 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/worker-terminal-release-lease.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/worker-terminal-release-lease.ts @@ -1,6 +1,9 @@ import type { OrcaRuntimeService } from '../../../../orca-runtime' import type { OrchestrationDb } from '../../../../orchestration/db' import type { WorkerTerminalResourceRow } from '../../../../orchestration/worker-terminal-ownership' +import type { WorkerDispatchRow } from '../../../../orchestration/types' +import { resolveStructuredWorkerIdentity } from '../../../../structured-worker-authority' +import { isStructuredWorkerHandle } from '../../../../structured-worker-identity' export function workerTerminalLeaseIsCurrent( runtime: OrcaRuntimeService, @@ -9,6 +12,9 @@ export function workerTerminalLeaseIsCurrent( resource: WorkerTerminalResourceRow ): boolean { const worker = db.getWorkerDispatch(dispatchId) + if (isStructuredWorkerHandle(resource.terminal_handle)) { + return structuredWorkerTerminalLeaseIsCurrent(db, dispatchId, worker, resource) + } const authority = runtime.getOrchestrationDispatchAuthority(resource.terminal_handle) // Exited PTYs retain identity and host evidence but no longer mint launch authority. return Boolean( @@ -24,3 +30,30 @@ export function workerTerminalLeaseIsCurrent( !db.workerTerminalResourceHasIdentityConflict(resource.id) ) } + +/** + * IDENTITY, not liveness. The durable row plus the session-lineage incarnation say whether this is + * still the same worker; whether its child is alive is what the observation reports, honestly, as + * live / unverifiable / exited. Asking the record for identity would make a restart — where the + * host may not be installed yet — read as a different worker, turning a durably requested release + * into a permanent `retained/identity_unproven`. + */ +function structuredWorkerTerminalLeaseIsCurrent( + db: OrchestrationDb, + dispatchId: string, + worker: WorkerDispatchRow | undefined, + resource: WorkerTerminalResourceRow +): boolean { + const identity = resolveStructuredWorkerIdentity(resource.terminal_handle, db) + return Boolean( + worker?.agent_terminal_handle === resource.terminal_handle && + identity && + resource.host_scope === JSON.stringify(identity.hostScope) && + db.isDispatchProcessCurrent({ + dispatchId, + paneKey: identity.paneKey, + processIncarnation: identity.processIncarnation + }) && + !db.workerTerminalResourceHasIdentityConflict(resource.id) + ) +} diff --git a/src/main/runtime/rpc/methods/orchestration/worker/worker-topology.ts b/src/main/runtime/rpc/methods/orchestration/worker/worker-topology.ts index e189e246bc4..d7c526696a9 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/worker-topology.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/worker-topology.ts @@ -2,6 +2,8 @@ import type { AgentLaunchPreferences } from '../../../../../../shared/agent-sess import type { TuiAgent } from '../../../../../../shared/tui-agent' import type { OrcaRuntimeService } from '../../../../orca-runtime' import type { OrchestrationDb } from '../../../../orchestration/db' +import { OrchestrationError } from '../../../../orchestration/orchestration-error' +import { createStructuredWorkerSession } from '../../orchestration-structured-worker-session' export type WorkerEffect = { kind: 'worktree' | 'terminal' | 'setup' | 'dispatch_input' @@ -83,6 +85,43 @@ export async function createExistingWorktreeWorkerTerminal(args: { return { handle: terminal.handle, warning: terminal.warning } } +/** + * A worker that IS a structured chat session, in the same shape the terminal path returns. + * + * `requireWorkerAuthority` needs no branch: the runtime's pane-key and process-incarnation getters + * consult the structured registry, so the handle minted here answers exactly like a PTY handle. + */ +export async function createStructuredWorkerSessionForWorktree(args: { + runtime: OrcaRuntimeService + worktreeId: string + agent: TuiAgent + dispatchId: string + effects: WorkerEffect[] +}): Promise>> { + if (args.agent !== 'claude' && args.agent !== 'codex') { + throw new OrchestrationError( + 'agent_unconfigured', + `Structured workers support claude and codex; ${args.agent} has no structured session.` + ) + } + const created = await createStructuredWorkerSession({ + runtime: args.runtime, + worktreeId: args.worktreeId, + agent: args.agent, + dispatchId: args.dispatchId, + onJournalActivity: (sessionId) => + args.runtime.notifyStructuredSessionJournalActivity?.(sessionId) + }) + args.effects.push({ + kind: 'terminal', + role: 'agent', + action: 'created', + id: created.identity.handle, + surface: 'background' + }) + return created +} + export function applyWaitForSetupOutcome( receipt: WorkerSetupReceipt, effects: WorkerEffect[], diff --git a/src/main/runtime/rpc/methods/orchestration/worker/workers.ts b/src/main/runtime/rpc/methods/orchestration/worker/workers.ts index dd9586abc42..6ccac1dea9e 100644 --- a/src/main/runtime/rpc/methods/orchestration/worker/workers.ts +++ b/src/main/runtime/rpc/methods/orchestration/worker/workers.ts @@ -2,6 +2,10 @@ import { OrchestrationError } from '../../../../orchestration/orchestration-erro import { defineMethod, type RpcMethod } from '../../../core' import { startFederatedWorker } from '../federation/federated-worker-start' import { startLocalWorker } from './local-worker-start' +import { + decideWorkerStartMode, + readWorkerStartModeSettings +} from '../../orchestration-worker-start-mode' import { resolveOrchestrationCaller } from '../runs/run-scope' import { WorkerStartParams } from './worker-start-schema' import { @@ -45,8 +49,15 @@ export const ORCHESTRATION_WORKER_START_METHODS: RpcMethod[] = [ ) } await assertWorkerStartTaskSpecWithinPromptBudget(params.spec ?? existingTask!.spec) + const mode = decideWorkerStartMode({ + params, + settings: readWorkerStartModeSettings(runtime), + platform: process.platform + }) if (params.on) { - return startFederatedWorker({ + // A remote worker is always a terminal agent; the mode receipt rides along so the + // coordinator still learns why its structured default did not apply. + const receipt = await startFederatedWorker({ params, runtime, db, @@ -54,6 +65,7 @@ export const ORCHESTRATION_WORKER_START_METHODS: RpcMethod[] = [ task: existingTask, orchestrationMutation }) + return receipt && typeof receipt === 'object' ? { ...receipt, mode } : receipt } return startLocalWorker({ params: { ...params, timeoutMs: readinessTimeoutMs }, @@ -62,7 +74,8 @@ export const ORCHESTRATION_WORKER_START_METHODS: RpcMethod[] = [ run, coordinatorPane, existingTask, - orchestrationMutation + orchestrationMutation, + mode }) } }) diff --git a/src/main/runtime/rpc/methods/structured-agent-session-create.ts b/src/main/runtime/rpc/methods/structured-agent-session-create.ts new file mode 100644 index 00000000000..75a13ba6af9 --- /dev/null +++ b/src/main/runtime/rpc/methods/structured-agent-session-create.ts @@ -0,0 +1,131 @@ +/** + * Creating a structured session for a worktree: resolve the create intent, attach it under the + * host-computed fingerprint, then publish its tab. + * + * Extracted from `agentSession.create` so orchestration can start a native-born structured worker + * on exactly the same path. `activate` is the only knob the two callers differ on: a chat the user + * asked for takes the surface, a background dispatch must not steal it (the terminal worker path's + * `surfaceOwner: false`). + * + * The prepare/commit split is the pre-commit boundary, not a style choice: nothing before `attach` + * commits a session, so that span answers with a refusal, and nothing after it may be folded back + * in. Both callers run the same two halves, so orchestration gets that guarantee too. + */ + +import { computeAgentSessionPayloadFingerprint } from '../../../../shared/agent-session-mutation-envelope' +import type { + AgentSessionAttachResult, + AgentSessionMutationEnvelope, + AgentSessionMutationResult +} from '../../../../shared/agent-session-wire' +import { + attachFingerprintFields, + type AgentSessionAttachParams +} from '../../../native-chat/agent-session-wire/structured-agent-session-attach' +import type { StructuredAgentSessionHost } from '../../../native-chat/agent-session-wire/structured-agent-session-host' +import type { StructuredAgentSessionCaller } from '../../../native-chat/agent-session-wire/structured-agent-session-host-types' +import type { OrcaRuntimeService } from '../../orca-runtime' +import { + resolveUncommittedStructuredCreate, + type StructuredCreateRefused +} from './structured-agent-session-precommit-refusal' + +export type PreparedStructuredAgentSessionCreate = { + host: StructuredAgentSessionHost + attachParams: AgentSessionAttachParams + /** Null when the caller supplied its own location; only a resolved worktree publishes a tab. */ + tab: { workspaceId: string; agent: 'claude' | 'codex' } | null +} + +/** The pre-commit half. Throws; the caller is expected to run it inside + * `resolveUncommittedStructuredCreate` so a failure reaches the client as a refusal. */ +export async function prepareStructuredAgentSessionCreateForWorktree(args: { + runtime: OrcaRuntimeService + /** Installs the host lazily; called at the same point the RPC handler always installed it. */ + ensureHost: () => Promise + envelope: AgentSessionMutationEnvelope + worktree: string + agent: 'claude' | 'codex' +}): Promise { + const resolved = await args.runtime.resolveStructuredAgentSessionCreateIntent({ + envelope: args.envelope, + worktree: args.worktree, + agent: args.agent + }) + const hostFingerprint = computeAgentSessionPayloadFingerprint({ + method: 'agentSession.attach', + sessionId: args.envelope.sessionId, + fields: attachFingerprintFields({ ...resolved, envelope: args.envelope }) + }) + const host = await args.ensureHost() + const { agent: _resolvedAgent, provider: _resolvedProvider, ...resolvedAttach } = resolved + return { + host, + attachParams: { + ...resolvedAttach, + provider: resolved.provider as 'claude' | 'codex', + agent: resolved.agent as 'claude' | 'codex', + envelope: { ...args.envelope, payloadFingerprint: hostFingerprint } + }, + tab: { + workspaceId: resolved.location.workspaceId, + agent: resolved.agent as 'claude' | 'codex' + } + } +} + +/** The commit half. Past `attach`, a failure no longer proves the session does not exist. */ +export async function commitStructuredAgentSessionCreate(args: { + runtime: OrcaRuntimeService + caller: StructuredAgentSessionCaller + prepared: PreparedStructuredAgentSessionCreate + activate: boolean +}): Promise> { + const { prepared } = args + const result = await prepared.host.attach(args.caller, prepared.attachParams) + if (!result.ok || !prepared.tab) { + return result + } + try { + await args.runtime.publishStructuredAgentSessionTab({ + workspaceId: prepared.tab.workspaceId, + sessionId: result.value.sessionId, + agent: prepared.tab.agent, + activate: args.activate + }) + } catch (error) { + console.warn('[agent-session] create committed before tab publication failed', error) + return { + ok: false, + refusal: { + code: 'agent_session_operation_unknown', + message: 'The chat may have been created, but its tab could not be confirmed.' + } + } + } + return result +} + +export async function createStructuredAgentSessionForWorktree(args: { + runtime: OrcaRuntimeService + ensureHost: () => Promise + caller: StructuredAgentSessionCaller + envelope: AgentSessionMutationEnvelope + worktree: string + agent: 'claude' | 'codex' + activate: boolean +}): Promise> { + const prepared: PreparedStructuredAgentSessionCreate | StructuredCreateRefused = + await resolveUncommittedStructuredCreate(() => + prepareStructuredAgentSessionCreateForWorktree(args) + ) + if ('refusal' in prepared) { + return { ok: false, refusal: prepared.refusal } + } + return commitStructuredAgentSessionCreate({ + runtime: args.runtime, + caller: args.caller, + prepared, + activate: args.activate + }) +} diff --git a/src/main/runtime/rpc/methods/structured-agent-session.test.ts b/src/main/runtime/rpc/methods/structured-agent-session.test.ts index 8defafb4433..fe24a11d047 100644 --- a/src/main/runtime/rpc/methods/structured-agent-session.test.ts +++ b/src/main/runtime/rpc/methods/structured-agent-session.test.ts @@ -98,6 +98,7 @@ function statusFeed(): StructuredAgentSessionStatusFeed { { journal: { isReadOnly: false, + lastActivityAt: () => 2, snapshot: () => ({ items: STATUS_ITEMS }) } as unknown as AgentSessionJournal, params: { location: { workspaceId: 'workspace-1' }, provider: 'codex' as const } @@ -387,7 +388,7 @@ describe('capability gating', () => { } // Bump deliberately: the whole agentSession.* surface is behind the structured capability, // so an additive method is invisible to old clients and needs no protocol bump. - expect(STRUCTURED_AGENT_SESSION_METHODS).toHaveLength(19) + expect(STRUCTURED_AGENT_SESSION_METHODS).toHaveLength(20) }) it('hides the surface from a declared client that did not advertise it', async () => { @@ -815,7 +816,8 @@ describe('agentSession.subscribeStatus', () => { workspaceId: 'workspace-1', agent: 'codex', status: 'working', - latestPrompt: 'write a poem' + latestPrompt: 'write a poem', + updatedAt: 2 } ] } diff --git a/src/main/runtime/rpc/methods/structured-agent-session.ts b/src/main/runtime/rpc/methods/structured-agent-session.ts index 9ca3c632a83..e9829d2945f 100644 --- a/src/main/runtime/rpc/methods/structured-agent-session.ts +++ b/src/main/runtime/rpc/methods/structured-agent-session.ts @@ -18,10 +18,11 @@ import { structuredCallerFor as callerFor, supportsStructuredSessions } from './structured-agent-session-gate' +import type { AgentSessionAttachParams } from '../../../native-chat/agent-session-wire/structured-agent-session-attach' import { - attachFingerprintFields, - type AgentSessionAttachParams -} from '../../../native-chat/agent-session-wire/structured-agent-session-attach' + commitStructuredAgentSessionCreate, + prepareStructuredAgentSessionCreateForWorktree +} from './structured-agent-session-create' import { STRUCTURED_AGENT_SESSION_HOLD_METHODS } from './structured-agent-session-hold' import { STRUCTURED_AGENT_SESSION_REVEAL_METHODS } from './structured-agent-session-reveal' import { resolveUncommittedStructuredCreate } from './structured-agent-session-precommit-refusal' @@ -110,28 +111,16 @@ export const STRUCTURED_AGENT_SESSION_METHODS: RpcAnyMethod[] = [ if (conflict) { return { refusal: conflict } } - const resolved = await ctx.runtime.resolveStructuredAgentSessionCreateIntent(params) - const hostFingerprint = computeAgentSessionPayloadFingerprint({ - method: 'agentSession.attach', - sessionId: params.envelope.sessionId, - fields: attachFingerprintFields({ ...resolved, envelope: params.envelope }) + return prepareStructuredAgentSessionCreateForWorktree({ + runtime: ctx.runtime, + ensureHost: async () => { + await ensureHostInstalled(ctx) + return requireHost(ctx) + }, + envelope: params.envelope, + worktree: params.worktree, + agent: params.agent as 'claude' | 'codex' }) - await ensureHostInstalled(ctx) - const { agent: _resolvedAgent, provider: _resolvedProvider, ...resolvedAttach } = resolved - const attachParams: AgentSessionAttachParams = { - ...resolvedAttach, - provider: resolved.provider as 'claude' | 'codex', - agent: resolved.agent as 'claude' | 'codex', - envelope: { ...params.envelope, payloadFingerprint: hostFingerprint } - } - return { - host: requireHost(ctx), - attachParams, - tab: { - workspaceId: resolved.location.workspaceId, - agent: resolved.agent as 'claude' | 'codex' - } - } } const { host, attachParams } = await resolveClientSuppliedAttach(params, ctx) return { host, attachParams, tab: null } @@ -139,27 +128,12 @@ export const STRUCTURED_AGENT_SESSION_METHODS: RpcAnyMethod[] = [ if ('refusal' in prepared) { return { ok: false, refusal: prepared.refusal } } - const result = await prepared.host.attach(callerFor(ctx), prepared.attachParams) - if (result.ok && prepared.tab) { - try { - await ctx.runtime.publishStructuredAgentSessionTab({ - workspaceId: prepared.tab.workspaceId, - sessionId: result.value.sessionId, - agent: prepared.tab.agent, - activate: true - }) - } catch (error) { - console.warn('[agent-session] create committed before tab publication failed', error) - return { - ok: false, - refusal: { - code: 'agent_session_operation_unknown', - message: 'The chat may have been created, but its tab could not be confirmed.' - } - } - } - } - return result + return commitStructuredAgentSessionCreate({ + runtime: ctx.runtime, + caller: callerFor(ctx), + prepared, + activate: true + }) } }), defineMethod({ @@ -224,6 +198,11 @@ export const STRUCTURED_AGENT_SESSION_METHODS: RpcAnyMethod[] = [ params: OptionsParams, handler: async (params, ctx) => requireHost(ctx).readOptions(params.sessionId) }), + defineMethod({ + name: 'agentSession.commands', + params: OptionsParams, + handler: async (params, ctx) => requireHost(ctx).readCommands(params.sessionId) + }), defineMethod({ name: 'agentSession.history', params: HistoryParams, diff --git a/src/main/runtime/rpc/methods/structured-worker-read-cursor.test.ts b/src/main/runtime/rpc/methods/structured-worker-read-cursor.test.ts new file mode 100644 index 00000000000..1a97223bc39 --- /dev/null +++ b/src/main/runtime/rpc/methods/structured-worker-read-cursor.test.ts @@ -0,0 +1,146 @@ +/** + * The structured `worker-read --source transcript` cursor across a MUTATING journal. + * + * The journal is a reduced, mutable timeline, and the old `source_changed` anchor fingerprinted + * only the oldest item's id. It fired when the window slid off the front and could not fire when + * the page's contents changed under a stable oldest item — the normal case. Two silent failures + * followed, both returning ok: an already-delivered item revised in place was never redelivered + * (omission), and a pending approval resolving into the MIDDLE of the array shifted the caller's + * saved index back onto content it already had (duplication). + * + * A static-journal test passes either way, so every case here mutates between reads. + */ + +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentJournalRenderItem } from '../../../../shared/agent-session-journal-types' +import type { StructuredWorkerIdentity } from '../../structured-worker-identity' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../../../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { readStructuredWorkerJournal } = await import('./orchestration-structured-worker-lifecycle') + +const IDENTITY: StructuredWorkerIdentity = { + handle: 'structworker_1', + sessionId: 'session-1', + agent: 'claude', + paneKey: 'structured-agent-session-session-1:11111111-1111-4111-a111-111111111111', + processIncarnation: 'structured:session-1', + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } +} + +function message(itemId: string, text: string, revision = 1): AgentJournalRenderItem { + return { + itemId, + revision, + sequence: Number(itemId.slice(1)), + observedAt: 1, + body: { kind: 'message', role: 'assistant', blocks: [{ type: 'text', text }] } + } as unknown as AgentJournalRenderItem +} + +/** Projects to null while pending, and to a system message once resolved — mid-array. */ +function approval(itemId: string, resolved: boolean, revision = 1): AgentJournalRenderItem { + return { + itemId, + revision, + sequence: Number(itemId.slice(1)), + observedAt: 1, + body: { + kind: 'approval', + title: 'run it?', + detail: null, + resolution: { state: resolved ? 'approved' : 'pending' } + } + } as unknown as AgentJournalRenderItem +} + +function installJournal(items: AgentJournalRenderItem[]): void { + hostRef.current = { + deps: { store: { getRecord: () => null } }, + hasSession: () => true, + history: () => ({ page: { items, hasOlder: false } }) + } +} + +function read(cursor?: string, limit?: number) { + return readStructuredWorkerJournal({ + identity: IDENTITY, + dispatchId: 'd1', + workerState: 'ready', + liveness: 'live', + agent: 'claude', + ...(cursor === undefined ? {} : { cursor }), + ...(limit === undefined ? {} : { limit }) + }) +} + +function textsOf(result: ReturnType): string[] { + return result.transcript.messages.map((entry) => + entry.blocks.map((block) => ('text' in block ? block.text : '')).join('') + ) +} + +describe('the structured worker-read cursor over a mutating journal', () => { + beforeEach(() => { + hostRef.current = null + }) + + it('refuses to resume when an already-delivered item was revised in place', () => { + // The `"hel"` / `"hello"` defect. The caller is handed a coalesced snapshot, resumes past it, + // and the item is later revised at its original sequence — under the old anchor the resume was + // accepted and that revision was never delivered to anyone. + installJournal([message('i1', 'hel'), message('i2', 'second')]) + const first = read(undefined, 1) + expect(textsOf(first)).toEqual(['hel']) + + installJournal([message('i1', 'hello world', 2), message('i2', 'second')]) + expect(() => read(first.cursor)).toThrow(/source changed/i) + }) + + it('refuses to resume when a resolved prompt inserts ahead of the caller position', () => { + // Duplication. A pending approval projects to null, so resolving it inserts a message in the + // MIDDLE; the oldest item never moved, so the old anchor accepted a now-stale index and the + // caller re-read content it already had. + installJournal([message('i1', 'first'), approval('i2', false), message('i3', 'second')]) + const first = read(undefined, 2) + expect(textsOf(first)).toEqual(['first', 'second']) + + installJournal([message('i1', 'first'), approval('i2', true, 2), message('i3', 'second')]) + expect(() => read(first.cursor)).toThrow(/source changed/i) + }) + + it('still resumes across a page boundary when only unread tail items change', () => { + // The reason this is prefix-scoped and not whole-page: during an active turn the coalescer + // revises the streaming item every 60ms. Fingerprinting the whole page would invalidate the + // cursor continuously — a useless verb — while the worker is working. + installJournal([message('i1', 'first'), message('i2', 'streaming')]) + const first = read(undefined, 1) + expect(textsOf(first)).toEqual(['first']) + + installJournal([message('i1', 'first'), message('i2', 'streaming more', 7)]) + const second = read(first.cursor) + expect(textsOf(second)).toEqual(['streaming more']) + }) + + it('delivers every message exactly once when nothing below the cursor changes', () => { + // The property the two refusals above protect: no omission, no duplication. + installJournal([message('i1', 'a'), message('i2', 'b'), message('i3', 'c')]) + const first = read(undefined, 2) + const second = read(first.cursor, 2) + expect([...textsOf(first), ...textsOf(second)]).toEqual(['a', 'b', 'c']) + }) + + it('still refuses when the window slides off the front', () => { + // The case the old anchor DID catch, and which the prefix scoping must not lose: a slide + // shifts every index. + installJournal([message('i1', 'a'), message('i2', 'b')]) + const first = read(undefined, 1) + installJournal([message('i2', 'b'), message('i3', 'c')]) + expect(() => read(first.cursor)).toThrow(/source changed/i) + }) +}) diff --git a/src/main/runtime/rpc/methods/structured-worker-stop-receipt.test.ts b/src/main/runtime/rpc/methods/structured-worker-stop-receipt.test.ts new file mode 100644 index 00000000000..82cc53ff735 --- /dev/null +++ b/src/main/runtime/rpc/methods/structured-worker-stop-receipt.test.ts @@ -0,0 +1,123 @@ +/** + * What `worker-stop` may claim it did to a structured worker. + * + * A runtime generation with no structured host installed cannot reach the session at all. Saying + * `closed_agent_terminal` there credits this runtime with an action it never took, and a + * coordinator reading the receipt treats the worker's chat tab as gone. + */ + +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { OrcaRuntimeService } from '../../orca-runtime' +import { OrchestrationDb } from '../../orchestration/db' +import { setStructuredAgentSessionHost } from '../../../native-chat/agent-session-wire/structured-agent-session-registry' +import { + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} from '../../structured-worker-identity' +import { ORCHESTRATION_METHODS } from './orchestration' + +const SESSION = 'session-stop-receipt' +const HANDLE = 'structworker_22222222-2222-4222-a222-222222222222' +const WORKTREE = 'repo::worktree' + +describe('worker-stop on a structured worker this runtime cannot reach', () => { + let db: OrchestrationDb + let runtime: OrcaRuntimeService + + beforeEach(() => { + structuredWorkerIdentities.clear() + setStructuredAgentSessionHost(null) + db = new OrchestrationDb(':memory:') + runtime = new OrcaRuntimeService() + runtime.setOrchestrationDb(db) + // The install is what release already does; here it is a no-op so the host stays absent. + vi.spyOn(runtime, 'ensureStructuredAgentSessionHost').mockResolvedValue(undefined) + }) + + afterEach(() => { + db.close() + structuredWorkerIdentities.clear() + setStructuredAgentSessionHost(null) + vi.restoreAllMocks() + }) + + async function call(name: string, params: Record) { + const method = ORCHESTRATION_METHODS.find((candidate) => candidate.name === name) + if (!method) { + throw new Error(`Method not found: ${name}`) + } + return method.handler(method.params!.parse(params), { runtime }) + } + + function startStructuredWorker(): string { + const paneKey = mintStructuredWorkerPaneKey(SESSION) + const processIncarnation = structuredWorkerProcessIncarnation(SESSION) + structuredWorkerIdentities.register({ + handle: HANDLE, + sessionId: SESSION, + agent: 'claude', + paneKey, + processIncarnation, + worktreeId: WORKTREE, + hostScope: { kind: 'local', hostId: 'local' } + }) + const task = db.createTask({ spec: 'stop a structured worker' }) + const started = db.createStartingWorkerDispatch({ + creator: { kind: 'system' }, + maxDepth: Number.MAX_SAFE_INTEGER, + taskId: task.id, + startOptions: {}, + runtimeEpoch: runtime.getRuntimeId() + }) + db.prepareStartingWorkerAuthority({ + dispatchId: started.dispatch.id, + handle: HANDLE, + paneKey, + processIncarnation, + worktreeId: WORKTREE, + setupState: 'not_applicable', + effects: [{ kind: 'terminal', action: 'created', id: HANDLE }], + terminalOwnership: 'created' + }) + db.markWorkerDispatchReady(started.dispatch.id) + return started.dispatch.id + } + + it('keeps a restarted worker unsettled when close finds no attached session', async () => { + const dispatchId = startStructuredWorker() + const close = vi.fn(async () => {}) + setStructuredAgentSessionHost({ + close, + setSessionTabVisibility: async () => {}, + hasSession: () => false, + deps: { + store: { + getRecord: () => ({ + location: { executionHostId: 'local', wslDistro: null }, + lease: { runtimeKind: 'native', claimStatus: 'live', deathEvidence: null } + }) + } + } + } as never) + + await expect(call('orchestration.workerStop', { dispatch: dispatchId })).resolves.toMatchObject( + { + processAction: 'closed_agent_terminal', + state: 'stop_unknown' + } + ) + expect(close).toHaveBeenCalledWith(SESSION) + expect(db.getWorkerDispatch(dispatchId)?.state).toBe('stop_unknown') + }) + + it('reports that nothing was closed', async () => { + const dispatchId = startStructuredWorker() + await expect(call('orchestration.workerStop', { dispatch: dispatchId })).resolves.toMatchObject( + { + processAction: 'none', + state: 'stop_unknown' + } + ) + }) +}) diff --git a/src/main/runtime/rpc/methods/structured-worker-tab-retirement.test.ts b/src/main/runtime/rpc/methods/structured-worker-tab-retirement.test.ts new file mode 100644 index 00000000000..b48dccf52a3 --- /dev/null +++ b/src/main/runtime/rpc/methods/structured-worker-tab-retirement.test.ts @@ -0,0 +1,329 @@ +/** + * Every structured-worker settlement has to retire the chat tab the worker start published. + * + * `setSessionTabVisibility(false)` only clears the DURABLE restore index. Without the snapshot + * prune, a coordinator that dispatches and releases five structured workers leaves five dead + * "Claude Chat" tabs in the worktree's tab bar, and opening one re-attaches the released session + * outside orchestration's hold and eviction accounting. + */ + +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { setStructuredAgentSessionHost } from '../../../native-chat/agent-session-wire/structured-agent-session-registry' +import { OrcaRuntimeService } from '../../orca-runtime' +import type { OrchestrationDb } from '../../orchestration/db' +import type { WorkerTerminalResourceRow } from '../../orchestration/worker-terminal-ownership' +import { + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation, + type StructuredWorkerIdentity +} from '../../structured-worker-identity' + +const createSpy = vi.fn() +vi.mock('./structured-agent-session-create', () => ({ + createStructuredAgentSessionForWorktree: (...args: unknown[]) => createSpy(...args) +})) + +const { stopStructuredWorker } = await import('./orchestration-structured-worker-lifecycle') +const { createStructuredWorkerSession } = await import('./orchestration-structured-worker-session') +const { completeWorkerTerminalRelease } = + await import('./orchestration/worker/worker-release-completion') + +const WORKTREE = 'workspace-1' +const SESSION = 'session-1' +const HANDLE = 'structworker_11111111-1111-4111-a111-111111111111' +const HOST_SCOPE = { kind: 'local', hostId: 'local' } as const + +function installHost(options: { closeThrows?: boolean; lease?: Record } = {}) { + let attached = true + const setSessionTabVisibility = vi.fn(async () => {}) + const close = vi.fn(async () => { + if (options.closeThrows) { + throw new Error('close is queued for retry') + } + attached = false + }) + setStructuredAgentSessionHost({ + setSessionTabVisibility, + close, + hasSession: () => attached, + hold: async () => {}, + release: () => {}, + subscribe: () => () => {}, + deps: { + store: { + getRecord: () => ({ + location: { executionHostId: 'local', wslDistro: null }, + lease: options.lease ?? { + runtimeKind: 'native', + claimStatus: attached ? 'live' : 'released', + deathEvidence: attached + ? null + : { kind: 'exit-observed', detail: 'closed', observedAt: 1 }, + runtimeFence: 2 + } + }) + } + } + } as never) + return { close, setSessionTabVisibility } +} + +type RuntimeInternals = { + ensureStructuredAgentSessionHost(): Promise + notifyMessageArrived(...args: unknown[]): void + emitMobileSessionTabsSnapshot(snapshot: unknown): void +} + +async function runtimeShowingStructuredTab(): Promise<{ + runtime: OrcaRuntimeService + emit: ReturnType +}> { + const runtime = new OrcaRuntimeService() + const internal = runtime as unknown as RuntimeInternals + internal.ensureStructuredAgentSessionHost = async () => undefined + internal.notifyMessageArrived = vi.fn() + await runtime.publishStructuredAgentSessionTab({ + workspaceId: WORKTREE, + sessionId: SESSION, + agent: 'claude', + activate: true + }) + const emit = vi.fn() + const original = internal.emitMobileSessionTabsSnapshot.bind(runtime) + internal.emitMobileSessionTabsSnapshot = (snapshot: unknown) => { + emit(snapshot) + original(snapshot) + } + return { runtime, emit } +} + +async function structuredTabIds(runtime: OrcaRuntimeService): Promise { + const snapshot = await runtime.listMobileSessionTabs(`id:${WORKTREE}`) + return snapshot.tabs.map((tab) => tab.id) +} + +function registerIdentity(): StructuredWorkerIdentity { + return structuredWorkerIdentities.register({ + handle: HANDLE, + sessionId: SESSION, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION), + processIncarnation: structuredWorkerProcessIncarnation(SESSION), + worktreeId: WORKTREE, + hostScope: HOST_SCOPE + }) +} + +beforeEach(() => { + structuredWorkerIdentities.clear() + createSpy.mockReset() +}) + +afterEach(() => { + setStructuredAgentSessionHost(null) + structuredWorkerIdentities.clear() + vi.restoreAllMocks() +}) + +describe('structured worker stop retires the chat tab', () => { + it('prunes the tab from the live snapshot and re-emits it', async () => { + installHost() + const identity = registerIdentity() + const { runtime, emit } = await runtimeShowingStructuredTab() + expect(await structuredTabIds(runtime)).toEqual([`agent-session:${SESSION}`]) + + await expect(stopStructuredWorker(identity, 'd1', runtime)).resolves.toEqual({ + stopped: true, + closeAttempted: true + }) + + expect(await structuredTabIds(runtime)).toEqual([]) + const published = await runtime.listMobileSessionTabs(`id:${WORKTREE}`) + expect(published.tabGroups?.[0]?.tabOrder ?? []).toEqual([]) + expect(published.activeTabId).toBeNull() + expect(published.activeTabType).toBeNull() + expect(emit).toHaveBeenCalled() + }) + + it('leaves the tab alone when the close was NOT proven', async () => { + installHost({ closeThrows: true }) + const identity = registerIdentity() + const { runtime } = await runtimeShowingStructuredTab() + + const stop = await stopStructuredWorker(identity, 'd1', runtime) + + expect(stop.stopped).toBe(false) + expect(await structuredTabIds(runtime)).toEqual([`agent-session:${SESSION}`]) + }) + + it('cannot turn a proven stop into a retained one when the prune throws', async () => { + installHost() + const identity = registerIdentity() + const runtime = { + forgetStructuredSessionMail: vi.fn(), + retireStructuredAgentSessionTabFromSnapshot: vi.fn(() => { + throw new Error('snapshot is wedged') + }) + } as unknown as OrcaRuntimeService + + await expect(stopStructuredWorker(identity, 'd1', runtime)).resolves.toEqual({ + stopped: true, + closeAttempted: true + }) + expect(runtime.retireStructuredAgentSessionTabFromSnapshot).toHaveBeenCalledWith(SESSION) + }) + + it('settles a runtime that has no tab surface at all', async () => { + installHost() + const identity = registerIdentity() + await expect(stopStructuredWorker(identity, 'd1')).resolves.toEqual({ + stopped: true, + closeAttempted: true + }) + }) +}) + +describe('structured worker release retires the chat tab', () => { + it('prunes the tab once the release settles', async () => { + installHost() + const identity = registerIdentity() + const { runtime } = await runtimeShowingStructuredTab() + const resource = { + id: 'resource-1', + terminal_handle: HANDLE, + host_scope: JSON.stringify(HOST_SCOPE), + archive_source: 'transcript', + archive_status: 'captured', + ownership_state: 'owned', + release_state: 'requested' + } as WorkerTerminalResourceRow + const db = { + getWorkerDispatch: () => ({ + agent_terminal_handle: HANDLE, + created_at: '2026-09-05 00:00:00' + }), + getDispatchContextById: () => null, + isDispatchProcessCurrent: (args: { paneKey: string; processIncarnation: string }) => + args.paneKey === identity.paneKey && + args.processIncarnation === identity.processIncarnation, + workerTerminalResourceHasIdentityConflict: () => false, + getWorkerTerminalArchive: () => ({ kind: 'transcript_pin' }), + commitWorkerTerminalArchiveForRelease: () => ({ + ...resource, + release_state: 'releasing' + }), + settleWorkerTerminalRelease: () => ({ ...resource, release_state: 'released' }), + markWorkerTerminalReleaseUnknown: (_id: string, error: string) => ({ + ...resource, + release_state: 'unknown', + release_error: error + }) + } as unknown as OrchestrationDb + + await expect( + completeWorkerTerminalRelease({ runtime, db, dispatchId: 'd1', resource }) + ).resolves.toMatchObject({ state: 'released' }) + + expect(await structuredTabIds(runtime)).toEqual([]) + }) + + it('settles rather than wedging when the user already closed the worker chat tab', async () => { + // Closing the tab evicts the child and detaches the journal for good. Throwing archive_failed + // there retained the worker forever on evidence that could never arrive, and `worker-abandon` + // was the only way out of a release the coordinator had every right to complete. + installHost({ + lease: { + runtimeKind: 'native', + claimStatus: 'released', + deathEvidence: { kind: 'exit-observed', detail: 'surface released', observedAt: 1 }, + runtimeFence: 2 + } + }) + const identity = registerIdentity() + const resource = { + id: 'resource-2', + terminal_handle: HANDLE, + host_scope: JSON.stringify(HOST_SCOPE), + archive_source: null, + archive_status: null, + ownership_state: 'owned', + release_state: 'requested' + } as unknown as WorkerTerminalResourceRow + let stored: { kind?: string; content?: string } = {} + const db = { + getWorkerDispatch: () => ({ + agent_terminal_handle: HANDLE, + created_at: '2026-09-05 00:00:00' + }), + getDispatchContextById: () => null, + isDispatchProcessCurrent: (args: { paneKey: string; processIncarnation: string }) => + args.paneKey === identity.paneKey && + args.processIncarnation === identity.processIncarnation, + workerTerminalResourceHasIdentityConflict: () => false, + // No archive yet: the capture is what release has to get past. + getWorkerTerminalArchive: () => undefined, + commitWorkerTerminalArchiveForRelease: (args: { kind?: string; content?: string }) => { + stored = args + return { ...resource, release_state: 'releasing' } + }, + settleWorkerTerminalRelease: () => ({ ...resource, release_state: 'released' }), + markWorkerTerminalReleaseUnknown: (_id: string, error: string) => ({ + ...resource, + release_state: 'unknown', + release_error: error + }) + } as unknown as OrchestrationDb + + await expect( + completeWorkerTerminalRelease({ + runtime: { + ensureStructuredAgentSessionHost: async () => {}, + notifyMessageArrived: vi.fn(), + forgetStructuredSessionMail: vi.fn(), + retireStructuredAgentSessionTabFromSnapshot: vi.fn() + } as unknown as OrcaRuntimeService, + db, + dispatchId: 'd2', + resource + }) + ).resolves.toMatchObject({ state: 'released', processAction: 'closed_agent_terminal' }) + expect(stored.kind).toBe('structured_journal') + expect(stored.content).toContain('could not be preserved') + }) +}) + +describe('structured worker discard retires the chat tab', () => { + it('prunes the tab a half-started worker published', async () => { + const { close } = installHost() + const runtime = new OrcaRuntimeService() + const internal = runtime as unknown as RuntimeInternals + internal.ensureStructuredAgentSessionHost = async () => undefined + let createdSessionId = '' + createSpy.mockImplementation(async (args: { envelope: { sessionId: string } }) => { + // The create is what publishes the background tab, and it publishes BEFORE the start can + // fail — which is exactly the tab the discard has to take back. + createdSessionId = args.envelope.sessionId + await runtime.publishStructuredAgentSessionTab({ + workspaceId: WORKTREE, + sessionId: createdSessionId, + agent: 'claude', + activate: false + }) + return { ok: false, refusal: { code: 'agent_session_operation_unknown', message: 'unknown' } } + }) + + await expect( + createStructuredWorkerSession({ + runtime, + worktreeId: WORKTREE, + agent: 'claude', + dispatchId: 'd_discard', + onJournalActivity: () => {} + }) + ).rejects.toThrow(/was refused/) + + expect(close).toHaveBeenCalledWith(createdSessionId) + expect(await structuredTabIds(runtime)).toEqual([]) + }) +}) diff --git a/src/main/runtime/rpc/methods/terminal-manifest-characterization.test.ts b/src/main/runtime/rpc/methods/terminal-manifest-characterization.test.ts index da255066a34..ccdcf5fb7b1 100644 --- a/src/main/runtime/rpc/methods/terminal-manifest-characterization.test.ts +++ b/src/main/runtime/rpc/methods/terminal-manifest-characterization.test.ts @@ -13,6 +13,7 @@ const METHOD_CASES: readonly (readonly [string, unknown, boolean])[] = [ ['terminal.resolvePane', { paneKey: 'pane' }, false], ['terminal.recoverPane', { paneKey: 'pane', worktreeId: 'worktree' }, false], ['terminal.show', { terminal: 'term' }, false], + ['terminal.resolveIdentity', { terminal: 'term' }, false], ['terminal.read', { terminal: 'term' }, false], ['terminal.inspectProcess', { terminal: 'term' }, false], ['terminal.isRunningAgent', { terminal: 'term' }, false], @@ -65,11 +66,11 @@ async function invoke(name: string, params: unknown, runtime: Partial { it('preserves all method names, order, streaming flags, and parseable minimum inputs', () => { - expect(TERMINAL_METHODS).toHaveLength(34) + expect(TERMINAL_METHODS).toHaveLength(35) expect(TERMINAL_METHODS.map((method) => [method.name, 'stream' in method])).toEqual( METHOD_CASES.map(([name, _params, stream]) => [name, stream]) ) - expect(new Set(TERMINAL_METHODS.map((method) => method.name)).size).toBe(34) + expect(new Set(TERMINAL_METHODS.map((method) => method.name)).size).toBe(35) for (const [name, params] of METHOD_CASES) { expect(() => schemaFor(name).parse(params), name).not.toThrow() } diff --git a/src/main/runtime/rpc/methods/terminal/terminal-query-methods.ts b/src/main/runtime/rpc/methods/terminal/terminal-query-methods.ts index bf7b4a5bd87..82edd55cd79 100644 --- a/src/main/runtime/rpc/methods/terminal/terminal-query-methods.ts +++ b/src/main/runtime/rpc/methods/terminal/terminal-query-methods.ts @@ -25,7 +25,10 @@ export const TERMINAL_QUERY_METHODS: RpcAnyMethod[] = [ name: 'terminal.resolveActive', params: TerminalResolveActive, handler: async (params, { runtime }) => ({ - handle: await runtime.resolveActiveTerminal(params.worktree) + handle: await runtime.resolveActiveTerminal( + params.worktree, + params.requireUnambiguous ? { requireUnambiguous: true } : {} + ) }) }), defineMethod({ @@ -53,6 +56,15 @@ export const TERMINAL_QUERY_METHODS: RpcAnyMethod[] = [ terminal: await runtime.showTerminal(params.terminal) }) }), + defineMethod({ + // Read-only identity probe. Deliberately NOT `terminal.show`: this one resolves a structured + // worker too, and must therefore never hand back anything that looks writable. + name: 'terminal.resolveIdentity', + params: TerminalHandle, + handler: async (params, { runtime }) => ({ + identity: runtime.resolveTerminalIdentity(params.terminal) + }) + }), defineMethod({ name: 'terminal.read', params: TerminalRead, diff --git a/src/main/runtime/rpc/methods/terminal/unary-schemas.ts b/src/main/runtime/rpc/methods/terminal/unary-schemas.ts index 2734de0af1e..89928128e69 100644 --- a/src/main/runtime/rpc/methods/terminal/unary-schemas.ts +++ b/src/main/runtime/rpc/methods/terminal/unary-schemas.ts @@ -38,7 +38,9 @@ export const TerminalListParams = z.object({ }) export const TerminalResolveActive = z.object({ - worktree: OptionalString + worktree: OptionalString, + /** Refuse instead of guessing when several leaves could be the caller's own terminal. */ + requireUnambiguous: z.boolean().optional() }) export const TerminalResolvePane = z.object({ diff --git a/src/main/runtime/runtime-agent-orchestration-projection.ts b/src/main/runtime/runtime-agent-orchestration-projection.ts index 1faca13dde1..db8dd462b9f 100644 --- a/src/main/runtime/runtime-agent-orchestration-projection.ts +++ b/src/main/runtime/runtime-agent-orchestration-projection.ts @@ -50,7 +50,11 @@ export class RuntimeAgentOrchestrationProjection { const handle = this.deps.issueLeafHandle(leaf) queriedHandles.add(handle) const paneKey = this.deps.makePaneKey(leaf) - const context = this.getForHandle(handle, db, evidenceByPaneKey.get(paneKey), batchAttention) + const context = this.getForHandle(handle, db, { + paneKey, + evidence: evidenceByPaneKey.get(paneKey), + deferAttention: batchAttention + }) if (context) { contexts[paneKey] = context } @@ -64,12 +68,11 @@ export class RuntimeAgentOrchestrationProjection { continue } queriedHandles.add(handle) - const context = this.getForHandle( - handle, - db, - evidenceByPaneKey.get(pty.paneKey), - batchAttention - ) + const context = this.getForHandle(handle, db, { + paneKey: pty.paneKey, + evidence: evidenceByPaneKey.get(pty.paneKey), + deferAttention: batchAttention + }) if (context) { contexts[pty.paneKey] = context } @@ -105,10 +108,16 @@ export class RuntimeAgentOrchestrationProjection { getForHandle( handle: string, db = this.deps.getDb(), - evidence?: FleetAgentStatusEvidence, - deferAttention = false + options: { + // Why: handles are minted per process; after a restart only the pane identity still names the dispatch. + paneKey?: string + evidence?: FleetAgentStatusEvidence + deferAttention?: boolean + } = {} ): AgentStatusOrchestrationContext | undefined { - const dispatch = db?.getActiveDispatchForTerminal?.(handle) ?? this.getRecent(handle, db) + const { paneKey, evidence, deferAttention = false } = options + const dispatch = + db?.getActiveDispatchForTerminal?.(handle, paneKey) ?? this.getRecent(handle, db) if (!dispatch) { return undefined } @@ -166,21 +175,42 @@ export class RuntimeAgentOrchestrationProjection { task.creator_dispatch_process_incarnation === task.created_by_process_incarnation && parsePaneKey(task.creator_dispatch_pane_key)?.leafId === storedCreatorPane?.leafId ) - const currentCreatorHandle = + // Why: durable Run membership is what makes this pane the child's creator; the live + // process-incarnation and handle checks below only decide mutation authority. + const creatorLineageInRun = Boolean( owningRun?.legacy === 0 && task?.created_by_run_generation === owningRun.consumer_generation && - task.created_by_process_incarnation === creatorAuthority?.processIncarnation && - sameCreatorPane && + creatorPaneKey && (paneRun ? paneRun.id === owningRun.id && paneRun.consumer_generation === task.created_by_run_generation : sameRunCreatorDispatch) + ) + const currentCreatorHandle = + creatorLineageInRun && + task?.created_by_process_incarnation === creatorAuthority?.processIncarnation && + sameCreatorPane ? (creatorPaneHandle ?? undefined) : undefined - const parentHandle = - currentCreatorHandle ?? - (coordinatorHandle && coordinatorHandle !== handle ? coordinatorHandle : undefined) - const parentPaneKey = parentHandle ? this.deps.getPaneKey(parentHandle) : undefined + const coordinator = this.resolveLivePane( + coordinatorHandle, + owningRun?.legacy === 0 ? owningRun.coordinator_pane_key : null + ) + const creator = currentCreatorHandle + ? { + handle: currentCreatorHandle, + paneKey: this.deps.getPaneKey(currentCreatorHandle) ?? undefined + } + : creatorLineageInRun + ? this.resolveLivePane(creatorPaneHandle, creatorPaneKey ?? null) + : undefined + const coordinatorIsSelf = + coordinator.handle === handle || + (paneKey !== undefined && + coordinator.paneKey !== undefined && + coordinator.paneKey === paneKey) + // Why: a creator whose pane is gone still has a coordinator to nest under. + const parent = creator?.handle ? creator : coordinatorIsSelf ? {} : coordinator const attention = !deferAttention && db && typeof db.getWorkerAttentionFacts === 'function' ? buildWorkerAttentionContext({ db, dispatch, task, evidence }) @@ -191,14 +221,40 @@ export class RuntimeAgentOrchestrationProjection { dispatchStatus: dispatch.status, ...(display.taskTitle ? { taskTitle: display.taskTitle } : {}), ...(display.displayName ? { displayName: display.displayName } : {}), - ...(parentHandle ? { parentTerminalHandle: parentHandle } : {}), - ...(parentPaneKey ? { parentPaneKey } : {}), - ...(coordinatorHandle ? { coordinatorHandle } : {}), + ...(parent.handle ? { parentTerminalHandle: parent.handle } : {}), + ...(parent.paneKey ? { parentPaneKey: parent.paneKey } : {}), + ...(coordinator.handle ? { coordinatorHandle: coordinator.handle } : {}), ...(orchestrationRunId ? { orchestrationRunId } : {}), ...(attention ? { attention } : {}) } } + /** + * Resolves a stored (handle, pane key) pair to what this process can address now. A handle + * this process never minted is stale and must not reach the renderer; the pane key is the + * remint-stable identity, so it is re-resolved to the live pane and published even when no + * pane is live yet, so the row nests again as soon as that pane is restored. + */ + private resolveLivePane( + storedHandle: string | null | undefined, + storedPaneKey: string | null + ): { handle?: string; paneKey?: string } { + if (storedHandle && this.deps.getWorktreeId(storedHandle) !== null) { + return { + handle: storedHandle, + paneKey: this.deps.getPaneKey(storedHandle) ?? storedPaneKey ?? undefined + } + } + if (!storedPaneKey) { + return {} + } + const liveHandle = this.deps.getHandleForPaneKey(storedPaneKey) + if (!liveHandle) { + return { paneKey: storedPaneKey } + } + return { handle: liveHandle, paneKey: this.deps.getPaneKey(liveHandle) ?? storedPaneKey } + } + private getRecent(handle: string, db: OrchestrationDb | null) { const dispatch = db?.getLatestDispatchForTerminal?.(handle) if ( diff --git a/src/main/runtime/runtime-browser-client-page-recovery.test.ts b/src/main/runtime/runtime-browser-client-page-recovery.test.ts index 4845493e1c0..133b02aaf66 100644 --- a/src/main/runtime/runtime-browser-client-page-recovery.test.ts +++ b/src/main/runtime/runtime-browser-client-page-recovery.test.ts @@ -43,6 +43,25 @@ describe('runtime browser client page recovery', () => { expect(notifyWorkspace).toHaveBeenCalledOnce() }) + it('does not apply an attach inventory to a replacement placed after capture', async () => { + const { authority, commands, notifyWorkspace, pages, placements } = harness() + const pagePlacementsAtAttach = new Map([['page-a', oldPlacement]]) + pages.replaceClientPagePlacement('page-a', oldPlacement, newPlacement) + placements.set('page-a', newPlacement) + await recoverUnavailableRuntimeBrowserClientPages({ + lease: lease([]), + authority, + pages, + notifyWorkspace, + pagePlacementsAtAttach + }) + expect(authority.getPlacement('page-a')).toEqual(newPlacement) + expect(pages.getPage('page-a')?.placement).toEqual(newPlacement) + expect(authority.createClientPage).not.toHaveBeenCalled() + expect(commands).toEqual([]) + expect(notifyWorkspace).not.toHaveBeenCalled() + }) + it('retains an exact active generation without commands or metadata churn', async () => { const { authority, commands, notifyWorkspace, pages } = harness() diff --git a/src/main/runtime/runtime-browser-client-page-recovery.ts b/src/main/runtime/runtime-browser-client-page-recovery.ts index 8390db8686a..ad0f062f03e 100644 --- a/src/main/runtime/runtime-browser-client-page-recovery.ts +++ b/src/main/runtime/runtime-browser-client-page-recovery.ts @@ -45,6 +45,8 @@ export async function recoverUnavailableRuntimeBrowserClientPages(options: { } authority: RecoveryAuthority pages: RuntimeBrowserPageRegistry + /** Placements as of the attach inventory. Omitting it recovers against unfenced live state. */ + pagePlacementsAtAttach?: ReadonlyMap notifyWorkspace(workspaceId: string): void /** Drops a page whose placement recovery destroyed without replacing it. */ releaseUnrecoverablePage?: (page: RuntimeBrowserClientPage) => void @@ -77,6 +79,7 @@ export async function recoverUnavailableRuntimeBrowserClientPages(options: { .listPages() .filter( (page) => + isPlacedAsObservedAtAttach(page, options.pagePlacementsAtAttach) && !options.adoptedPageIds?.has(page.browserPageId) && isRecoverableByLease(page, options.lease) && !isActiveExactPage(page, inventoryByPageId.get(page.browserPageId), options.lease) @@ -101,6 +104,23 @@ export async function recoverUnavailableRuntimeBrowserClientPages(options: { ) } +/** + * Whether the attach inventory can still speak for this page. + * + * Readiness is published before recovery runs, so the client can place a page the inventory predates + * -- and absence from the inventory means "recreate". Those are left to the attach that can see them. + */ +function isPlacedAsObservedAtAttach( + page: RuntimeBrowserClientPage, + pagePlacementsAtAttach: ReadonlyMap | undefined +): boolean { + if (!pagePlacementsAtAttach) { + return true + } + const observed = pagePlacementsAtAttach.get(page.browserPageId) + return observed !== undefined && sameRuntimeBrowserPlacement(observed, page.placement) +} + /** * Whether this lease is the one allowed to take a page back. * diff --git a/src/main/runtime/runtime-client-settings.ts b/src/main/runtime/runtime-client-settings.ts index b9e6959c8da..fc80c924156 100644 --- a/src/main/runtime/runtime-client-settings.ts +++ b/src/main/runtime/runtime-client-settings.ts @@ -10,6 +10,8 @@ import { } from '../../shared/terminal-quick-commands' import { haveSameDisabledTuiAgents } from '../../shared/tui-agent-selection' import type { GlobalSettings } from '../../shared/global-settings-types' +import { applyNativeChatSessionOptionSettingsMutation } from '../../shared/native-chat-session-option-defaults' +import type { NativeChatSessionOptionSettingsMutation } from '../../shared/native-chat-session-options' import { getHostDisplayLabelOverrides } from '../../shared/host-setting-overrides' import type { ExecutionHostId } from '../../shared/execution-host' import type { TerminalQuickCommand } from '../../shared/terminal-quick-command-types' @@ -32,6 +34,8 @@ export type RuntimeClientSettings = Pick< | 'defaultLinearTeamSelection' | 'githubProjects' | 'experimentalNewWorktreeCardStyle' + | 'experimentalNativeChat' + | 'openAgentTabsInChatByDefault' | 'experimentalStructuredNativeChat' | 'compactWorktreeCards' | 'minimaxGroupId' @@ -98,6 +102,10 @@ export class RuntimeClientSettingsController { defaultLinearTeamSelection: settings.defaultLinearTeamSelection ?? null, githubProjects: settings.githubProjects, experimentalNewWorktreeCardStyle: settings.experimentalNewWorktreeCardStyle === true, + // The three that decide whether a new agent tab -- and so an orchestration worker -- is a + // structured chat session rather than a terminal agent. + experimentalNativeChat: settings.experimentalNativeChat === true, + openAgentTabsInChatByDefault: settings.openAgentTabsInChatByDefault === true, experimentalStructuredNativeChat: settings.experimentalStructuredNativeChat === true, compactWorktreeCards: settings.compactWorktreeCards === true, minimaxGroupId: settings.minimaxGroupId ?? '', @@ -172,6 +180,19 @@ export class RuntimeClientSettingsController { return this.get() } + updateNativeChatSessionOptions(mutation: NativeChatSessionOptionSettingsMutation): void { + if (!this.store?.getSettings || !this.store.updateSettings) { + throw new Error('runtime_unavailable') + } + const next = applyNativeChatSessionOptionSettingsMutation( + this.store.getSettings().nativeChatSessionOptions, + mutation + ) + if (next) { + this.store.updateSettings({ nativeChatSessionOptions: next }, { notifyListeners: true }) + } + } + private reconcileManagedAgentHooks(): Promise { const generation = ++this.reconciliationGeneration const reconciliation = this.reconciliationTail.then(async () => { diff --git a/src/main/runtime/runtime-mobile-notification-controller.ts b/src/main/runtime/runtime-mobile-notification-controller.ts index a9c1d437f95..9b061a3690f 100644 --- a/src/main/runtime/runtime-mobile-notification-controller.ts +++ b/src/main/runtime/runtime-mobile-notification-controller.ts @@ -1,9 +1,16 @@ +import type { AgentStatusState } from '../../shared/agent-status-types' +import type { + MobilePushRegisterInput, + MobilePushRegisterResult +} from '../../shared/mobile-push-contract' import { MobileNotificationReplayBuffer } from './mobile-notification-replay' import { notifyRuntimeListeners } from './runtime-async-boundaries' import { getRuntimeDesktopSurface } from './runtime-desktop-surface' export type MobileNotificationDispatchEvent = { type: 'notification' + desktopAllowed?: boolean + emittedAt?: number source: 'agent-task-complete' | 'terminal-bell' | 'test' | 'plugin' title: string body: string @@ -11,6 +18,9 @@ export type MobileNotificationDispatchEvent = { notificationId?: string notificationSeq?: number notificationEpoch?: string + // Why: background push must tell "needs input" from "finished" without re-deriving + // it from the title. Optional and additive — old clients ignore it. + agentState?: AgentStatusState } export type MobileNotificationDismissEvent = { @@ -24,9 +34,33 @@ export type MobileNotificationEvent = | MobileNotificationDispatchEvent | MobileNotificationDismissEvent +/** The desktop push service, once it exists; absent on hosts that never started one. */ +export type MobilePushRegistrar = { + register(input: MobilePushRegisterInput): Promise + unregister(deviceId: string): Promise<{ unregistered: boolean }> +} + export class RuntimeMobileNotificationController { private readonly listeners = new Set<(event: MobileNotificationEvent) => void>() private readonly replay = new MobileNotificationReplayBuffer() + private pushRegistrar: MobilePushRegistrar | null = null + + setPushRegistrar(registrar: MobilePushRegistrar | null): void { + this.pushRegistrar = registrar + } + + async registerPushDevice(input: MobilePushRegisterInput): Promise { + return ( + (await this.pushRegistrar?.register(input)) ?? { + registered: false, + reason: 'gateway_unreachable' + } + ) + } + + async unregisterPushDevice(deviceId: string): Promise<{ unregistered: boolean }> { + return (await this.pushRegistrar?.unregister(deviceId)) ?? { unregistered: false } + } onDispatched(listener: (event: MobileNotificationEvent) => void): () => void { this.listeners.add(listener) diff --git a/src/main/runtime/runtime-owned-terminal-publication-lineage.test.ts b/src/main/runtime/runtime-owned-terminal-publication-lineage.test.ts new file mode 100644 index 00000000000..2df5abf00da --- /dev/null +++ b/src/main/runtime/runtime-owned-terminal-publication-lineage.test.ts @@ -0,0 +1,57 @@ +import { expect, it, vi } from 'vitest' +import type { RuntimeMobileSessionTabsResult } from '../../shared/runtime-types' + +// Fragments stay side-effect ordered: mocks, then lifecycle, then fixtures. +const { OrcaRuntimeService } = await import('./orca-runtime-test-mocks.spec') +await import('./orca-runtime-test-lifecycle.spec') +const { store, TEST_WORKTREE_ID } = await import('./orca-runtime-test-fixtures.spec') + +it.each(['renderer:active-generation', 'headless:active-generation'])( + 'keeps %s live when runtime-owned creation supplements its inventory', + async (publicationEpoch) => { + const runtime = new OrcaRuntimeService(store) + runtime.setPtyController({ + spawn: vi.fn().mockResolvedValue({ id: 'pty-runtime-fallback' }), + write: () => true, + kill: () => true, + getForegroundProcess: async () => null + }) + runtime.syncWindowGraph(0, { + tabs: [], + leaves: [], + mobileSessionTabs: [ + { + worktree: TEST_WORKTREE_ID, + publicationEpoch, + snapshotVersion: 7, + activeGroupId: null, + activeTabId: null, + activeTabType: null, + tabs: [] + } + ] + }) + const events: RuntimeMobileSessionTabsResult[] = [] + const unsubscribe = runtime.onMobileSessionTabsChanged( + (snapshot) => events.push(snapshot), + 'paired-client' + ) + try { + const created = await runtime.createMobileSessionTerminal(`id:${TEST_WORKTREE_ID}`, { + activate: false, + select: false, + navigation: 'caller', + clientNavigationId: 'paired-client' + }) + expect(created.tab.status).toBe('ready') + expect(created.publicationEpoch).toBe(publicationEpoch) + expect(created.snapshotVersion).toBeGreaterThan(7) + expect(events.at(-1)).toMatchObject({ + publicationEpoch: `${publicationEpoch}:client-navigation`, + tabs: [expect.objectContaining({ id: created.tab.id, status: 'ready' })] + }) + } finally { + unsubscribe() + } + } +) diff --git a/src/main/runtime/runtime-rpc-mobile-native-chat-settings.test.ts b/src/main/runtime/runtime-rpc-mobile-native-chat-settings.test.ts new file mode 100644 index 00000000000..d295e1f922f --- /dev/null +++ b/src/main/runtime/runtime-rpc-mobile-native-chat-settings.test.ts @@ -0,0 +1,8 @@ +import { describe, expect, it } from 'vitest' +import { MOBILE_RPC_METHOD_ALLOWLIST } from './runtime-rpc/runtime-rpc-mobile-method-allowlist' + +describe('mobile native-chat settings RPC', () => { + it('allows a paired phone to persist a structured option pick', () => { + expect(MOBILE_RPC_METHOD_ALLOWLIST.has('settings.mutateNativeChatSessionOptions')).toBe(true) + }) +}) diff --git a/src/main/runtime/runtime-rpc/runtime-rpc-mobile-method-allowlist.ts b/src/main/runtime/runtime-rpc/runtime-rpc-mobile-method-allowlist.ts index 05666add0bc..4e49c658960 100644 --- a/src/main/runtime/runtime-rpc/runtime-rpc-mobile-method-allowlist.ts +++ b/src/main/runtime/runtime-rpc/runtime-rpc-mobile-method-allowlist.ts @@ -172,7 +172,9 @@ export const MOBILE_RPC_METHOD_ALLOWLIST = new Set([ 'markdown.readTab', 'markdown.saveTab', 'notifications.getMissedSince', + 'notifications.registerPush', 'notifications.subscribe', + 'notifications.unregisterPush', 'notifications.unsubscribe', 'pairing.getEndpoints', 'pairing.provisionRelay', @@ -214,6 +216,7 @@ export const MOBILE_RPC_METHOD_ALLOWLIST = new Set([ 'agentSession.setOption', 'agentSession.handoffStatus', 'agentSession.options', + 'agentSession.commands', 'agentSession.history', 'agentSession.subscribe', 'agentSession.unsubscribe', @@ -224,6 +227,7 @@ export const MOBILE_RPC_METHOD_ALLOWLIST = new Set([ 'nativeChat.unsubscribe', 'settings.get', 'settings.getTerminalQuickCommands', + 'settings.mutateNativeChatSessionOptions', 'settings.update', 'settings.updateTerminalQuickCommands', 'ssh.connect', diff --git a/src/main/runtime/runtime-rpc/runtime-rpc-pairing.ts b/src/main/runtime/runtime-rpc/runtime-rpc-pairing.ts index 592131779eb..7d8bba9f958 100644 --- a/src/main/runtime/runtime-rpc/runtime-rpc-pairing.ts +++ b/src/main/runtime/runtime-rpc/runtime-rpc-pairing.ts @@ -6,6 +6,7 @@ import type { RelayRevokeOutbox, RelayRevokeOutboxItem } from '../relay/relay-revoke-outbox' +import type { PushUnregisterOutbox } from '../push/push-unregister-outbox' import { encodePairingOffer, PAIRING_OFFER_VERSION } from '../../../shared/pairing' import type { RuntimePairingReach } from '../../../shared/runtime-pairing-reach' import { resolveAdvertisedPairingEndpoint } from '../pairing-endpoint' @@ -20,6 +21,8 @@ import { } from './runtime-rpc-pairing-types' export class RuntimeRpcPairing extends RuntimeRpcNetworkExposure { + private onPushUnregisterQueued?: () => void + getDeviceRegistry(): DeviceRegistry | null { return this.deviceRegistry } @@ -44,6 +47,10 @@ export class RuntimeRpcPairing extends RuntimeRpcNetworkExposure { return this.relayRevokeOutbox } + getPushUnregisterOutbox(): PushUnregisterOutbox { + return this.pushUnregisterOutbox + } + setMobileRelayBinding(deviceId: string, binding: RelayDeviceBinding): boolean { const current = this.deviceRegistry?.getDevice(deviceId) if ( @@ -88,6 +95,9 @@ export class RuntimeRpcPairing extends RuntimeRpcNetworkExposure { return false } } + // Why: unpairing must delete the phone's push token at the gateway too, and the + // registration id is only readable while the device row still exists. + this.queuePushUnregister(deviceId, device.pushRegistration?.registrationId) if (!this.deviceRegistry?.removeDevice(deviceId)) { return false } @@ -182,6 +192,23 @@ export class RuntimeRpcPairing extends RuntimeRpcNetworkExposure { } } + /** Best-effort: a failed enqueue must never block the revoke the user asked for. */ + protected queuePushUnregister(deviceId: string, registrationId: string | undefined): void { + if (!registrationId) { + return + } + try { + this.pushUnregisterOutbox.enqueue({ registrationId, deviceId }) + this.onPushUnregisterQueued?.() + } catch (error) { + console.error('[runtime] Failed to persist a push token cleanup:', error) + } + } + + setOnPushUnregisterQueued(callback: (() => void) | null): void { + this.onPushUnregisterQueued = callback ?? undefined + } + protected queueOrRetainRelayDeviceRevoke(deviceId: string, binding: RelayDeviceBinding): void { if (this.queueRelayDeviceRevoke(binding)) { return diff --git a/src/main/runtime/runtime-rpc/runtime-rpc-state.ts b/src/main/runtime/runtime-rpc/runtime-rpc-state.ts index ca9ab173feb..dc54275dcde 100644 --- a/src/main/runtime/runtime-rpc/runtime-rpc-state.ts +++ b/src/main/runtime/runtime-rpc/runtime-rpc-state.ts @@ -10,6 +10,7 @@ import type { E2EEKeypair } from '../e2ee-keypair' import type { UnpairedDeviceAuthThrottle } from '../rpc/unpaired-device-auth-throttle' import type { MobileSocketWiring } from '../rpc/mobile-socket-wiring' import { RelayRevokeOutbox } from '../relay/relay-revoke-outbox' +import { PushUnregisterOutbox } from '../push/push-unregister-outbox' import { RuntimeBinaryMessageRouter } from '../runtime-binary-message-router' import type { RuntimeMetadataOwnershipWatch } from '../runtime-metadata-ownership-watch' import { RUNTIME_METADATA_OWNERSHIP_POLL_MS } from '../runtime-metadata-ownership-watch' @@ -56,6 +57,7 @@ export class RuntimeRpcState { protected readonly browserHostLongPollCapPerDevice: number protected readonly specializedLongPollCap: number protected readonly relayRevokeOutbox: RelayRevokeOutbox + protected readonly pushUnregisterOutbox: PushUnregisterOutbox protected deviceRegistry: DeviceRegistry | null = null protected e2eeKeypair: E2EEKeypair | null = null protected pairingInitializationFailure: PairingOfferUnavailable | null = null @@ -129,5 +131,6 @@ export class RuntimeRpcState { this.browserHostLongPollCapPerDevice = Math.max(1, Math.floor(this.browserHostLongPollCap / 2)) this.specializedLongPollCap = Math.max(1, Math.floor(longPollCap * SPECIALIZED_LONG_POLL_SHARE)) this.relayRevokeOutbox = new RelayRevokeOutbox(userDataPath) + this.pushUnregisterOutbox = new PushUnregisterOutbox(userDataPath) } } diff --git a/src/main/runtime/runtime-service-command-surface.ts b/src/main/runtime/runtime-service-command-surface.ts index 19545cc76e6..23d5b9e0686 100644 --- a/src/main/runtime/runtime-service-command-surface.ts +++ b/src/main/runtime/runtime-service-command-surface.ts @@ -30,6 +30,9 @@ export type RuntimeServiceCommandSurface = { getMobileNotificationEpoch: RuntimeMobileNotificationController['getEpoch'] dismissMobileNotification: RuntimeMobileNotificationController['dismiss'] dispatchPluginNotification: RuntimeMobileNotificationController['dispatchPlugin'] + setMobilePushRegistrar: RuntimeMobileNotificationController['setPushRegistrar'] + registerMobilePushDevice: RuntimeMobileNotificationController['registerPushDevice'] + unregisterMobilePushDevice: RuntimeMobileNotificationController['unregisterPushDevice'] setAccountServices: RuntimeAccountController['setServices'] setCommitMessageAgentEnvironmentResolvers: RuntimeAccountController['setCommitMessageAgentEnvironment'] getCommitMessageAgentEnvironmentResolvers: RuntimeAccountController['getCommitMessageAgentEnvironment'] @@ -110,6 +113,9 @@ export function installRuntimeServiceCommandSurface( getMobileNotificationEpoch: notifications.getEpoch.bind(notifications), dismissMobileNotification: notifications.dismiss.bind(notifications), dispatchPluginNotification: notifications.dispatchPlugin.bind(notifications), + setMobilePushRegistrar: notifications.setPushRegistrar.bind(notifications), + registerMobilePushDevice: notifications.registerPushDevice.bind(notifications), + unregisterMobilePushDevice: notifications.unregisterPushDevice.bind(notifications), setAccountServices: accounts.setServices.bind(accounts), setCommitMessageAgentEnvironmentResolvers: accounts.setCommitMessageAgentEnvironment.bind(accounts), diff --git a/src/main/runtime/runtime-store-contract.ts b/src/main/runtime/runtime-store-contract.ts index 6b9858bda0c..854d52bd0ba 100644 --- a/src/main/runtime/runtime-store-contract.ts +++ b/src/main/runtime/runtime-store-contract.ts @@ -87,6 +87,8 @@ export type RuntimeStore = { terminalWindowsShell?: GlobalSettings['terminalWindowsShell'] floatingTerminalEnabled?: GlobalSettings['floatingTerminalEnabled'] agentStatusHooksEnabled?: GlobalSettings['agentStatusHooksEnabled'] + experimentalNativeChat?: GlobalSettings['experimentalNativeChat'] + openAgentTabsInChatByDefault?: GlobalSettings['openAgentTabsInChatByDefault'] experimentalStructuredNativeChat?: GlobalSettings['experimentalStructuredNativeChat'] defaultTaskSource?: GlobalSettings['defaultTaskSource'] defaultTaskViewPreset?: GlobalSettings['defaultTaskViewPreset'] @@ -115,6 +117,7 @@ export type RuntimeStore = { worktreeVisibilityDefaults?: GlobalSettings['worktreeVisibilityDefaults'] hostSettingOverrides?: GlobalSettings['hostSettingOverrides'] agentSkillSharingEnabled?: GlobalSettings['agentSkillSharingEnabled'] + nativeChatSessionOptions?: GlobalSettings['nativeChatSessionOptions'] } // Why: narrow to `unknown` return so test mocks can return void without // a cast. The runtime never reads the return value — the persisted value diff --git a/src/main/runtime/runtime-terminal-agent-presence.ts b/src/main/runtime/runtime-terminal-agent-presence.ts index e87520fccc8..aad71c1a08a 100644 --- a/src/main/runtime/runtime-terminal-agent-presence.ts +++ b/src/main/runtime/runtime-terminal-agent-presence.ts @@ -20,6 +20,8 @@ const WRAPPER_RETRY_INTERVAL_MS = 150 const WRAPPER_RETRY_TIMEOUT_MS = 6_500 type RuntimeTerminalAgentPresenceDependencies = { + /** A structured agent session of this runtime; it has no pane, so no PTY probe can see it. */ + isLiveStructuredAgent?(handle: string): boolean getLivePty(handle: string): RuntimePtyWorktreeRecord | null getLiveLeaf(handle: string): RuntimeLeafRecord getPrimaryLeaf(ptyId: string): RuntimeLeafRecord | null @@ -41,6 +43,13 @@ export class RuntimeTerminalAgentPresence { handle: string, options: RuntimeTerminalAgentPresenceOptions = {} ): Promise { + // Before every PTY probe below, because none of them can answer for a session that has no + // pane: `getLiveLeaf` threw, the catch turned that into `false`, and a coordinator running + // `dispatch --inject` concluded its structured worker was a bare shell — `no_agent_detected`. + // A structured session IS the agent; there is no foreground process to recognise. + if (this.deps.isLiveStructuredAgent?.(handle)) { + return true + } try { const pty = this.deps.getLivePty(handle) if (pty) { diff --git a/src/main/runtime/structured-agent-session-close.ts b/src/main/runtime/structured-agent-session-close.ts new file mode 100644 index 00000000000..756dbdeaef4 --- /dev/null +++ b/src/main/runtime/structured-agent-session-close.ts @@ -0,0 +1,81 @@ +/** + * Closing a structured agent session's provider child, and proving it went. + * + * Extracted from `stopStructuredWorker` so that orchestration settlement and worktree teardown + * close a session the SAME way rather than one of them inventing a shorter version. Everything + * dispatch-shaped — dropping the hold, the redrive subscription and the parked mail — stays with + * the caller that has a dispatch; this is only the child. + * + * `host.close` returns void and keeps a failed close indexed for retry, so the only settlement + * evidence is the observation AFTER it: a session the host no longer holds and whose lease is no + * longer live is proven gone. Anything else is retained rather than settled. + */ + +import { getStructuredAgentSessionHost } from '../native-chat/agent-session-wire/structured-agent-session-registry' +import type { OrcaRuntimeService } from './orca-runtime' +import { retireSettledStructuredWorkerTab } from './structured-agent-session-tab-retirement' +import { observeStructuredWorker } from './structured-worker-authority' + +export type StructuredAgentSessionCloseOutcome = { + stopped: boolean + /** Whether a close was actually issued; a receipt must not claim one that never happened. */ + closeAttempted: boolean + reason?: string +} + +export type StructuredAgentSessionCloseOptions = { + runtime?: Pick< + OrcaRuntimeService, + 'forgetStructuredSessionMail' | 'retireStructuredAgentSessionTabFromSnapshot' + > + /** + * Runs after the close is issued and BEFORE the proof is read. + * + * Not after: an unsettled close returns early, so a dispatch that released its hold there would + * keep the child un-evictable for the life of the app. Every settlement has to reach it. + */ + afterClose?: () => void +} + +export async function closeStructuredAgentSessionChild( + sessionId: string, + options: StructuredAgentSessionCloseOptions = {} +): Promise { + const host = getStructuredAgentSessionHost() + if (!host) { + // Nothing was reached, so nothing was acted on; the receipt must not claim a close. + return { + stopped: false, + closeAttempted: false, + reason: 'The structured agent-session host is not installed; no session was closed.' + } + } + // Set only once the close is actually issued: `setSessionTabVisibility` throwing first leaves a + // running child, and a receipt that still said `closed_agent_terminal` for it would be the + // close-that-never-happened this flag exists to rule out. + let closeAttempted = false + try { + await host.setSessionTabVisibility?.(sessionId, false) + closeAttempted = true + await host.close(sessionId) + } catch (error) { + return { + stopped: false, + closeAttempted, + reason: error instanceof Error ? error.message : String(error) + } + } + options.afterClose?.() + const observation = observeStructuredWorker({ sessionId }) + if (observation.status !== 'exited') { + return { + stopped: false, + closeAttempted: true, + reason: observation.reason ?? 'The structured session is still attached after close.' + } + } + // Only past the proof, and structurally unable to throw: the session's chat tab is retired from + // the live snapshot, which `setSessionTabVisibility(false)` above does not do. + retireSettledStructuredWorkerTab(sessionId, options.runtime) + return { stopped: true, closeAttempted: true } +} diff --git a/src/main/runtime/structured-agent-session-runtime.ts b/src/main/runtime/structured-agent-session-runtime.ts index d670c16f47d..d9b3e59186a 100644 --- a/src/main/runtime/structured-agent-session-runtime.ts +++ b/src/main/runtime/structured-agent-session-runtime.ts @@ -16,7 +16,10 @@ import { type CodexStructuredSessionAdapterDeps } from '../codex/codex-structured-session-adapter' import type { ClaudeStructuredSessionAdapterDeps } from '../claude/claude-structured-session-adapter' -import { StructuredAgentSessionHost } from '../native-chat/agent-session-wire/structured-agent-session-host' +import { + StructuredAgentSessionHost, + type StructuredAgentSessionHostDeps +} from '../native-chat/agent-session-wire/structured-agent-session-host' import { StructuredAgentSessionAdapterRouter } from '../native-chat/agent-session-wire/structured-agent-session-adapter-router' import type { StructuredAgentSessionHandoffTransport } from '../native-chat/agent-session-wire/structured-agent-session-handoff-types' import { setStructuredAgentSessionHost } from '../native-chat/agent-session-wire/structured-agent-session-registry' @@ -76,6 +79,9 @@ export type StructuredAgentSessionRuntimeDeps = { resolveEnvironment?: () => Promise resolveCodexOverrides?: () => NodeJS.ProcessEnv onError?: (input: { scope: string; error: unknown }) => void + /** Every structured-session status projection, for host-side reactions such as the first-work + * workspace rename that CLI agents get from their hooks. */ + onSessionStatusChanged?: StructuredAgentSessionHostDeps['onSessionStatusChanged'] handoffTransport?: StructuredAgentSessionHandoffTransport reapOrphanChildren?: typeof stopOrphanAgentSessionChildren } @@ -260,6 +266,14 @@ async function install(deps: StructuredAgentSessionRuntimeDeps): Promise host?.publishBackgroundTaskState(sessionId, state), + onDispatchSettledLate: (settlement) => { + void host?.settleLateDispatch(settlement).catch((error) => + deps.onError?.({ + scope: `structured-agent-session-late-settlement:${settlement.sessionId}`, + error + }) + ) + }, ...(deps.openClaudeConnection ? { openClaudeConnection: deps.openClaudeConnection } : {}), ...(deps.readProcessStartTime ? { readProcessStartTime: deps.readProcessStartTime } : {}) }) @@ -281,6 +295,9 @@ async function install(deps: StructuredAgentSessionRuntimeDeps): Promise deps.onError?.({ scope: `structured-agent-session-journal:${sessionId}`, error }), + ...(deps.onSessionStatusChanged + ? { onSessionStatusChanged: deps.onSessionStatusChanged } + : {}), persistTuiProviderHandle: async ({ sessionId, link, now }) => { await store.transitionHandoff(sessionId, (record) => recordAgentSessionProviderHandle({ record, fence: record.lease.runtimeFence, link, now }) diff --git a/src/main/runtime/structured-agent-session-tab-retirement.ts b/src/main/runtime/structured-agent-session-tab-retirement.ts new file mode 100644 index 00000000000..aee5ef6d719 --- /dev/null +++ b/src/main/runtime/structured-agent-session-tab-retirement.ts @@ -0,0 +1,79 @@ +/** + * Removing a structured agent session's chat tab from a live workspace tab snapshot. + * + * Extracted from `closeStructuredAgentSessionTab` so that user-initiated tab closes and + * orchestration settlements (stop / release / discard) retire the same tab the same way, rather + * than orchestration leaving a dead chat tab behind that re-attaches the session when opened. + */ + +import type { + RuntimeMobileSessionSnapshotTab, + RuntimeMobileSessionTabsSnapshot +} from '../../shared/runtime-types' +import { structuredAgentSessionTabId } from '../../shared/structured-agent-session-projection' + +/** The snapshot's tab for a structured session, matched by session id and by published tab id. */ +export function findStructuredAgentSessionTab( + snapshot: RuntimeMobileSessionTabsSnapshot, + sessionId: string +): RuntimeMobileSessionSnapshotTab | null { + const tabId = structuredAgentSessionTabId(sessionId) + return ( + snapshot.tabs.find( + (candidate) => + candidate.type === 'agent-session' && + (candidate.sessionId === sessionId || candidate.id === tabId) + ) ?? null + ) +} + +/** + * The snapshot with that session's tab pruned, or null when it holds no such tab. + * + * Pure: the caller owns storing and emitting, so nothing here can fail a settlement. + */ +export function retireStructuredAgentSessionTabFrom( + snapshot: RuntimeMobileSessionTabsSnapshot, + sessionId: string +): RuntimeMobileSessionTabsSnapshot | null { + const tab = findStructuredAgentSessionTab(snapshot, sessionId) + if (!tab) { + return null + } + const nextTabs = snapshot.tabs.filter((candidate) => candidate.id !== tab.id) + const active = nextTabs.find((candidate) => candidate.isActive) ?? nextTabs[0] ?? null + return { + ...snapshot, + snapshotVersion: snapshot.snapshotVersion + 1, + activeTabId: active?.id ?? null, + activeTabType: active?.type ?? null, + tabGroups: (snapshot.tabGroups ?? []).map((group) => ({ + ...group, + tabOrder: group.tabOrder.filter((id) => id !== tab.id), + activeTabId: group.activeTabId === tab.id ? null : group.activeTabId, + recentTabIds: group.recentTabIds?.filter((id) => id !== tab.id) + })), + tabs: nextTabs + } +} + +/** + * Retires a settled structured worker's chat tab, and cannot fail the settlement that called it. + * + * Every caller runs this AFTER it has already proven the session's close, so a snapshot problem + * here must never be able to turn a proven stop into `release_unknown`: the runtime method is + * called optionally (a runtime double or an older surface may not have it) and any throw is + * swallowed. It talks to no renderer, so the startup release reconciler can call it too. + */ +export function retireSettledStructuredWorkerTab( + sessionId: string, + runtime: + | { retireStructuredAgentSessionTabFromSnapshot?: (sessionId: string) => boolean } + | undefined +): void { + try { + runtime?.retireStructuredAgentSessionTabFromSnapshot?.(sessionId) + } catch (error) { + console.warn('[orchestration] structured worker tab retirement failed', sessionId, error) + } +} diff --git a/src/main/runtime/structured-claude-runtime-adapter.ts b/src/main/runtime/structured-claude-runtime-adapter.ts index 95151f1dae9..26c9922bb07 100644 --- a/src/main/runtime/structured-claude-runtime-adapter.ts +++ b/src/main/runtime/structured-claude-runtime-adapter.ts @@ -34,6 +34,7 @@ export type StructuredClaudeRuntimeAdapterDeps = { sessionId: string, state: AgentSessionBackgroundTaskState | null ) => void + onDispatchSettledLate?: ClaudeStructuredSessionAdapterDeps['onDispatchSettledLate'] } export function createStructuredClaudeRuntimeAdapter( @@ -100,6 +101,7 @@ export function createStructuredClaudeRuntimeAdapter( ...(deps.onBackgroundTasksChanged ? { onBackgroundTasksChanged: deps.onBackgroundTasksChanged } : {}), + ...(deps.onDispatchSettledLate ? { onDispatchSettledLate: deps.onDispatchSettledLate } : {}), ...(deps.openClaudeConnection ? { openConnection: deps.openClaudeConnection } : {}), ...(deps.readProcessStartTime ? { readProcessStartTime: deps.readProcessStartTime } : {}) }) diff --git a/src/main/runtime/structured-session-worktree-teardown.test.ts b/src/main/runtime/structured-session-worktree-teardown.test.ts new file mode 100644 index 00000000000..a9bdf6aa45c --- /dev/null +++ b/src/main/runtime/structured-session-worktree-teardown.test.ts @@ -0,0 +1,192 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentSessionRecord } from '../../shared/agent-session-record' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { killAllProcessesForWorktree } = await import('./worktree-teardown') +const { classifyWorktreeForceDeleteReason } = await import('../../shared/worktree/removal') +const { listLiveStructuredSessionsForWorktree } = + await import('./structured-session-worktree-teardown') + +const WORKTREE = 'repo_1::/tmp/wt-a' +const OTHER_WORKTREE = 'repo_1::/tmp/wt-b' + +function record(sessionId: string, workspaceId: string): AgentSessionRecord { + return { + sessionId, + provider: 'claude', + location: { executionHostId: 'local', wslDistro: null, workspaceId, workspaceKind: 'folder' }, + lease: { + sessionId, + runtimeKind: 'native', + claimStatus: 'live', + handoffStage: null, + runtimeFence: 1, + deathEvidence: null + } + } as unknown as AgentSessionRecord +} + +function installHost(options: { + records: AgentSessionRecord[] + /** Sessions the host still holds; a close removes one unless it is listed as stuck. */ + stuck?: Set +}): { closed: string[] } { + const held = new Set(options.records.map((entry) => entry.sessionId)) + const closed: string[] = [] + hostRef.current = { + deps: { store: { listRecords: () => options.records, getRecord: () => null } }, + hasSession: (sessionId: string) => held.has(sessionId), + setSessionTabVisibility: async () => {}, + close: async (sessionId: string) => { + closed.push(sessionId) + if (!options.stuck?.has(sessionId)) { + held.delete(sessionId) + const record = options.records.find((entry) => entry.sessionId === sessionId) + if (record) { + record.lease.claimStatus = 'released' + record.lease.deathEvidence = { kind: 'exit-observed', detail: 'closed', observedAt: 1 } + } + } + } + } + // `observeStructuredWorker` reads the record through the same host, so keep them consistent. + ;( + hostRef.current as { deps: { store: { getRecord: (id: string) => unknown } } } + ).deps.store.getRecord = (sessionId: string) => + options.records.find((entry) => entry.sessionId === sessionId) ?? null + return { closed } +} + +const localProvider = { + listProcesses: async () => [], + shutdown: async () => {} +} as never + +function destructiveDeps(extra: { allowUnverifiedStop?: boolean } = {}) { + return { + localProvider, + requirePhysicalStop: true, + includeProviderInventory: false as const, + includeLocalRegistry: false as const, + ...extra + } +} + +describe('worktree teardown and structured agent sessions', () => { + beforeEach(() => { + hostRef.current = null + }) + + it('finds sessions by workspace, and ignores a sibling worktree', () => { + installHost({ records: [record('s1', WORKTREE), record('s2', OTHER_WORKTREE)] }) + expect(listLiveStructuredSessionsForWorktree(WORKTREE)).toEqual([ + { sessionId: 's1', agent: 'claude' } + ]) + }) + + it('refuses a destructive removal rather than deleting the checkout under a live child', async () => { + // The defect this pins: all three PTY sweeps enumerate leaves, provider sessions and the local + // registry, and a structured session is on NONE of them. Every sweep answered zero, nothing + // errored, and removal proceeded — leaving the provider child running with its `cwd` deleted + // and the dispatch still reporting the worker live and exact. + installHost({ records: [record('s1', WORKTREE)] }) + await expect(killAllProcessesForWorktree(WORKTREE, destructiveDeps())).rejects.toThrow( + /1 running agent session/ + ) + }) + + it('names the force escape hatch in the refusal, like the unstopped-PTY gate', async () => { + installHost({ records: [record('s1', WORKTREE)] }) + await expect(killAllProcessesForWorktree(WORKTREE, destructiveDeps())).rejects.toThrow(/force/i) + }) + + it('classifies for the desktop Force Delete button, not just the CLI', async () => { + // The #11960 dead end, and the shape this file's own comments warn about: the desktop + // affordance comes ONLY from the classifier, and an ordinary delete already passes force:true + // for the dirty-file skip — so a refusal with no matcher shows raw CLI wording with no button. + installHost({ records: [record('s1', WORKTREE)] }) + const error = await killAllProcessesForWorktree(WORKTREE, destructiveDeps()).catch( + (thrown: Error) => thrown.message + ) + expect(classifyWorktreeForceDeleteReason(error as string, true)).toBe('running-agent-session') + // Nulled once the waiver is spent, exactly as `unstopped-pty` is, so the button does not + // reappear on a delete the user already forced. + expect(classifyWorktreeForceDeleteReason(error as string, true, true)).toBeNull() + }) + + it('keeps session ids out of a message users and agents read', async () => { + // A session id is one tab-id hop from the random pane key that gates a worker's mailbox, and + // this string reaches CLI output and a desktop toast. A count and the providers are what a + // user deciding whether to force actually needs. + installHost({ records: [record('s1', WORKTREE)] }) + const error = await killAllProcessesForWorktree(WORKTREE, destructiveDeps()).catch( + (thrown: Error) => thrown.message + ) + expect(error).not.toContain('s1') + expect(error).toContain('1 running agent session') + }) + + it('closes best-effort for a folder-workspace removal, which requires no stop proof', async () => { + // Those paths sweep and kill PTYs without `requirePhysicalStop`, so the structured sweep used + // to no-op there and left a live session bound to a workspace Orca was about to forget. They + // do not refuse: the root is shared so no checkout vanishes, and one of them is a never-throw + // forget that a refusal would wedge. + const host = installHost({ records: [record('s1', WORKTREE)] }) + await expect( + killAllProcessesForWorktree(WORKTREE, { + localProvider, + includeProviderInventory: false, + includeLocalRegistry: false, + closeStructuredSessions: true + }) + ).resolves.toMatchObject({ structuredStopped: 1 }) + expect(host.closed).toEqual(['s1']) + }) + + it('closes them under force instead of orphaning the child', async () => { + const host = installHost({ records: [record('s1', WORKTREE), record('s2', WORKTREE)] }) + const result = await killAllProcessesForWorktree( + WORKTREE, + destructiveDeps({ allowUnverifiedStop: true }) + ) + expect(host.closed).toEqual(['s1', 's2']) + expect(result.structuredStopped).toBe(2) + }) + + it('still removes under force when a close does not settle, and says so', async () => { + const warn = vi.spyOn(console, 'warn').mockImplementation(() => {}) + installHost({ records: [record('s1', WORKTREE)], stuck: new Set(['s1']) }) + const result = await killAllProcessesForWorktree( + WORKTREE, + destructiveDeps({ allowUnverifiedStop: true }) + ) + expect(result.structuredStopped).toBeUndefined() + expect(warn).toHaveBeenCalledWith(expect.stringContaining('still attached')) + warn.mockRestore() + }) + + it('leaves the best-effort reconciliation paths alone', async () => { + // Those callers repair state and delete nothing, so a refusal there would wedge a repair. + installHost({ records: [record('s1', WORKTREE)] }) + await expect( + killAllProcessesForWorktree(WORKTREE, { + localProvider, + includeProviderInventory: false, + includeLocalRegistry: false + }) + ).resolves.toMatchObject({ runtimeStopped: 0 }) + }) + + it('does not block removal when no structured host is installed', async () => { + // Not being able to look is not evidence a child is there, and reading the persisted store + // directly would force-install the host as a side effect of a teardown. + await expect(killAllProcessesForWorktree(WORKTREE, destructiveDeps())).resolves.toMatchObject({ + runtimeStopped: 0 + }) + }) +}) diff --git a/src/main/runtime/structured-session-worktree-teardown.ts b/src/main/runtime/structured-session-worktree-teardown.ts new file mode 100644 index 00000000000..226f785f039 --- /dev/null +++ b/src/main/runtime/structured-session-worktree-teardown.ts @@ -0,0 +1,107 @@ +/** + * The structured half of worktree teardown. + * + * `killAllProcessesForWorktree` sweeps three PTY surfaces — the renderer graph, the provider's + * session list, and the local pty-registry — and a structured agent session appears on NONE of + * them. It has no PTY, no leaf, and no provider session row. So every sweep counted zero, no error + * was raised, and removal deleted the checkout out from under a running provider child: the child + * kept running with its `cwd` gone, the durable record and chat tab survived to republish at the + * next launch pointing at a deleted worktree, and `worker-show` still reported the worker live. + * + * Membership is `location.workspaceId`, which every structured session carries — so this covers a + * plain chat session in the worktree as well as a dispatched worker. Liveness is + * `observeStructuredWorker`, the same `live` / `unverifiable` / `exited` vocabulary the rest of the + * structured surface uses; only a PROVEN live child is worth refusing a removal over. + */ + +import { getStructuredAgentSessionHost } from '../native-chat/agent-session-wire/structured-agent-session-registry' +import { observeStructuredWorker } from './structured-worker-authority' +import { closeStructuredAgentSessionChild } from './structured-agent-session-close' +import type { OrcaRuntimeService } from './orca-runtime' + +export type LiveStructuredSessionInWorkspace = { + sessionId: string + agent: 'claude' | 'codex' +} + +export type StructuredWorktreeSweepRuntime = Pick< + OrcaRuntimeService, + 'forgetStructuredSessionMail' | 'retireStructuredAgentSessionTabFromSnapshot' +> + +/** + * Structured sessions with a proven-live child in this worktree. + * + * An uninstalled host answers empty rather than throwing: no host in this generation means no + * provider child was started by this process, and the three PTY sweeps fall through the same way + * when their surface is unavailable. It is deliberately NOT read through the persisted store + * directly — that would force-install the host, which is itself a side effect on a teardown path. + */ +export function listLiveStructuredSessionsForWorktree( + worktreeId: string +): LiveStructuredSessionInWorkspace[] { + const host = getStructuredAgentSessionHost() + if (!host) { + return [] + } + let records: ReturnType + try { + records = host.deps.store.listRecords() + } catch { + return [] + } + return records + .filter( + (record) => + record.location.workspaceId === worktreeId && + observeStructuredWorker({ sessionId: record.sessionId }).status === 'live' + ) + .map((record) => ({ sessionId: record.sessionId, agent: record.provider })) +} + +/** + * Counts and providers, never session ids. + * + * A session id is one tab-id hop from the random pane key that gates a worker's mailbox, and this + * string reaches agent-readable CLI output and a desktop toast. The count and the providers are + * what a user deciding whether to force actually needs; the ids identify nothing they can act on. + */ +export function describeLiveStructuredSessions( + sessions: readonly LiveStructuredSessionInWorkspace[] +): string { + const noun = sessions.length === 1 ? 'agent session' : 'agent sessions' + const providers = [...new Set(sessions.map((session) => session.agent))].sort().join(', ') + return `${sessions.length} running ${noun} (${providers})` +} + +/** + * Closes every live structured session in the worktree, and reports what stayed. + * + * Force is the documented escape hatch, so it closes rather than orphaning: a child left running + * against a deleted `cwd` is the exact outcome this whole sweep exists to prevent. + */ +export async function closeStructuredSessionsForWorktree( + worktreeId: string, + runtime?: StructuredWorktreeSweepRuntime +): Promise<{ closed: number; unstopped: LiveStructuredSessionInWorkspace[] }> { + // No `afterClose` for a dispatched worker: `host.close` drops the holds, so nothing keeps a + // provider child un-evictable, but the dispatch's redrive subscription and registry entry do + // survive until it settles by another verb. That is a bounded leak, not a hazard — and passing + // one here would mean resolving a dispatch id per session on a teardown path that must stay + // inside the sweep deadline. + const sessions = listLiveStructuredSessionsForWorktree(worktreeId) + const unstopped: LiveStructuredSessionInWorkspace[] = [] + let closed = 0 + for (const session of sessions) { + const outcome = await closeStructuredAgentSessionChild( + session.sessionId, + runtime ? { runtime } : {} + ) + if (outcome.stopped) { + closed += 1 + } else { + unstopped.push(session) + } + } + return { closed, unstopped } +} diff --git a/src/main/runtime/structured-worker-agent-presence.test.ts b/src/main/runtime/structured-worker-agent-presence.test.ts new file mode 100644 index 00000000000..ca31e891cc9 --- /dev/null +++ b/src/main/runtime/structured-worker-agent-presence.test.ts @@ -0,0 +1,46 @@ +/** + * `isTerminalRunningAgent` for a worker that IS a structured agent session. + * + * This seam had no test at all: nothing in the repo referenced `isLiveStructuredAgent`, so the + * early return could be deleted and every suite stayed green. `dispatch --to --inject` + * depends on it — without it `getLiveLeaf` throws, the catch returns false, and a coordinator is + * told its worker is a bare shell (`no_agent_detected`). + */ + +import { describe, expect, it, vi } from 'vitest' +import { RuntimeTerminalAgentPresence } from './runtime-terminal-agent-presence' + +function presence(isLiveStructuredAgent: (handle: string) => boolean) { + const getLiveLeaf = vi.fn(() => { + // Exactly what the runtime does for a handle with no pane, and the reason the catch below + // used to swallow the question into `false`. + throw new Error('terminal_handle_stale') + }) + return { + getLiveLeaf, + presence: new RuntimeTerminalAgentPresence({ + isLiveStructuredAgent, + getLivePty: () => null, + getLiveLeaf: getLiveLeaf as never, + getPrimaryLeaf: () => null, + getTrackedPty: () => null, + getTabTitle: () => null, + getForegroundProcess: () => null + }) + } +} + +describe('agent presence for a structured worker', () => { + it('reports the session as running an agent without probing a pane', async () => { + const { presence: subject, getLiveLeaf } = presence(() => true) + await expect(subject.isRunning('structworker_1')).resolves.toBe(true) + // A structured session IS the agent; there is no foreground process to recognise, and the + // leaf probe would only throw. + expect(getLiveLeaf).not.toHaveBeenCalled() + }) + + it('still answers false for a handle that is not a live structured worker', async () => { + const { presence: subject } = presence(() => false) + await expect(subject.isRunning('term_gone')).resolves.toBe(false) + }) +}) diff --git a/src/main/runtime/structured-worker-authority.test.ts b/src/main/runtime/structured-worker-authority.test.ts new file mode 100644 index 00000000000..d65f50451e8 --- /dev/null +++ b/src/main/runtime/structured-worker-authority.test.ts @@ -0,0 +1,88 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { resolveStructuredWorkerIdentity, structuredWorkerAgent } = + await import('./structured-worker-authority') +const { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} = await import('./structured-worker-identity') + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + +function installRecordProvider(provider: 'claude' | 'codex' | null): void { + hostRef.current = { + deps: { store: { getRecord: () => (provider ? { provider } : null) } } + } +} + +/** The durable worker-terminal row is all a restarted runtime has; it carries no provider. */ +function durableRow(handle: string): { + terminal_handle: string + pane_key: string + process_incarnation: string + worktree_id: string + host_scope: string +} { + return { + terminal_handle: handle, + pane_key: mintStructuredWorkerPaneKey(SESSION_ID), + process_incarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktree_id: 'wt_1', + host_scope: JSON.stringify({ kind: 'local', hostId: 'local' }) + } +} + +function rehydratedIdentity(): NonNullable> { + const handle = mintStructuredWorkerHandle() + const row = durableRow(handle) + const identity = resolveStructuredWorkerIdentity(handle, { + getWorkerTerminalResourceByHandle: () => row + } as never) + if (!identity) { + throw new Error('the durable row should rehydrate') + } + return identity +} + +describe('structuredWorkerAgent', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + it('reads a rehydrated worker provider off the durable record', () => { + installRecordProvider('codex') + const identity = rehydratedIdentity() + expect(identity.agent).toBeNull() + // Defaulting here is what stamped a restarted Codex worker's frozen archive as Claude. + expect(structuredWorkerAgent(identity)).toBe('codex') + }) + + it('keeps the provider this process registered, without consulting the record', () => { + installRecordProvider('claude') + const handle = mintStructuredWorkerHandle() + const identity = structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'codex', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + expect(structuredWorkerAgent(identity)).toBe('codex') + }) + + it('falls back to claude only when no record can name the provider', () => { + installRecordProvider(null) + expect(structuredWorkerAgent(rehydratedIdentity())).toBe('claude') + }) +}) diff --git a/src/main/runtime/structured-worker-authority.ts b/src/main/runtime/structured-worker-authority.ts new file mode 100644 index 00000000000..2dbe9dabd0d --- /dev/null +++ b/src/main/runtime/structured-worker-authority.ts @@ -0,0 +1,133 @@ +/** + * Resolves a structured worker handle to the same authority facts a live PTY supplies. + * + * The registry holds the handle→session mapping for this process; the durable worker-terminal + * resource row is what survives a restart, so a miss falls back to rehydrating from it. The + * durable agent-session record is the liveness half: a session handed to a TUI owner, released, or + * pinned to another execution host is no longer this runtime's structured worker. + */ + +import type { AgentSessionRecord } from '../../shared/agent-session-record' +import type { RuntimeTerminalState } from '../../shared/runtime-types' +import { getStructuredAgentSessionHost } from '../native-chat/agent-session-wire/structured-agent-session-registry' +import type { OrchestrationDb } from './orchestration/db' +import { + isStructuredWorkerHandle, + structuredWorkerIdentities, + structuredWorkerRecordIsCurrent, + type StructuredWorkerIdentity +} from './structured-worker-identity' + +export type StructuredWorkerAuthority = { + identity: StructuredWorkerIdentity + record: AgentSessionRecord +} + +export function readStructuredAgentSessionRecord(sessionId: string): AgentSessionRecord | null { + try { + return getStructuredAgentSessionHost()?.deps.store.getRecord(sessionId) ?? null + } catch { + return null + } +} + +/** Registry entry for a handle, rehydrated from the durable row when this process restarted. */ +export function resolveStructuredWorkerIdentity( + handle: string, + db: OrchestrationDb | null | undefined +): StructuredWorkerIdentity | null { + if (!isStructuredWorkerHandle(handle)) { + return null + } + const known = structuredWorkerIdentities.get(handle) + if (known) { + return known + } + const row = db?.getWorkerTerminalResourceByHandle?.(handle) + return row ? structuredWorkerIdentities.rehydrate(row) : null +} + +/** Identity plus a record that still proves this runtime owns the session. */ +export function resolveStructuredWorkerAuthority( + handle: string, + db: OrchestrationDb | null | undefined +): StructuredWorkerAuthority | null { + const identity = resolveStructuredWorkerIdentity(handle, db) + if (!identity) { + return null + } + const record = readStructuredAgentSessionRecord(identity.sessionId) + return record && structuredWorkerRecordIsCurrent(record) ? { identity, record } : null +} + +/** + * Which provider this worker actually talks to. + * + * The registry carries it only for a session THIS process started; a rehydrated entry has null, + * because the durable worker-terminal row does not record a provider. The durable agent-session + * record does, and it is the only source that survives a restart — defaulting instead would + * relabel every restarted Codex worker as Claude, permanently, because the startup release + * reconciler stamps the frozen journal archive with whatever it is told here. + */ +export function structuredWorkerAgent(identity: StructuredWorkerIdentity): 'claude' | 'codex' { + return ( + identity.agent ?? readStructuredAgentSessionRecord(identity.sessionId)?.provider ?? 'claude' + ) +} + +export type StructuredWorkerObservation = { + status: 'live' | 'unverifiable' | 'exited' + reason?: string +} + +/** + * The observation as the terminal state every read result reports. + * + * `unverifiable` must never render as `running`: losing sight of the structured host is not + * evidence its child is alive, and the PTY sibling maps the same verdict to `unknown`. + */ +export function structuredWorkerTerminalState( + liveness: StructuredWorkerObservation['status'] +): RuntimeTerminalState { + return liveness === 'exited' ? 'exited' : liveness === 'live' ? 'running' : 'unknown' +} + +/** + * Only the session id is needed: the durable agent-session record is the authority, and it + * outlives both the in-memory identity registry and this process. Callers that hold nothing but a + * process incarnation therefore do not have to resolve a registry entry first — after `forget` + * there is none, and gating on one answers `unverifiable` forever. + */ +export function observeStructuredWorker( + identity: Pick +): StructuredWorkerObservation { + const host = getStructuredAgentSessionHost() + if (!host) { + // Reading the persisted record store here would force-install the host, which is itself a side + // effect; not being able to look is not evidence the child is gone. + return { + status: 'unverifiable', + reason: 'The structured agent-session host is not installed in this runtime generation.' + } + } + const record = host.deps.store.getRecord(identity.sessionId) + if (!record) { + return { status: 'unverifiable', reason: 'No durable record backs this structured session.' } + } + if (record.lease.claimStatus === 'released' && record.lease.deathEvidence) { + return { status: 'exited' } + } + if (record.lease.runtimeKind !== 'native') { + return { + status: 'unverifiable', + reason: 'The session lease is held by a terminal owner, not this structured host.' + } + } + if (host.hasSession(identity.sessionId) && record.lease.claimStatus === 'live') { + return { status: 'live' } + } + return { + status: 'unverifiable', + reason: 'The session has no attached provider child in this runtime generation.' + } +} diff --git a/src/main/runtime/structured-worker-child-identity-env.test.ts b/src/main/runtime/structured-worker-child-identity-env.test.ts new file mode 100644 index 00000000000..e966dd55824 --- /dev/null +++ b/src/main/runtime/structured-worker-child-identity-env.test.ts @@ -0,0 +1,146 @@ +import { join } from 'node:path' +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { installFakeAppEnvironment } from '../../../config/scripts/vitest-host-ports-setup' + +const shim = vi.hoisted(() => ({ ensureLinuxTerminalOrcaCliShimDir: vi.fn() })) +vi.mock('../cli/linux-terminal-orca-cli-shim', () => shim) + +import { structuredWorkerChildIdentityEnv } from './structured-worker-child-identity-env' +import { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerHostScope, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} from './structured-worker-identity' + +const SESSION_ID = 'f7a1c0de-1111-4222-8333-444455556666' +const USER_DATA = '/data/orca' +const RESOURCES = '/app/Resources' +const SHIM_DIR = join(USER_DATA, 'linux-orca-cli-shim') + +const platformDescriptor = Object.getOwnPropertyDescriptor(process, 'platform')! +const resourcesDescriptor = Object.getOwnPropertyDescriptor(process, 'resourcesPath') + +function pinPlatform(platform: NodeJS.Platform): void { + Object.defineProperty(process, 'platform', { configurable: true, value: platform }) +} + +function registerWorker(): string { + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + return handle +} + +beforeEach(() => { + shim.ensureLinuxTerminalOrcaCliShimDir.mockReset() + shim.ensureLinuxTerminalOrcaCliShimDir.mockReturnValue(SHIM_DIR) + Object.defineProperty(process, 'resourcesPath', { configurable: true, value: RESOURCES }) +}) + +afterEach(() => { + structuredWorkerIdentities.clear() + Object.defineProperty(process, 'platform', platformDescriptor) + if (resourcesDescriptor) { + Object.defineProperty(process, 'resourcesPath', resourcesDescriptor) + } else { + Reflect.deleteProperty(process, 'resourcesPath') + } +}) + +describe('structuredWorkerChildIdentityEnv', () => { + it('marks an ordinary chat session as having NO identity, and grants it nothing', () => { + // The marker names nothing — no handle, no pane key, no session id, no token — so it cannot be + // replayed or impersonated, and it does not reach the hook, agent-row or mobile-projection + // pipelines a pane key would. Its only job is to let the CLI REFUSE instead of guessing: this + // session has no pane, so every implicit-terminal guess resolved to a sibling, and a + // destructive `check` then consumed that sibling's mail. + pinPlatform('linux') + installFakeAppEnvironment({ isPackaged: () => true, getPath: () => USER_DATA }) + const childEnv = { PATH: '/usr/bin' } + const env = structuredWorkerChildIdentityEnv(SESSION_ID, childEnv) + expect(env).toEqual({ PATH: '/usr/bin', ORCA_STRUCTURED_SESSION: '1' }) + expect(env.ORCA_TERMINAL_HANDLE).toBeUndefined() + expect(env.ORCA_PANE_KEY).toBeUndefined() + expect(env.ORCA_CLI_COMMAND).toBeUndefined() + // Still no CLI reachability granted, so packaged builds keep today's exposure. + expect(childEnv.PATH).toBe('/usr/bin') + expect(shim.ensureLinuxTerminalOrcaCliShimDir).not.toHaveBeenCalled() + }) + + it('gives a packaged-Linux worker the bare-orca shim its ORCA_CLI_COMMAND assumes', () => { + // Without this the child's first `orca orchestration check` execs GNOME Orca — the CLI + // installs as `orca-ide` on Linux (stablyai/orca#7904) — and the dispatch hangs to timeout. + pinPlatform('linux') + installFakeAppEnvironment({ isPackaged: () => true, getPath: () => USER_DATA }) + const handle = registerWorker() + const env = structuredWorkerChildIdentityEnv(SESSION_ID, { PATH: '/usr/bin:/bin' }) + expect(env.ORCA_TERMINAL_HANDLE).toBe(handle) + expect(env.ORCA_CLI_COMMAND).toBe('orca') + expect(env.PATH).toBe(`${SHIM_DIR}:/usr/bin:/bin`) + }) + + it('gives a packaged-macOS worker the bundled CLI dir', () => { + pinPlatform('darwin') + installFakeAppEnvironment({ isPackaged: () => true, getPath: () => USER_DATA }) + registerWorker() + const env = structuredWorkerChildIdentityEnv(SESSION_ID, { PATH: '/usr/bin' }) + expect(env.PATH).toBe(`${join(RESOURCES, 'bin')}:/usr/bin`) + }) + + it('gives a packaged-Windows worker the bundled CLI dir under the env block spelling', () => { + pinPlatform('win32') + installFakeAppEnvironment({ isPackaged: () => true, getPath: () => USER_DATA }) + registerWorker() + const env = structuredWorkerChildIdentityEnv(SESSION_ID, { Path: 'C:\\Windows' }) + expect(env.Path).toBe(`${join(RESOURCES, 'bin')};C:\\Windows`) + expect(env.PATH).toBeUndefined() + }) + + it('gives an unpackaged worker the dev launcher dir', () => { + pinPlatform('darwin') + installFakeAppEnvironment({ isPackaged: () => false, getPath: () => USER_DATA }) + registerWorker() + const env = structuredWorkerChildIdentityEnv(SESSION_ID, { PATH: '/usr/bin' }) + expect(env.PATH).toBe(`${join(USER_DATA, 'cli', 'bin')}:/usr/bin`) + }) + + it('never puts a pane key in the child environment', () => { + // A pane key here flows into hook-emitted agent statuses and the attestation, agent-row and + // mobile-projection pipelines, all of which assume it names a live PTY leaf. + pinPlatform('linux') + installFakeAppEnvironment({ isPackaged: () => true, getPath: () => USER_DATA }) + registerWorker() + const env = structuredWorkerChildIdentityEnv(SESSION_ID, { PATH: '/usr/bin' }) + expect(env.ORCA_PANE_KEY).toBeUndefined() + expect(Object.keys(env).filter((key) => key.includes('PANE'))).toEqual([]) + }) + + it('never names the WSL-scoped launcher, because a structured worker cannot run in WSL', () => { + // `orca-ide` is the literal the PTY lane exports for WSL only. A structured session that + // resolves to a WSL distro is refused a host scope, so it never becomes a worker at all — + // which is why the bare-`orca` shim, not the literal, is the right fix on Linux. + expect( + structuredWorkerHostScope({ + executionHostId: 'local', + workspaceId: 'wt_1', + workspaceKind: 'git-worktree', + wslDistro: 'Ubuntu' + }) + ).toBeNull() + pinPlatform('linux') + installFakeAppEnvironment({ isPackaged: () => true, getPath: () => USER_DATA }) + registerWorker() + expect( + structuredWorkerChildIdentityEnv(SESSION_ID, { PATH: '/usr/bin' }).ORCA_CLI_COMMAND + ).not.toBe('orca-ide') + }) +}) diff --git a/src/main/runtime/structured-worker-child-identity-env.ts b/src/main/runtime/structured-worker-child-identity-env.ts new file mode 100644 index 00000000000..6cf9f46e910 --- /dev/null +++ b/src/main/runtime/structured-worker-child-identity-env.ts @@ -0,0 +1,74 @@ +/** + * The orchestration identity — and the CLI reachability — a structured worker's own child needs + * to speak for itself. + * + * Without `ORCA_TERMINAL_HANDLE` the worker's Bash tool has nothing to pass as `--from`, and + * `resolveOrchestrationTerminalHandle` falls back to a cwd lookup that returns whichever leaf in + * the worktree comes first. Two attacks follow from that: a bare `check` reads and consumes a + * SIBLING's dispatch mailbox, and a bare `send --type worker_done` can settle a sibling's + * context-only dispatch, a tier that has no capability token to reject on. + * + * `ORCA_CLI_COMMAND: 'orca'` is honest ONLY because of the PATH prepend below. Orca's Linux CLI + * installs as `orca-ide` so it never claims GNOME Orca's /usr/bin/orca (stablyai/orca#7904), and + * on packaged macOS/Windows the bundled launcher is reachable only from the app's own resources + * dir. A PTY worker gets that treatment from `buildPtyHostEnv`; a structured worker has no PTY, + * so it applies the SAME function here rather than a second, drifting copy of the rule. + * + * Deliberately NOT `ORCA_PANE_KEY`. Claude structured sessions run hooks, and a pane key in their + * environment starts flowing into hook-emitted agent-status payloads and the hook-attestation, + * agent-row and mobile-projection pipelines, every one of which assumes a pane key names a live + * PTY leaf. It would also open `selectExactWorkerProviderSession`, which is fail-closed today + * precisely because a structured session emits no hook agent status. The CLI needs none of it once + * the handle is present. + * + * A session that is not a dispatched worker gets ONE variable, `ORCA_STRUCTURED_SESSION`, and it + * names nothing: no handle, no pane key, no session id, no token. Its only meaning is "this child + * is a structured session with no orchestration identity", which is what a verb needs in order to + * REFUSE rather than guess one. Because it names nothing it cannot be replayed, cannot impersonate, + * and cannot flow into the hook, agent-row or mobile-projection pipelines the way a pane key would + * — which is why it is a different decision from withholding `ORCA_PANE_KEY`, not a reversal of it. + * Without it, `check` fell through to the active-terminal guess and destructively consumed a + * SIBLING pane's oldest unread batch; `requireUnambiguous` only narrows that, because with exactly + * one terminal pane in the worktree the guess still resolves — to a sibling. + * + * The handle is read from the registry at spawn time, so an in-host recovery respawn re-bakes the + * SAME handle rather than a stale or fresh one. + */ + +import { getAppEnvironment, hasAppEnvironment } from '../../shared/app-environment' +import { prependOrcaCliDirToChildPath } from '../cli/orca-cli-child-path' +import { ORCA_STRUCTURED_SESSION_ENV } from '../../shared/structured-session-marker' +import { structuredWorkerIdentities } from './structured-worker-identity' + +export function structuredWorkerChildIdentityEnv( + sessionId: string, + childEnv: Record +): Record { + const identity = structuredWorkerIdentities.getBySessionId(sessionId) + if (!identity) { + return { ...childEnv, [ORCA_STRUCTURED_SESSION_ENV]: '1' } + } + const env: Record = { + ...childEnv, + ORCA_TERMINAL_HANDLE: identity.handle, + ORCA_CLI_COMMAND: 'orca' + } + applyOrcaCliPath(env) + return env +} + +/** + * A host with no app environment installed — a plain-Node fork, or a unit test — has no userData + * root to resolve, and inventing one would write a shim into the wrong directory. + */ +function applyOrcaCliPath(env: Record): void { + if (!hasAppEnvironment()) { + return + } + const app = getAppEnvironment() + prependOrcaCliDirToChildPath(env, { + isPackaged: app.isPackaged(), + userDataPath: app.getPath('userData'), + resourcesPath: process.resourcesPath ?? null + }) +} diff --git a/src/main/runtime/structured-worker-hook-attestation.test.ts b/src/main/runtime/structured-worker-hook-attestation.test.ts new file mode 100644 index 00000000000..2f3566c5249 --- /dev/null +++ b/src/main/runtime/structured-worker-hook-attestation.test.ts @@ -0,0 +1,127 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { OrcaRuntimeWithGetOrchestrationDispatchAuthority } = + await import('./orca-runtime-get-orchestration-dispatch-authority') +const { OrcaRuntimeWithVerifyOrchestrationCompatibilityCaller } = + await import('./orca-runtime-verify-orchestration-compatibility-caller') +const { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} = await import('./structured-worker-identity') + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + +const getAuthority = + OrcaRuntimeWithGetOrchestrationDispatchAuthority.prototype.getOrchestrationDispatchAuthority +// Both borrowed from the real prototype through their public surface: a stubbed copy of the +// method under test would pin nothing. +const verifyCaller = + OrcaRuntimeWithVerifyOrchestrationCompatibilityCaller.prototype + .verifyOrchestrationCompatibilityCaller + +function registerStructuredWorker(): string { + hostRef.current = { + deps: { + store: { + getRecord: () => ({ + location: { executionHostId: 'local', wslDistro: null }, + lease: { runtimeKind: 'native', claimStatus: 'live', runtimeFence: 1 } + }) + } + } + } + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + return handle +} + +function runtimeStub(overrides: Record = {}) { + return { + runtimeId: 'runtime-1', + getOrchestrationDbIfAvailable: () => null, + restoredOrchestrationAuthorityByPtyId: new Map(), + getOrchestrationDispatchAuthority: (handle: string) => + getAuthority.call(runtimeStub(overrides), handle), + orchestrationCompatibilityHostMatches: () => true, + attestAgentHookCompatibilityAuthorityFn: undefined, + // `freeze...` is protected and is reached only on the SUCCESS path, which every test here + // asserts is never taken. Leaving it off the stub means a regression that does reach it fails + // loudly instead of quietly returning a frozen authority. + ...overrides + } +} + +describe('structured worker hook attestation stays closed', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + it('leaves both the launch token hash and the pty id empty', () => { + const handle = registerStructuredWorker() + const authority = getAuthority.call(runtimeStub(), handle) + expect(authority).not.toBeNull() + expect(authority!.launchTokenHash).toBeNull() + // Non-empty would make the restored-authority receipt lookup reachable. + expect(authority!.ptyId).toBe('') + }) + + it('refuses to attest a structured handle as a compatibility caller', () => { + const handle = registerStructuredWorker() + const stub = runtimeStub() + expect( + verifyCaller.call(stub, { + terminalHandle: handle, + paneKey: structuredWorkerIdentities.get(handle)!.paneKey, + launchToken: 'anything-the-caller-claims' + }) + ).toBeNull() + }) + + it('still refuses when a restored receipt exists under an empty pty id', () => { + const handle = registerStructuredWorker() + const identity = structuredWorkerIdentities.get(handle)! + // Fabricate the exact receipt the fallback would accept, keyed by the empty pty id. + const stub = runtimeStub({ + restoredOrchestrationAuthorityByPtyId: new Map([ + [ + '', + { + ptyId: '', + worktreeId: identity.worktreeId, + terminalHandle: handle, + paneKey: identity.paneKey, + processIncarnation: identity.processIncarnation, + hostScope: identity.hostScope + } + ] + ]), + orchestrationCompatibilityHostScopesEqual: () => true, + attestAgentHookCompatibilityAuthorityFn: undefined + }) + // Even then, attestation is required and there is no hook to provide it. + expect( + verifyCaller.call(stub, { + terminalHandle: handle, + paneKey: identity.paneKey, + launchToken: 'anything-the-caller-claims' + }) + ).toBeNull() + }) +}) diff --git a/src/main/runtime/structured-worker-identity.test.ts b/src/main/runtime/structured-worker-identity.test.ts new file mode 100644 index 00000000000..9b678ebb0eb --- /dev/null +++ b/src/main/runtime/structured-worker-identity.test.ts @@ -0,0 +1,247 @@ +import { describe, expect, it, beforeEach } from 'vitest' +import { isTerminalLeafId, parsePaneKey } from '../../shared/stable-pane-id' +import { structuredAgentSessionPaneKey } from '../../shared/structured-agent-session-projection' +import { selectExactWorkerProviderSession } from './orchestration/worker-provider-session' +import { structuredWorkerChildIdentityEnv } from './structured-worker-child-identity-env' +import { + StructuredWorkerIdentityRegistry, + isStructuredWorkerHandle, + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + sessionIdFromStructuredWorkerIncarnation, + structuredWorkerHostScope, + structuredWorkerPaneKeyBelongsToSession, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation, + structuredWorkerRecordIsCurrent +} from './structured-worker-identity' +import type { AgentSessionRecord } from '../../shared/agent-session-record' + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + +function record(overrides: { + runtimeKind?: 'native' | 'tui' + claimStatus?: AgentSessionRecord['lease']['claimStatus'] + executionHostId?: string + wslDistro?: string | null + runtimeFence?: number +}): AgentSessionRecord { + return { + schemaVersion: 2, + sessionId: SESSION_ID, + location: { + executionHostId: overrides.executionHostId ?? 'local', + wslDistro: overrides.wslDistro ?? null, + workspaceId: 'wt_1', + workspaceKind: 'git-worktree' + }, + provider: 'claude', + providerHandleChain: [], + accountHome: { variable: 'CLAUDE_CONFIG_DIR', path: '/home/me/.claude' }, + lease: { + sessionId: SESSION_ID, + runtimeKind: overrides.runtimeKind ?? 'native', + runtimeFence: overrides.runtimeFence ?? 1, + handoffStage: null, + provenHandleLinkId: null, + ownerProcess: null, + reservedSpawnToken: null, + leaseDeadlineAt: 0, + lastRenewedAt: 0, + handoffOperationId: null, + journalCheckpoint: null, + claimKeyId: 'k', + claimStatus: overrides.claimStatus ?? 'live', + unreconciled: false, + deathEvidence: null + }, + createdAt: 0, + updatedAt: 0 + } as AgentSessionRecord +} + +describe('structured worker identity', () => { + it('mints a random bearer handle that is never derived from the session id', () => { + const first = mintStructuredWorkerHandle() + const second = mintStructuredWorkerHandle() + expect(first).not.toBe(second) + expect(isStructuredWorkerHandle(first)).toBe(true) + expect(first).not.toContain(SESSION_ID) + expect(first.startsWith('term_')).toBe(false) + }) + + it('mints an UNGUESSABLE pane key, because check accepts a caller-supplied one', () => { + // A derivable pane key would let anyone who learns a session id read that worker's mailbox: + // orchestration.check falls back to params.terminalPaneKey and matches assignee_pane_key. + const first = mintStructuredWorkerPaneKey(SESSION_ID) + const second = mintStructuredWorkerPaneKey(SESSION_ID) + expect(first).not.toBe(second) + // The TAB half legitimately names the session; it is the LEAF that must be unguessable, + // because both dispatch lookups key on the leaf (exact match, then leaf-suffix equivalence). + expect(parsePaneKey(first)!.leafId).not.toContain(SESSION_ID.slice(0, 8)) + // Specifically not the sha256-of-session-id helper the chat tab projection uses. + expect(first).not.toBe( + structuredAgentSessionPaneKey(`structured-agent-session-${SESSION_ID}`, SESSION_ID) + ) + }) + + it("accepts a persisted pane key for its own session and rejects another session's", () => { + const paneKey = mintStructuredWorkerPaneKey(SESSION_ID) + expect(structuredWorkerPaneKeyBelongsToSession(paneKey, SESSION_ID)).toBe(true) + expect(structuredWorkerPaneKeyBelongsToSession(paneKey, 'another-session-id')).toBe(false) + expect(structuredWorkerPaneKeyBelongsToSession('not-a-pane-key', SESSION_ID)).toBe(false) + expect(structuredWorkerPaneKeyBelongsToSession(null, SESSION_ID)).toBe(false) + }) + + it('derives a pane key whose leaf passes the terminal leaf check', () => { + const paneKey = mintStructuredWorkerPaneKey(SESSION_ID) + const parsed = parsePaneKey(paneKey) + expect(parsed).not.toBeNull() + expect(isTerminalLeafId(parsed!.leafId)).toBe(true) + expect(parsed!.tabId).toBe(`structured-agent-session-${SESSION_ID}`) + }) + + it('round-trips the session id through the process incarnation', () => { + const incarnation = structuredWorkerProcessIncarnation(SESSION_ID) + expect(sessionIdFromStructuredWorkerIncarnation(incarnation)).toBe(SESSION_ID) + expect(sessionIdFromStructuredWorkerIncarnation('ptyid:3')).toBeNull() + }) + + it('claims local authority only for a local, non-WSL session', () => { + expect(structuredWorkerHostScope(record({}).location)).toEqual({ + kind: 'local', + hostId: 'local' + }) + expect(structuredWorkerHostScope(record({ wslDistro: 'Ubuntu' }).location)).toBeNull() + expect(structuredWorkerHostScope(record({ executionHostId: 'ssh-1' }).location)).toBeNull() + }) + + it('keeps a recovered session current across a fence bump', () => { + // The host bumps the fence on its own transparent crash recovery; fencing identity on it + // would wedge the SAME worker as identity_unproven forever. + expect(structuredWorkerRecordIsCurrent(record({ runtimeFence: 1 }))).toBe(true) + expect(structuredWorkerRecordIsCurrent(record({ runtimeFence: 9 }))).toBe(true) + expect(structuredWorkerProcessIncarnation(SESSION_ID)).toBe( + structuredWorkerProcessIncarnation(SESSION_ID) + ) + }) + + it('refuses a session handed to a TUI owner or released', () => { + expect(structuredWorkerRecordIsCurrent(record({ runtimeKind: 'tui' }))).toBe(false) + expect(structuredWorkerRecordIsCurrent(record({ claimStatus: 'released' }))).toBe(false) + expect(structuredWorkerRecordIsCurrent(null)).toBe(false) + }) +}) + +describe('structured worker identity registry', () => { + let registry: StructuredWorkerIdentityRegistry + + beforeEach(() => { + registry = new StructuredWorkerIdentityRegistry() + }) + + it('rehydrates a durable row whose persisted pane key belongs to its session', () => { + const handle = mintStructuredWorkerHandle() + const paneKey = mintStructuredWorkerPaneKey(SESSION_ID) + const identity = registry.rehydrate({ + terminal_handle: handle, + pane_key: paneKey, + process_incarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktree_id: 'wt_1', + host_scope: JSON.stringify({ kind: 'local', hostId: 'local' }) + }) + expect(identity?.sessionId).toBe(SESSION_ID) + // The leaf is random, so the durable row is the ONLY place it survives a restart. + expect(identity?.paneKey).toBe(paneKey) + expect(registry.get(handle)?.handle).toBe(handle) + expect(registry.getBySessionId(SESSION_ID)?.handle).toBe(handle) + }) + + it('refuses a row whose pane key does not match its own session id', () => { + expect( + registry.rehydrate({ + terminal_handle: mintStructuredWorkerHandle(), + pane_key: mintStructuredWorkerPaneKey('some-other-session-id'), + process_incarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktree_id: 'wt_1', + host_scope: JSON.stringify({ kind: 'local', hostId: 'local' }) + }) + ).toBeNull() + }) + + it('forgets both indexes', () => { + const handle = mintStructuredWorkerHandle() + registry.register({ + handle, + sessionId: SESSION_ID, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + registry.forget(handle) + expect(registry.get(handle)).toBeNull() + expect(registry.getBySessionId(SESSION_ID)).toBeNull() + }) +}) + +describe('structured workers stay outside the PTY-only fail-closed paths', () => { + it('keeps the selector shut by never letting a structured pane key reach a hook status', () => { + // The selector matches on pane key, so it is fail-closed for a structured worker only while + // ORCA_PANE_KEY is absent from its child's environment. That absence IS the guard: put the key + // back and the first assertion below is what an attacker gets. + // The PROCESS registry, because that is the one the spawn path reads. + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + try { + const env = structuredWorkerChildIdentityEnv(SESSION_ID, {}) + // Registered, so this is a populated env — not the empty one an unregistered session gets, + // which would satisfy the pane-key assertion for the wrong reason. + expect(env.ORCA_TERMINAL_HANDLE).toBe(handle) + expect(Object.keys(env)).not.toContain('ORCA_PANE_KEY') + } finally { + structuredWorkerIdentities.forget(handle) + } + + const paneKey = mintStructuredWorkerPaneKey(SESSION_ID) + expect( + selectExactWorkerProviderSession({ + paneKey, + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + connectionId: null, + launchToken: null, + observedAfter: 0, + statuses: [ + { + paneKey, + connectionId: null, + launchToken: null, + receivedAt: 10, + agentType: 'claude', + providerSession: { id: 'p1', transcriptPath: null } + } as never + ] + }) + ).not.toBeNull() + // With no hook status at all — the real structured case — it is null. + expect( + selectExactWorkerProviderSession({ + paneKey, + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + connectionId: null, + launchToken: null, + observedAfter: 0, + statuses: [] + }) + ).toBeNull() + }) +}) diff --git a/src/main/runtime/structured-worker-identity.ts b/src/main/runtime/structured-worker-identity.ts new file mode 100644 index 00000000000..161ae55dd5d --- /dev/null +++ b/src/main/runtime/structured-worker-identity.ts @@ -0,0 +1,203 @@ +/** + * Orchestration identity for a NATIVE-BORN structured agent session. + * + * Orchestration derives a worker's identity and its lifecycle authority from a live PTY. A + * structured session has none, so this registry is the second authority source: it maps a session + * id onto the same three facts the PTY path supplies — a bearer handle, a stable pane key, and a + * host scope — and nothing else about dispatch changes. + * + * The handle AND the pane key are both RANDOM on purpose. `orchestration.check` is identity-gated, + * not capability-gated: it falls back to a caller-supplied `terminalPaneKey` + * (`orchestration-check-methods.ts`) and dispatch lookup matches `assignee_pane_key` directly, so a + * derivable pane key alone would let anyone who learns a session id read and consume that worker's + * mailbox — and session ids are embedded in tab ids. PTY pane keys are safe only because their leaf + * is a random UUID; these match that. + */ + +import { randomUUID } from 'node:crypto' +import type { + AgentSessionExecutionLocation, + AgentSessionRecord +} from '../../shared/agent-session-record' +import { LOCAL_EXECUTION_HOST_ID } from '../../shared/execution-host' +import { structuredAgentSessionTabId } from '../../shared/structured-agent-session-projection' +import { isTerminalLeafId, makePaneKey, parsePaneKey } from '../../shared/stable-pane-id' +import { + parseWorkerTerminalHostScope, + type WorkerTerminalHostScope +} from './orchestration/worker-terminal-process-liveness' + +// Deliberately not `term_`: `issueHandle` revalidates the renderer graph epoch against the +// renderer-driven leaves map, so a main-minted `term_` leaf evaporates on the next window reload. +const STRUCTURED_WORKER_HANDLE_PREFIX = 'structworker_' +const STRUCTURED_WORKER_INCARNATION_PREFIX = 'structured:' + +export type StructuredWorkerIdentity = { + handle: string + sessionId: string + /** Null when the entry was rehydrated from the durable row, which does not carry the provider. */ + agent: 'claude' | 'codex' | null + paneKey: string + processIncarnation: string + worktreeId: string + hostScope: WorkerTerminalHostScope +} + +export function isStructuredWorkerHandle(handle: string | null | undefined): boolean { + return typeof handle === 'string' && handle.startsWith(STRUCTURED_WORKER_HANDLE_PREFIX) +} + +export function mintStructuredWorkerHandle(): string { + return `${STRUCTURED_WORKER_HANDLE_PREFIX}${randomUUID()}` +} + +/** + * A RANDOM leaf, minted once per worker and persisted with the rest of the identity. + * + * Emphatically not `structuredAgentSessionPaneKey`, which is a sha256 of the session id. A pane + * key is an identity credential on its own: `orchestration.check` is identity-gated, not + * capability-gated, and accepts a caller-supplied `terminalPaneKey` that `getActiveDispatchForIdentity` + * matches by leaf suffix. A derivable pane key would therefore let anyone who learns a session id — + * which the tab id embeds in plain text — read and consume that worker's mailbox with no token. + * PTY pane keys are safe only because their leaf UUID is random; this one has to be too. + * + * Restart stability comes from persisting the minted key, not from re-deriving it. + */ +export function mintStructuredWorkerPaneKey(sessionId: string): string { + return makePaneKey(structuredAgentSessionTabId(sessionId), randomUUID()) +} + +/** Integrity check for a persisted pane key: same session's tab, and a real terminal leaf. */ +export function structuredWorkerPaneKeyBelongsToSession( + paneKey: string | null | undefined, + sessionId: string +): boolean { + const parsed = paneKey ? parsePaneKey(paneKey) : null + return Boolean( + parsed && + parsed.tabId === structuredAgentSessionTabId(sessionId) && + isTerminalLeafId(parsed.leafId) + ) +} + +/** + * Process continuity for a structured worker. + * + * NOT the runtime fence: the fence is an owner-generation counter that the host bumps during its + * own transparent crash recovery, so fencing identity on it would make a recovered — but same — + * worker fail `verifyDispatchCapability` forever and wedge release as `identity_unproven`. The + * session id is minted once per dispatch and survives that recovery, so it is the lineage. + */ +export function structuredWorkerProcessIncarnation(sessionId: string): string { + return `${STRUCTURED_WORKER_INCARNATION_PREFIX}${sessionId}` +} + +export function sessionIdFromStructuredWorkerIncarnation( + processIncarnation: string | null | undefined +): string | null { + if (!processIncarnation?.startsWith(STRUCTURED_WORKER_INCARNATION_PREFIX)) { + return null + } + const sessionId = processIncarnation.slice(STRUCTURED_WORKER_INCARNATION_PREFIX.length) + return sessionId.length > 0 ? sessionId : null +} + +/** Structured sessions can only exist local and outside WSL; anything else is not our authority. */ +export function structuredWorkerHostScope( + location: AgentSessionExecutionLocation +): WorkerTerminalHostScope | null { + return location.executionHostId === LOCAL_EXECUTION_HOST_ID && !location.wslDistro + ? { kind: 'local', hostId: 'local' } + : null +} + +/** Whether the durable record still describes THIS worker under this host. */ +export function structuredWorkerRecordIsCurrent( + record: AgentSessionRecord | null | undefined +): boolean { + return Boolean( + record && + record.lease.runtimeKind === 'native' && + record.lease.claimStatus !== 'released' && + structuredWorkerHostScope(record.location) + ) +} + +export class StructuredWorkerIdentityRegistry { + private readonly byHandle = new Map() + private readonly bySessionId = new Map() + + register(identity: StructuredWorkerIdentity): StructuredWorkerIdentity { + this.byHandle.set(identity.handle, identity) + this.bySessionId.set(identity.sessionId, identity) + return identity + } + + get(handle: string): StructuredWorkerIdentity | null { + return this.byHandle.get(handle) ?? null + } + + getBySessionId(sessionId: string): StructuredWorkerIdentity | null { + return this.bySessionId.get(sessionId) ?? null + } + + /** Every worker this process knows about; callers apply their own liveness gate. */ + list(): StructuredWorkerIdentity[] { + return [...this.byHandle.values()] + } + + forget(handle: string): void { + const identity = this.byHandle.get(handle) + if (!identity) { + return + } + this.byHandle.delete(handle) + if (this.bySessionId.get(identity.sessionId) === identity) { + this.bySessionId.delete(identity.sessionId) + } + } + + /** + * Rebuilds an entry from the durable worker-terminal resource row after a restart, which is the + * only place a structured worker's pane key and host scope outlive this process. A row whose + * pane key does not belong to its own recorded session is refused rather than trusted. + */ + rehydrate(row: { + terminal_handle: string + pane_key: string | null + process_incarnation: string | null + worktree_id: string | null + host_scope: string | null + }): StructuredWorkerIdentity | null { + const sessionId = sessionIdFromStructuredWorkerIncarnation(row.process_incarnation) + const hostScope = parseWorkerTerminalHostScope(row.host_scope) + if ( + !sessionId || + !hostScope || + !row.worktree_id || + !isStructuredWorkerHandle(row.terminal_handle) || + // The leaf is random, so the row IS the only source for it; verify only that it is a real + // leaf under this session's tab rather than trying to re-derive it. + !structuredWorkerPaneKeyBelongsToSession(row.pane_key, sessionId) + ) { + return null + } + return this.register({ + handle: row.terminal_handle, + sessionId, + // The row does not carry the provider; callers that need it read the durable record. + agent: null, + paneKey: row.pane_key as string, + processIncarnation: structuredWorkerProcessIncarnation(sessionId), + worktreeId: row.worktree_id, + hostScope + }) + } + + clear(): void { + this.byHandle.clear() + this.bySessionId.clear() + } +} + +export const structuredWorkerIdentities = new StructuredWorkerIdentityRegistry() diff --git a/src/main/runtime/structured-worker-mail-routing.test.ts b/src/main/runtime/structured-worker-mail-routing.test.ts new file mode 100644 index 00000000000..d6a1caee8e6 --- /dev/null +++ b/src/main/runtime/structured-worker-mail-routing.test.ts @@ -0,0 +1,144 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { OrcaRuntimeWithAdoptTerminalOrphansFromInventory } = + await import('./orca-runtime-adopt-terminal-orphans-from-inventory') +const { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} = await import('./structured-worker-identity') + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' +const prototype = OrcaRuntimeWithAdoptTerminalOrphansFromInventory.prototype +const getLivePaneKey = prototype.getLiveTerminalPaneKey +const resolveActiveTerminal = prototype.resolveActiveTerminal + +function installRecord(lease: { runtimeKind: string; claimStatus: string } | null): void { + hostRef.current = lease + ? { + deps: { + store: { + getRecord: () => ({ + location: { executionHostId: 'local', wslDistro: null }, + lease: { ...lease, runtimeFence: 1, deathEvidence: null } + }) + } + }, + hasSession: () => lease.claimStatus === 'live' + } + : null +} + +function registerWorker(): string { + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + return handle +} + +const paneKeyStub = { + getOrchestrationDbIfAvailable: () => null, + getLivePtyForHandle: () => null, + resolveLiveLeafForHandle: () => null, + ptysById: new Map(), + getPaneKeyForTerminalHandle: () => null +} + +describe('bare-handle direct mail to a structured session', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + it('resolves a live pane key, so recipient routing does not answer terminal_not_found', () => { + // resolveBareOrchestrationRecipient reads this getter, not getTerminalPaneKey. + const handle = registerWorker() + installRecord({ runtimeKind: 'native', claimStatus: 'live' }) + expect(getLivePaneKey.call(paneKeyStub, handle)).toBe( + structuredWorkerIdentities.get(handle)!.paneKey + ) + }) + + it('withholds the pane key when the session is not proven live', () => { + // The PTY branch is connected-gated so mail is never routed to a corpse; so is this one. + const handle = registerWorker() + installRecord({ runtimeKind: 'native', claimStatus: 'reserved' }) + expect(getLivePaneKey.call(paneKeyStub, handle)).toBeNull() + }) + + it('withholds the pane key when the lease moved to a terminal owner', () => { + const handle = registerWorker() + installRecord({ runtimeKind: 'tui', claimStatus: 'live' }) + expect(getLivePaneKey.call(paneKeyStub, handle)).toBeNull() + }) +}) + +describe('implicit sender resolution refuses to guess', () => { + function senderStub(leafIds: readonly string[]) { + return { + graphStatus: 'ready', + assertGraphReady: () => {}, + resolveWorktreeSelector: async () => ({ id: 'wt_1' }), + tabs: new Map(), + leaves: new Map( + leafIds.map((leafId) => [leafId, { tabId: 'tab_1', leafId, worktreeId: 'wt_1' }]) + ), + issueHandle: (leaf: { leafId: string }) => `term_${leaf.leafId}` + } + } + + it('returns the only candidate leaf', async () => { + await expect( + resolveActiveTerminal.call(senderStub(['leaf_a']), 'id:wt_1', { requireUnambiguous: true }) + ).resolves.toBe('term_leaf_a') + }) + + it('refuses rather than picking the first of several', async () => { + // An arbitrary pick lets a bare `send --type worker_done` settle a SIBLING's context-only + // dispatch, a tier that has no capability token to reject on. + await expect( + resolveActiveTerminal.call(senderStub(['leaf_a', 'leaf_b']), 'id:wt_1', { + requireUnambiguous: true + }) + ).rejects.toThrow('no_active_terminal') + }) + + it('refuses the same arbitrary pick before the terminal graph is ready', async () => { + // The snapshot carries a focused terminal on purpose: without it the refusal below would come + // from the ambiguous `listTerminals` fallback alone and would still hold with the pre-ready + // focus guess left in, proving nothing about it. + const preReady = { + graphStatus: 'starting', + resolveWorktreeSelector: async () => ({ id: 'wt_1' }), + getMobileSessionTabsForWorktree: () => ({ + tabs: [{ type: 'terminal', isActive: true, status: 'ready', terminal: 'term_focused' }] + }), + listTerminals: async () => ({ terminals: [{ handle: 'term_a' }, { handle: 'term_b' }] }) + } + await expect( + resolveActiveTerminal.call(preReady, 'id:wt_1', { requireUnambiguous: true }) + ).rejects.toThrow('no_active_terminal') + // The same stub still answers the focus guess for a caller that is not claiming an identity. + await expect(resolveActiveTerminal.call(preReady, 'id:wt_1')).resolves.toBe('term_focused') + }) + + it('still picks arbitrarily for callers that are not claiming an identity', async () => { + await expect( + resolveActiveTerminal.call(senderStub(['leaf_a', 'leaf_b']), 'id:wt_1') + ).resolves.toBe('term_leaf_a') + }) +}) diff --git a/src/main/runtime/structured-worker-takeover-pane-key.test.ts b/src/main/runtime/structured-worker-takeover-pane-key.test.ts new file mode 100644 index 00000000000..709bdbc8db4 --- /dev/null +++ b/src/main/runtime/structured-worker-takeover-pane-key.test.ts @@ -0,0 +1,83 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentSessionRecord } from '../../shared/agent-session-record' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { OrcaRuntimeWithGetPtyRecordForPaneKey } = + await import('./orca-runtime-get-pty-record-for-pane-key') +const { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} = await import('./structured-worker-identity') + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + +function installRecord(lease: { runtimeKind: string; claimStatus: string }): void { + hostRef.current = { + deps: { + store: { + getRecord: (sessionId: string) => + ({ + sessionId, + location: { executionHostId: 'local', wslDistro: null }, + lease: { ...lease, runtimeFence: 1, deathEvidence: null } + }) as unknown as AgentSessionRecord + } + }, + hasSession: () => true + } +} + +function registerWorker(): string { + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + return handle +} + +function runtime() { + return Object.assign(Object.create(OrcaRuntimeWithGetPtyRecordForPaneKey.prototype), { + _orchestrationDb: null + }) as { getStructuredWorkerPaneKeyForSession: (sessionId: string) => string | null } +} + +describe('resolving a structured worker takeover by session', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + it('resolves the session to the persisted pane key that worker owns', () => { + const handle = registerWorker() + installRecord({ runtimeKind: 'native', claimStatus: 'live' }) + expect(runtime().getStructuredWorkerPaneKeyForSession(SESSION_ID)).toBe( + structuredWorkerIdentities.get(handle)!.paneKey + ) + }) + + it('answers nothing for a session this runtime no longer owns', () => { + registerWorker() + installRecord({ runtimeKind: 'tui', claimStatus: 'live' }) + expect(runtime().getStructuredWorkerPaneKeyForSession(SESSION_ID)).toBeNull() + }) + + it('answers nothing for a session that is not an orchestration worker', () => { + // A plain chat session owns no worker-terminal resource, so there is no ownership to + // relinquish and nothing to mark. + installRecord({ runtimeKind: 'native', claimStatus: 'live' }) + expect(runtime().getStructuredWorkerPaneKeyForSession(SESSION_ID)).toBeNull() + }) +}) diff --git a/src/main/runtime/structured-worker-terminal-read.test.ts b/src/main/runtime/structured-worker-terminal-read.test.ts new file mode 100644 index 00000000000..97859a988e0 --- /dev/null +++ b/src/main/runtime/structured-worker-terminal-read.test.ts @@ -0,0 +1,188 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentJournalRenderItem } from '../../shared/agent-session-journal-types' +import type { AgentSessionRecord } from '../../shared/agent-session-record' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { readStructuredWorkerTerminal } = await import('./structured-worker-terminal-read') +const { OrcaRuntimeWithResolveTerminalPane } = await import('./orca-runtime-resolve-terminal-pane') +const { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} = await import('./structured-worker-identity') + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + +function message(id: string, text: string): AgentJournalRenderItem { + return { + itemId: id, + observedAt: 1, + body: { kind: 'message', role: 'assistant', blocks: [{ type: 'text', text }] } + } as unknown as AgentJournalRenderItem +} + +function installHost(options: { + items?: readonly AgentJournalRenderItem[] | 'unreadable' + hasOlder?: boolean + lease?: { runtimeKind: string; claimStatus: string } + hasSession?: boolean +}): void { + const lease = options.lease ?? { runtimeKind: 'native', claimStatus: 'live' } + hostRef.current = { + deps: { + store: { + getRecord: (sessionId: string) => + ({ + sessionId, + location: { executionHostId: 'local', wslDistro: null }, + lease: { ...lease, runtimeFence: 1, deathEvidence: null } + }) as unknown as AgentSessionRecord + } + }, + hasSession: () => options.hasSession ?? true, + history: () => { + if (options.items === 'unreadable') { + throw new Error('agent_session_ownership_unknown') + } + return { page: { items: options.items ?? [], hasOlder: options.hasOlder ?? false } } + } + } +} + +function registerWorker(): string { + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + return handle +} + +describe('reading a structured worker through the terminal-read path', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + it('serves the journal as terminal lines, with no dispatch and no capability', () => { + // The defect this pins: a peer has no dispatch id and no coordinator standing, so `worker-read` + // is closed to it, and `terminal read` threw `terminal_handle_stale` for a perfectly live + // worker. A peer could not see a structured agent's recent output at all. + const handle = registerWorker() + installHost({ items: [message('i1', 'first line\nsecond line'), message('i2', 'done')] }) + const read = readStructuredWorkerTerminal({ handle, db: null }) + expect(read?.tail).toEqual(['[assistant] first line', 'second line', '[assistant] done']) + expect(read?.status).toBe('running') + expect(read?.truncated).toBe(false) + }) + + it('honours limit, and claims no cursor space it cannot honour', () => { + const handle = registerWorker() + installHost({ items: [message('i1', 'a'), message('i2', 'b'), message('i3', 'c')] }) + const read = readStructuredWorkerTerminal({ handle, db: null, limit: 2 }) + expect(read?.tail).toEqual(['[assistant] b', '[assistant] c']) + // No index is advertised: the next read re-projects a sliding window, so 0/length would name + // positions that address different lines by then. + expect(read?.nextCursor).toBeNull() + expect(read?.oldestCursor).toBeUndefined() + expect(read?.latestCursor).toBeUndefined() + }) + + it('refuses a cursor read rather than silently misdelivering lines', () => { + // The PTY cursor indexes an append-only completed-line buffer with a monotone count. This + // window is a bounded tail re-projected every read, so a saved index addresses different lines + // as the journal grows — and `truncated` could never fire to say so, because it tests + // `cursor < oldestCursor` and `oldestCursor` was always 0. A poller would get wrong or + // duplicated lines with `truncated:false`. + const handle = registerWorker() + installHost({ items: [message('i1', 'a')] }) + const refusal = (() => { + try { + readStructuredWorkerTerminal({ handle, db: null, cursor: 0 }) + return '' + } catch (error) { + return (error as Error).message + } + })() + expect(refusal).toMatch(/not line-addressable/) + // Tells the caller what DOES work here. Polling a bounded newest-last tail and diffing fails + // safe — a harmless re-read — where a broken cursor fails unsafe, as a silent hole. + expect(refusal).toMatch(/poll it and diff/) + // And names no paging alternative, because there is none. It must never send a peer to + // `worker-read`: that verb needs a dispatch id and coordinator standing this caller does not + // have, and it is a window index over the same bounded page rather than an append-only anchor. + expect(refusal).not.toContain('worker-read') + }) + + it('reports dropped history as truncated rather than pretending the page is whole', () => { + const handle = registerWorker() + installHost({ items: [message('i1', 'tail only')], hasOlder: true }) + expect(readStructuredWorkerTerminal({ handle, db: null })?.truncated).toBe(true) + }) + + it('redacts dispatch capability tokens the same way the archive path does', () => { + const handle = registerWorker() + const token = `dcap_${'a'.repeat(32)}` + installHost({ items: [message('i1', `token is ${token} here`)] }) + const tail = readStructuredWorkerTerminal({ handle, db: null })?.tail.join('\n') ?? '' + expect(tail).not.toContain(token) + expect(tail).toContain('[dispatch capability redacted]') + }) + + it('refuses when the session is not attached rather than answering an empty tail', () => { + // An empty tail is the claim "this worker has produced no output", which is a different and + // false statement — and the one a caller cannot tell apart from a real silence. + const handle = registerWorker() + installHost({ items: 'unreadable' }) + expect(() => readStructuredWorkerTerminal({ handle, db: null })).toThrow( + 'agent_session_ownership_unknown' + ) + }) + + it('reports a session it cannot verify as unknown, never as running', () => { + const handle = registerWorker() + installHost({ items: [message('i1', 'said something')], hasSession: false }) + expect(readStructuredWorkerTerminal({ handle, db: null })?.status).toBe('unknown') + }) + + it('is what `terminal read` answers with, ahead of the PTY lookup', async () => { + // The real method through the real prototype, because the wiring IS the fix: the module below + // could be perfect and a peer would still get `terminal_handle_stale` if nothing called it. + const handle = registerWorker() + installHost({ items: [message('i1', 'hello')] }) + const runtime = Object.assign(Object.create(OrcaRuntimeWithResolveTerminalPane.prototype), { + getOrchestrationDbIfAvailable: () => null, + getLivePtyForHandle: () => { + throw new Error('the PTY lookup must never be reached for a structured worker') + } + }) as { readTerminal: (handle: string, opts?: object) => Promise<{ tail: string[] }> } + await expect(runtime.readTerminal(handle)).resolves.toMatchObject({ + tail: ['[assistant] hello'], + source: 'stream' + }) + // There is no rendered grid to screenshot, and saying so beats inventing one. + await expect(runtime.readTerminal(handle, { screen: true })).resolves.toMatchObject({ + source: 'screen-unavailable' + }) + }) + + it('leaves every handle that is not a live structured worker to the PTY path', () => { + const handle = registerWorker() + installHost({ items: [message('i1', 'x')] }) + expect(readStructuredWorkerTerminal({ handle: 'term_abc', db: null })).toBeNull() + // A lease handed to a TUI owner is no longer this runtime's structured worker. + installHost({ items: [message('i1', 'x')], lease: { runtimeKind: 'tui', claimStatus: 'live' } }) + expect(readStructuredWorkerTerminal({ handle, db: null })).toBeNull() + }) +}) diff --git a/src/main/runtime/structured-worker-terminal-read.ts b/src/main/runtime/structured-worker-terminal-read.ts new file mode 100644 index 00000000000..9b4a118b8ae --- /dev/null +++ b/src/main/runtime/structured-worker-terminal-read.ts @@ -0,0 +1,110 @@ +/** + * `terminal read` for a worker that IS a structured agent session. + * + * Peers peek at each other's recent output constantly, and for a PTY worker that is `terminal + * read`. A structured worker had no answer at all: `worker-read` demands a dispatch id and + * coordinator standing a peer does not have, so the only agent-to-agent read verb refused to + * resolve the handle. This serves the same verb from the session's journal. + * + * The result is a plain `RuntimeTerminalRead` — the journal is projected to LINES and bounded by + * the very reader the PTY tail uses — so nothing an agent reads reveals which kind of worker + * answered. `limit` and `truncated` keep their existing meanings. + * + * `cursor` does NOT, and is refused rather than approximated. The PTY contract is an index into an + * append-only completed-line buffer with a monotone count. A session journal is a REDUCED, MUTABLE + * timeline: an item's projected text changes at its original sequence after later items exist, the + * delta coalescer revises items repeatedly, settlement can rewrite one smaller, a pending approval + * renders as nothing and then as something, and `sequence` resets on epoch rollover — so no index, + * numeric or opaque, stays valid. `worker-read --source transcript` is a window index over the same + * bounded page, not an append-only anchor; do not point callers at it as one. + * + * The refusal is therefore permanent, not a stopgap, and no windowed alternative should be built: + * a broken cursor fails UNSAFE (a silent hole in a poller's output) while diffing a bounded tail + * fails safe (a harmless re-read), and a second paging-shaped verb would invite the PTY assumptions + * this one cannot honour. + * + * READ ONLY, deliberately. `terminal.show` still refuses a structured handle: synthesising a + * `ptyId`/`leafId`/`paneRuntimeId` would hand every public terminal verb something that looks + * writable and is not. + */ + +import type { RuntimeTerminalRead } from '../../shared/runtime-types' +import { formatWorkerTranscriptMessage } from '../../shared/worker-transcript-text' +import { AGENT_SESSION_NOT_ATTACHED } from '../native-chat/agent-session-wire/structured-agent-session-mutation-admission' +import type { OrchestrationDb } from './orchestration/db' +import { boundStructuredJournalTail } from './orchestration/structured-worker-journal-archive' +import { readStructuredJournalPage } from './orchestration/structured-worker-journal-page' +import { + observeStructuredWorker, + resolveStructuredWorkerAuthority, + structuredWorkerTerminalState +} from './structured-worker-authority' +import { readTerminalTail } from './terminal-tail-read' + +/** + * The recent output of a structured worker, or null when this handle is not one. + * + * Null is the "not mine" answer, so the PTY path keeps every handle it already owned. A handle that + * IS a structured worker never falls through: an unreadable journal refuses rather than answering + * an empty tail, which a caller cannot tell from a worker that has said nothing. + */ +export function readStructuredWorkerTerminal(args: { + handle: string + db: OrchestrationDb | null + cursor?: number + limit?: number +}): RuntimeTerminalRead | null { + const identity = resolveStructuredWorkerAuthority(args.handle, args.db)?.identity + if (!identity) { + return null + } + if (args.cursor !== undefined) { + // No index can be re-anchored here, so this refusal names no paging alternative — there is + // none. `terminal.read`'s cursor indexes an append-only completed-line buffer with a monotone + // count; this window is a bounded tail re-projected every read over a MUTABLE timeline, so the + // same index means different lines as items are revised in place, and `truncated` + // (`cursor < oldestCursor`) could never fire to say so because `oldestCursor` is always 0. + // Serving it would silently return wrong or duplicated lines to a poller. + // + // It must NOT redirect to `worker-read --source transcript`: a peer reaching this verb has + // neither a dispatch id nor coordinator standing (see the header), so it cannot run that one — + // and that verb is a window index over the same bounded page, so it would not be a paging + // answer even if it could. + throw new Error( + `${args.handle} serves recent output without a cursor; its history is not line-addressable. ` + + 'Read it without --cursor: the tail is bounded and newest-last, so poll it and diff. ' + + 'A structured session has no durable line anchor to page from — nothing else does either.' + ) + } + const page = readStructuredJournalPage(identity.sessionId) + if (!page) { + // Honest refusal, and the same one the send lane reports: an empty tail would read as "this + // worker has produced no output", which is a different and false claim. + throw new Error(AGENT_SESSION_NOT_ATTACHED.code) + } + // Redacts dispatch capabilities and clips oversized blocks under the archive path's byte bound. + const bounded = boundStructuredJournalTail(page.items) + const lines = bounded.messages.flatMap((message) => + formatWorkerTranscriptMessage(message).split('\n') + ) + const read = readTerminalTail({ + handle: args.handle, + status: structuredWorkerTerminalState(observeStructuredWorker(identity).status), + previewLines: lines, + // Unreachable without a cursor, and deliberately empty rather than a copy of `lines`: a + // running turn's text is still growing, so calling it "completed" is the `"hel"`/`"hello"` + // hazard the PTY reader guards against. + completedLines: [], + partialLine: '', + completedLineCount: 0, + // Older items really were dropped, by the page limit or the byte bound; `truncated` is how the + // PTY read already says exactly that. + bufferTruncated: page.hasOlder || bounded.limited, + ...(args.limit === undefined ? {} : { limit: args.limit }) + }) + // No cursor space is claimed, because none exists here. `nextCursor: null` is the contract's own + // "nothing to continue from"; emitting 0/length would advertise an index the next read cannot + // honour. + const { oldestCursor: _oldest, latestCursor: _latest, ...withoutCursorSpace } = read + return { ...withoutCursorSpace, nextCursor: null } +} diff --git a/src/main/runtime/structured-worker-terminal-refusal.test.ts b/src/main/runtime/structured-worker-terminal-refusal.test.ts new file mode 100644 index 00000000000..413e7b3ee5d --- /dev/null +++ b/src/main/runtime/structured-worker-terminal-refusal.test.ts @@ -0,0 +1,90 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentSessionRecord } from '../../shared/agent-session-record' + +const hostRef: { current: unknown } = { current: null } + +vi.mock('../native-chat/agent-session-wire/structured-agent-session-registry', () => ({ + getStructuredAgentSessionHost: () => hostRef.current +})) + +const { structuredWorkerTerminalRefusal } = await import('./structured-worker-terminal-refusal') +const { + mintStructuredWorkerHandle, + mintStructuredWorkerPaneKey, + structuredWorkerIdentities, + structuredWorkerProcessIncarnation +} = await import('./structured-worker-identity') + +const SESSION_ID = 'a1b2c3d4-e5f6-4a7b-8c9d-0e1f2a3b4c5d' + +function installRecord(): void { + hostRef.current = { + deps: { + store: { + getRecord: (sessionId: string) => + ({ + sessionId, + location: { executionHostId: 'local', wslDistro: null }, + lease: { + runtimeKind: 'native', + claimStatus: 'live', + runtimeFence: 1, + deathEvidence: null + } + }) as unknown as AgentSessionRecord + } + }, + hasSession: () => true + } +} + +function registerWorker(): string { + const handle = mintStructuredWorkerHandle() + structuredWorkerIdentities.register({ + handle, + sessionId: SESSION_ID, + agent: 'claude', + paneKey: mintStructuredWorkerPaneKey(SESSION_ID), + processIncarnation: structuredWorkerProcessIncarnation(SESSION_ID), + worktreeId: 'wt_1', + hostScope: { kind: 'local', hostId: 'local' } + }) + return handle +} + +describe('the refusal a terminal verb gives a structured worker handle', () => { + beforeEach(() => { + structuredWorkerIdentities.clear() + hostRef.current = null + }) + + it('says the handle is an agent session, not that it went stale', () => { + // `terminal_handle_stale` is a claim the handle died. It never did — the session is live and + // has no terminal — so callers went looking for a remint that cannot exist. + const handle = registerWorker() + installRecord() + const error = structuredWorkerTerminalRefusal(handle, null) + expect(error.message).not.toContain('terminal_handle_stale') + expect((error as { code?: string }).code).toBe('terminal_unsupported_for_agent_session') + }) + + it('points at the structured equivalents rather than just failing', () => { + const handle = registerWorker() + installRecord() + const message = structuredWorkerTerminalRefusal(handle, null).message + expect(message).toContain('orca terminal read') + expect(message).toContain('worker-read --source transcript') + expect(message).toContain('orca orchestration send') + }) + + it('keeps the stale error for a PTY handle, which really can go stale', () => { + expect(structuredWorkerTerminalRefusal('term_gone', null).message).toBe('terminal_handle_stale') + }) + + it('keeps the stale error for a session this runtime no longer owns', () => { + // Once the lease moves or the session is released the handle IS dead, and saying so is right. + registerWorker() + hostRef.current = null + expect(structuredWorkerTerminalRefusal('term_gone', null).message).toBe('terminal_handle_stale') + }) +}) diff --git a/src/main/runtime/structured-worker-terminal-refusal.ts b/src/main/runtime/structured-worker-terminal-refusal.ts new file mode 100644 index 00000000000..5f934980a02 --- /dev/null +++ b/src/main/runtime/structured-worker-terminal-refusal.ts @@ -0,0 +1,33 @@ +/** + * What a terminal verb should say when handed a structured worker's handle. + * + * `terminal_handle_stale` is a claim that the handle went dead, and for a structured worker it is + * simply false: the session is live, it has no terminal, and it never had one. Callers acting on + * that claim went looking for a remint that cannot exist. The refusal names the structured + * equivalent instead, so an agent that lands here knows what to run rather than what failed. + * + * `terminal.show` stays non-resolving on purpose: synthesising a `ptyId`/`leafId`/`paneRuntimeId` + * would hand every public terminal verb something that looks writable and is not. + */ + +import type { OrchestrationDb } from './orchestration/db' +import { resolveStructuredWorkerAuthority } from './structured-worker-authority' + +const TERMINAL_HANDLE_STALE = 'terminal_handle_stale' +const AGENT_SESSION_HAS_NO_TERMINAL = 'terminal_unsupported_for_agent_session' + +export function structuredWorkerTerminalRefusal( + handle: string, + db: OrchestrationDb | null | undefined +): Error { + if (!resolveStructuredWorkerAuthority(handle, db)) { + return new Error(TERMINAL_HANDLE_STALE) + } + const error = new Error( + `${handle} is an agent session, not a terminal, so terminal commands cannot address it. ` + + 'Read its output with `orca terminal read` or `orca orchestration worker-read --source transcript`, ' + + 'send it work with `orca orchestration send`, and open it from its chat tab.' + ) + Object.assign(error, { code: AGENT_SESSION_HAS_NO_TERMINAL }) + return error +} diff --git a/src/main/runtime/terminal-identity-probe.test.ts b/src/main/runtime/terminal-identity-probe.test.ts new file mode 100644 index 00000000000..cf356bb1bbc --- /dev/null +++ b/src/main/runtime/terminal-identity-probe.test.ts @@ -0,0 +1,64 @@ +import { describe, expect, it, vi } from 'vitest' +import { + resolveTerminalIdentityFromProbes, + TERMINAL_HANDLE_STALE_ERROR +} from './terminal-identity-probe' + +function probes(overrides: { structured?: boolean; livePty?: boolean; leafError?: Error | null }) { + const assertLiveLeaf = vi.fn(() => { + if (overrides.leafError) { + throw overrides.leafError + } + }) + return { + calls: { assertLiveLeaf }, + probes: { + isLiveStructuredWorker: () => overrides.structured ?? false, + hasLivePty: () => overrides.livePty ?? false, + assertLiveLeaf + } + } +} + +describe('the terminal identity probe', () => { + it('answers live for a structured worker without touching the PTY graph', () => { + // The defect this pins: the sender validator asked `terminal.show`, whose leaf lookup misses + // for a session that never had a pane, and reported a live worker's own handle as stale. + const { calls, probes: p } = probes({ structured: true }) + expect(resolveTerminalIdentityFromProbes('structworker_1', p)).toEqual({ + handle: 'structworker_1', + live: true + }) + expect(calls.assertLiveLeaf).not.toHaveBeenCalled() + }) + + it('answers live for a PTY handle the runtime still holds', () => { + const { probes: p } = probes({ livePty: true }) + expect(resolveTerminalIdentityFromProbes('term_1', p).live).toBe(true) + }) + + it('runs the full leaf check for a handle with no live PTY', () => { + // `getLiveLeafForHandle` is the one that re-checks `rendererGraphEpoch`, and that check is the + // entire reason the sender is validated: a long-lived shell keeps a stale + // `ORCA_TERMINAL_HANDLE` across a window reload. A cheaper probe would start passing it. + const { calls, probes: p } = probes({}) + expect(resolveTerminalIdentityFromProbes('term_1', p).live).toBe(true) + expect(calls.assertLiveLeaf).toHaveBeenCalledTimes(1) + }) + + it('answers not-live for a stale handle', () => { + const { probes: p } = probes({ leafError: new Error(TERMINAL_HANDLE_STALE_ERROR) }) + expect(resolveTerminalIdentityFromProbes('term_1', p)).toEqual({ + handle: 'term_1', + live: false + }) + }) + + it('propagates "could not look" rather than reporting it as a dead handle', () => { + // A graph that is not ready yet is not evidence the handle died, and `terminal.show` lets that + // error through today. Answering `live: false` here would make a command refuse its own sender + // during startup instead of failing loudly. + const { probes: p } = probes({ leafError: new Error('graph_not_ready') }) + expect(() => resolveTerminalIdentityFromProbes('term_1', p)).toThrow('graph_not_ready') + }) +}) diff --git a/src/main/runtime/terminal-identity-probe.ts b/src/main/runtime/terminal-identity-probe.ts new file mode 100644 index 00000000000..43aca59cb1b --- /dev/null +++ b/src/main/runtime/terminal-identity-probe.ts @@ -0,0 +1,55 @@ +/** + * "Is this handle a live orchestration identity?" — answered for BOTH lanes. + * + * The CLI asked that question by calling `terminal.show`, which is a PTY verb: it resolves a pane, + * a ptyId and a preview. A structured worker has none of those, so `showTerminal` missed, threw + * `terminal_handle_stale`, and the caller concluded the handle the child was BORN with was dead — + * failing twelve coordinator verbs for a worker whose own preamble tells it to run them. + * + * So the identity question gets its own probe, returning a handle and a boolean and nothing + * writable. `terminal.show` deliberately still refuses a structured handle: synthesising + * `ptyId`/`leafId`/`paneRuntimeId` would hand every public terminal verb something that looks + * writable and is not. + * + * The PTY half is EXACTLY today's `terminal.show` liveness test, `getLiveLeafForHandle` included, + * so its `rendererGraphEpoch` re-check still runs. That check is the whole point of validating at + * all — a long-lived shell keeps a stale `ORCA_TERMINAL_HANDLE` across a window reload — and a + * cheaper probe that skipped it (`getPaneKeyForTerminalHandle`, say) would quietly start passing + * handles that fail today. + */ + +export type RuntimeTerminalIdentity = { + handle: string + live: boolean +} + +/** The one error code that means "not live" rather than "could not look". */ +export const TERMINAL_HANDLE_STALE_ERROR = 'terminal_handle_stale' + +export type TerminalIdentityProbes = { + /** A structured worker of THIS runtime, proven through its durable record. */ + isLiveStructuredWorker: () => boolean + hasLivePty: () => boolean + /** Today's leaf check; throws `terminal_handle_stale` for a stale or reloaded handle. */ + assertLiveLeaf: () => void +} + +export function resolveTerminalIdentityFromProbes( + handle: string, + probes: TerminalIdentityProbes +): RuntimeTerminalIdentity { + if (probes.isLiveStructuredWorker() || probes.hasLivePty()) { + return { handle, live: true } + } + try { + probes.assertLiveLeaf() + return { handle, live: true } + } catch (error) { + if (error instanceof Error && error.message === TERMINAL_HANDLE_STALE_ERROR) { + return { handle, live: false } + } + // Anything else — a graph that is not ready yet — is "could not look", and must propagate + // exactly as it does through `terminal.show` today rather than being read as a dead handle. + throw error + } +} diff --git a/src/main/runtime/worktree-pty-surface-sweeps.ts b/src/main/runtime/worktree-pty-surface-sweeps.ts new file mode 100644 index 00000000000..4c663086a39 --- /dev/null +++ b/src/main/runtime/worktree-pty-surface-sweeps.ts @@ -0,0 +1,140 @@ +/** + * The two PTY-surface sweeps `killAllProcessesForWorktree` fans out to. + * + * Split from the teardown entry point so that file stays under the line ceiling once the + * structured-session sweep joined it. Each function owns one registration surface: the installed + * provider's session list, and the local pty-registry. + */ + +import type { IPtyProvider } from '../providers/types' +import { listRegisteredPtys } from '../memory/pty-registry' +import { isPathInsideOrEqual } from '../../shared/cross-platform-path' +import { splitWorktreeId, splitWorktreeIdForFilesystem } from '../../shared/worktree/id' +import { mapWithConcurrency } from '../../shared/map-with-concurrency' +import { teardownRpcDeadline } from './worktree-teardown-deadline' + +// Why: normal inventories still coalesce into one process scan, while a stale +// or pathological inventory cannot fan out unbounded provider/RPC shutdowns. +const WORKTREE_TEARDOWN_CONCURRENCY = 32 + +export type WorktreeTeardownStopPty = ( + ptyId: string, + stop: () => Promise +) => Promise<{ stopped: boolean; owner: boolean }> + +export async function sweepProviderByPrefix( + worktreeId: string, + provider: IPtyProvider, + deadline: number, + stopPty: ( + ptyId: string, + stop: () => Promise + ) => Promise<{ stopped: boolean; owner: boolean }>, + onPtyStopped?: (ptyId: string) => void, + failClosed = false +): Promise { + const prefix = `${worktreeId}@@` + // Why (#10252): the cwd fallback only proves ownership when the filesystem path + // is the *whole* worktree path. A folder-workspace instance strips its + // `::workspace:` suffix to a checkout dir shared with sibling instances, + // so leave the fallback unset whenever stripping shortened the path — else + // deleting one instance would sweep the others. + const fullWorktreePath = splitWorktreeId(worktreeId)?.worktreePath + const cwdFallbackPath = + splitWorktreeIdForFilesystem(worktreeId)?.worktreePath === fullWorktreePath + ? fullWorktreePath + : undefined + const rpcDeadline = teardownRpcDeadline(deadline) + const sessions = failClosed + ? await provider.listProcesses({ deadlineMs: rpcDeadline }) + : await provider.listProcesses({ deadlineMs: rpcDeadline }).catch(() => []) + const ownedSessions = sessions.filter((session) => { + // Why: older daemon/relay process rows may omit cwd; their established ID + // and authoritative worktree ownership must remain usable during teardown. + const cwdOwned = + cwdFallbackPath !== undefined && + session.worktreeId === undefined && + typeof session.cwd === 'string' && + session.cwd.length > 0 && + isPathInsideOrEqual(cwdFallbackPath, session.cwd) + return session.id.startsWith(prefix) || session.worktreeId === worktreeId || cwdOwned + }) + // Why: agent shutdown snapshots coalesce only when requests begin together; + // bounded concurrency avoids serial process scans without unbounded fanout. + const stopped = await mapWithConcurrency( + ownedSessions, + WORKTREE_TEARDOWN_CONCURRENCY, + async (session) => { + if (Date.now() >= deadline) { + return 0 + } + const stopResult = await stopPty(session.id, async () => { + if (Date.now() >= deadline) { + return false + } + try { + await provider.shutdown(session.id, { immediate: true, deadlineMs: rpcDeadline }) + return Date.now() < deadline + } catch { + return false + } + }) + if (stopResult.owner && Date.now() < deadline) { + clearStoppedPtyState(session.id, onPtyStopped) + return 1 + } + return 0 + } + ) + return stopped.reduce((count, value) => count + value, 0) +} + +export async function sweepRegistryForWorktree( + worktreeId: string, + localProvider: IPtyProvider, + deadline: number, + stopPty: ( + ptyId: string, + stop: () => Promise + ) => Promise<{ stopped: boolean; owner: boolean }>, + onPtyStopped?: (ptyId: string) => void +): Promise { + const rpcDeadline = teardownRpcDeadline(deadline) + const entries = listRegisteredPtys().filter((r) => r.worktreeId === worktreeId) + const stopped = await mapWithConcurrency( + entries, + WORKTREE_TEARDOWN_CONCURRENCY, + async (entry) => { + if (Date.now() >= deadline) { + return 0 + } + const stopResult = await stopPty(entry.ptyId, async () => { + if (Date.now() >= deadline) { + return false + } + try { + await localProvider.shutdown(entry.ptyId, { immediate: true, deadlineMs: rpcDeadline }) + return Date.now() < deadline + } catch { + return false + } + }) + if (stopResult.owner && Date.now() < deadline) { + clearStoppedPtyState(entry.ptyId, onPtyStopped) + return 1 + } + return 0 + } + ) + return stopped.reduce((count, value) => count + value, 0) +} + +export function clearStoppedPtyState(ptyId: string, onPtyStopped?: (ptyId: string) => void): void { + try { + // Why: daemon shutdown does not always fan a local pty:exit event back + // through pty.ts, but removed worktrees must immediately drop memory rows. + onPtyStopped?.(ptyId) + } catch { + /* cleanup is best-effort and must not block git-level removal */ + } +} diff --git a/src/main/runtime/worktree-teardown-deadline.ts b/src/main/runtime/worktree-teardown-deadline.ts new file mode 100644 index 00000000000..8dcbfb4b3dd --- /dev/null +++ b/src/main/runtime/worktree-teardown-deadline.ts @@ -0,0 +1,19 @@ +/** + * The one deadline arithmetic worktree teardown shares. + * + * Its own module because both the teardown entry point and the PTY-surface sweeps need it, and a + * sweep importing the entry point back would be a cycle. + */ + +// Why: keep each bounded stop RPC settling before the sweep deadline itself, so +// a wedged provider surfaces as a stop failure rather than as the outer timeout. +// (The recheck this margin once also reserved time for now runs on its own +// budget — see verifyUnstoppedPtys — because sharing this one wedged #11960.) +export const WORKTREE_TEARDOWN_RPC_MARGIN_MS = 500 + +// Absolute deadline (epoch ms) threaded into provider RPCs on the destructive +// path; each RPC leaf converts it to the remaining time when it actually issues, +// so sequential RPCs share one budget without any relative-timeout bookkeeping. +export function teardownRpcDeadline(sweepDeadline: number): number { + return sweepDeadline - WORKTREE_TEARDOWN_RPC_MARGIN_MS +} diff --git a/src/main/runtime/worktree-teardown.ts b/src/main/runtime/worktree-teardown.ts index dfe3d4ae5a1..82fa055cc3a 100644 --- a/src/main/runtime/worktree-teardown.ts +++ b/src/main/runtime/worktree-teardown.ts @@ -1,15 +1,23 @@ import type { IPtyProvider } from '../providers/types' import type { OrcaRuntimeService } from './orca-runtime' -import { listRegisteredPtys } from '../memory/pty-registry' -import { isPathInsideOrEqual } from '../../shared/cross-platform-path' -import { splitWorktreeId, splitWorktreeIdForFilesystem } from '../../shared/worktree/id' -import { mapWithConcurrency } from '../../shared/map-with-concurrency' import { isUnstoppedPtyRemovalError, + RUNNING_AGENT_SESSION_REMOVAL_PREFIX, + UNSTOPPED_PTY_DETAIL_SEPARATOR, WORKTREE_TEARDOWN_FORCE_HINT, WORKTREE_TEARDOWN_TIMEOUT_PREFIX } from '../../shared/worktree/removal' import { settleBeforeDeadline } from './settle-before-deadline' +import { + clearStoppedPtyState, + sweepProviderByPrefix, + sweepRegistryForWorktree +} from './worktree-pty-surface-sweeps' +import { + closeStructuredSessionsForWorktree, + describeLiveStructuredSessions, + listLiveStructuredSessionsForWorktree +} from './structured-session-worktree-teardown' import { createWorktreeSweepTracker, settleSweepsForForcedRemoval } from './forced-sweep-settlement' import { describeError, @@ -18,10 +26,6 @@ import { resolveUnstoppedPtyVerdict } from './unstopped-pty-verification' -// Why: normal inventories still coalesce into one process scan, while a stale -// or pathological inventory cannot fan out unbounded provider/RPC shutdowns. -const WORKTREE_TEARDOWN_CONCURRENCY = 32 - export type WorktreeTeardownDeps = { runtime?: OrcaRuntimeService /** Authoritative id for callers whose selector no longer resolves (orphaned workspace). */ @@ -38,28 +42,28 @@ export type WorktreeTeardownDeps = { allowUnverifiedStop?: boolean includeProviderInventory?: boolean includeLocalRegistry?: boolean + /** + * Close structured agent sessions best-effort, for a destructive removal that does NOT require + * PTY-stop proof — the folder-workspace paths, which sweep and kill PTYs the same way. + * + * Separate from `requirePhysicalStop` because the two questions are different: that one asks + * whether a stop must be PROVEN before files are touched, and it is what licenses a refusal. + * Reconciliation sweeps set neither; they repair state and must never close anything. + */ + closeStructuredSessions?: boolean } export type WorktreeTeardownResult = { runtimeStopped: number providerStopped: number registryStopped: number + /** Structured agent sessions closed by the force path; absent when none were found. */ + structuredStopped?: number } export const WORKTREE_PROCESS_SWEEP_TIMEOUT_MS = 10_000 -// Why: keep each bounded stop RPC settling before the sweep deadline itself, so -// a wedged provider surfaces as a stop failure rather than as the outer timeout. -// (The recheck this margin once also reserved time for now runs on its own -// budget — see verifyUnstoppedPtys — because sharing this one wedged #11960.) -export const WORKTREE_TEARDOWN_RPC_MARGIN_MS = 500 - -// Absolute deadline (epoch ms) threaded into provider RPCs on the destructive -// path; each RPC leaf converts it to the remaining time when it actually issues, -// so sequential RPCs share one budget without any relative-timeout bookkeeping. -export function teardownRpcDeadline(sweepDeadline: number): number { - return sweepDeadline - WORKTREE_TEARDOWN_RPC_MARGIN_MS -} +export { WORKTREE_TEARDOWN_RPC_MARGIN_MS, teardownRpcDeadline } from './worktree-teardown-deadline' /** * Kills every PTY we can prove belongs to `worktreeId`, across all three @@ -95,6 +99,11 @@ export async function killAllProcessesForWorktree( const deadlineError = new Error( `${WORKTREE_TEARDOWN_TIMEOUT_PREFIX} ${worktreeId}. ${WORKTREE_TEARDOWN_FORCE_HINT}` ) + // FIRST, and before a single PTY sweep starts: a structured agent session is registered on none + // of the three surfaces below, so all three answered zero and removal deleted the checkout out + // from under a running provider child. Refusing costs nothing when there are none, and the check + // is synchronous, so a destructive removal fails fast instead of after the whole sweep budget. + const structuredStopped = await sweepStructuredSessions(worktreeId, deps, deadline, deadlineError) const sweeps = createWorktreeSweepTracker() const stopAttempts = new Map>() const stopPty = ( @@ -245,7 +254,10 @@ export async function killAllProcessesForWorktree( } } else { const summary = describeUnstoppedPtys(worktreeId, failedPtyIds, verdict) - if (!deps.allowUnverifiedStop) { + // Only a proof-requiring removal may refuse. A folder-workspace removal shares its root, so no + // checkout disappears under the child — the harm is a session left pointing at a workspace Orca + // has forgotten — and one of those paths is a never-throw forget, which a refusal would wedge. + if (deps.requirePhysicalStop && !deps.allowUnverifiedStop) { throw new Error(`${summary}. ${WORKTREE_TEARDOWN_FORCE_HINT}`) } // Why: force is the documented escape hatch, so removal continues — but the @@ -256,122 +268,68 @@ export async function killAllProcessesForWorktree( } } - return { runtimeStopped: runtimeResult.stopped, providerStopped, registryStopped } -} - -async function sweepProviderByPrefix( - worktreeId: string, - provider: IPtyProvider, - deadline: number, - stopPty: ( - ptyId: string, - stop: () => Promise - ) => Promise<{ stopped: boolean; owner: boolean }>, - onPtyStopped?: (ptyId: string) => void, - failClosed = false -): Promise { - const prefix = `${worktreeId}@@` - // Why (#10252): the cwd fallback only proves ownership when the filesystem path - // is the *whole* worktree path. A folder-workspace instance strips its - // `::workspace:` suffix to a checkout dir shared with sibling instances, - // so leave the fallback unset whenever stripping shortened the path — else - // deleting one instance would sweep the others. - const fullWorktreePath = splitWorktreeId(worktreeId)?.worktreePath - const cwdFallbackPath = - splitWorktreeIdForFilesystem(worktreeId)?.worktreePath === fullWorktreePath - ? fullWorktreePath - : undefined - const rpcDeadline = teardownRpcDeadline(deadline) - const sessions = failClosed - ? await provider.listProcesses({ deadlineMs: rpcDeadline }) - : await provider.listProcesses({ deadlineMs: rpcDeadline }).catch(() => []) - const ownedSessions = sessions.filter((session) => { - // Why: older daemon/relay process rows may omit cwd; their established ID - // and authoritative worktree ownership must remain usable during teardown. - const cwdOwned = - cwdFallbackPath !== undefined && - session.worktreeId === undefined && - typeof session.cwd === 'string' && - session.cwd.length > 0 && - isPathInsideOrEqual(cwdFallbackPath, session.cwd) - return session.id.startsWith(prefix) || session.worktreeId === worktreeId || cwdOwned - }) - // Why: agent shutdown snapshots coalesce only when requests begin together; - // bounded concurrency avoids serial process scans without unbounded fanout. - const stopped = await mapWithConcurrency( - ownedSessions, - WORKTREE_TEARDOWN_CONCURRENCY, - async (session) => { - if (Date.now() >= deadline) { - return 0 - } - const stopResult = await stopPty(session.id, async () => { - if (Date.now() >= deadline) { - return false - } - try { - await provider.shutdown(session.id, { immediate: true, deadlineMs: rpcDeadline }) - return Date.now() < deadline - } catch { - return false - } - }) - if (stopResult.owner && Date.now() < deadline) { - clearStoppedPtyState(session.id, onPtyStopped) - return 1 - } - return 0 - } - ) - return stopped.reduce((count, value) => count + value, 0) -} - -async function sweepRegistryForWorktree( - worktreeId: string, - localProvider: IPtyProvider, - deadline: number, - stopPty: ( - ptyId: string, - stop: () => Promise - ) => Promise<{ stopped: boolean; owner: boolean }>, - onPtyStopped?: (ptyId: string) => void -): Promise { - const rpcDeadline = teardownRpcDeadline(deadline) - const entries = listRegisteredPtys().filter((r) => r.worktreeId === worktreeId) - const stopped = await mapWithConcurrency( - entries, - WORKTREE_TEARDOWN_CONCURRENCY, - async (entry) => { - if (Date.now() >= deadline) { - return 0 - } - const stopResult = await stopPty(entry.ptyId, async () => { - if (Date.now() >= deadline) { - return false - } - try { - await localProvider.shutdown(entry.ptyId, { immediate: true, deadlineMs: rpcDeadline }) - return Date.now() < deadline - } catch { - return false - } - }) - if (stopResult.owner && Date.now() < deadline) { - clearStoppedPtyState(entry.ptyId, onPtyStopped) - return 1 - } - return 0 - } - ) - return stopped.reduce((count, value) => count + value, 0) -} - -function clearStoppedPtyState(ptyId: string, onPtyStopped?: (ptyId: string) => void): void { - try { - // Why: daemon shutdown does not always fan a local pty:exit event back - // through pty.ts, but removed worktrees must immediately drop memory rows. - onPtyStopped?.(ptyId) - } catch { - /* cleanup is best-effort and must not block git-level removal */ + return { + runtimeStopped: runtimeResult.stopped, + providerStopped, + registryStopped, + ...(structuredStopped > 0 ? { structuredStopped } : {}) } } + +/** + * The fourth sweep: structured agent sessions bound to this worktree. + * + * Refuses rather than auto-closing on the ordinary destructive path. `worktree rm` is the verb + * that deletes a user's work, and a running agent session is exactly the thing they would want to + * be told about before it goes — the same bargain the unstopped-PTY gate already strikes, using + * the same `--force` escape hatch. Force closes them properly instead of orphaning a child against + * a `cwd` that is about to disappear. + * + * Two callers participate, for different reasons. A proof-requiring removal (`requirePhysicalStop`) + * refuses, then closes under force. A folder-workspace removal (`closeStructuredSessions`) closes + * best-effort without refusing: it shares its root so no checkout vanishes under the child, and one + * of those paths is a never-throw forget that a refusal would wedge. Reconciliation sweeps set + * neither — they repair state, delete nothing, and must never close a session. + */ +async function sweepStructuredSessions( + worktreeId: string, + deps: WorktreeTeardownDeps, + deadline: number, + deadlineError: Error +): Promise { + if (!deps.requirePhysicalStop && !deps.closeStructuredSessions) { + return 0 + } + const live = listLiveStructuredSessionsForWorktree(worktreeId) + if (live.length === 0) { + return 0 + } + // Only a proof-requiring removal may refuse. A folder-workspace removal shares its root, so no + // checkout disappears under the child — the harm is a session left pointing at a workspace Orca + // has forgotten — and one of those paths is a never-throw forget, which a refusal would wedge. + if (deps.requirePhysicalStop && !deps.allowUnverifiedStop) { + // The prefix is what the desktop classifier matches on; without it the toast shows raw CLI + // wording and hides the Force Delete button — the #11960 dead end this file already documents. + throw new Error( + `${RUNNING_AGENT_SESSION_REMOVAL_PREFIX} ${worktreeId}${UNSTOPPED_PTY_DETAIL_SEPARATOR}${describeLiveStructuredSessions(live)}. ${WORKTREE_TEARDOWN_FORCE_HINT}` + ) + } + // Raced against the same sweep budget every PTY surface is bounded by: `host.close` awaits a + // provider round trip, and a wedged one would otherwise hang `worktree rm --force` forever with + // no timeout error at all. On expiry the force path reports the timeout exactly as the PTY + // sweeps do rather than proceeding as if the sessions had closed. + const { closed, unstopped } = await settleBeforeDeadline( + () => closeStructuredSessionsForWorktree(worktreeId, deps.runtime), + { closed: 0, unstopped: live }, + deadline, + deadlineError + ) + if (unstopped.length > 0) { + // Force is the documented escape hatch, so removal continues — but say so, because the child + // outliving its `cwd` is the failure this sweep exists to make visible. + console.warn( + `[worktree-teardown] forcing removal of ${worktreeId} with ${describeLiveStructuredSessions(unstopped)} still attached` + ) + } + return closed +} diff --git a/src/main/startup/branch-rename-hook-structured-session.test.ts b/src/main/startup/branch-rename-hook-structured-session.test.ts new file mode 100644 index 00000000000..de85e824db6 --- /dev/null +++ b/src/main/startup/branch-rename-hook-structured-session.test.ts @@ -0,0 +1,98 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import type { AgentSessionStatusSummary } from '../../shared/agent-session-wire' + +// Why the mocks: this file only proves the structured-session seam, and the real orchestrator's +// import graph reaches git, electron, and the agent-hook installers. +const { renameCalls } = vi.hoisted(() => ({ renameCalls: [] as unknown[][] })) +vi.mock('../agent-hooks/first-work-branch-rename', () => ({ + maybeAutoRenameBranchOnFirstWork: (...args: unknown[]) => { + renameCalls.push(args) + return Promise.resolve() + } +})) +vi.mock('../agent-hooks/branch-rename-failure-output', () => ({ + rememberBranchRenameFailureOutput: vi.fn() +})) +vi.mock('../agent-hooks/first-work-folder-rename', () => ({ + renameWorktreeFolderOnFirstWork: vi.fn() +})) +vi.mock('../git/worktree', () => ({ moveWorktree: vi.fn() })) +vi.mock('electron', () => ({ app: { getPath: () => '', on: vi.fn(), isReady: () => true } })) + +import { maybeAutoRenameWorkspaceOnFirstStructuredTurn } from '../agent-hooks/first-work-structured-session-rename' +import { firstWorkRenameDeps } from '../agent-hooks/first-work-rename-runtime' +import { mainProcessState } from './main-process-state' + +let renameDeps: ReturnType + +const WORKSPACE_ID = 'repo1::/repo/wt' + +function summary(overrides: Partial = {}): AgentSessionStatusSummary { + return { + sessionId: 'session-1', + workspaceId: WORKSPACE_ID, + agent: 'claude', + status: 'working', + latestPrompt: 'Fix the auth bug', + updatedAt: 1, + ...overrides + } +} + +beforeEach(() => { + renameCalls.length = 0 + mainProcessState.store = { + getSettings: () => ({}), + getRepo: () => undefined, + getWorktreeMeta: () => undefined, + getWorktreeIdForTab: () => undefined + } as unknown as typeof mainProcessState.store + mainProcessState.runtime = { + getCommitMessageAgentEnvironmentResolvers: () => undefined + } as unknown as typeof mainProcessState.runtime + renameDeps = firstWorkRenameDeps(mainProcessState.store!, mainProcessState.runtime!) +}) + +describe('maybeAutoRenameWorkspaceOnFirstStructuredTurn', () => { + it('drives the first-work rename from the session workspace, with no pane to resolve', () => { + maybeAutoRenameWorkspaceOnFirstStructuredTurn(summary(), { replay: false }, renameDeps) + + expect(renameCalls).toHaveLength(1) + expect(renameCalls[0]?.[0]).toEqual({ + paneKey: '', + tabId: undefined, + worktreeId: WORKSPACE_ID, + state: 'working', + prompt: 'Fix the auth bug', + assistantMessage: undefined, + isReplay: false + }) + }) + + it('marks a re-projected summary as a replay so restore cannot rename on old state', () => { + maybeAutoRenameWorkspaceOnFirstStructuredTurn(summary(), { replay: true }, renameDeps) + + expect(renameCalls[0]?.[0]).toMatchObject({ isReplay: true }) + }) + + it('ignores every status that is not a running turn', () => { + for (const status of ['idle', 'attention', null] as const) { + maybeAutoRenameWorkspaceOnFirstStructuredTurn( + summary({ status }), + { replay: false }, + renameDeps + ) + } + + expect(renameCalls).toEqual([]) + }) + + it('uses the owning runtime even when desktop singletons do not exist', () => { + mainProcessState.store = null + mainProcessState.runtime = null + maybeAutoRenameWorkspaceOnFirstStructuredTurn(summary(), { replay: false }, renameDeps) + + expect(renameCalls).toHaveLength(1) + expect(renameCalls[0]?.[1]).toBe(renameDeps) + }) +}) diff --git a/src/main/startup/branch-rename-hook.ts b/src/main/startup/branch-rename-hook.ts index 578935132fb..e2d5dcfcd19 100644 --- a/src/main/startup/branch-rename-hook.ts +++ b/src/main/startup/branch-rename-hook.ts @@ -1,15 +1,7 @@ -import { existsSync } from 'node:fs' -import { parseWorkspaceKey } from '../../shared/workspace-scope' -import { getRepoIdFromWorktreeId } from '../../shared/worktree/id' import { maybeAutoRenameBranchOnFirstWork } from '../agent-hooks/first-work-branch-rename' -import { rememberBranchRenameFailureOutput } from '../agent-hooks/branch-rename-failure-output' -import { renameWorktreeFolderOnFirstWork } from '../agent-hooks/first-work-folder-rename' -import { moveWorktree } from '../git/worktree' +import { firstWorkRenameDeps } from '../agent-hooks/first-work-rename-runtime' import { mainProcessState as state } from './main-process-state' -// Kill switch for the first-work on-disk folder rename; the renderer reconciles the id change (migrateWorktreeIdentity) so it isn't mistaken for a deletion. -const ENABLE_FIRST_WORK_FOLDER_RENAME = false - // Why: inject the index.ts store/runtime singletons so the rename orchestrator stays module-state-free and unit-testable. export function maybeAutoRenameBranchOnFirstWorkFromHook(event: { paneKey: string @@ -33,103 +25,6 @@ export function maybeAutoRenameBranchOnFirstWorkFromHook(event: { assistantMessage: event.payload.lastAssistantMessage, isReplay: event.isReplay }, - { - getSettings: () => store.getSettings(), - getRepo: (repoId) => store.getRepo(repoId), - getAgentEnvResolvers: () => runtime.getCommitMessageAgentEnvironmentResolvers(), - getCurrentDisplayName: (worktreeId) => { - const scope = parseWorkspaceKey(worktreeId) - return scope?.type === 'folder' - ? store.getFolderWorkspace(scope.folderWorkspaceId)?.name - : store.getWorktreeMeta(worktreeId)?.displayName - }, - getFolderWorkspacePath: (worktreeId) => { - const scope = parseWorkspaceKey(worktreeId) - return scope?.type === 'folder' - ? store.getFolderWorkspace(scope.folderWorkspaceId)?.folderPath - : undefined - }, - isPendingFirstAgentMessageRename: (worktreeId) => { - const scope = parseWorkspaceKey(worktreeId) - return scope?.type === 'folder' - ? store.getFolderWorkspace(scope.folderWorkspaceId)?.pendingFirstAgentMessageRename === - true - : store.getWorktreeMeta(worktreeId)?.pendingFirstAgentMessageRename === true - }, - canRenameOrcaCreatedBranch: (worktreeId) => { - const meta = store.getWorktreeMeta(worktreeId) - // Why: a user branch could coincidentally match a creature name; only Orca-stamped worktrees are safe to auto-rename. - return !!meta?.orcaCreationSource && meta.preserveBranchOnDelete !== true - }, - setDisplayName: (worktreeId, displayName) => { - rememberBranchRenameFailureOutput(worktreeId, null) - const scope = parseWorkspaceKey(worktreeId) - if (scope?.type === 'folder') { - store.updateFolderWorkspace(scope.folderWorkspaceId, { - name: displayName, - pendingFirstAgentMessageRename: false, - firstAgentMessageRenameError: null - }) - runtime.notifyFolderWorkspaceChanged() - return - } - store.setWorktreeMeta(worktreeId, { - displayName, - // The first-agent title is an intentional user-facing label; keep it stable after the - // generated branch is renamed and across subsequent catalog refreshes. - displayNameIsPinned: true, - pendingFirstAgentMessageRename: false, - // Success clears the failure badge (redundant with the explicit setRenameError(null)). - firstAgentMessageRenameError: null - }) - }, - renameWorktreeFolder: ENABLE_FIRST_WORK_FOLDER_RENAME - ? (worktreeId, newLeaf) => - renameWorktreeFolderOnFirstWork(worktreeId, newLeaf, { - getRepo: (repoId) => store.getRepo(repoId), - getSettings: () => store.getSettings(), - migrateWorktreeIdentity: (oldId, newId) => - store.migrateWorktreeIdentity(oldId, newId), - notifyWorktreeRenamed: (repoId, oldId, newId) => - runtime.notifyWorktreeFolderRenamed(repoId, oldId, newId), - pathExists: async (candidate) => existsSync(candidate), - moveWorktree - }) - : undefined, - setRenameError: (worktreeId, error, failureOutput) => { - // Refresh the full-output capture before the dedupe below — a repeat error string is still a fresh run. - rememberBranchRenameFailureOutput(worktreeId, error === null ? null : failureOutput) - // Skip the write + push when unchanged — most settled worktrees never had an error to clear. - const scope = parseWorkspaceKey(worktreeId) - if (scope?.type === 'folder') { - const current = store.getFolderWorkspace( - scope.folderWorkspaceId - )?.firstAgentMessageRenameError - if ((current ?? null) === (error ?? null)) { - return - } - store.updateFolderWorkspace(scope.folderWorkspaceId, { - firstAgentMessageRenameError: error - }) - runtime.notifyFolderWorkspaceChanged() - return - } - const current = store.getWorktreeMeta(worktreeId)?.firstAgentMessageRenameError - if ((current ?? null) === (error ?? null)) { - return - } - store.setWorktreeMeta(worktreeId, { firstAgentMessageRenameError: error }) - // Why: the hook only knows the worktreeId, so derive the repoId notifyBranchRenamed expects. - runtime.notifyBranchRenamed(getRepoIdFromWorktreeId(worktreeId)) - }, - resolveWorktreeIdForTab: (tabId) => store.getWorktreeIdForTab(tabId), - onRenamed: (repoIdOrWorktreeId) => { - if (parseWorkspaceKey(repoIdOrWorktreeId)?.type === 'folder') { - runtime.notifyFolderWorkspaceChanged() - return - } - runtime.notifyBranchRenamed(repoIdOrWorktreeId) - } - } + firstWorkRenameDeps(store, runtime) ) } diff --git a/src/main/startup/main-process-account-services.ts b/src/main/startup/main-process-account-services.ts index c4575c7a0a3..ebfdd73f2f8 100644 --- a/src/main/startup/main-process-account-services.ts +++ b/src/main/startup/main-process-account-services.ts @@ -14,6 +14,7 @@ import { getInitialCodexRateLimitTarget } from '../rate-limits/codex-rate-limit- import { getInitialClaudeRateLimitTarget } from '../rate-limits/claude-rate-limit-target' import { getKimiRuntimeTarget, resolveKimiHome } from '../kimi/kimi-runtime-home' import { readMiniMaxSessionCookie } from '../minimax/minimax-cookie-store' +import { readMiniMaxApiKey } from '../minimax/minimax-api-key-store' import { createAccountRuntimeTargetSettingsSync } from '../rate-limits/account-runtime-target-sync' import { normalizeCodexRuntimeSelection } from '../codex-accounts/runtime-selection' import { normalizeClaudeRuntimeSelection } from '../claude-accounts/runtime-selection' @@ -102,10 +103,13 @@ export function initializeMainProcessAccountServices(): void { }) state.rateLimits.setMiniMaxConfigResolver(() => { const settings = store.getSettings() + const apiKey = readMiniMaxApiKey() ?? '' return { - sessionCookie: readMiniMaxSessionCookie() ?? '', + sessionCookie: apiKey ? '' : (readMiniMaxSessionCookie() ?? ''), groupId: settings.minimaxGroupId, - models: settings.minimaxUsageModels + models: settings.minimaxUsageModels, + endpoint: settings.minimaxEndpoint, + apiKey } }) state.rateLimits.setGeminiCliOAuthEnabledResolver(() => store.getSettings().geminiCliOAuthEnabled) diff --git a/src/main/startup/main-process-push-startup.ts b/src/main/startup/main-process-push-startup.ts new file mode 100644 index 00000000000..6d1b9fda1bd --- /dev/null +++ b/src/main/startup/main-process-push-startup.ts @@ -0,0 +1,32 @@ +import { getOrcaPushGatewayUrl } from '../orca-profiles/profile-cloud-auth-config' +import { DesktopPushService } from '../runtime/push/desktop-push-service' +import type { OrcaRuntimeService } from '../runtime/orca-runtime' +import type { OrcaRuntimeRpcServer } from '../runtime/runtime-rpc' +import { mainProcessState as state } from './main-process-state' + +// Why: deliberately not gated on cloud sign-in like the relay is — the push gateway +// authenticates with the host keypair, so an accountless host registers phones on +// exactly the same path. The runtime is read from shared state because both launch +// modes have already stored it there; threading it as a parameter would push the +// launch module past its line budget for no gain. +export function startDesktopPushService(runtimeRpc: OrcaRuntimeRpcServer): void { + const runtime: OrcaRuntimeService | null = state.runtime + if (!runtime) { + console.warn('[push] Background push startup skipped: runtime not started') + return + } + try { + const pushService = DesktopPushService.create({ + runtime, + runtimeRpc, + gatewayUrl: getOrcaPushGatewayUrl() + }) + pushService?.start() + state.desktopPushService = pushService + } catch (error) { + console.warn( + '[push] Background push startup unavailable:', + error instanceof Error ? error.message : String(error) + ) + } +} diff --git a/src/main/startup/main-process-quit.ts b/src/main/startup/main-process-quit.ts index a4149e13ba7..de580bf7d95 100644 --- a/src/main/startup/main-process-quit.ts +++ b/src/main/startup/main-process-quit.ts @@ -72,6 +72,9 @@ function installBeforeQuitHandler(): void { } state.isQuitting = true state.desktopRelayService?.fenceAndCloseNow() + // Why: drops the notification subscription so a late dispatch cannot start a + // push (and its unref'd outbox retry) on the way out. + state.desktopPushService?.stop() state.runtimeRpc?.setMobileRelayPairingProvider(null) state.unsubscribeAgentAwakeStatusChanges?.() state.unsubscribeAgentAwakeStatusChanges = null diff --git a/src/main/startup/main-process-runtime-launch.ts b/src/main/startup/main-process-runtime-launch.ts index 5f2691d6f31..849ce9061da 100644 --- a/src/main/startup/main-process-runtime-launch.ts +++ b/src/main/startup/main-process-runtime-launch.ts @@ -35,6 +35,7 @@ import { CliInstaller } from '../cli/cli-installer' import { installLinuxBareOrcaDispatcher } from '../cli/linux-bare-orca-dispatcher' import { scheduleAllPendingHistoryTreeRemovals } from '../terminal-history-deletion' import { triggerStartupNotificationRegistration } from '../ipc/startup-notification-registration' +import { startDesktopPushService } from './main-process-push-startup' import { mainProcessState as state } from './main-process-state' import { logStartupMilestone } from './startup-diagnostics' @@ -158,6 +159,9 @@ async function launchServeMode( console.error('[runtime] Failed to start headless RPC transport:', error) throw error }) + // Why: a phone paired to a headless host still registers and unregisters its token; + // it simply never receives a push, because nothing dispatches notifications here. + startDesktopPushService(runtimeRpc) settleDesktopActivation() // Why: every attempt must reach app.quit(); a page beforeunload can veto an earlier signal. registerServeSignalHandlers(process, () => app.quit()) @@ -241,6 +245,9 @@ async function launchDesktopMode( // fetcher until the persisted proxy lands, so this only has to keep the launch phase itself // ordered ahead of the relay — it must not gate the renderer. await state.initialProxyApplicationReady + // Why after the proxy await: the push gateway client is an app-owned fetcher, so it must not + // issue its first request ahead of the persisted proxy. + startDesktopPushService(runtimeRpc) const cloudAuth = getOrcaCloudAuthConfig() if (cloudAuth.configured) { try { diff --git a/src/main/startup/main-process-state.ts b/src/main/startup/main-process-state.ts index c88d5a66c48..a194d36aebd 100644 --- a/src/main/startup/main-process-state.ts +++ b/src/main/startup/main-process-state.ts @@ -13,6 +13,7 @@ import type { OrcaRuntimeService } from '../runtime/orca-runtime' import type { RateLimitService } from '../rate-limits/service' import type { OrcaRuntimeRpcServer } from '../runtime/runtime-rpc' import type { DesktopRelayService } from '../runtime/relay/desktop-relay-service' +import type { DesktopPushService } from '../runtime/push/desktop-push-service' import type { StarNagService } from '../star-nag/service' import type { AgentAwakeService } from '../agent-awake-service' import type { CrashReportStore } from '../crash-reporting/crash-report-store' @@ -65,6 +66,7 @@ export const mainProcessState = { runtimeRpc: null as OrcaRuntimeRpcServer | null, serveReadinessPublisher: new ServeReadinessPublisher(), desktopRelayService: null as DesktopRelayService | null, + desktopPushService: null as DesktopPushService | null, desktopRelayStatus: 'offline' as RelayBrokerStatus, pendingUnpairedDeviceAuthFailure: false, // Why: gates whether headless serve installs the offscreen browser backend (and advertises browser pane support). diff --git a/src/main/windows-descendant-exit-verification.test.ts b/src/main/windows-descendant-exit-verification.test.ts index 392c44399e7..d1944f5ef86 100644 --- a/src/main/windows-descendant-exit-verification.test.ts +++ b/src/main/windows-descendant-exit-verification.test.ts @@ -19,6 +19,101 @@ function snapshot( } describe('captureWindowsDescendantSnapshot', () => { + it('does not claim an older process whose former parent PID was reused by the root', async () => { + const olderProcess = { pid: 50244, ppid: 36084, creationTimeMs: 1788659167395 } + const captured = await captureWindowsDescendantSnapshot(36084, { + readTable: async () => [ + { pid: 36084, ppid: 60976, creationTimeMs: 1788733587893 }, + olderProcess + ] + }) + + expect(captured?.descendants).toEqual([]) + await expect( + verifyWindowsDescendantSnapshotExit(captured!, { readTable: async () => [olderProcess] }) + ).resolves.toBe('exited') + }) + + it('prunes a stale parent link and its subtree at any depth', async () => { + const captured = await captureWindowsDescendantSnapshot(100, { + readTable: async () => [ + { pid: 100, ppid: 1, creationTimeMs: 5 }, + { pid: 200, ppid: 100, creationTimeMs: 10 }, + { pid: 300, ppid: 200, creationTimeMs: 7 }, + { pid: 400, ppid: 300, creationTimeMs: 12 }, + { pid: 500, ppid: 100, creationTimeMs: 4 }, + { pid: 600, ppid: 500, creationTimeMs: 13 }, + { pid: 700, ppid: 200, creationTimeMs: 10 } + ] + }) + + expect(captured?.descendants).toEqual([ + { pid: 700, creationTimeMs: 10 }, + { pid: 200, creationTimeMs: 10 } + ]) + }) + + it('keeps the root when its own parent PID was reused by a newer process', async () => { + // The root's retained ppid now names a process created after it. Pruning the + // root drops the whole snapshot, so its own link is never evidence about it. + const captured = await captureWindowsDescendantSnapshot(100, { + readTable: async () => [ + { pid: 100, ppid: 900, creationTimeMs: 5 }, + { pid: 900, ppid: 1, creationTimeMs: 50 }, + { pid: 200, ppid: 100, creationTimeMs: 7 } + ], + now: () => 42 + }) + + expect(captured).toEqual({ + root: { pid: 100, creationTimeMs: 5 }, + descendants: [{ pid: 200, creationTimeMs: 7 }], + unidentifiedCount: 0, + capturedAtMs: 42 + }) + }) + + it('bounds a link by the root when the claimed parent denied its creation time', async () => { + // 300 has no creation time for a child to be compared against, so the root's + // start is the only bound left: 350 ties with it, which a same-millisecond + // spawn does routinely, while 360 predates the whole tree. + const captured = await captureWindowsDescendantSnapshot(100, { + readTable: async () => [ + { pid: 100, ppid: 1, creationTimeMs: 5 }, + { pid: 300, ppid: 100 }, + { pid: 350, ppid: 300, creationTimeMs: 5 }, + { pid: 360, ppid: 300, creationTimeMs: 2 } + ], + now: () => 42 + }) + + expect(captured).toEqual({ + root: { pid: 100, creationTimeMs: 5 }, + descendants: [{ pid: 350, creationTimeMs: 5 }], + unidentifiedCount: 1, + capturedAtMs: 42 + }) + }) + + it('drops an unidentified row whose parent link was pruned', async () => { + // 250 denied its creation time, but 200's claim on the root is impossible, so + // 250 was never in this tree: counting it would cap the verdict at + // unverifiable over a process the root does not own. + const captured = await captureWindowsDescendantSnapshot(100, { + readTable: async () => [ + { pid: 100, ppid: 1, creationTimeMs: 10 }, + { pid: 200, ppid: 100, creationTimeMs: 5 }, + { pid: 250, ppid: 200 } + ] + }) + + expect(captured?.descendants).toEqual([]) + expect(captured?.unidentifiedCount).toBe(0) + await expect( + verifyWindowsDescendantSnapshotExit(captured!, { readTable: async () => [] }) + ).resolves.toBe('exited') + }) + it('walks the whole subtree and keeps only rows a later read can re-identify', async () => { const captured = await captureWindowsDescendantSnapshot(100, { // 400 is a grandchild; 300 denied a creation-time query, so no later read diff --git a/src/main/windows-descendant-exit-verification.ts b/src/main/windows-descendant-exit-verification.ts index 079833a2bd6..5e365a57e5a 100644 --- a/src/main/windows-descendant-exit-verification.ts +++ b/src/main/windows-descendant-exit-verification.ts @@ -1,3 +1,4 @@ +import { getProcessTableIndex } from '../shared/process-table-index' import type { DescendantTreeVerdict } from './pty-descendant-exit-verification' import { windowsDescendantsFromRows } from './providers/windows-foreground-process-rows' import { readWindowsProcessTableFresh } from './windows/windows-process-table' @@ -57,6 +58,9 @@ function delay(ms: number): Promise { * Snapshot a Windows root's descendants while it is still alive. Resolves null * (never rejects) when the table is unreadable or the root is absent — the same * contract as the POSIX walk, because "cannot see" is never "nothing is there". + * + * Stale parent links are pruned by creation time, so a backwards clock step + * between two spawns can drop a live descendant — accepted over a certain stall. */ export async function captureWindowsDescendantSnapshot( rootPid: number, @@ -69,9 +73,35 @@ export async function captureWindowsDescendantSnapshot( // One table read, not a walk plus an identity read: each is bounded in // seconds, and this runs inside the close ladder's budget. const table = await (deps.readTable ?? readWindowsProcessTableFresh)().catch(() => null) - const descendants = table && windowsDescendantsFromRows(table, rootPid) - const root = table?.find((row) => row.pid === rootPid) - if (!descendants || typeof root?.creationTimeMs !== 'number') { + if (!table) { + return null + } + // One index for both lookups, so a repeated pid resolves to the same row for + // the root and for a parent link: `byPid` is first-wins, a Map is not. + const rowsByPid = getProcessTableIndex(table).byPid + const root = rowsByPid.get(rootPid) + if (typeof root?.creationTimeMs !== 'number') { + return null + } + const rootCreationTimeMs = root.creationTimeMs + // Windows keeps a process's original parent PID after that parent exits, so a + // reused PID is not ancestry: no real child predates the parent it claims. + // The root's start backstops the undefined-time bypass, which admits a row + // unchecked and leaves its children no parent time to compare against. Ties + // pass -- FILETIMEs truncated to ms make a same-millisecond parent and child + // collide exactly, so `>` would drop true descendants. + const currentRows = table.filter((row) => { + const parentCreationTimeMs = rowsByPid.get(row.ppid)?.creationTimeMs + return ( + // Its own ppid can be recycled too, and a pruned root loses the snapshot. + row.pid === rootPid || + row.creationTimeMs === undefined || + (row.creationTimeMs >= rootCreationTimeMs && + (parentCreationTimeMs === undefined || row.creationTimeMs >= parentCreationTimeMs)) + ) + }) + const descendants = windowsDescendantsFromRows(currentRows, rootPid) + if (!descendants) { return null } return { diff --git a/src/main/windows/windows-process-table-native-addon.win32.test.ts b/src/main/windows/windows-process-table-native-addon.win32.test.ts new file mode 100644 index 00000000000..1120d8173b4 --- /dev/null +++ b/src/main/windows/windows-process-table-native-addon.win32.test.ts @@ -0,0 +1,26 @@ +import { expect, it } from 'vitest' +import { + isWindowsProcessStartTimeAvailable, + readWindowsProcessTableFresh +} from './windows-process-table' + +it.runIf(process.platform === 'win32')( + 'reads creation times from the real Windows process-tree addon', + async () => { + expect(isWindowsProcessStartTimeAvailable()).toBe(true) + + const rows = await readWindowsProcessTableFresh() + const rowsWithCreationTime = rows.filter((row) => typeof row.creationTimeMs === 'number').length + expect(rowsWithCreationTime).toBeGreaterThan(0) + + // Why our own row and not merely a count: a single stray row satisfies a + // count, and an addon that forwards the raw FILETIME satisfies it too. We + // opened our own handle, so this row is the one the addon can never fail to + // answer, and its value is bounded on both sides -- a 1601-epoch stamp lands + // below the floor, an unconverted 100ns tick lands astronomically above now. + const self = rows.find((row) => row.pid === process.pid) + expect(typeof self?.creationTimeMs).toBe('number') + expect(self?.creationTimeMs).toBeGreaterThan(Date.parse('2020-01-01T00:00:00Z')) + expect(self?.creationTimeMs).toBeLessThanOrEqual(Date.now()) + } +) diff --git a/src/main/windows/windows-process-table.test.ts b/src/main/windows/windows-process-table.test.ts index f4fd514d319..16c411ceb71 100644 --- a/src/main/windows/windows-process-table.test.ts +++ b/src/main/windows/windows-process-table.test.ts @@ -149,6 +149,7 @@ describe('windows process table', () => { Object.defineProperty(process, 'platform', { configurable: true, value: 'win32' }) __setWindowsProcessTreeLoaderForTests(() => ({ ProcessDataFlag: { None: 0, Memory: 1, CommandLine: 2, CreationTime: 4 }, + supportedProcessDataFlags: 7, getAllProcesses })) }) @@ -327,10 +328,23 @@ describe('windows process table', () => { vi.useRealTimers() }) - it('only advertises PID-safe ownership when the native creation-time field exists', () => { + it('only advertises PID-safe ownership when the BINARY reports creation-time support', () => { expect(isWindowsProcessStartTimeAvailable()).toBe(true) + + // The shape CI produced: pnpm patched the source tree, so the enum carries + // CreationTime, while the tarball's prebuilt .node still ignores flag 4. + // Believing the enum here is what let structured chat run with a reaper + // that can never identify a PID. __setWindowsProcessTreeLoaderForTests(() => ({ - ProcessDataFlag: { None: 0, Memory: 1, CommandLine: 2 }, + ProcessDataFlag: { None: 0, Memory: 1, CommandLine: 2, CreationTime: 4 }, + supportedProcessDataFlags: 3, + getAllProcesses + })) + expect(isWindowsProcessStartTimeAvailable()).toBe(false) + + // An addon predating the export at all reports nothing, which is also false. + __setWindowsProcessTreeLoaderForTests(() => ({ + ProcessDataFlag: { None: 0, Memory: 1, CommandLine: 2, CreationTime: 4 }, getAllProcesses })) expect(isWindowsProcessStartTimeAvailable()).toBe(false) @@ -654,12 +668,21 @@ describe('resolving the native reader', () => { } }) - function addonReturning(rows: unknown): { getProcessList: ReturnType } { + function addonReturning(rows: unknown): { + getProcessList: ReturnType + supportedProcessDataFlags: number + } { return { - getProcessList: vi.fn((cb: (r: unknown) => void) => cb(rows)) + getProcessList: vi.fn((cb: (r: unknown) => void) => cb(rows)), + supportedProcessDataFlags: 7 } } + /** An addon built before the creation-time patch: no capability export at all. */ + function staleAddonReturning(rows: unknown): { getProcessList: ReturnType } { + return { getProcessList: vi.fn((cb: (r: unknown) => void) => cb(rows)) } + } + it('prefers the npm package where the desktop app installs it', async () => { const resolve = vi.fn((specifier: string) => { if (specifier === PACKAGE_SPECIFIER) { @@ -695,7 +718,7 @@ describe('resolving the native reader', () => { expect(isWindowsProcessTableAvailable()).toBe(true) }) - it('asks the addon for the command line, as the package path does', async () => { + it('asks the addon for the command line and creation time, as the package path does', async () => { const addon = addonReturning(NATIVE) __setWindowsProcessTreeRequireForTests((specifier: string) => { if (specifier === ADDON_SPECIFIER) { @@ -704,13 +727,15 @@ describe('resolving the native reader', () => { throw new Error('MODULE_NOT_FOUND') }) await readWindowsProcessTableFresh() - // CommandLine alone: a bare snapshot would silently drop the command line - // every agent-recognition caller matches on first, and Memory would add a - // second per-process handle nothing reads. - expect(addon.getProcessList).toHaveBeenCalledWith(expect.any(Function), 2) + // Same flag set as the package path (6). Dropping CreationTime would strand + // the relay's own teardown on bare pids: every Windows descendant identity + // is a pid plus a creation time, so a table without one can never prove a + // tree exited. Memory stays off -- a second per-process handle nothing reads. + expect(addon.getProcessList).toHaveBeenCalledWith(expect.any(Function), 6) + expect(isWindowsProcessStartTimeAvailable()).toBe(true) }) - it('asks the addon for nothing per-process on the identity path', async () => { + it('asks the addon for the creation time alone on the identity path', async () => { const addon = addonReturning(NATIVE) __setWindowsProcessTreeRequireForTests((specifier: string) => { if (specifier === ADDON_SPECIFIER) { @@ -719,9 +744,25 @@ describe('resolving the native reader', () => { throw new Error('MODULE_NOT_FOUND') }) await readWindowsProcessIdentityTableFresh() - // The relay addon exposes no CreationTime bit, so this is a bare Toolhelp32 - // walk: zero OpenProcess calls. - expect(addon.getProcessList).toHaveBeenCalledWith(expect.any(Function), 0) + // CreationTime (4) and nothing else: no CommandLine, so the only per-process + // handle is the PROCESS_QUERY_LIMITED_INFORMATION one GetProcessTimes needs. + expect(addon.getProcessList).toHaveBeenCalledWith(expect.any(Function), 4) + }) + + it('trusts the staged addon on its own report, not on ours', async () => { + // A relay carrying an addon built before the creation-time patch still + // enumerates, so the table stays usable -- but it cannot prove identity, + // and saying otherwise would hand teardown a PID it can never re-check. + const addon = staleAddonReturning(NATIVE) + __setWindowsProcessTreeRequireForTests((specifier: string) => { + if (specifier === ADDON_SPECIFIER) { + return addon + } + throw new Error('MODULE_NOT_FOUND') + }) + await expect(readWindowsProcessTableFresh()).resolves.toHaveLength(2) + expect(isWindowsProcessTableAvailable()).toBe(true) + expect(isWindowsProcessStartTimeAvailable()).toBe(false) }) it('reaches the CIM scan when neither the package nor the addon is present', async () => { diff --git a/src/main/windows/windows-process-table.ts b/src/main/windows/windows-process-table.ts index 0a1acd7ae1c..e3064560174 100644 --- a/src/main/windows/windows-process-table.ts +++ b/src/main/windows/windows-process-table.ts @@ -33,8 +33,13 @@ import { readWindowsProcessRowsWithCim } from './windows-process-table-cim-scan' * * Dropping Memory removed the second per-process handle: it took an * OpenProcess(...|VM_READ) it never read through. CommandLine's own read is no - * longer a PEB walk either -- the patched addon asks the kernel, so identity is - * now the only flag set that opens nothing at all. + * longer a PEB walk either -- the patched addon asks the kernel. + * + * Both Toolhelp32 rows predate `CreationTime`, which both flag sets now also + * ask for and which is unmeasured here: it costs one + * OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION) plus GetProcessTimes per + * process, so identity no longer opens nothing at all -- but that pair is far + * cheaper than either handle the rows above measure. * * All Toolhelp32 rows assume the optional `windows-process-tree.node` addon. * The desktop bundles it; no released relay carries it, so on an SSH host the @@ -70,6 +75,13 @@ type WindowsProcessTreeModule = { CommandLine: number CreationTime?: number } + /** + * Flag bits the COMPILED addon reports, straight from `addon.cc`. Absent on a + * build that predates the patch — which is not the same question as the enum + * above, because pnpm patches the source tree and leaves the tarball's + * prebuilt `.node` in place. + */ + supportedProcessDataFlags?: number getAllProcesses: ( callback: (processes: NativeProcessInfo[] | undefined) => void, flags?: number @@ -104,14 +116,18 @@ type WindowsProcessTreeAddon = { callback: (processes: NativeProcessInfo[] | undefined) => void, flags: number ) => void + supportedProcessDataFlags?: number } /** * Mirrors the package's enum; the addon takes the raw bit field. `Memory` (1) * is listed for completeness and is deliberately never set — see the projections * below. + * + * Naming `CreationTime` here only decides what we ASK for; whether the binary + * answers is `supportedProcessDataFlags`, which the addon reports itself. */ -const PROCESS_DATA_FLAG = { None: 0, Memory: 1, CommandLine: 2 } as const +const PROCESS_DATA_FLAG = { None: 0, Memory: 1, CommandLine: 2, CreationTime: 4 } as const /** Staged beside the relay bundle by build-relay; see RELAY_ARTIFACTS. */ const RELAY_ADDON_FILENAME = './windows-process-tree.node' @@ -160,6 +176,7 @@ let cimScan: () => Promise = readWindowsProcessRowsWithCim function adaptAddon(addon: WindowsProcessTreeAddon): WindowsProcessTreeModule { return { ProcessDataFlag: PROCESS_DATA_FLAG, + supportedProcessDataFlags: addon.supportedProcessDataFlags, getAllProcesses: (callback, flags) => addon.getProcessList(callback, flags ?? 0) } } @@ -492,13 +509,23 @@ export function isWindowsProcessTableAvailable(): boolean { /** * PID-reuse-safe ownership needs the native creation-time field, not merely a - * process list. Older addon builds expose the table without that field; keep - * structured ownership unavailable on those hosts instead of fabricating proof - * from a PID. + * process list. + * + * Why the binary's own answer and not the enum: pnpm patches the package's + * source tree but leaves the tarball's prebuilt `.node` at the same + * `build/Release/` path, so a host can hold a patched `lib/index.js` — enum and + * all — over a binary that ignores flag 4. CI produced exactly that: the enum + * said available, and every row came back without `creationTimeMs`. Answering + * true there is worse than answering false: the descendant snapshot then + * returns null forever and the exit proof latches `unverifiable`, while + * structured chat believes it has a reaper. */ export function isWindowsProcessStartTimeAvailable(): boolean { const native = moduleLoader() - return native !== null && typeof native.ProcessDataFlag.CreationTime === 'number' + return ( + native !== null && + ((native.supportedProcessDataFlags ?? 0) & PROCESS_DATA_FLAG.CreationTime) !== 0 + ) } function resetSnapshotReaders(): void { diff --git a/src/main/windows/windows-process-tree-command-line-patch.test.ts b/src/main/windows/windows-process-tree-command-line-patch.test.ts index 1eaa4459c9e..051ca85786e 100644 --- a/src/main/windows/windows-process-tree-command-line-patch.test.ts +++ b/src/main/windows/windows-process-tree-command-line-patch.test.ts @@ -94,7 +94,9 @@ describe('windows-process-tree command line patch', () => { expect(source).not.toMatch(/ReadProcessMemory\(/) } // Memory and CPU counters kept VM_READ and never read an address space. - expect(processSource.match(/OpenProcess\(PROCESS_QUERY_LIMITED_INFORMATION/g)).toHaveLength(2) + // Three sites now: those two plus GetProcessCreationTime, which needs the + // same limited handle for GetProcessTimes. + expect(processSource.match(/OpenProcess\(PROCESS_QUERY_LIMITED_INFORMATION/g)).toHaveLength(3) }) it('value-initializes ProcessInfo so memory is not stack garbage', () => { diff --git a/src/main/wsl-availability-missing-kernel.test.ts b/src/main/wsl-availability-missing-kernel.test.ts new file mode 100644 index 00000000000..3ab7b6d4a26 --- /dev/null +++ b/src/main/wsl-availability-missing-kernel.test.ts @@ -0,0 +1,98 @@ +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import { execFile, execFileSync } from 'node:child_process' +import { runProcess, runProcessSync, type ProcessResult } from '../shared/child-process/run-process' +import { + _resetWslAvailabilityCacheForTests, + isWslAvailable, + isWslAvailableAsync +} from './wsl-availability' + +vi.mock('node:child_process', () => ({ execFile: vi.fn(), execFileSync: vi.fn() })) +vi.mock('../shared/child-process/run-process', () => ({ + runProcess: vi.fn(), + runProcessSync: vi.fn() +})) +vi.mock('./wsl-interop-spawn-directory', () => ({ + resolveWslInteropSpawnCwd: () => 'C:\\Windows' +})) + +const originalPlatform = process.platform +const success: ProcessResult = { code: 0, signal: null, stdout: '', stderr: '', timedOut: false } + +beforeEach(() => { + vi.resetAllMocks() + Object.defineProperty(process, 'platform', { value: 'win32' }) + _resetWslAvailabilityCacheForTests() +}) +afterEach(() => { + Object.defineProperty(process, 'platform', { value: originalPlatform }) + _resetWslAvailabilityCacheForTests() +}) + +for (const mode of ['sync', 'async'] as const) { + describe(`${mode} WSL1 availability without WSL2 kernel`, () => { + const probe = () => (mode === 'sync' ? isWslAvailable() : isWslAvailableAsync()) + const guestRunner = () => (mode === 'sync' ? runProcessSync : runProcess) + + function failStatus(code: number): void { + vi.mocked(execFileSync).mockImplementation(() => { + throw { status: code } + }) + vi.mocked(execFile).mockImplementation((...args: unknown[]) => { + const callback = args.at(-1) as (error: unknown) => void + callback({ code }) + return {} as ReturnType + }) + } + function guestResult(result: ProcessResult): void { + vi.mocked(runProcess).mockResolvedValue(result) + vi.mocked(runProcessSync).mockReturnValue(result) + } + + // Node reports the Windows DWORD; the console prints its signed equivalent. + for (const status of [-444, 4_294_966_852]) { + it(`requires guest execution and caches its success for ${status}`, async () => { + failStatus(status) + guestResult(success) + expect(await probe()).toBe(true) + expect(await probe()).toBe(true) + expect(guestRunner()).toHaveBeenCalledTimes(1) + expect(guestRunner()).toHaveBeenCalledWith( + expect.objectContaining({ + program: 'wsl.exe', + args: ['--exec', '/bin/true'], + timeoutMs: 5000, + cwd: 'C:\\Windows' + }) + ) + }) + } + + for (const result of [ + { ...success, code: 1 }, + { ...success, code: null, timedOut: true } + ]) { + it(`keeps a failed guest unavailable: ${JSON.stringify(result)}`, async () => { + failStatus(-444) + guestResult(result) + expect(await probe()).toBe(false) + }) + } + + it('stays unavailable when the guest probe cannot be spawned', async () => { + failStatus(-444) + vi.mocked(runProcess).mockRejectedValue(new Error('EPERM')) + vi.mocked(runProcessSync).mockImplementation(() => { + throw new Error('EPERM') + }) + expect(await probe()).toBe(false) + }) + + it('does not probe a guest for unrelated status failures', async () => { + failStatus(1) + expect(await probe()).toBe(false) + expect(runProcess).not.toHaveBeenCalled() + expect(runProcessSync).not.toHaveBeenCalled() + }) + }) +} diff --git a/src/main/wsl-availability.ts b/src/main/wsl-availability.ts index 1d14f526413..ad5e5645c30 100644 --- a/src/main/wsl-availability.ts +++ b/src/main/wsl-availability.ts @@ -1,4 +1,6 @@ import { execFile, execFileSync } from 'node:child_process' +import { runProcess, runProcessSync, type ProcessSpec } from '../shared/child-process/run-process' +import { buildWslExecArgs } from '../shared/wsl-login-shell-command' import { resolveWslInteropSpawnCwd } from './wsl-interop-spawn-directory' type WslAvailabilityCache = @@ -94,6 +96,55 @@ function cacheWslAvailabilityProbeResult(error: unknown, startedAtGeneration: nu return !error } +// `wsl --status` exits 0x1bc when the WSL2 kernel package is missing -- a package +// a WSL1 distro never needed. Node keeps the Windows DWORD; the console prints the +// signed form, and either spelling can reach us. +function isMissingWsl2KernelStatus(error: unknown): boolean { + const failure = error as { status?: unknown; code?: unknown } | null + return [failure?.status, failure?.code].some((code) => code === -444 || code === 4_294_966_852) +} + +// Cheapest proof the default guest runs: no login shell, no output to parse. +function defaultGuestExecutionProbe(): ProcessSpec { + return { + program: 'wsl.exe', + args: buildWslExecArgs(undefined, ['/bin/true']), + cwd: resolveWslInteropSpawnCwd(), + timeoutMs: WSL_AVAILABILITY_PROBE_TIMEOUT_MS, + maxOutputBytes: 4096 + } +} + +/** + * The `--status` error still worth caching, or null once the guest ran anyway. + * + * Why it returns that error rather than a fresh negative: a guest probe that could not + * spawn means "could not ask", and minting an answer for that is the bug this subsystem + * keeps re-shipping (docs/reference/wsl-probe-failure-semantics.md). + */ +function wslStatusErrorAfterGuestProbe(error: unknown): unknown { + if (!isMissingWsl2KernelStatus(error)) { + return error + } + try { + return runProcessSync(defaultGuestExecutionProbe()).code === 0 ? null : error + } catch { + return error + } +} + +/** Async twin of `wslStatusErrorAfterGuestProbe`; the sync/async pair share one cache. */ +async function wslStatusErrorAfterGuestProbeAsync(error: unknown): Promise { + if (!isMissingWsl2KernelStatus(error)) { + return error + } + try { + return (await runProcess(defaultGuestExecutionProbe())).code === 0 ? null : error + } catch { + return error + } +} + function probeWslStatus(): Promise { return new Promise((resolve, reject) => { execFile( @@ -147,7 +198,10 @@ export function isWslAvailable(): boolean { }) return cacheWslAvailabilityProbeResult(null, startedAtGeneration) } catch (error) { - return cacheWslAvailabilityProbeResult(error, startedAtGeneration) + return cacheWslAvailabilityProbeResult( + wslStatusErrorAfterGuestProbe(error), + startedAtGeneration + ) } } @@ -176,7 +230,12 @@ export function isWslAvailableAsync(): Promise { const startedAtGeneration = wslAvailabilityCacheGeneration wslAvailabilityProbeInFlight = probeWslStatus() .then(() => cacheWslAvailabilityProbeResult(null, startedAtGeneration)) - .catch((error: unknown) => cacheWslAvailabilityProbeResult(error, startedAtGeneration)) + .catch(async (error: unknown) => + cacheWslAvailabilityProbeResult( + await wslStatusErrorAfterGuestProbeAsync(error), + startedAtGeneration + ) + ) .finally(() => { wslAvailabilityProbeInFlight = null }) diff --git a/src/preload/api/agent-account-api.ts b/src/preload/api/agent-account-api.ts index ae75cbf1c35..ff98244299d 100644 --- a/src/preload/api/agent-account-api.ts +++ b/src/preload/api/agent-account-api.ts @@ -59,9 +59,18 @@ export type GrokAccountsApi = { } export type MinimaxCredentialsApi = { - getStatus: () => Promise<{ configured: boolean }> - saveCookie: (cookie: string) => Promise<{ configured: boolean }> - clearCookie: () => Promise<{ configured: boolean }> + // Why: cookie + API key each live in their own safeStorage file, so the + // status separates them. 'configured' stays as the OR so existing callers + // that only care about "anything saved" keep working unchanged. + getStatus: () => Promise<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }> + saveCookie: (cookie: string) => Promise<{ cookieConfigured: boolean }> + clearCookie: () => Promise<{ cookieConfigured: boolean }> + saveApiKey: (key: string) => Promise<{ apiKeyConfigured: boolean }> + clearApiKey: () => Promise<{ apiKeyConfigured: boolean }> } export type CodexConfigSyncApi = { diff --git a/src/preload/api/minimax-credentials-bridge.ts b/src/preload/api/minimax-credentials-bridge.ts index e99bd843909..f49e32d42ec 100644 --- a/src/preload/api/minimax-credentials-bridge.ts +++ b/src/preload/api/minimax-credentials-bridge.ts @@ -2,10 +2,17 @@ import { ipcRenderer } from 'electron' import type { PreloadApi } from '../api-types' export const minimaxCredentialsApi = { - getStatus: (): Promise<{ configured: boolean }> => - ipcRenderer.invoke('minimaxCredentials:getStatus'), - saveCookie: (cookie: string): Promise<{ configured: boolean }> => + getStatus: (): Promise<{ + configured: boolean + cookieConfigured: boolean + apiKeyConfigured: boolean + }> => ipcRenderer.invoke('minimaxCredentials:getStatus'), + saveCookie: (cookie: string): Promise<{ cookieConfigured: boolean }> => ipcRenderer.invoke('minimaxCredentials:saveCookie', cookie), - clearCookie: (): Promise<{ configured: boolean }> => - ipcRenderer.invoke('minimaxCredentials:clearCookie') + clearCookie: (): Promise<{ cookieConfigured: boolean }> => + ipcRenderer.invoke('minimaxCredentials:clearCookie'), + saveApiKey: (key: string): Promise<{ apiKeyConfigured: boolean }> => + ipcRenderer.invoke('minimaxCredentials:saveApiKey', key), + clearApiKey: (): Promise<{ apiKeyConfigured: boolean }> => + ipcRenderer.invoke('minimaxCredentials:clearApiKey') } satisfies PreloadApi['minimaxCredentials'] diff --git a/src/relay/relay-pty-source-activation.ts b/src/relay/relay-pty-source-activation.ts index d4ab4e06726..3898dda6df1 100644 --- a/src/relay/relay-pty-source-activation.ts +++ b/src/relay/relay-pty-source-activation.ts @@ -4,7 +4,10 @@ import type { PtySourceRecoveryResult } from '../shared/pty-source-recovery-contract' import type { PtySourceReceivingActivation } from '../shared/pty-source-receiving-activation' -import type { PtySourceDeliveryIdentity } from '../shared/pty-source-credit-contract' +import type { + PtySourceDeliveryIdentity, + PtySourceDeliverySnapshot +} from '../shared/pty-source-credit-contract' import type { RequestContext } from './dispatcher' import type { RelayPtySourceDeliveryRecord, @@ -12,6 +15,16 @@ import type { } from './relay-pty-source-send-scheduler' import type { SshPtyConsumerSessionAdapter } from './ssh-pty-consumer-session-adapter' +// Takes the post-rotation snapshot: creditedEndSu is the accepted checkpoint and windowSu the +// reconnecting client's window. The pre-rotation snapshot would fence below the checkpoint. +export function boundedPtyRecoveryEnd( + snapshot: Pick +): number { + const { receivedEndSu, creditedEndSu, windowSu } = snapshot + // Oversized quarantine cannot earn credit; fence at the checkpoint and drain it live. + return receivedEndSu - creditedEndSu > windowSu ? creditedEndSu : receivedEndSu +} + export function createPtySourceReceivingActivation( identity: PtySourceDeliveryIdentity, checkpointSourceEndSu: number, diff --git a/src/relay/relay-pty-source-publication.ts b/src/relay/relay-pty-source-publication.ts index 16b6e81c61b..c1959fd24d1 100644 --- a/src/relay/relay-pty-source-publication.ts +++ b/src/relay/relay-pty-source-publication.ts @@ -7,6 +7,7 @@ import type { PtySourceReceivingActivation } from '../shared/pty-source-receivin import { createPtySourceReceivingActivation, pendingPtySourceRecoveryResult, + boundedPtyRecoveryEnd, registerCanceledPtySourceRetirement, registerPtySourceActivationSettlement, samePtySourceRecoveryRequest @@ -66,9 +67,7 @@ export class RelayPtySourcePublication { recovery?: PtySourceRecoveryRequest ): false | 'opened' | 'rotated' | 'existing' | PtySourceRecoveryResult { let current = this.deliveries.get(id) - // A superseded request can find the delivery its own replacement opened: releasing that fence - // resumes a send the replacement is still rotating, and cancelling it blanks the pane that owns - // it. So every bail-out below acts only on a record this caller still owns. + // Only release this caller's delivery; its replacement may still be rotating. const owned = current?.clientId === context?.clientId ? current : undefined if (!context?.onResponseSettled) { this.sender.releaseRotationFence(owned) @@ -141,7 +140,7 @@ export class RelayPtySourcePublication { identity = rotation.identity displayEnd = current.displayEnd recoveryCheckpointSourceEndSu = recovery.acceptedSourceEndSu - recoveryEndSu = snapshot.receivedEndSu + recoveryEndSu = boundedPtyRecoveryEnd(this.session.sourceDeliverySnapshot(identity)) recoveryWasSealed = snapshot.state === 'sealed-unsettled' this.counters.rotated++ } catch (error) { diff --git a/src/relay/relay-pty-source-recovery-window.test.ts b/src/relay/relay-pty-source-recovery-window.test.ts new file mode 100644 index 00000000000..6f3f8186178 --- /dev/null +++ b/src/relay/relay-pty-source-recovery-window.test.ts @@ -0,0 +1,181 @@ +import { afterEach, expect, it } from 'vitest' +import { RelayDispatcher, type RelayClientSessionIdentity } from './dispatcher' +import { boundedPtyRecoveryEnd } from './relay-pty-source-activation' +import { encodeJsonRpcFrame, MessageType } from './protocol' +import { RelayPtySourcePublication } from './relay-pty-source-publication' +import { SshPtyConsumerSessionAdapter } from './ssh-pty-consumer-session-adapter' + +const endpointIdentity: RelayClientSessionIdentity = { + principal: 'endpoint-principal', + authenticated: true, + allowSessionOwner: true, + authenticationKind: 'endpoint-credential' +} + +type Frame = { + id?: number + method?: string + params?: Record + result?: Record +} + +function decode(buffer: Buffer): Frame | null { + return buffer[0] === MessageType.Regular + ? JSON.parse(buffer.subarray(13, 13 + buffer.readUInt32BE(9)).toString('utf8')) + : null +} + +const flushRequests = (): Promise => new Promise((resolve) => setImmediate(resolve)) +let dispatcher: RelayDispatcher | undefined + +afterEach(() => dispatcher?.dispose()) + +it.each([0, 4])( + 'drains a retained tail larger than the window from checkpoint %i', + async (checkpoint) => { + const original: Frame[] = [] + dispatcher = new RelayDispatcher( + (data, settled) => { + const frame = decode(data) + if (frame) { + original.push(frame) + } + settled({ ok: true }) + return true + }, + { supportsWriteCallback: true }, + endpointIdentity + ) + const mux = dispatcher + let publication: RelayPtySourcePublication + const adapter = new SshPtyConsumerSessionAdapter(mux, 'build', undefined, (id) => + publication.onCreditAvailable(id) + ) + publication = new RelayPtySourcePublication(mux, adapter, () => {}) + const open = (clientId: number, id: number, resume?: Record): void => { + mux.feedClient( + clientId, + encodeJsonRpcFrame( + { + jsonrpc: '2.0', + id, + method: 'pty.openClient', + params: { + protocolVersion: 1, + clientInstanceId: 'client', + requestedRole: 'session-owner', + resume, + capabilities: { outputFlowControl: { versions: [1], requestedWindowSu: 4 } } + } + }, + id, + 0 + ) + ) + } + open(1, 1) + await flushRequests() + publication.activate('pty', 'incarnation', { + clientId: 1, + isStale: () => false, + sessionIdentity: endpointIdentity, + onResponseSettled: (settle) => queueMicrotask(() => settle({ ok: true })) + }) + await flushRequests() + expect(publication.publish('pty', { data: 'abcdefghijkl' }, false)).toBe(true) + const oldFrame = original.find((frame) => frame.method === 'pty.data')!.params! + const grant = original.find((frame) => frame.id === 1)!.result! + mux.invalidateClient() + const replacement: Frame[] = [] + const clientId = mux.attachClient( + (data, settled) => { + const frame = decode(data) + if (frame) { + replacement.push(frame) + } + settled({ ok: true }) + return true + }, + { supportsWriteCallback: true }, + endpointIdentity + ) + open(clientId, 2, { ownerGeneration: grant.ownerGeneration, ownerLease: grant.ownerLease }) + await flushRequests() + const recovery = publication.activate( + 'pty', + 'incarnation', + { + clientId, + isStale: () => false, + sessionIdentity: endpointIdentity, + onResponseSettled: (settle) => queueMicrotask(() => settle({ ok: true })) + }, + { + status: 'checkpoint', + clientGeneration: Number(oldFrame.clientGeneration), + ownerGeneration: Number(oldFrame.ownerGeneration), + deliveryToken: String(oldFrame.deliveryToken), + ptyIncarnation: 'incarnation', + acceptedSourceEndSu: checkpoint + } + ) + // The fence lands on the checkpoint itself: the tail drains live rather than behind it. + expect(recovery).toMatchObject({ + status: 'pending', + checkpointSourceEndSu: checkpoint, + recoveryEndSu: checkpoint + }) + await flushRequests() + // The receiver cannot ACK quarantined data until this fence arrives. + expect(replacement.filter((frame) => frame.method === 'pty.recoveryComplete')).toHaveLength(1) + let accepted = checkpoint + let output = '' + for (let turn = 0; accepted < 12 && turn < 4; turn++) { + const frames = replacement.filter( + (frame) => frame.method === 'pty.data' && Number(frame.params!.sourceEndSu) > accepted + ) + expect(frames.length).toBeGreaterThan(0) + for (const frame of frames) { + const params = frame.params! + expect(Number(params.sourceEndSu) - Number(params.sourceLengthSu)).toBe(accepted) + accepted = Number(params.sourceEndSu) + output += String(params.data) + } + expect(publication.getDebugSnapshot().outstandingSourceUnits).toBeLessThanOrEqual(4) + const params = frames.at(-1)!.params! + mux.feedClient( + clientId, + encodeJsonRpcFrame( + { + jsonrpc: '2.0', + method: 'pty.ackData', + params: { + acknowledgements: [ + { + id: 'pty', + clientGeneration: params.clientGeneration, + ownerGeneration: params.ownerGeneration, + deliveryToken: params.deliveryToken, + creditedEndSu: accepted + } + ] + } + }, + 3 + turn, + 0 + ) + ) + await flushRequests() + } + expect(accepted).toBe(12) + expect(output).toBe('abcdefghijkl'.slice(checkpoint)) + expect(publication.getDebugSnapshot().outstandingSourceUnits).toBe(0) + } +) + +it('fences at the checkpoint only once the tail outgrows the window', () => { + const tail = (receivedEndSu: number) => ({ receivedEndSu, creditedEndSu: 4, windowSu: 4 }) + // Exactly one window is still deliverable without credit, so it keeps the ordinary fence. + expect(boundedPtyRecoveryEnd(tail(8))).toBe(8) + expect(boundedPtyRecoveryEnd(tail(9))).toBe(4) +}) diff --git a/src/renderer/src/components/WorktreeJumpPalette.linear-url.test.tsx b/src/renderer/src/components/WorktreeJumpPalette.linear-url.test.tsx index 0637cad8d42..ff7d4198f92 100644 --- a/src/renderer/src/components/WorktreeJumpPalette.linear-url.test.tsx +++ b/src/renderer/src/components/WorktreeJumpPalette.linear-url.test.tsx @@ -13,6 +13,7 @@ import type { Repo } from '../../../shared/repo-types' import { projectHostSetupProjectionFromRepos } from '../../../shared/project-host-setup-projection' import { resolveWorkspaceCreationTarget } from '@/lib/project-host-workspace-target' import { WORKTREE_PALETTE_QUERY_MAX_BYTES } from '@/lib/worktree-palette-query-bounds' +import { encodePaletteIdentity } from '@/lib/palette-match/palette-ranking' import WorktreeJumpPalette from './WorktreeJumpPalette' import { makeRecentTabState, makeRepo, makeWorktree } from './worktree-jump-palette-test-fixtures' @@ -637,10 +638,10 @@ describe('WorktreeJumpPalette Linear URL intent', () => { await flushEffects() expect(getRenderedRowIds().filter(Boolean)).toEqual([ - 'worktree:wt-linked', + encodePaletteIdentity(['worktree', '|wt-linked']), '__create_worktree__' ]) - expect(getCommandValue()).toBe('worktree:wt-linked') + expect(getCommandValue()).toBe(encodePaletteIdentity(['worktree', '|wt-linked'])) expect(testContainer.querySelector('[data-cmd-j-linear-issue-preview="true"]')).not.toBeNull() }) @@ -731,10 +732,10 @@ describe('WorktreeJumpPalette Linear URL intent', () => { await flushEffects() expect(getRenderedRowIds().filter(Boolean)).toEqual([ - 'worktree:wt-linked', + encodePaletteIdentity(['worktree', '|wt-linked']), '__create_worktree__' ]) - expect(getCommandValue()).toBe('worktree:wt-linked') + expect(getCommandValue()).toBe(encodePaletteIdentity(['worktree', '|wt-linked'])) expect( testContainer.querySelector('[data-cmd-j-task-url-preview="true"]')?.dataset .cmdJTaskUrlProvider diff --git a/src/renderer/src/components/WorktreeJumpPalette.recent-tabs.behavior.test.tsx b/src/renderer/src/components/WorktreeJumpPalette.recent-tabs.behavior.test.tsx index a83b4b63201..b5650c72000 100644 --- a/src/renderer/src/components/WorktreeJumpPalette.recent-tabs.behavior.test.tsx +++ b/src/renderer/src/components/WorktreeJumpPalette.recent-tabs.behavior.test.tsx @@ -8,6 +8,7 @@ import { useAppStore } from '@/store' import type { AppState } from '@/store/types' import { emitCmdJRowIndexJump } from '@/lib/cmd-j-row-index-jump' import WorktreeJumpPalette from './WorktreeJumpPalette' +import { encodePaletteIdentity } from '@/lib/palette-match/palette-ranking' import { makePaneKey } from '../../../shared/stable-pane-id' import { LEAF_ID, @@ -177,9 +178,20 @@ function getRenderedRowIds(): string[] { } function getTabRowIds(): string[] { - return [...testContainer.querySelectorAll('[data-command-item^="workspace-tab:"]')] + return [ + ...testContainer.querySelectorAll( + `[data-command-item^="${encodePaletteIdentity(['workspace-tab'])}"]` + ) + ] .map((node) => node.dataset.commandItem ?? '') - .map((id) => id.replace('workspace-tab:', '')) + .map( + (id) => + Object.values(useAppStore.getState().unifiedTabsByWorktree) + .flat() + .find( + (tab) => encodePaletteIdentity(['workspace-tab', '', tab.worktreeId, tab.id]) === id + )?.id ?? '' + ) } describe('WorktreeJumpPalette recent chats & terminals', () => { @@ -351,7 +363,20 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { it('activates the row a digit chord addresses while open', async () => { await renderPalette( makeRecentTabState({ - lastVisitedAtByWorktreeId: { 'wt-beta': Date.now() } + unifiedTabsByWorktree: { + 'wt-alpha': [ + { + ...makeUnifiedTab('tab-alpha', 'wt-alpha', 'term-alpha', 'Alpha chat'), + lastFocusedAt: Date.now() - 3 * 86400_000 + } + ], + 'wt-beta': [ + { + ...makeUnifiedTab('tab-beta', 'wt-beta', 'term-beta', 'Beta chat'), + lastFocusedAt: Date.now() + } + ] + } }) ) @@ -417,7 +442,7 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { expect(getTabRowIds()).toContain('tab-alpha') expect(getTabRowIds()).not.toContain('tab-beta') const alphaRow = testContainer.querySelector( - '[data-command-item="workspace-tab:tab-alpha"]' + `[data-command-item="${encodePaletteIdentity(['workspace-tab', '', 'wt-alpha', 'tab-alpha'])}"]` ) expect(alphaRow?.querySelector('[data-slot=tooltip-trigger]')?.textContent).toContain('Working') }) diff --git a/src/renderer/src/components/WorktreeJumpPalette.recent-tabs.test.tsx b/src/renderer/src/components/WorktreeJumpPalette.recent-tabs.test.tsx index 52cdec1c683..0571cead58f 100644 --- a/src/renderer/src/components/WorktreeJumpPalette.recent-tabs.test.tsx +++ b/src/renderer/src/components/WorktreeJumpPalette.recent-tabs.test.tsx @@ -8,6 +8,7 @@ import { useAppStore } from '@/store' import type { AppState } from '@/store/types' import { emitCmdJRowIndexJump } from '@/lib/cmd-j-row-index-jump' import WorktreeJumpPalette from './WorktreeJumpPalette' +import { encodePaletteIdentity } from '@/lib/palette-match/palette-ranking' import { makePaneKey } from '../../../shared/stable-pane-id' import { LEAF_ID, @@ -176,9 +177,11 @@ async function renderPalette(overrides: Partial): Promise { } function getWorktreeRows(): string[] { - return [...testContainer.querySelectorAll('[data-command-item^="worktree:"]')].map( - (node) => node.textContent ?? '' - ) + return [ + ...testContainer.querySelectorAll( + `[data-command-item^="${encodePaletteIdentity(['worktree'])}"]` + ) + ].map((node) => node.textContent ?? '') } function getRenderedRowIds(): string[] { @@ -195,13 +198,26 @@ function getCommandValue(): string { } function getTabRowIds(): string[] { - return [...testContainer.querySelectorAll('[data-command-item^="workspace-tab:"]')] + return [ + ...testContainer.querySelectorAll( + `[data-command-item^="${encodePaletteIdentity(['workspace-tab'])}"]` + ) + ] .map((node) => node.dataset.commandItem ?? '') - .map((id) => id.replace('workspace-tab:', '')) + .map( + (id) => + Object.values(useAppStore.getState().unifiedTabsByWorktree) + .flat() + .find( + (tab) => encodePaletteIdentity(['workspace-tab', '', tab.worktreeId, tab.id]) === id + )?.id ?? '' + ) } function getTabRowShortcutDigits(): string[] { return [ - ...testContainer.querySelectorAll('[data-command-item^="workspace-tab:"]') + ...testContainer.querySelectorAll( + `[data-command-item^="${encodePaletteIdentity(['workspace-tab'])}"]` + ) ].flatMap((row) => [...row.querySelectorAll('span')] .map((node) => node.textContent ?? '') @@ -238,8 +254,8 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { await renderPalette(makeRecentTabState()) const rows = getRenderedRowIds().filter((id) => id.length > 0) - expect(rows[0]).toMatch(/^workspace-tab:/) - expect(rows.some((id) => id.startsWith('worktree:'))).toBe(true) + expect(rows[0].startsWith(encodePaletteIdentity(['workspace-tab']))).toBe(true) + expect(rows.some((id) => id.startsWith(encodePaletteIdentity(['worktree'])))).toBe(true) expect(testContainer.textContent).toContain('Recent Chats & Terminals') expect(testContainer.textContent).toContain('Recent Worktrees') }) @@ -248,10 +264,11 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { await renderPalette(makeDuplicateRecentTabState()) expect( - getRenderedRowIds().filter( - (id) => id === 'workspace-tab:tab-duplicate' || id.includes(':workspace-tab:tab-duplicate') - ) - ).toEqual(['workspace-tab:tab-duplicate', 'palette-dup:1:workspace-tab:tab-duplicate']) + getRenderedRowIds().filter((id) => id.startsWith(encodePaletteIdentity(['workspace-tab']))) + ).toEqual([ + encodePaletteIdentity(['workspace-tab', 'ssh:alpha', 'wt-alpha', 'tab-duplicate']), + encodePaletteIdentity(['workspace-tab', 'ssh:beta', 'wt-beta', 'tab-duplicate']) + ]) await act(async () => { emitCmdJRowIndexJump(1) @@ -358,9 +375,11 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { await flushEffects() const rows = getRenderedRowIds().filter((id) => id.length > 0) - expect(rows[0]).toBe('workspace-tab:tab-host') - expect(rows).toContain('worktree:wt-weak') - expect(getCommandValue()).toBe('workspace-tab:tab-host') + expect(rows[0]).toBe(encodePaletteIdentity(['workspace-tab', '', 'wt-host', 'tab-host'])) + expect(rows).toContain(encodePaletteIdentity(['worktree', '|wt-weak'])) + expect(getCommandValue()).toBe( + encodePaletteIdentity(['workspace-tab', '', 'wt-host', 'tab-host']) + ) }) it('selects the new first result when cmdk reports the deferred list selection', async () => { @@ -370,16 +389,20 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { setCommandQuery?.('improve') }) await flushEffects() - expect(getCommandValue()).toBe('worktree:wt-weak') + expect(getCommandValue()).toBe(encodePaletteIdentity(['worktree', '|wt-weak'])) await act(async () => { setCommandQuery?.('perf') - setCommandSelection?.('worktree:wt-weak') + setCommandSelection?.(encodePaletteIdentity(['worktree', '|wt-weak'])) }) await flushEffects() - expect(getRenderedRowIds().find((id) => id.length > 0)).toBe('workspace-tab:tab-host') - expect(getCommandValue()).toBe('workspace-tab:tab-host') + expect(getRenderedRowIds().find((id) => id.length > 0)).toBe( + encodePaletteIdentity(['workspace-tab', '', 'wt-host', 'tab-host']) + ) + expect(getCommandValue()).toBe( + encodePaletteIdentity(['workspace-tab', '', 'wt-host', 'tab-host']) + ) }) // Why: after typing, arrow moves must stick. Dropping onValueChange while cmdk already @@ -391,7 +414,9 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { setCommandQuery?.('perf') }) await flushEffects() - expect(getCommandValue()).toBe('workspace-tab:tab-host') + expect(getCommandValue()).toBe( + encodePaletteIdentity(['workspace-tab', '', 'wt-host', 'tab-host']) + ) const rows = getRenderedRowIds().filter((id) => id.length > 0) expect(rows.length).toBeGreaterThan(1) @@ -421,7 +446,7 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { await flushEffects() const firstRow = getRenderedRowIds().find((id) => id.length > 0) - expect(firstRow).toBe('worktree:wt-strong') + expect(firstRow).toBe(encodePaletteIdentity(['worktree', '|wt-strong'])) }) it('ranks a typed query by match position inside the worktree section', async () => { @@ -445,10 +470,12 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { // Why word-b beats word-a despite input order: `perf` is a whole word in // `rc-perf-update-channels` but only a prefix of `performance`. - expect(getRenderedRowIds().filter((id) => id.startsWith('worktree:'))).toEqual([ - 'worktree:wt-prefix', - 'worktree:wt-word-b', - 'worktree:wt-word-a' + expect( + getRenderedRowIds().filter((id) => id.startsWith(encodePaletteIdentity(['worktree']))) + ).toEqual([ + encodePaletteIdentity(['worktree', '|wt-prefix']), + encodePaletteIdentity(['worktree', '|wt-word-b']), + encodePaletteIdentity(['worktree', '|wt-word-a']) ]) }) @@ -479,7 +506,9 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { expect(getTabRowIds()).toEqual([]) // Why: cmdk claims the first row it sees, which before hydration is a worktree. - const firstWorktreeId = getRenderedRowIds().find((id) => id.startsWith('worktree:')) + const firstWorktreeId = getRenderedRowIds().find((id) => + id.startsWith(encodePaletteIdentity(['worktree'])) + ) expect(firstWorktreeId).toBeDefined() await act(async () => { setCommandSelection?.(firstWorktreeId ?? '') @@ -497,7 +526,9 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { const [topRowId] = getTabRowIds() expect(getTabRowIds()).toHaveLength(2) // Enter has to follow the rows up: ⌘1 already points at the first recent chat. - expect(getCommandValue()).toBe(`workspace-tab:${topRowId}`) + expect(getCommandValue()).toBe( + getRenderedRowIds().find((id) => id.startsWith(encodePaletteIdentity(['workspace-tab']))) + ) // Why here: an empty snapshot also left the digit chords addressing nothing until reopen. await act(async () => { @@ -518,7 +549,9 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { unifiedTabsByWorktree: {} }) - const worktreeIds = getRenderedRowIds().filter((id) => id.startsWith('worktree:')) + const worktreeIds = getRenderedRowIds().filter((id) => + id.startsWith(encodePaletteIdentity(['worktree'])) + ) expect(worktreeIds.length).toBeGreaterThan(1) // Why the second row: only a selection that differs from the auto-picked head proves the user moved it. const movedTo = worktreeIds[1] @@ -539,18 +572,31 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { expect(getCommandValue()).toBe(movedTo) }) - it('re-ranks once when terminal entities hydrate after unified tabs', async () => { - // Why split hydration: unified tabs can land before tabsByWorktree; without a re-capture every - // row ranks IDLE. A deliberate second-row highlight must survive that one re-rank. + it('preserves visit ordering and selection when terminal entities hydrate', async () => { const hydrated = makeRecentTabState({ agentStatusByPaneKey: { [makePaneKey('term-alpha', LEAF_ID)]: makeAgentEntry('term-alpha', 'blocked', Date.now()) }, - lastVisitedAtByWorktreeId: { 'wt-beta': Date.now() } + unifiedTabsByWorktree: { + 'wt-alpha': [ + { + ...makeUnifiedTab('tab-alpha', 'wt-alpha', 'term-alpha', 'Alpha chat'), + lastFocusedAt: Date.now() - 3 * 86400_000 + } + ], + 'wt-beta': [ + { + ...makeUnifiedTab('tab-beta', 'wt-beta', 'term-beta', 'Beta chat'), + lastFocusedAt: Date.now() + } + ] + } }) await renderPalette({ ...hydrated, tabsByWorktree: {} }) expect(getTabRowIds()).toEqual(['tab-beta', 'tab-alpha']) - const movedTo = `workspace-tab:${getTabRowIds()[1]}` + const movedTo = getRenderedRowIds().filter((id) => + id.startsWith(encodePaletteIdentity(['workspace-tab'])) + )[1] await act(async () => { setCommandSelection?.(movedTo) }) @@ -559,7 +605,7 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { useAppStore.setState({ tabsByWorktree: hydrated.tabsByWorktree } as Partial) }) await flushEffects() - expect(getTabRowIds()).toEqual(['tab-alpha', 'tab-beta']) + expect(getTabRowIds()).toEqual(['tab-beta', 'tab-alpha']) expect(getCommandValue()).toBe(movedTo) }) @@ -587,23 +633,49 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { expect(getTabRowIds()).toEqual(['tab-alpha', 'tab-beta']) }) - it('ranks a blocked agent above a more recently visited idle tab', async () => { + it('ranks a recently visited idle tab above a three-day-old blocked tab', async () => { await renderPalette( makeRecentTabState({ agentStatusByPaneKey: { [makePaneKey('term-alpha', LEAF_ID)]: makeAgentEntry('term-alpha', 'blocked', Date.now()) }, - lastVisitedAtByWorktreeId: { 'wt-beta': Date.now() } + unifiedTabsByWorktree: { + 'wt-alpha': [ + { + ...makeUnifiedTab('tab-alpha', 'wt-alpha', 'term-alpha', 'Alpha chat'), + lastFocusedAt: Date.now() - 3 * 86400_000 + } + ], + 'wt-beta': [ + { + ...makeUnifiedTab('tab-beta', 'wt-beta', 'term-beta', 'Beta chat'), + lastFocusedAt: Date.now() + } + ] + } }) ) - expect(getTabRowIds()).toEqual(['tab-alpha', 'tab-beta']) + expect(getTabRowIds()).toEqual(['tab-beta', 'tab-alpha']) }) it('freezes the order captured on open while statuses keep changing', async () => { await renderPalette( makeRecentTabState({ - lastVisitedAtByWorktreeId: { 'wt-beta': Date.now() } + unifiedTabsByWorktree: { + 'wt-alpha': [ + { + ...makeUnifiedTab('tab-alpha', 'wt-alpha', 'term-alpha', 'Alpha chat'), + lastFocusedAt: Date.now() - 3 * 86400_000 + } + ], + 'wt-beta': [ + { + ...makeUnifiedTab('tab-beta', 'wt-beta', 'term-beta', 'Beta chat'), + lastFocusedAt: Date.now() + } + ] + } }) ) @@ -670,13 +742,26 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { agentStatusByPaneKey: { [makePaneKey('term-alpha', LEAF_ID)]: makeAgentEntry('term-alpha', 'blocked', Date.now()) }, - lastVisitedAtByWorktreeId: { 'wt-beta': Date.now() } + unifiedTabsByWorktree: { + 'wt-alpha': [ + { + ...makeUnifiedTab('tab-alpha', 'wt-alpha', 'term-alpha', 'Alpha chat'), + lastFocusedAt: Date.now() - 3 * 86400_000 + } + ], + 'wt-beta': [ + { + ...makeUnifiedTab('tab-beta', 'wt-beta', 'term-beta', 'Beta chat'), + lastFocusedAt: Date.now() + } + ] + } }) ) // Why: high-signal current tabs stay scannable (ask-question / permission badge) even though // idle "where you are" rows are still dropped. - expect(getTabRowIds()).toEqual(['tab-alpha', 'tab-beta']) + expect(getTabRowIds()).toEqual(['tab-beta', 'tab-alpha']) expect(testContainer.textContent).toContain('Current Tab') }) @@ -762,4 +847,30 @@ describe('WorktreeJumpPalette recent chats & terminals', () => { expect(getTabRowIds()).toContain('tab-alpha') }) + + it('keeps the recent section under a sidebar-seeded repository filter and refreshes on clear', async () => { + const secondRepo = { ...makeRepo(), id: 'repo-2', path: '/repos/repo-2', displayName: 'Repo 2' } + await renderPalette({ + ...makeRecentTabState(), + repos: [makeRepo(), secondRepo], + worktreesByRepo: { + 'repo-1': [makeWorktree('wt-alpha', 'Alpha workspace')], + 'repo-2': [makeWorktree('wt-beta', 'Beta workspace', { repoId: 'repo-2' })] + }, + filterRepoIds: ['repo-1'] + }) + + // The seeded filter narrows the recent rows instead of dropping the section. + expect(testContainer.textContent).toContain('Recent Chats & Terminals') + expect(getTabRowIds()).toEqual(['tab-alpha']) + + await act(async () => { + ;[...testContainer.querySelectorAll('button')] + .find((button) => button.textContent?.includes('Clear all')) + ?.click() + }) + await flushEffects() + + expect(getTabRowIds()).toEqual(expect.arrayContaining(['tab-alpha', 'tab-beta'])) + }) }) diff --git a/src/renderer/src/components/WorktreeJumpPalette.test.tsx b/src/renderer/src/components/WorktreeJumpPalette.test.tsx index e9c560dfc4b..0881dd10d78 100644 --- a/src/renderer/src/components/WorktreeJumpPalette.test.tsx +++ b/src/renderer/src/components/WorktreeJumpPalette.test.tsx @@ -7,6 +7,7 @@ import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' import type * as ReactI18Next from 'react-i18next' import { useAppStore } from '@/store' import type { AppState } from '@/store/types' +import { encodePaletteIdentity } from '@/lib/palette-match/palette-ranking' import WorktreeJumpPalette from './WorktreeJumpPalette' import { makeRepo, makeWorktree } from './worktree-jump-palette-test-fixtures' @@ -181,9 +182,11 @@ async function renderPalette(overrides: Partial): Promise { } function getWorktreeRows(): string[] { - return [...testContainer.querySelectorAll('[data-command-item*="worktree:"]')].map( - (node) => node.textContent ?? '' - ) + return [ + ...testContainer.querySelectorAll( + `[data-command-item^="${encodePaletteIdentity(['worktree'])}"]` + ) + ].map((node) => node.textContent ?? '') } describe('WorktreeJumpPalette', () => { @@ -361,6 +364,37 @@ describe('WorktreeJumpPalette', () => { expect(testContainer.textContent).toContain('Feature workspace') }) + it('reseeds the repository filter when reopened during the close linger', async () => { + const secondRepo = { + ...makeRepo(), + id: 'repo-2', + path: '/repos/repo-2', + displayName: 'Repo 2' + } + const first = makeWorktree('first', 'First repository workspace') + const second = makeWorktree('second', 'Second repository workspace', { repoId: 'repo-2' }) + + await renderPalette({ + repos: [makeRepo(), secondRepo], + worktreesByRepo: { 'repo-1': [first], 'repo-2': [second] }, + filterRepoIds: ['repo-1'], + showSleepingWorkspaces: true + }) + + expect(testContainer.textContent).toContain('First repository workspace') + expect(testContainer.textContent).not.toContain('Second repository workspace') + + await act(async () => { + useAppStore.setState({ activeModal: 'none', filterRepoIds: ['repo-2'] }) + }) + await flushEffects() + await act(async () => useAppStore.getState().openModal('worktree-palette')) + await flushEffects() + + expect(testContainer.textContent).not.toContain('First repository workspace') + expect(testContainer.textContent).toContain('Second repository workspace') + }) + // STA-4343 closed: two workspaces sharing `repoId::path` across hosts are two distinct // rows. The documents map and worktreeMap are keyed by host identity, so each row resolves // to its OWN worktree, and render keys keep the two apart for React and cmdk. @@ -374,15 +408,14 @@ describe('WorktreeJumpPalette', () => { await renderPalette(state) - // Both rows render; the second carries a disambiguated command value so the two never - // share a React key. + // Host-qualified command values keep both rows independently selectable. const rows = testContainer.querySelectorAll( - '[data-command-item$="worktree:shared"]' + `[data-command-item^="${encodePaletteIdentity(['worktree'])}"]` ) expect(rows).toHaveLength(2) expect([...rows].map((candidate) => candidate.getAttribute('data-command-item'))).toEqual([ - 'worktree:shared', - 'palette-dup:1:worktree:shared' + encodePaletteIdentity(['worktree', 'local|shared']), + encodePaletteIdentity(['worktree', 'ssh:box|shared']) ]) // The first row names ITS OWN host — the wrong-host open is gone. @@ -404,7 +437,7 @@ describe('WorktreeJumpPalette', () => { }) const rows = testContainer.querySelectorAll( - '[data-command-item$="worktree:shared"]' + `[data-command-item^="${encodePaletteIdentity(['worktree'])}"]` ) expect(rows).toHaveLength(2) @@ -414,16 +447,50 @@ describe('WorktreeJumpPalette', () => { }) }) - it('keeps a lone host-qualified row on its clean command value', async () => { + it('keeps the host in a lone row command value', async () => { const ssh = makeWorktree('single', 'SSH workspace', { hostId: 'ssh:box' }) await renderPalette({ worktreesByRepo: { 'repo-1': [ssh] }, showSleepingWorkspaces: true }) expect( - testContainer.querySelector('[data-command-item="worktree:single"]')?.textContent + testContainer.querySelector( + `[data-command-item="${encodePaletteIdentity(['worktree', 'ssh:box|single'])}"]` + )?.textContent ).toContain('SSH workspace') }) + it('routes same-target SSH rows through their paired runtime owner', async () => { + const hubA = makeWorktree('shared-runtime', 'Hub A workspace', { + hostId: 'ssh:same-private-target', + runtimeOwnerEnvironmentId: 'hub-a' + }) + const hubB = makeWorktree('shared-runtime', 'Hub B workspace', { + hostId: 'ssh:same-private-target', + runtimeOwnerEnvironmentId: 'hub-b' + }) + + await renderPalette({ + worktreesByRepo: { 'repo-1': [hubA, hubB] }, + showSleepingWorkspaces: true + }) + await act(async () => setCommandQuery?.('workspace')) + await flushEffects() + + const hubARow = testContainer.querySelector( + `[data-command-item="${encodePaletteIdentity(['worktree', 'runtime:hub-a|shared-runtime'])}"]` + ) + const hubBRow = testContainer.querySelector( + `[data-command-item="${encodePaletteIdentity(['worktree', 'runtime:hub-b|shared-runtime'])}"]` + ) + expect(hubARow).not.toBeNull() + expect(hubBRow).not.toBeNull() + + await act(async () => fireEvent.click(hubARow!)) + expect(activateAndRevealWorktree).toHaveBeenLastCalledWith('shared-runtime', { + executionHostId: 'runtime:hub-a' + }) + }) + it('does not badge a runtime-owned row with its physical SSH repo', async () => { const worktree = makeWorktree('runtime-repo', 'Runtime workspace', { hostId: 'ssh:box', @@ -436,7 +503,9 @@ describe('WorktreeJumpPalette', () => { showSleepingWorkspaces: true }) - const row = testContainer.querySelector('[data-command-item="worktree:runtime-repo"]') + const row = testContainer.querySelector( + `[data-command-item="${encodePaletteIdentity(['worktree', 'runtime:missing-runtime|runtime-repo'])}"]` + ) expect(row?.textContent).toContain('Runtime workspace') expect(row?.textContent).not.toContain('Physical SSH repo') }) @@ -467,14 +536,16 @@ describe('WorktreeJumpPalette', () => { showSleepingWorkspaces: true }) - const activeRow = testContainer.querySelector('[data-command-item="worktree:active-wt"]') + const activeRow = testContainer.querySelector( + `[data-command-item="${encodePaletteIdentity(['worktree', '|active-wt'])}"]` + ) expect(activeRow?.textContent).toContain('23d') const activeSpan = activeRow?.querySelector('span[aria-label="Last active 23d ago"]') expect(activeSpan).not.toBeNull() expect(activeSpan?.textContent).toBe('23d') const noActivityRow = testContainer.querySelector( - '[data-command-item="worktree:no-activity-wt"]' + `[data-command-item="${encodePaletteIdentity(['worktree', '|no-activity-wt'])}"]` ) expect(noActivityRow?.querySelector('span[aria-label*="Last active"]')).toBeNull() }) diff --git a/src/renderer/src/components/browser-pane/assemble-chrome/BrowserAddressBar.tsx b/src/renderer/src/components/browser-pane/assemble-chrome/BrowserAddressBar.tsx index a91d9cb7744..18a75ef85cd 100644 --- a/src/renderer/src/components/browser-pane/assemble-chrome/BrowserAddressBar.tsx +++ b/src/renderer/src/components/browser-pane/assemble-chrome/BrowserAddressBar.tsx @@ -53,6 +53,7 @@ export default function BrowserAddressBar({ const browserDefaultSearchEngine = useAppStore((s) => s.browserDefaultSearchEngine) const browserKagiSessionLink = useAppStore((s) => s.browserKagiSessionLink) const closingRef = useRef(false) + const initialMouseDownRef = useRef(false) const openedAtRef = useRef(0) const blurCloseTimerRef = useRef(null) const closingResetTimerRef = useRef(null) @@ -241,12 +242,15 @@ export default function BrowserAddressBar({ window.clearTimeout(blurCloseTimerRef.current) blurCloseTimerRef.current = null } - inputRef.current?.select() + if (!initialMouseDownRef.current) { + inputRef.current?.select() + } openedAtRef.current = Date.now() setOpen(true) }, [inputRef]) const handleBlur = useCallback(() => { + initialMouseDownRef.current = false // Why: delay close so that clicking a suggestion item registers before // the popover unmounts. Without this, onSelect never fires because the // mousedown on PopoverContent triggers input blur first. @@ -424,6 +428,18 @@ export default function BrowserAddressBar({ ref={inputRef} value={value} onFocus={handleFocus} + onMouseDown={(event) => { + initialMouseDownRef.current = + event.button === 0 && document.activeElement !== event.currentTarget + }} + onClick={(event) => { + const input = event.currentTarget + // Preserve native drag selection; only expand a collapsed initial click. + if (initialMouseDownRef.current && input.selectionStart === input.selectionEnd) { + input.select() + } + initialMouseDownRef.current = false + }} onBlur={handleBlur} onKeyDown={handleKeyDown} data-orca-browser-address-bar="true" diff --git a/src/renderer/src/components/browser-pane/host-guest/browser-guest-page-id-identity.test.ts b/src/renderer/src/components/browser-pane/host-guest/browser-guest-page-id-identity.test.ts new file mode 100644 index 00000000000..386be8b8220 --- /dev/null +++ b/src/renderer/src/components/browser-pane/host-guest/browser-guest-page-id-identity.test.ts @@ -0,0 +1,30 @@ +import { describe, expect, it } from 'vitest' +import { collectBrowserPageIds } from './browser-guest-paint-retention' + +describe('collectBrowserPageIds identity', () => { + it('returns one shared reference for every empty input', () => { + // useWorktreeBrowserPageIds runs this on every store write, and a worktree with + // no browser tabs is the common case; a fresh [] there is pure allocation. + const fromUndefined = collectBrowserPageIds(undefined) + + expect(collectBrowserPageIds(null)).toBe(fromUndefined) + expect(collectBrowserPageIds([])).toBe(fromUndefined) + expect(fromUndefined).toEqual([]) + }) + + it('still collects page ids, preferring pageIds over the active page', () => { + const ids = collectBrowserPageIds([ + { id: 'tab-1', pageIds: ['page-a', 'page-b'] }, + { id: 'tab-2', activePageId: 'page-c' }, + { id: 'tab-3' } + ]) + + expect(ids).toEqual(['page-a', 'page-b', 'page-c', 'tab-3']) + }) + + it('falls back to the active page when pageIds is present but empty', () => { + expect(collectBrowserPageIds([{ id: 'tab-1', pageIds: [], activePageId: 'page-a' }])).toEqual([ + 'page-a' + ]) + }) +}) diff --git a/src/renderer/src/components/browser-pane/host-guest/browser-guest-paint-retention.ts b/src/renderer/src/components/browser-pane/host-guest/browser-guest-paint-retention.ts index 5e15f002722..0d1364c139b 100644 --- a/src/renderer/src/components/browser-pane/host-guest/browser-guest-paint-retention.ts +++ b/src/renderer/src/components/browser-pane/host-guest/browser-guest-paint-retention.ts @@ -29,19 +29,23 @@ type BrowserTabPageIdSource = { pageIds?: readonly string[] | null } +// Why: a stable identity keeps the disabled branch from re-running downstream shallow compares. +const NO_BROWSER_PAGE_IDS: readonly string[] = [] + export function collectBrowserPageIds( tabs: readonly BrowserTabPageIdSource[] | null | undefined -): string[] { - return (tabs ?? []).flatMap((tab) => +): readonly string[] { + // Why the early return: no browser tabs is the common case, and this runs on every store write. + if (!tabs || tabs.length === 0) { + return NO_BROWSER_PAGE_IDS + } + return tabs.flatMap((tab) => tab.pageIds && tab.pageIds.length > 0 ? tab.pageIds : [tab.activePageId ?? tab.id] ) } - -// Why: a stable identity keeps the disabled branch from re-running downstream shallow compares. -const NO_BROWSER_PAGE_IDS: string[] = [] const NO_BROWSER_TABS_BY_WORKTREE: Record = {} -export function useWorktreeBrowserPageIds(worktreeId: string): string[] { +export function useWorktreeBrowserPageIds(worktreeId: string): readonly string[] { return useAppStore( useShallow((state) => collectBrowserPageIds(state.browserTabsByWorktree[worktreeId])) ) diff --git a/src/renderer/src/components/cmd-j/PaletteFilterChips.tsx b/src/renderer/src/components/cmd-j/PaletteFilterChips.tsx index 4261e0e2e8e..1ac3e71e989 100644 --- a/src/renderer/src/components/cmd-j/PaletteFilterChips.tsx +++ b/src/renderer/src/components/cmd-j/PaletteFilterChips.tsx @@ -23,22 +23,24 @@ export default function PaletteFilterChips({ }): React.JSX.Element | null { const chips = useMemo(() => { const hostLabels = new Map(model.hosts.map((host) => [host.id, host.label])) - const projectLabels = new Map(model.projects.map((project) => [project.id, project.label])) + const repositoryLabels = new Map( + model.repositories.map((repository) => [repository.id, repository.label]) + ) return [ ...filter.hostIds.map((id) => ({ field: 'host' as const, id, label: hostLabels.get(id) ?? id })), - ...filter.projectKeys.map((id) => ({ - field: 'project' as const, + ...filter.repoIds.map((id) => ({ + field: 'repository' as const, id, - label: projectLabels.get(id) ?? id + label: repositoryLabels.get(id) ?? id })) ] - }, [filter.hostIds, filter.projectKeys, model.hosts, model.projects]) + }, [filter.hostIds, filter.repoIds, model.hosts, model.repositories]) - if (!isPaletteFilterActive(filter) || chips.length === 0) { + if (!isPaletteFilterActive(filter)) { return null } diff --git a/src/renderer/src/components/cmd-j/PaletteFilterMenu.tsx b/src/renderer/src/components/cmd-j/PaletteFilterMenu.tsx index d9f8b77ac7a..fc151933b23 100644 --- a/src/renderer/src/components/cmd-j/PaletteFilterMenu.tsx +++ b/src/renderer/src/components/cmd-j/PaletteFilterMenu.tsx @@ -78,7 +78,7 @@ export default function PaletteFilterMenu({ const groups = useMemo(() => { const entries: PaletteFilterGroup[] = [] - // Why: a single host (or single project) is nothing to disambiguate between, + // Why: a single host (or single repository) is nothing to disambiguate between, // so that axis stays hidden rather than offering a no-op checkbox. if (model.hosts.length > 1) { entries.push({ @@ -88,16 +88,17 @@ export default function PaletteFilterMenu({ selected: filter.hostIds }) } - if (model.projects.length > 1) { + if (model.repositories.length > 1) { entries.push({ - field: 'project', + field: 'repository', + // "Projects" is the user-facing term for repository-granular choices; see filter.emptySubtitle. heading: translate('worktreeJumpPalette.filter.projects', 'Projects'), - options: model.projects, - selected: filter.projectKeys + options: model.repositories, + selected: filter.repoIds }) } return entries - }, [filter.hostIds, filter.projectKeys, model.hosts, model.projects]) + }, [filter.hostIds, filter.repoIds, model.hosts, model.repositories]) // Stale field falls back to root if its group disappeared mid-session. const activeGroup = diff --git a/src/renderer/src/components/cmd-j/palette-filter-options.test.ts b/src/renderer/src/components/cmd-j/palette-filter-options.test.ts index 021e92a2359..64bd5740bba 100644 --- a/src/renderer/src/components/cmd-j/palette-filter-options.test.ts +++ b/src/renderer/src/components/cmd-j/palette-filter-options.test.ts @@ -5,11 +5,7 @@ import type { Project, ProjectHostSetup } from '../../../../shared/project-types import type { Repo } from '../../../../shared/repo-types' import type { Worktree } from '../../../../shared/worktree/types' import { buildSidebarHostOptions } from '../sidebar/sidebar-host-options' -import { - buildPaletteFilterModel, - resolveRepoFilterHostId, - resolveWorktreeFilterHostId -} from './palette-filter-options' +import { buildPaletteFilterModel, resolveWorktreeFilterHostId } from './palette-filter-options' function repo(id: string, displayName: string, connectionId: string | null = null): Repo { return { @@ -66,15 +62,17 @@ const buildModel = (worktrees: readonly Worktree[]) => buildPaletteFilterModel({ repos, worktrees, hostOptions, projects, projectHostSetups }) describe('buildPaletteFilterModel', () => { - it('collapses the repos of one project into a single row', () => { + it('keeps filter options repo-granular while retaining project-row membership', () => { const model = buildModel([worktree('w1', 'r1'), worktree('w2', 'r2'), worktree('w3', 'r3')]) expect(model.repoIdsByProjectKey.get('project:p1')).toEqual(['r1', 'r2']) - expect(model.projects.map((option) => [option.id, option.label, option.count])).toEqual([ - ['project:p1', 'Orca', 2], - ['repo:r3', 'Solo', 1] + expect(model.repositories.map((option) => [option.id, option.label, option.count])).toEqual([ + ['r1', 'Orca', 1], + ['r2', 'Orca (builder)', 1], + ['r3', 'Solo', 1] ]) - expect(model.projects[0]?.searchText).toBe('orca') + expect(model.repositories[0]?.searchText).toContain('orca') + expect(model.repositories[0]?.searchText).toContain(path.join('/repos', 'r1')) }) it('counts a worktree against its own host stamp, not its repo host', () => { @@ -88,8 +86,8 @@ describe('buildPaletteFilterModel', () => { ['local', 1], ['ssh:ssh-1', 2] ]) - // Host stamp does not move the workspace out of its project row. - expect(model.projects.find((option) => option.id === 'project:p1')?.count).toBe(3) + expect(model.repositories.find((option) => option.id === 'r1')?.count).toBe(2) + expect(model.repositories.find((option) => option.id === 'r2')?.count).toBe(1) }) it('omits archived worktrees from every count', () => { @@ -99,29 +97,85 @@ describe('buildPaletteFilterModel', () => { worktree('w3', 'r3', { isArchived: true }) ]) - expect(model.hosts.map((option) => option.id)).toEqual(['local']) - expect(model.hosts[0]?.count).toBe(1) - expect(model.projects.map((option) => option.id)).toEqual(['project:p1']) + expect(model.hosts.map((option) => [option.id, option.count])).toEqual([ + ['local', 1], + ['ssh:ssh-1', 0] + ]) + expect(model.repositories.map((option) => [option.id, option.count])).toEqual([ + ['r1', 1], + ['r2', 0], + ['r3', 0] + ]) }) - it('offers no options at all when there is nothing to narrow', () => { + it('retains options while worktrees are loading', () => { const model = buildModel([]) - expect(model.hosts).toEqual([]) - expect(model.projects).toEqual([]) - // The mapping still resolves so a lingering selection prunes cleanly. + expect(model.hosts.map((option) => [option.id, option.count])).toEqual([ + ['local', 0], + ['ssh:ssh-1', 0] + ]) + expect(model.repositories.map((option) => [option.id, option.count])).toEqual([ + ['r1', 0], + ['r2', 0], + ['r3', 0] + ]) expect(model.repoIdsByProjectKey.get('project:p1')).toEqual(['r1', 'r2']) - expect(model.hostIdByRepoId.get('r2')).toBe('ssh:ssh-1') + expect(model.hostIdsByRepoId.get('r2')).toEqual(new Set(['ssh:ssh-1'])) }) - it('sorts project rows by workspace count then label', () => { + it('deduplicates a repository ID shared by multiple hosts', () => { + const duplicateRepos = [repo('shared', 'Shared'), repo('shared', 'Shared remote', 'ssh-1')] + const model = buildPaletteFilterModel({ + repos: duplicateRepos, + worktrees: [ + worktree('local', 'shared', { hostId: 'local' }), + worktree('remote', 'shared', { hostId: 'ssh:ssh-1' }) + ], + hostOptions: buildSidebarHostOptions({ + repos: duplicateRepos, + sshTargetLabels: new Map([['ssh-1', 'Builder']]), + settings: { activeRuntimeEnvironmentId: null } + }), + projects: [], + projectHostSetups: [] + }) + + expect(model.repositories.map((option) => [option.id, option.count])).toEqual([['shared', 2]]) + expect(model.hostIdsByRepoId.get('shared')).toEqual(new Set(['local', 'ssh:ssh-1'])) + expect(model.repoIdsByProjectKey.get('repo:shared')).toEqual(['shared']) + }) + + it('disambiguates repositories with the same display name', () => { + const duplicateNames = [ + { ...repo('payments', 'api'), path: path.join('/repos', 'payments', 'api') }, + { ...repo('billing', 'api'), path: path.join('/repos', 'billing', 'api') } + ] + const model = buildPaletteFilterModel({ + repos: duplicateNames, + worktrees: [], + hostOptions: [], + projects: [], + projectHostSetups: [] + }) + + expect(model.repositories.map((option) => option.label)).toEqual([ + 'billing/api', + 'payments/api' + ]) + }) + + it('sorts repository options by workspace count then label', () => { const model = buildModel([worktree('w1', 'r3'), worktree('w2', 'r1'), worktree('w3', 'r2')]) - // Orca has 2 workspaces, Solo has 1 — popularity beats alpha. - expect(model.projects.map((option) => option.label)).toEqual(['Orca', 'Solo']) + expect(model.repositories.map((option) => option.label)).toEqual([ + 'Orca', + 'Orca (builder)', + 'Solo' + ]) }) - it('prefers a busier project ahead of an alphabetically earlier quiet one', () => { + it('prefers a busier repository ahead of an alphabetically earlier quiet one', () => { const model = buildModel([ worktree('w1', 'r3'), worktree('w2', 'r3'), @@ -129,26 +183,41 @@ describe('buildPaletteFilterModel', () => { worktree('w4', 'r1') ]) - expect(model.projects.map((option) => [option.label, option.count])).toEqual([ + expect(model.repositories.map((option) => [option.label, option.count])).toEqual([ ['Solo', 3], - ['Orca', 1] + ['Orca', 1], + ['Orca (builder)', 0] ]) }) }) describe('resolveWorktreeFilterHostId', () => { - const hostIdByRepoId = new Map([['r2', 'ssh:ssh-1']]) + const repoById = new Map([['r2', repo('r2', 'Remote', 'ssh-1')]]) it('prefers the worktree stamp, then the repo host, then the default host', () => { - expect( - resolveWorktreeFilterHostId({ repoId: 'r2', hostId: 'local' }, hostIdByRepoId, 'local') - ).toBe('local') - expect(resolveWorktreeFilterHostId({ repoId: 'r2' }, hostIdByRepoId, 'local')).toBe('ssh:ssh-1') - expect(resolveWorktreeFilterHostId({ repoId: 'unknown' }, hostIdByRepoId, 'local')).toBe( + expect(resolveWorktreeFilterHostId({ repoId: 'r2', hostId: 'local' }, repoById, 'local')).toBe( 'local' ) + expect(resolveWorktreeFilterHostId({ repoId: 'r2' }, repoById, 'local')).toBe('ssh:ssh-1') + expect(resolveWorktreeFilterHostId({ repoId: 'unknown' }, repoById, 'local')).toBe('local') + expect(resolveWorktreeFilterHostId({ repoId: 'unknown' }, repoById, 'runtime:env-1')).toBe( + 'runtime:env-1' + ) + }) + + it('uses the same last repository row as the sidebar for a shared legacy ID', () => { + const duplicateRepos = [repo('shared', 'Shared'), repo('shared', 'Shared remote', 'ssh-1')] + const sidebarRepoMap = new Map(duplicateRepos.map((entry) => [entry.id, entry])) + + expect(resolveWorktreeFilterHostId({ repoId: 'shared' }, sidebarRepoMap, 'local')).toBe( + 'ssh:ssh-1' + ) expect( - resolveWorktreeFilterHostId({ repoId: 'unknown' }, hostIdByRepoId, 'runtime:env-1') + resolveWorktreeFilterHostId( + { repoId: 'shared', hostId: 'runtime:env-1' }, + sidebarRepoMap, + 'local' + ) ).toBe('runtime:env-1') }) @@ -174,20 +243,10 @@ describe('resolveWorktreeFilterHostId', () => { defaultHostId }) for (const entry of cases) { - expect(resolveWorktreeFilterHostId(entry, model.hostIdByRepoId, model.defaultHostId)).toBe( + expect(resolveWorktreeFilterHostId(entry, model.repoById, model.defaultHostId)).toBe( getWorktreeExecutionHostId(entry, repoMap.get(entry.repoId), defaultHostId) ) } } }) }) - -describe('resolveRepoFilterHostId', () => { - it('falls back to the default host when the repo has no stamp', () => { - const hostIdByRepoId = new Map([['r2', 'ssh:ssh-1']]) - expect(resolveRepoFilterHostId('r2', hostIdByRepoId, 'local')).toBe('ssh:ssh-1') - expect(resolveRepoFilterHostId('missing', hostIdByRepoId, 'runtime:env-1')).toBe( - 'runtime:env-1' - ) - }) -}) diff --git a/src/renderer/src/components/cmd-j/palette-filter-options.ts b/src/renderer/src/components/cmd-j/palette-filter-options.ts index b455354886c..1dcb2c74962 100644 --- a/src/renderer/src/components/cmd-j/palette-filter-options.ts +++ b/src/renderer/src/components/cmd-j/palette-filter-options.ts @@ -1,5 +1,7 @@ +import { getRepoDisplayLabelKey, getRepoDisplayLabelsByPath } from '@/lib/repo-display-labels' import { getRepoExecutionHostId, + getWorktreeExecutionHostId, LOCAL_EXECUTION_HOST_ID, type ExecutionHostId } from '../../../../shared/execution-host' @@ -42,73 +44,59 @@ function toFilterOption({ export type PaletteFilterModel = { hosts: readonly PaletteFilterOption[] - projects: readonly PaletteFilterOption[] - /** A project row can span several repos (Project.sourceRepoIds), so selection resolves through this. */ + repositories: readonly PaletteFilterOption[] + /** Repository IDs represented by each project row in the sidebar grouping. */ repoIdsByProjectKey: ReadonlyMap - /** Only repos that carry a host stamp; absent means "inherit defaultHostId". */ - hostIdByRepoId: ReadonlyMap + /** Every execution host that owns a repository ID. */ + hostIdsByRepoId: ReadonlyMap> + /** Same last-row-wins repository index used by the sidebar. */ + repoById: ReadonlyMap> /** The focused runtime host, which host-less repos and worktrees inherit. */ defaultHostId: ExecutionHostId } -/** - * Precomputes only the repos that actually carry a host stamp so the lookup miss - * below stays equivalent to getWorktreeExecutionHostId's `defaultHostId` branch. - * Collapsing host-less repos to `local` here would disagree with the sidebar - * whenever a runtime environment is focused. - */ -function buildRepoHostIndex(repos: readonly Repo[]): Map { - const hostIdByRepoId = new Map() +function buildRepoHostIndex( + repos: readonly Repo[], + defaultHostId: ExecutionHostId +): Map> { + const hostIdsByRepoId = new Map>() for (const repo of repos) { - if (repo.connectionId || repo.executionHostId) { - hostIdByRepoId.set(repo.id, getRepoExecutionHostId(repo)) - } + const hostIds = hostIdsByRepoId.get(repo.id) ?? new Set() + hostIds.add( + repo.connectionId || repo.executionHostId ? getRepoExecutionHostId(repo) : defaultHostId + ) + hostIdsByRepoId.set(repo.id, hostIds) } - return hostIdByRepoId + return hostIdsByRepoId } export function resolveWorktreeFilterHostId( worktree: Pick, - hostIdByRepoId: ReadonlyMap, + repoById: ReadonlyMap>, defaultHostId: ExecutionHostId ): ExecutionHostId { - // Why: same precedence as getWorktreeExecutionHostId without re-resolving the - // repo per worktree — the repo host is precomputed once for the whole pass. - return worktree.hostId ?? hostIdByRepoId.get(worktree.repoId) ?? defaultHostId + return getWorktreeExecutionHostId(worktree, repoById.get(worktree.repoId), defaultHostId) } -/** Repo-derived host for a project row, which owns no worktree of its own. */ -export function resolveRepoFilterHostId( - repoId: string, - hostIdByRepoId: ReadonlyMap, - defaultHostId: ExecutionHostId -): ExecutionHostId { - return hostIdByRepoId.get(repoId) ?? defaultHostId -} - -type ProjectRow = { key: string; label: string; repoIds: string[] } - -function buildProjectRows( +function buildRepoIdsByProjectKey( repos: readonly Repo[], - repoMap: Map, + repoById: Map, grouping: ProjectGroupingModel -): { rows: ProjectRow[]; keyByRepoId: Map } { - const rows = new Map() - const keyByRepoId = new Map() +): Map { + const repoIdsByProjectKey = new Map() for (const repo of repos) { - const target = getProjectHeaderRevealTarget(repo.id, repoMap, grouping) + const target = getProjectHeaderRevealTarget(repo.id, repoById, grouping) if (!target.repo) { continue } - const existing = rows.get(target.key) - if (existing) { - existing.repoIds.push(repo.id) + const repoIds = repoIdsByProjectKey.get(target.key) + if (repoIds) { + repoIds.push(repo.id) } else { - rows.set(target.key, { key: target.key, label: target.label, repoIds: [repo.id] }) + repoIdsByProjectKey.set(target.key, [repo.id]) } - keyByRepoId.set(repo.id, target.key) } - return { rows: [...rows.values()], keyByRepoId } + return repoIdsByProjectKey } export function buildPaletteFilterModel({ @@ -126,60 +114,56 @@ export function buildPaletteFilterModel({ projectHostSetups: readonly ProjectHostSetup[] defaultHostId?: ExecutionHostId }): PaletteFilterModel { - const repoMap = new Map(repos.map((repo) => [repo.id, repo])) - const hostIdByRepoId = buildRepoHostIndex(repos) - const { rows, keyByRepoId } = buildProjectRows(repos, repoMap, { projects, projectHostSetups }) + const repoById = new Map(repos.map((repo) => [repo.id, repo])) + const hostIdsByRepoId = buildRepoHostIndex(repos, defaultHostId) + const repoIdsByProjectKey = buildRepoIdsByProjectKey([...repoById.values()], repoById, { + projects, + projectHostSetups + }) const worktreeCountByHostId = new Map() - const worktreeCountByProjectKey = new Map() + const worktreeCountByRepoId = new Map() for (const worktree of worktrees) { if (worktree.isArchived) { continue } - const hostId = resolveWorktreeFilterHostId(worktree, hostIdByRepoId, defaultHostId) + const hostId = resolveWorktreeFilterHostId(worktree, repoById, defaultHostId) worktreeCountByHostId.set(hostId, (worktreeCountByHostId.get(hostId) ?? 0) + 1) - const projectKey = keyByRepoId.get(worktree.repoId) - if (projectKey) { - worktreeCountByProjectKey.set( - projectKey, - (worktreeCountByProjectKey.get(projectKey) ?? 0) + 1 - ) - } + worktreeCountByRepoId.set( + worktree.repoId, + (worktreeCountByRepoId.get(worktree.repoId) ?? 0) + 1 + ) } - // Why: options are gated on a live workspace count, not on configuration — an - // option that can only ever yield an empty list is a trap, and it also keeps - // stale selections self-healing through reconcilePaletteFilter. // Registry order (local first, then SSH/runtime) matches the sidebar host headers. - const hosts = hostOptions - .filter((host) => (worktreeCountByHostId.get(host.id) ?? 0) > 0) - .map((host) => - toFilterOption({ - id: host.id, - label: host.label, - detail: host.detail, - count: worktreeCountByHostId.get(host.id) ?? 0 - }) - ) + const hosts = hostOptions.map((host) => + toFilterOption({ + id: host.id, + label: host.label, + detail: host.detail, + count: worktreeCountByHostId.get(host.id) ?? 0 + }) + ) - // Popularity first so a long project list surfaces busy workspaces without search. - const projectOptions = rows - .filter((row) => (worktreeCountByProjectKey.get(row.key) ?? 0) > 0) - .map((row) => + // Keep repository IDs aligned with the sidebar; project grouping remains a row concern. + const repositoryLabels = getRepoDisplayLabelsByPath([...repoById.values()]) + const repositories = [...repoById.values()] + .map((repo) => toFilterOption({ - id: row.key, - label: row.label, - detail: '', - count: worktreeCountByProjectKey.get(row.key) ?? 0 + id: repo.id, + label: repositoryLabels.get(getRepoDisplayLabelKey(repo)) ?? repo.displayName, + detail: repo.path, + count: worktreeCountByRepoId.get(repo.id) ?? 0 }) ) .sort((a, b) => b.count - a.count || a.label.localeCompare(b.label) || a.id.localeCompare(b.id)) return { hosts, - projects: projectOptions, - repoIdsByProjectKey: new Map(rows.map((row) => [row.key, row.repoIds])), - hostIdByRepoId, + repositories, + repoIdsByProjectKey, + hostIdsByRepoId, + repoById, defaultHostId } } diff --git a/src/renderer/src/components/cmd-j/palette-filter.test.ts b/src/renderer/src/components/cmd-j/palette-filter.test.ts index fcb62102f82..18db5178c39 100644 --- a/src/renderer/src/components/cmd-j/palette-filter.test.ts +++ b/src/renderer/src/components/cmd-j/palette-filter.test.ts @@ -1,14 +1,14 @@ import { describe, expect, it } from 'vitest' import type { ExecutionHostId } from '../../../../shared/execution-host' +import type { Repo } from '../../../../shared/repo-types' import { addPaletteFilterValues, + buildPaletteFilterFromSidebarScope, buildPaletteFilterPredicate, clearPaletteFilterField, EMPTY_PALETTE_FILTER, getPaletteFilterSelectionCount, isPaletteFilterActive, - PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD, - reconcilePaletteFilter, togglePaletteFilterValue, type PaletteFilterState } from './palette-filter' @@ -27,30 +27,35 @@ const option = (id: string, count = 1) => ({ // r1 + r2 are two repos behind one project row; r3 is a standalone repo row. const model: PaletteFilterModel = { hosts: [option('local'), option('ssh:builder'), option('runtime:env-1')], - projects: [option('project:p1'), option('repo:r3')], + repositories: [option('r1'), option('r2'), option('r3')], repoIdsByProjectKey: new Map([ ['project:p1', ['r1', 'r2']], ['repo:r3', ['r3']] ]), - hostIdByRepoId: new Map([ - ['r1', 'local'], - ['r2', 'ssh:builder'], - ['r3', 'runtime:env-1'] + hostIdsByRepoId: new Map>([ + ['r1', new Set(['local'])], + ['r2', new Set(['ssh:builder'])], + ['r3', new Set(['runtime:env-1'])] + ]), + repoById: new Map>([ + ['r1', {}], + ['r2', { connectionId: 'builder' }], + ['r3', { executionHostId: 'runtime:env-1' }] ]), defaultHostId: LOCAL_EXECUTION_HOST_ID } -const filterOf = (hostIds: string[], projectKeys: string[]): PaletteFilterState => ({ +const filterOf = (hostIds: string[], repoIds: string[]): PaletteFilterState => ({ hostIds, - projectKeys + repoIds }) describe('palette filter state', () => { it('reports activity and selection count across both fields', () => { expect(isPaletteFilterActive(EMPTY_PALETTE_FILTER)).toBe(false) expect(getPaletteFilterSelectionCount(EMPTY_PALETTE_FILTER)).toBe(0) - expect(isPaletteFilterActive(filterOf([], ['project:p1']))).toBe(true) - expect(getPaletteFilterSelectionCount(filterOf(['local'], ['project:p1']))).toBe(2) + expect(isPaletteFilterActive(filterOf([], ['r1']))).toBe(true) + expect(getPaletteFilterSelectionCount(filterOf(['local'], ['r1']))).toBe(2) }) it('toggles values on and off, keeping each field sorted', () => { @@ -58,7 +63,7 @@ describe('palette filter state', () => { const withBothHosts = togglePaletteFilterValue(withHost, 'host', 'local') expect(withBothHosts.hostIds).toEqual(['local', 'ssh:builder']) - expect(withBothHosts.projectKeys).toEqual([]) + expect(withBothHosts.repoIds).toEqual([]) expect(togglePaletteFilterValue(withBothHosts, 'host', 'local').hostIds).toEqual([ 'ssh:builder' ]) @@ -67,70 +72,31 @@ describe('palette filter state', () => { it('keeps the two fields independent', () => { const filter = togglePaletteFilterValue( togglePaletteFilterValue(EMPTY_PALETTE_FILTER, 'host', 'local'), - 'project', - 'project:p1' + 'repository', + 'r1' ) - expect(clearPaletteFilterField(filter, 'project')).toEqual(filterOf(['local'], [])) - expect(clearPaletteFilterField(filter, 'host')).toEqual(filterOf([], ['project:p1'])) + expect(clearPaletteFilterField(filter, 'repository')).toEqual(filterOf(['local'], [])) + expect(clearPaletteFilterField(filter, 'host')).toEqual(filterOf([], ['r1'])) }) - it('refuses selections past the per-field cap', () => { - const saturated = filterOf( - Array.from({ length: PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD }, (_, i) => `ssh:host-${i}`), - [] - ) + it('bulk-adds every matching id without duplicating', () => { + const withOne = addPaletteFilterValues(EMPTY_PALETTE_FILTER, 'repository', ['r1', 'r3', 'r1']) + expect(withOne.repoIds).toEqual(['r1', 'r3']) - const next = togglePaletteFilterValue(saturated, 'host', 'ssh:one-too-many') - - expect(next.hostIds).toHaveLength(PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD) - expect(next.hostIds).not.toContain('ssh:one-too-many') - // Deselecting still works at the cap, so the user is never stuck. - expect(togglePaletteFilterValue(saturated, 'host', 'ssh:host-0').hostIds).toHaveLength( - PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD - 1 - ) + const manyIds = Array.from({ length: 501 }, (_, index) => `repo-${index}`) + expect( + addPaletteFilterValues(EMPTY_PALETTE_FILTER, 'repository', manyIds).repoIds + ).toHaveLength(501) }) - it('bulk-adds matching ids up to the per-field cap without duplicating', () => { - const withOne = addPaletteFilterValues(EMPTY_PALETTE_FILTER, 'project', [ - 'project:p1', - 'repo:r3', - 'project:p1' - ]) - expect(withOne.projectKeys).toEqual(['project:p1', 'repo:r3']) + it('preserves state identity when bulk-add and clear are no-ops', () => { + const filter = filterOf(['local'], ['r1']) + const repoOnly = filterOf([], ['r1']) - const nearCap = filterOf( - Array.from( - { length: PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD - 1 }, - (_, i) => `ssh:host-${i}` - ), - [] - ) - const filled = addPaletteFilterValues(nearCap, 'host', ['ssh:a', 'ssh:b']) - expect(filled.hostIds).toHaveLength(PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD) - expect(filled.hostIds).toContain('ssh:a') - expect(filled.hostIds).not.toContain('ssh:b') - }) -}) - -describe('reconcilePaletteFilter', () => { - it('returns the same reference when every selection still exists', () => { - const filter = filterOf(['local'], ['project:p1']) - - expect(reconcilePaletteFilter(filter, model)).toBe(filter) - expect(reconcilePaletteFilter(EMPTY_PALETTE_FILTER, model)).toBe(EMPTY_PALETTE_FILTER) - }) - - it('drops selections whose host or project disappeared', () => { - const filter = filterOf(['local', 'ssh:deleted'], ['project:p1', 'repo:removed']) - - expect(reconcilePaletteFilter(filter, model)).toEqual(filterOf(['local'], ['project:p1'])) - }) - - it('empties a filter whose every selection is gone', () => { - const reconciled = reconcilePaletteFilter(filterOf(['ssh:deleted'], []), model) - - expect(isPaletteFilterActive(reconciled)).toBe(false) + expect(addPaletteFilterValues(filter, 'repository', ['r1'])).toBe(filter) + expect(clearPaletteFilterField(filter, 'host')).not.toBe(filter) + expect(clearPaletteFilterField(repoOnly, 'host')).toBe(repoOnly) }) }) @@ -155,11 +121,11 @@ describe('buildPaletteFilterPredicate', () => { expect(local?.matchesWorktree({ repoId: 'never-seen' })).toBe(true) }) - it('matches every repo behind a multi-repo project row', () => { - const predicate = buildPaletteFilterPredicate(filterOf([], ['project:p1']), model) + it('keeps repository filtering exact within a multi-repo project row', () => { + const predicate = buildPaletteFilterPredicate(filterOf([], ['r1']), model) expect(predicate?.matchesWorktree({ repoId: 'r1' })).toBe(true) - expect(predicate?.matchesWorktree({ repoId: 'r2' })).toBe(true) + expect(predicate?.matchesWorktree({ repoId: 'r2' })).toBe(false) expect(predicate?.matchesWorktree({ repoId: 'r3' })).toBe(false) expect(predicate?.matchesProjectRowKey('project:p1')).toBe(true) expect(predicate?.matchesProjectRowKey('repo:r3')).toBe(false) @@ -179,21 +145,38 @@ describe('buildPaletteFilterPredicate', () => { }) it('ORs within a field and ANDs across fields', () => { - const ored = buildPaletteFilterPredicate(filterOf([], ['project:p1', 'repo:r3']), model) + const ored = buildPaletteFilterPredicate(filterOf([], ['r1', 'r3']), model) expect(ored?.matchesWorktree({ repoId: 'r1' })).toBe(true) expect(ored?.matchesWorktree({ repoId: 'r3' })).toBe(true) - // Project p1 spans local (r1) and ssh:builder (r2); adding the host axis - // narrows to the intersection rather than widening the result set. - const anded = buildPaletteFilterPredicate(filterOf(['local'], ['project:p1']), model) + const anded = buildPaletteFilterPredicate(filterOf(['local'], ['r1', 'r2']), model) expect(anded?.matchesWorktree({ repoId: 'r1' })).toBe(true) expect(anded?.matchesWorktree({ repoId: 'r2' })).toBe(false) expect(anded?.matchesProjectRowKey('project:p1')).toBe(true) expect(anded?.matchesProjectRowKey('repo:r3')).toBe(false) + + const disjoint = buildPaletteFilterPredicate(filterOf(['local'], ['r2']), model) + expect(disjoint?.matchesProjectRowKey('project:p1')).toBe(false) }) - it('never matches a stale project key that resolves to no repos', () => { - const predicate = buildPaletteFilterPredicate(filterOf([], ['project:gone']), model) + it('matches every host that owns a shared repository ID', () => { + const sharedRepoModel: PaletteFilterModel = { + ...model, + repositories: [option('shared')], + repoIdsByProjectKey: new Map([['repo:shared', ['shared']]]), + hostIdsByRepoId: new Map([['shared', new Set(['local', 'ssh:builder'])]]) + } + + expect( + buildPaletteFilterPredicate( + filterOf(['ssh:builder'], ['shared']), + sharedRepoModel + )?.matchesProjectRowKey('repo:shared') + ).toBe(true) + }) + + it('never matches a stale repository id', () => { + const predicate = buildPaletteFilterPredicate(filterOf([], ['repo:gone']), model) expect(predicate?.matchesWorktree({ repoId: 'r1' })).toBe(false) expect(predicate?.matchesProjectRowKey('project:p1')).toBe(false) @@ -204,11 +187,68 @@ describe('buildPaletteFilterPredicate', () => { expect(hostOnly?.matchesGroupHostId('ssh:builder')).toBe(true) expect(hostOnly?.matchesGroupHostId('local')).toBe(false) - // A group header belongs to no project, so any project selection excludes it. - const withProject = buildPaletteFilterPredicate( - filterOf(['ssh:builder'], ['project:p1']), - model - ) + // A group header belongs to no repository, so any repository selection excludes it. + const withProject = buildPaletteFilterPredicate(filterOf(['ssh:builder'], ['r2']), model) expect(withProject?.matchesGroupHostId('ssh:builder')).toBe(false) }) }) + +describe('buildPaletteFilterFromSidebarScope', () => { + const allHosts = { workspaceHostScope: 'all', visibleWorkspaceHostIds: null } as const + + it('opens unfiltered when the sidebar shows every host and project', () => { + expect(buildPaletteFilterFromSidebarScope({ ...allHosts, filterRepoIds: [] })).toBe( + EMPTY_PALETTE_FILTER + ) + }) + + it('seeds the host chips from the sidebar host scope', () => { + expect( + buildPaletteFilterFromSidebarScope({ + workspaceHostScope: 'ssh:builder', + visibleWorkspaceHostIds: null, + filterRepoIds: [] + }) + ).toEqual(filterOf(['ssh:builder'], [])) + expect( + buildPaletteFilterFromSidebarScope({ + workspaceHostScope: 'all', + visibleWorkspaceHostIds: ['runtime:env-1', 'local'], + filterRepoIds: [] + }) + ).toEqual(filterOf(['local', 'runtime:env-1'], [])) + }) + + it('preserves sidebar repository picks exactly', () => { + expect(buildPaletteFilterFromSidebarScope({ ...allHosts, filterRepoIds: ['r2'] })).toEqual( + filterOf([], ['r2']) + ) + + const predicate = buildPaletteFilterPredicate(filterOf([], ['r2']), model) + expect(predicate?.matchesWorktree({ repoId: 'r1' })).toBe(false) + expect(predicate?.matchesWorktree({ repoId: 'r2' })).toBe(true) + }) + + it('preserves explicit selections even when they currently cover every known option', () => { + expect( + buildPaletteFilterFromSidebarScope({ + workspaceHostScope: 'all', + visibleWorkspaceHostIds: ['local', 'ssh:builder', 'runtime:env-1'], + filterRepoIds: ['r1', 'r2', 'r3'] + }) + ).toEqual(filterOf(['local', 'runtime:env-1', 'ssh:builder'], ['r1', 'r2', 'r3'])) + }) + + it('preserves empty or stale scopes instead of widening to a global search', () => { + const filter = buildPaletteFilterFromSidebarScope({ + workspaceHostScope: 'ssh:gone', + visibleWorkspaceHostIds: null, + filterRepoIds: ['r-gone'] + }) + + expect(filter).toEqual(filterOf(['ssh:gone'], ['r-gone'])) + expect(buildPaletteFilterPredicate(filter, model)?.matchesWorktree({ repoId: 'r1' })).toBe( + false + ) + }) +}) diff --git a/src/renderer/src/components/cmd-j/palette-filter.ts b/src/renderer/src/components/cmd-j/palette-filter.ts index f521a6ad92a..76319e5e9e1 100644 --- a/src/renderer/src/components/cmd-j/palette-filter.ts +++ b/src/renderer/src/components/cmd-j/palette-filter.ts @@ -1,12 +1,9 @@ import type { ExecutionHostId } from '../../../../shared/execution-host' import type { Worktree } from '../../../../shared/worktree/types' -import { - resolveRepoFilterHostId, - resolveWorktreeFilterHostId, - type PaletteFilterModel -} from './palette-filter-options' +import { getVisibleWorkspaceHostIdSet } from '../sidebar/visible-worktree-host-scope' +import { resolveWorktreeFilterHostId, type PaletteFilterModel } from './palette-filter-options' -export type PaletteFilterField = 'host' | 'project' +export type PaletteFilterField = 'host' | 'repository' /** * Sorted arrays rather than Sets: identity is stable across renders and the @@ -14,31 +11,23 @@ export type PaletteFilterField = 'host' | 'project' */ export type PaletteFilterState = { hostIds: readonly string[] - projectKeys: readonly string[] + repoIds: readonly string[] } -export const EMPTY_PALETTE_FILTER: PaletteFilterState = { hostIds: [], projectKeys: [] } - -/** Guard against a pathological selection blowing up the predicate's Set build. */ -export const PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD = 500 +export const EMPTY_PALETTE_FILTER: PaletteFilterState = { hostIds: [], repoIds: [] } export function isPaletteFilterActive(filter: PaletteFilterState): boolean { - return filter.hostIds.length > 0 || filter.projectKeys.length > 0 + return filter.hostIds.length > 0 || filter.repoIds.length > 0 } export function getPaletteFilterSelectionCount(filter: PaletteFilterState): number { - return filter.hostIds.length + filter.projectKeys.length + return filter.hostIds.length + filter.repoIds.length } function toggleValue(values: readonly string[], id: string): readonly string[] { if (values.includes(id)) { return values.filter((value) => value !== id) } - // Why: same reference on the capped no-op — a fresh array would invalidate - // every downstream search memo for a click that changed nothing. - if (values.length >= PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD) { - return values - } return [...values, id].sort() } @@ -49,74 +38,69 @@ export function togglePaletteFilterValue( ): PaletteFilterState { return field === 'host' ? { ...filter, hostIds: toggleValue(filter.hostIds, id) } - : { ...filter, projectKeys: toggleValue(filter.projectKeys, id) } + : { ...filter, repoIds: toggleValue(filter.repoIds, id) } } function addValues(values: readonly string[], ids: readonly string[]): readonly string[] { - if (ids.length === 0 || values.length >= PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD) { + if (ids.length === 0) { return values } const merged = new Set(values) const sizeBefore = merged.size for (const id of ids) { - if (merged.size >= PALETTE_FILTER_MAX_SELECTIONS_PER_FIELD) { - break - } merged.add(id) } - // Why: same reference when nothing new fit — keeps search memos stable. + // Why: same reference when nothing was added keeps search memos stable. if (merged.size === sizeBefore) { return values } return [...merged].sort() } -/** Bulk-add for "Select all matching"; respects the per-field cap and de-dupes. */ +/** Bulk-add for "Select all matching"; de-dupes while preserving stable no-ops. */ export function addPaletteFilterValues( filter: PaletteFilterState, field: PaletteFilterField, ids: readonly string[] ): PaletteFilterState { - return field === 'host' - ? { ...filter, hostIds: addValues(filter.hostIds, ids) } - : { ...filter, projectKeys: addValues(filter.projectKeys, ids) } + const values = field === 'host' ? filter.hostIds : filter.repoIds + const nextValues = addValues(values, ids) + if (nextValues === values) { + return filter + } + return field === 'host' ? { ...filter, hostIds: nextValues } : { ...filter, repoIds: nextValues } } export function clearPaletteFilterField( filter: PaletteFilterState, field: PaletteFilterField ): PaletteFilterState { - return field === 'host' ? { ...filter, hostIds: [] } : { ...filter, projectKeys: [] } + if ((field === 'host' ? filter.hostIds : filter.repoIds).length === 0) { + return filter + } + return field === 'host' ? { ...filter, hostIds: [] } : { ...filter, repoIds: [] } } -function pruneToAvailable(values: readonly string[], available: ReadonlySet): string[] { - return values.filter((value) => available.has(value)) +type SidebarScopeForPaletteFilter = Parameters[0] & { + filterRepoIds: readonly string[] } -/** - * Drops selections whose host or project disappeared (repo removed, SSH target - * deleted). Without this a stale id would silently empty the palette forever. - * Returns the same reference when nothing changed so memo deps stay stable. - */ -export function reconcilePaletteFilter( - filter: PaletteFilterState, - model: PaletteFilterModel +function sortedUnique(values: Iterable): string[] { + return [...new Set(values)].sort() +} + +/** Seeds the palette from the sidebar's exact host and repository scope. */ +export function buildPaletteFilterFromSidebarScope( + scope: SidebarScopeForPaletteFilter ): PaletteFilterState { - if (!isPaletteFilterActive(filter)) { - return filter + const visibleHostIds = getVisibleWorkspaceHostIdSet(scope) + const hostIds = visibleHostIds ? sortedUnique(visibleHostIds) : [] + const repoIds = sortedUnique(scope.filterRepoIds) + + if (hostIds.length === 0 && repoIds.length === 0) { + return EMPTY_PALETTE_FILTER } - const hostIds = pruneToAvailable(filter.hostIds, new Set(model.hosts.map((host) => host.id))) - const projectKeys = pruneToAvailable( - filter.projectKeys, - new Set(model.projects.map((project) => project.id)) - ) - if ( - hostIds.length === filter.hostIds.length && - projectKeys.length === filter.projectKeys.length - ) { - return filter - } - return { hostIds, projectKeys } + return { hostIds, repoIds } } export type PaletteFilterPredicate = { @@ -140,31 +124,29 @@ export function buildPaletteFilterPredicate( } const selectedHostIds = filter.hostIds.length > 0 ? new Set(filter.hostIds) : null - const selectedProjectKeys = filter.projectKeys.length > 0 ? new Set(filter.projectKeys) : null - let selectedRepoIds: Set | null = null - if (selectedProjectKeys) { - selectedRepoIds = new Set() - for (const projectKey of selectedProjectKeys) { - for (const repoId of model.repoIdsByProjectKey.get(projectKey) ?? []) { - selectedRepoIds.add(repoId) + const selectedRepoIds = filter.repoIds.length > 0 ? new Set(filter.repoIds) : null + const repoMatchesSelectedHost = (repoId: string): boolean => { + if (!selectedHostIds) { + return true + } + const repoHostIds = model.hostIdsByRepoId.get(repoId) + if (!repoHostIds) { + return selectedHostIds.has(model.defaultHostId) + } + for (const hostId of repoHostIds) { + if (selectedHostIds.has(hostId)) { + return true } } + return false } return { matchesProjectRowKey: (rowKey) => { - if (selectedProjectKeys && !selectedProjectKeys.has(rowKey)) { - return false - } - if (!selectedHostIds) { - return true - } - // Why: the row survives if *any* of its repos is on a selected host — a - // project checked out on both local and SSH is still reachable from either. - return (model.repoIdsByProjectKey.get(rowKey) ?? []).some((repoId) => - selectedHostIds.has( - resolveRepoFilterHostId(repoId, model.hostIdByRepoId, model.defaultHostId) - ) + const rowRepoIds = model.repoIdsByProjectKey.get(rowKey) ?? [] + return rowRepoIds.some( + (repoId) => + (!selectedRepoIds || selectedRepoIds.has(repoId)) && repoMatchesSelectedHost(repoId) ) }, matchesWorktree: (worktree) => { @@ -177,10 +159,10 @@ export function buildPaletteFilterPredicate( // Why: worktree.hostId wins over the repo fallback — a runtime-owned // workspace can live on a different host than the repo it came from. return selectedHostIds.has( - resolveWorktreeFilterHostId(worktree, model.hostIdByRepoId, model.defaultHostId) + resolveWorktreeFilterHostId(worktree, model.repoById, model.defaultHostId) ) }, - // Why: a group header is not a project, so an explicit project selection + // Why: a group header has no repository, so a repository selection // excludes every group row; only the host axis can keep one. matchesGroupHostId: (hostId) => selectedRepoIds === null && (!selectedHostIds || selectedHostIds.has(hostId)) diff --git a/src/renderer/src/components/cmd-j/palette-section-render-cap.test.ts b/src/renderer/src/components/cmd-j/palette-section-render-cap.test.ts index 65c7c31818f..c147c01cc20 100644 --- a/src/renderer/src/components/cmd-j/palette-section-render-cap.test.ts +++ b/src/renderer/src/components/cmd-j/palette-section-render-cap.test.ts @@ -51,6 +51,14 @@ describe('capPaletteSection', () => { it('supports an explicit cap of zero', () => { expect(capPaletteSection(range(3), 0)).toEqual({ visible: [], overflowCount: 3 }) }) + + it('keeps a retained eligible row when it crosses from 50th to 51st', () => { + const capped = capPaletteSection(range(60), PALETTE_SECTION_RENDER_CAP, (item) => item === 50) + + expect(capped.visible).toHaveLength(PALETTE_SECTION_RENDER_CAP) + expect(capped.visible.slice(-2)).toEqual([48, 50]) + expect(capped.overflowCount).toBe(10) + }) }) describe('softSplitPaletteSection', () => { diff --git a/src/renderer/src/components/cmd-j/palette-section-render-cap.ts b/src/renderer/src/components/cmd-j/palette-section-render-cap.ts index 4c83291e03f..29dee043a27 100644 --- a/src/renderer/src/components/cmd-j/palette-section-render-cap.ts +++ b/src/renderer/src/components/cmd-j/palette-section-render-cap.ts @@ -28,12 +28,27 @@ export type CappedPaletteSection = { export function capPaletteSection( items: readonly T[], - cap: number = PALETTE_SECTION_RENDER_CAP + cap: number = PALETTE_SECTION_RENDER_CAP, + retain?: (item: T) => boolean ): CappedPaletteSection { if (!Number.isFinite(cap) || cap < 0 || items.length <= cap) { return { visible: items, overflowCount: 0 } } - return { visible: items.slice(0, cap), overflowCount: items.length - cap } + const visible = items.slice(0, cap) + // Keep the selected match visible after reranking without increasing the DOM row cap. + let retained: T | undefined + if (retain) { + for (let index = cap; index < items.length; index += 1) { + if (retain(items[index])) { + retained = items[index] + break + } + } + } + if (retained !== undefined && cap > 0) { + visible.splice(cap - 1, 1, retained) + } + return { visible, overflowCount: items.length - visible.length } } /** @@ -50,9 +65,10 @@ export type SoftSplitSection = { export function softSplitPaletteSection( items: readonly T[], previewCount: number, - hardCap: number = PALETTE_SECTION_RENDER_CAP + hardCap: number = PALETTE_SECTION_RENDER_CAP, + retain?: (item: T) => boolean ): SoftSplitSection { - const capped = capPaletteSection(items, hardCap) + const capped = capPaletteSection(items, hardCap, retain) const previewSize = Math.max(0, Math.min(previewCount, capped.visible.length)) return { preview: capped.visible.slice(0, previewSize), @@ -95,7 +111,9 @@ export function layoutMultiPrimaryPaletteSections({ trailingFloorCount = TYPED_QUERY_TRAILING_FLOOR, hardCap, leadingHardCap = hardCap ?? PALETTE_SECTION_RENDER_CAP, - trailingHardCap = hardCap ?? PALETTE_SECTION_RENDER_CAP + trailingHardCap = hardCap ?? PALETTE_SECTION_RENDER_CAP, + leadingRetain, + trailingRetain }: { leadingItems: readonly T[] trailingItems: readonly T[] @@ -104,9 +122,21 @@ export function layoutMultiPrimaryPaletteSections({ hardCap?: number leadingHardCap?: number trailingHardCap?: number + leadingRetain?: (item: T) => boolean + trailingRetain?: (item: T) => boolean }): MultiPrimarySectionLayout { - const leading = softSplitPaletteSection(leadingItems, leadingPreviewCount, leadingHardCap) - const trailing = softSplitPaletteSection(trailingItems, trailingFloorCount, trailingHardCap) + const leading = softSplitPaletteSection( + leadingItems, + leadingPreviewCount, + leadingHardCap, + leadingRetain + ) + const trailing = softSplitPaletteSection( + trailingItems, + trailingFloorCount, + trailingHardCap, + trailingRetain + ) return { leadingPreview: leading.preview, leadingRest: leading.rest, diff --git a/src/renderer/src/components/dashboard/agent-row-lineage-model.test.ts b/src/renderer/src/components/dashboard/agent-row-lineage-model.test.ts index dcc5d754de2..45a9f0cbc37 100644 --- a/src/renderer/src/components/dashboard/agent-row-lineage-model.test.ts +++ b/src/renderer/src/components/dashboard/agent-row-lineage-model.test.ts @@ -97,6 +97,38 @@ describe('buildAgentRowLineageTree', () => { ]) }) + it('nests by parent pane key when the parent handles are stale after a restart', () => { + // Why: terminal handles are minted per process, so after an app restart the + // persisted coordinator handle names no live row; the durable pane key must win. + const parent = makeRow('parent:1', { terminalHandle: 'term-parent-reminted' }) + const child = makeRow('child:1', { + parentPaneKey: 'parent:1', + parentTerminalHandle: 'term-parent-stale', + coordinatorHandle: 'term-parent-stale' + }) + + const tree = buildAgentRowLineageTree([parent, child]) + + expect(tree.rootRows.map((row) => row.paneKey)).toEqual(['parent:1']) + expect(tree.childrenByParentPaneKey.get('parent:1')?.map((row) => row.paneKey)).toEqual([ + 'child:1' + ]) + expect(tree.childPaneKeys.has('child:1')).toBe(true) + }) + + it('keeps a child as a root when its parent pane key names no visible row', () => { + const unrelated = makeRow('other:1', { terminalHandle: 'term-other' }) + const orphan = makeRow('child:1', { + parentPaneKey: 'parent-closed:1', + coordinatorHandle: 'term-parent-stale' + }) + + const tree = buildAgentRowLineageTree([unrelated, orphan]) + + expect(tree.rootRows.map((row) => row.paneKey)).toEqual(['other:1', 'child:1']) + expect(tree.childrenByParentPaneKey.size).toBe(0) + }) + it('keeps cyclic lineage rows visible as flat roots', () => { const root = makeRow('root:1') const firstCycleRow = makeRow('cycle-a:1', { parentPaneKey: 'cycle-b:1' }) diff --git a/src/renderer/src/components/github-item-dialog/edit-item-fields/gh-edit-section.tsx b/src/renderer/src/components/github-item-dialog/edit-item-fields/gh-edit-section.tsx index 43b47323864..f92c99b2a09 100644 --- a/src/renderer/src/components/github-item-dialog/edit-item-fields/gh-edit-section.tsx +++ b/src/renderer/src/components/github-item-dialog/edit-item-fields/gh-edit-section.tsx @@ -27,6 +27,7 @@ import { } from './gh-edit-section-mutations' import { GHEditSectionTopColumns } from './gh-edit-section-top-columns' import { GHEditSectionHorizontal } from './gh-edit-section-horizontal' +import { useGitHubDuplicateIssueCandidates } from '@/components/github/github-duplicate-issue-candidates' export function GHEditSection({ item, @@ -74,27 +75,7 @@ export function GHEditSection({ const assigneesItemKey = `${item.repoId}\0${item.id}` const patchWorkItem = useAppStore((s) => s.patchWorkItem) const patchProjectRowContent = useAppStore((s) => s.patchProjectRowContent) - const duplicateIssueCandidates = useAppStore( - useShallow((s) => { - if (!duplicatePickerOpen) { - return [] - } - const deduped = new Map() - for (const entry of Object.values(s.workItemsCache)) { - for (const candidate of entry.data ?? []) { - if ( - candidate.type === 'issue' && - candidate.repoId === item.repoId && - candidate.number !== item.number && - !deduped.has(candidate.number) - ) { - deduped.set(candidate.number, candidate) - } - } - } - return Array.from(deduped.values()).sort((a, b) => b.number - a.number) - }) - ) + const duplicateIssueCandidates = useGitHubDuplicateIssueCandidates(item, duplicatePickerOpen) const repoOwnerSettings = useAppStore( useShallow((s) => getSettingsForRepoRuntimeOwner(s, item.repoId ?? null)) ) diff --git a/src/renderer/src/components/github/github-duplicate-issue-candidates.ts b/src/renderer/src/components/github/github-duplicate-issue-candidates.ts new file mode 100644 index 00000000000..301d9208046 --- /dev/null +++ b/src/renderer/src/components/github/github-duplicate-issue-candidates.ts @@ -0,0 +1,35 @@ +import { useShallow } from 'zustand/react/shallow' +import { useAppStore } from '@/store' +import type { GitHubWorkItem } from '../../../../shared/github/work-item-types' + +// Why a shared constant: the selector runs on every store write while the picker is +// closed, which is nearly always; a fresh [] there is pure allocation. +const NO_DUPLICATE_CANDIDATES: readonly GitHubWorkItem[] = [] + +/** Cached issues of `item`'s repo, newest first, for the close-as-duplicate picker. */ +export function useGitHubDuplicateIssueCandidates( + item: Pick, + pickerOpen: boolean +): readonly GitHubWorkItem[] { + return useAppStore( + useShallow((s) => { + if (!pickerOpen) { + return NO_DUPLICATE_CANDIDATES + } + const deduped = new Map() + for (const entry of Object.values(s.workItemsCache)) { + for (const candidate of entry.data ?? []) { + if ( + candidate.type === 'issue' && + candidate.repoId === item.repoId && + candidate.number !== item.number && + !deduped.has(candidate.number) + ) { + deduped.set(candidate.number, candidate) + } + } + } + return Array.from(deduped.values()).sort((a, b) => b.number - a.number) + }) + ) +} diff --git a/src/renderer/src/components/terminal-pane/TerminalLinkActionPopover.test.tsx b/src/renderer/src/components/link-actions/LinkActionPopover.test.tsx similarity index 81% rename from src/renderer/src/components/terminal-pane/TerminalLinkActionPopover.test.tsx rename to src/renderer/src/components/link-actions/LinkActionPopover.test.tsx index e4fab4bedb4..af92f5acab4 100644 --- a/src/renderer/src/components/terminal-pane/TerminalLinkActionPopover.test.tsx +++ b/src/renderer/src/components/link-actions/LinkActionPopover.test.tsx @@ -3,7 +3,7 @@ import type { ReactNode } from 'react' import { cleanup, fireEvent, render, screen, waitFor } from '@testing-library/react' import { afterEach, describe, expect, it, vi } from 'vitest' import { BROWSER_TERMINAL_LINK_ACTIONS_SETTINGS_TARGET_ID } from '@/lib/settings-navigation-types' -import type { TerminalLinkActionRequest } from './terminal-link-action-request' +import type { LinkActionRequest } from './link-action-request' const mocks = vi.hoisted(() => ({ openSettingsPage: vi.fn(), @@ -58,7 +58,7 @@ vi.mock('@/components/ui/popover', () => ({ ) })) -import { TerminalLinkActionPopover } from './TerminalLinkActionPopover' +import { LinkActionPopover } from './LinkActionPopover' afterEach(() => { cleanup() @@ -66,24 +66,23 @@ afterEach(() => { vi.unstubAllGlobals() }) -describe('TerminalLinkActionPopover', () => { +describe('LinkActionPopover', () => { it('shows the full destination and runs the selected action', () => { vi.stubGlobal('navigator', { userAgent: 'Macintosh' }) const onClose = vi.fn() - const focusTerminal = vi.fn() + const restoreFocus = vi.fn() const run = vi.fn() - const request: TerminalLinkActionRequest = { - paneId: 1, + const request: LinkActionRequest = { anchorX: 100, anchorY: 200, destination: 'https://example.com/full/hidden/destination?query=actual', kind: 'url', primary: { label: 'Open link', run }, alternate: { external: true, label: 'System Browser', run: vi.fn() }, - focusTerminal + restoreFocus } - render() + render() const destination = screen.getByText(request.destination) expect(destination.className).toContain('line-clamp-2') @@ -102,24 +101,23 @@ describe('TerminalLinkActionPopover', () => { fireEvent.click(screen.getByText('Open link')) expect(onClose).toHaveBeenCalledOnce() - expect(focusTerminal).toHaveBeenCalledOnce() + expect(restoreFocus).toHaveBeenCalledOnce() expect(run).toHaveBeenCalledOnce() }) it('identifies the dismissed request so a newer request can survive', () => { vi.stubGlobal('navigator', { userAgent: 'Macintosh' }) const onClose = vi.fn() - const request: TerminalLinkActionRequest = { - paneId: 1, + const request: LinkActionRequest = { anchorX: 100, anchorY: 200, destination: 'https://example.com', kind: 'url', primary: { label: 'Open link', run: vi.fn() }, - focusTerminal: vi.fn() + restoreFocus: vi.fn() } - render() + render() fireEvent.click(screen.getByTestId('dismiss-popover')) expect(onClose).toHaveBeenCalledWith(request) @@ -127,18 +125,17 @@ describe('TerminalLinkActionPopover', () => { it('uses distinct icons for system and Orca browser actions', () => { vi.stubGlobal('navigator', { userAgent: 'Macintosh' }) - const request: TerminalLinkActionRequest = { - paneId: 1, + const request: LinkActionRequest = { anchorX: 100, anchorY: 200, destination: 'https://example.com', kind: 'url', primary: { external: false, label: 'Orca Browser', run: vi.fn() }, alternate: { external: true, label: 'System Browser', run: vi.fn() }, - focusTerminal: vi.fn() + restoreFocus: vi.fn() } - render() + render() expect( screen.getByText('Orca Browser').closest('button')?.querySelector('.lucide-globe') @@ -153,25 +150,24 @@ describe('TerminalLinkActionPopover', () => { Object.assign(window, { api: { ui: { writeClipboardText: mocks.writeClipboardText } } }) mocks.writeClipboardText.mockResolvedValue(undefined) const onClose = vi.fn() - const focusTerminal = vi.fn() - const request: TerminalLinkActionRequest = { - paneId: 1, + const restoreFocus = vi.fn() + const request: LinkActionRequest = { anchorX: 100, anchorY: 200, destination: 'https://example.com/hidden-destination', kind: 'url', primary: { label: 'Open link', run: vi.fn() }, - focusTerminal + restoreFocus } - render() + render() fireEvent.click(screen.getByRole('button', { name: 'Copy link' })) await waitFor(() => expect(mocks.writeClipboardText).toHaveBeenCalledWith(request.destination)) await waitFor(() => expect(screen.getByRole('button', { name: 'Copied' })).toBeTruthy()) expect(mocks.toastSuccess).toHaveBeenCalledWith('Copied link') expect(onClose).not.toHaveBeenCalled() - expect(focusTerminal).not.toHaveBeenCalled() + expect(restoreFocus).not.toHaveBeenCalled() }) it('ignores duplicate copy clicks while the clipboard write is in flight', async () => { @@ -183,17 +179,16 @@ describe('TerminalLinkActionPopover', () => { resolveWrite = resolve }) ) - const request: TerminalLinkActionRequest = { - paneId: 1, + const request: LinkActionRequest = { anchorX: 100, anchorY: 200, destination: 'https://example.com/hidden-destination', kind: 'url', primary: { label: 'Open link', run: vi.fn() }, - focusTerminal: vi.fn() + restoreFocus: vi.fn() } - render() + render() const copyButton = screen.getByRole('button', { name: 'Copy link' }) fireEvent.click(copyButton) fireEvent.click(copyButton) @@ -209,17 +204,16 @@ describe('TerminalLinkActionPopover', () => { vi.stubGlobal('navigator', { userAgent: 'Macintosh' }) Object.assign(window, { api: { ui: { writeClipboardText: mocks.writeClipboardText } } }) mocks.writeClipboardText.mockRejectedValue(new Error('denied')) - const request: TerminalLinkActionRequest = { - paneId: 1, + const request: LinkActionRequest = { anchorX: 100, anchorY: 200, destination: 'https://example.com/hidden-destination', kind: 'url', primary: { label: 'Open link', run: vi.fn() }, - focusTerminal: vi.fn() + restoreFocus: vi.fn() } - render() + render() fireEvent.click(screen.getByRole('button', { name: 'Copy link' })) await waitFor(() => expect(mocks.toastError).toHaveBeenCalledWith('Failed to copy link')) @@ -229,17 +223,16 @@ describe('TerminalLinkActionPopover', () => { it('does not offer copy link for non-URL destinations', () => { vi.stubGlobal('navigator', { userAgent: 'Macintosh' }) - const request: TerminalLinkActionRequest = { - paneId: 1, + const request: LinkActionRequest = { anchorX: 100, anchorY: 200, destination: '/tmp/example.ts', kind: 'file', primary: { label: 'Open file', run: vi.fn() }, - focusTerminal: vi.fn() + restoreFocus: vi.fn() } - render() + render() expect(screen.queryByRole('button', { name: 'Copy link' })).toBeNull() }) @@ -247,18 +240,17 @@ describe('TerminalLinkActionPopover', () => { it('opens the terminal link setting from the compact settings button', () => { vi.stubGlobal('navigator', { userAgent: 'Macintosh' }) const onClose = vi.fn() - const focusTerminal = vi.fn() - const request: TerminalLinkActionRequest = { - paneId: 1, + const restoreFocus = vi.fn() + const request: LinkActionRequest = { anchorX: 100, anchorY: 200, destination: 'https://example.com', kind: 'url', primary: { label: 'System Browser', run: vi.fn() }, - focusTerminal + restoreFocus } - render() + render() fireEvent.click(screen.getByRole('button', { name: 'Terminal link settings' })) expect(onClose).toHaveBeenCalledOnce() @@ -268,6 +260,6 @@ describe('TerminalLinkActionPopover', () => { sectionId: BROWSER_TERMINAL_LINK_ACTIONS_SETTINGS_TARGET_ID }) expect(mocks.openSettingsPage).toHaveBeenCalledOnce() - expect(focusTerminal).not.toHaveBeenCalled() + expect(restoreFocus).not.toHaveBeenCalled() }) }) diff --git a/src/renderer/src/components/terminal-pane/TerminalLinkActionPopover.tsx b/src/renderer/src/components/link-actions/LinkActionPopover.tsx similarity index 90% rename from src/renderer/src/components/terminal-pane/TerminalLinkActionPopover.tsx rename to src/renderer/src/components/link-actions/LinkActionPopover.tsx index f04d7a93180..4bde5fdbae1 100644 --- a/src/renderer/src/components/terminal-pane/TerminalLinkActionPopover.tsx +++ b/src/renderer/src/components/link-actions/LinkActionPopover.tsx @@ -9,11 +9,11 @@ import { useClipboardTextCopyFeedback } from '@/hooks/use-clipboard-text-copy-fe import { translate } from '@/i18n/i18n' import { BROWSER_TERMINAL_LINK_ACTIONS_SETTINGS_TARGET_ID } from '@/lib/settings-navigation-types' import { useAppStore } from '@/store' -import type { TerminalLinkAction, TerminalLinkActionRequest } from './terminal-link-action-request' +import type { LinkAction, LinkActionRequest } from './link-action-request' -type TerminalLinkActionPopoverProps = { - request: TerminalLinkActionRequest | null - onClose: (dismissed?: TerminalLinkActionRequest) => void +type LinkActionPopoverProps = { + request: TRequest | null + onClose: (dismissed?: TRequest) => void } function ActionRow({ @@ -21,7 +21,7 @@ function ActionRow({ alternate, onRun }: { - action: TerminalLinkAction + action: LinkAction alternate: boolean onRun: () => void }): React.JSX.Element { @@ -46,10 +46,11 @@ function ActionRow({ ) } -export function TerminalLinkActionPopover({ +/** Shared by the terminal and native chat: pick where a clicked link opens. */ +export function LinkActionPopover({ request, onClose -}: TerminalLinkActionPopoverProps): React.JSX.Element { +}: LinkActionPopoverProps): React.JSX.Element { const openSettingsPage = useAppStore((state) => state.openSettingsPage) const openSettingsTarget = useAppStore((state) => state.openSettingsTarget) const copyableDestination = request?.kind === 'url' ? request.destination : '' @@ -64,9 +65,9 @@ export function TerminalLinkActionPopover({ [request?.anchorX, request?.anchorY] ) - const runAction = (action: TerminalLinkAction): void => { + const runAction = (action: LinkAction): void => { onClose() - request?.focusTerminal() + request?.restoreFocus() void action.run() } @@ -128,10 +129,11 @@ export function TerminalLinkActionPopover({ sideOffset={6} collisionPadding={8} className="w-max min-w-52 max-w-[min(21rem,calc(100vw-1rem))] p-1" + data-link-action-popover data-terminal-link-action-popover onOpenAutoFocus={(event) => event.preventDefault()} onCloseAutoFocus={(event) => event.preventDefault()} - onEscapeKeyDown={() => request.focusTerminal()} + onEscapeKeyDown={() => request.restoreFocus()} >
void +} + +export function closeLinkActionRequest( + current: T | null, + dismissed?: T +): T | null { + return dismissed && current !== dismissed ? current : null +} diff --git a/src/renderer/src/components/mobile/mobile-platform-copy.ts b/src/renderer/src/components/mobile/mobile-platform-copy.ts index cd6669891a3..f33f9a42196 100644 --- a/src/renderer/src/components/mobile/mobile-platform-copy.ts +++ b/src/renderer/src/components/mobile/mobile-platform-copy.ts @@ -22,7 +22,7 @@ const IOS_CHANNEL_COPY: Record = { const ANDROID_COPY: InstallCopy = { ctaLabel: 'Download APK', - url: 'https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk' + url: 'https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk' } export function getInstallCopy(platform: Platform, iosChannel: IosChannel): InstallCopy { diff --git a/src/renderer/src/components/native-chat/NativeChatComposer.test.tsx b/src/renderer/src/components/native-chat/NativeChatComposer.test.tsx index a6f233b96a3..55d13c088e8 100644 --- a/src/renderer/src/components/native-chat/NativeChatComposer.test.tsx +++ b/src/renderer/src/components/native-chat/NativeChatComposer.test.tsx @@ -288,7 +288,9 @@ describe('NativeChatComposer', () => { optionsSurface, optionSnapshot, onError: vi.fn(), - runtime: 'local' + runtime: 'local', + sessionId: 'session-test', + runtimeEnvironmentId: null }} /> ) @@ -328,7 +330,9 @@ describe('NativeChatComposer', () => { }, optionSnapshot: [], onError: vi.fn(), - runtime: 'local' + runtime: 'local', + sessionId: 'session-test', + runtimeEnvironmentId: null }} /> ) @@ -367,7 +371,9 @@ describe('NativeChatComposer', () => { optionSnapshot: [], worktreeId: 'wt-1', onError: vi.fn(), - runtime: 'local' + runtime: 'local', + sessionId: 'session-test', + runtimeEnvironmentId: null }} /> ) diff --git a/src/renderer/src/components/native-chat/NativeChatComposer.tsx b/src/renderer/src/components/native-chat/NativeChatComposer.tsx index 06ae0ec5c8c..e675739c9a3 100644 --- a/src/renderer/src/components/native-chat/NativeChatComposer.tsx +++ b/src/renderer/src/components/native-chat/NativeChatComposer.tsx @@ -1,9 +1,7 @@ -import { forwardRef, useCallback, useImperativeHandle, useMemo, useState } from 'react' +import { forwardRef, useCallback, useImperativeHandle, useState } from 'react' import { useAppStore } from '../../store' import { sendRuntimePtyInput } from '@/runtime/runtime-terminal-inspection' import { getSettingsForAgentTabRuntimeOwner } from '@/lib/agent-paste-draft' -import { getVerifiedNativeChatCommands } from '../../../../shared/native-chat-agent-profiles' -import { structuredSlashCommands } from '../../../../shared/structured-agent-session-composer' import { applyMentionSuggestion, EMPTY_HISTORY, @@ -22,6 +20,7 @@ import { useNativeChatSessionOptions } from './use-native-chat-session-options' import { useNativeChatFileAttachmentActions } from './use-native-chat-file-attachment-actions' import { useNativeChatDictationActions } from './use-native-chat-dictation-actions' import { useNativeChatSessionOptionCommand } from './use-native-chat-session-option-command' +import { useNativeChatComposerCatalog } from './use-native-chat-composer-catalog' import { useNativeChatPickerState } from './use-native-chat-picker-state' import { useNativeChatPickerCommandDispatch } from './use-native-chat-picker-command-dispatch' import { useNativeChatTypedInsertion } from './use-native-chat-typed-insertion' @@ -109,10 +108,9 @@ const NativeChatComposerPane = forwardRef - structuredTransport ? structuredSlashCommands(agent) : getVerifiedNativeChatCommands(agent), - [agent, structuredTransport] + const { agentCommands, sessionSkillNames } = useNativeChatComposerCatalog( + agent, + structuredTransport ) const picker = useNativeChatPickerState({ agent, @@ -121,6 +119,7 @@ const NativeChatComposerPane = forwardRef { - afterEach(cleanup) - - it('stays hidden while dispatch state is loading or settled', () => { - const { rerender } = render() - - expect(screen.queryByRole('status')).toBeNull() - - rerender() - expect(screen.queryByRole('status')).toBeNull() - }) - - it.each(['pending', 'dispatched'] as const)( - 'persists recovery guidance for an active %s Dispatch', - (dispatchStatus) => { - render() - - const notice = screen.getByRole('status') - expect(notice.textContent).toContain('Orchestration paused') - expect(notice.textContent).toContain('Structured Chat blocks terminal prompts and sends') - expect(notice.textContent).toContain('Orchestration messages remain queued') - expect(notice.textContent).toContain( - 'switch to Terminal, then check the Orca inbox with orca orchestration check' - ) - } - ) -}) diff --git a/src/renderer/src/components/native-chat/NativeChatOrchestrationPausedNotice.tsx b/src/renderer/src/components/native-chat/NativeChatOrchestrationPausedNotice.tsx deleted file mode 100644 index da3bfec5aaa..00000000000 --- a/src/renderer/src/components/native-chat/NativeChatOrchestrationPausedNotice.tsx +++ /dev/null @@ -1,42 +0,0 @@ -import { PauseCircle } from 'lucide-react' -import type { AgentStatusOrchestrationContext } from '../../../../shared/agent-status-types' -import { Badge } from '@/components/ui/badge' -import { translate } from '@/i18n/i18n' - -export function NativeChatOrchestrationPausedNotice({ - dispatchStatus -}: { - dispatchStatus?: AgentStatusOrchestrationContext['dispatchStatus'] -}): React.JSX.Element | null { - if (dispatchStatus !== 'pending' && dispatchStatus !== 'dispatched') { - return null - } - - return ( -
-
- ) -} diff --git a/src/renderer/src/components/native-chat/NativeChatResolvedView.tsx b/src/renderer/src/components/native-chat/NativeChatResolvedView.tsx index 6f934f66a6e..5f79c492fc9 100644 --- a/src/renderer/src/components/native-chat/NativeChatResolvedView.tsx +++ b/src/renderer/src/components/native-chat/NativeChatResolvedView.tsx @@ -48,10 +48,10 @@ import { } from './use-native-chat-context-menu' import { selectNativeChatRuntimeEnvironmentId } from './native-chat-runtime-owner' import { useNativeChatPasteBridge } from './use-native-chat-paste-bridge' -import { useNativeChatFileLinkClick } from './use-native-chat-file-link-click' +import { LinkActionPopover } from '@/components/link-actions/LinkActionPopover' +import { useNativeChatLinkActions } from './use-native-chat-link-actions' import type { NativeChatResolvedViewProps } from './native-chat-view-types' import { useNativeChatFileLinkContext } from './use-native-chat-file-link-context' -import { NativeChatOrchestrationPausedNotice } from './NativeChatOrchestrationPausedNotice' import { matchNativeChatSplitShortcut } from './native-chat-split-shortcut' import { getShortcutPlatform } from '@/lib/shortcut-platform' import { formatShortcutLabel } from '@/hooks/useShortcutLabel' @@ -68,8 +68,7 @@ export function NativeChatResolvedView({ ownsTabWideLaunchDraft, onSwitchToTerminal, readTerminalScreen, - contextMenuActions, - orchestrationDispatchStatus + contextMenuActions }: NativeChatResolvedViewProps): React.JSX.Element { // Primitive owner selection (no useShallow): routes the pane's read/subscribe to // the remote runtime host for a runtime-owned pane; null keeps the local path. @@ -321,7 +320,11 @@ export function NativeChatResolvedView({ setPending(writePendingSendCache(pendingScope, [])) interactiveSend.cancel() }, [interactiveSend, pendingScope]) - const nativeChatFileLinkClick = useNativeChatFileLinkClick(fileLinkContext) + const { onLinkClick, linkActionRequest, closeLinkActions } = useNativeChatLinkActions( + fileLinkContext, + rootRef, + { sessionId, isVisible } + ) // Chat-only font zoom via Cmd/Ctrl +/-/0, gated to the live conversation so // the chord is inert on the loading/empty/error states and elsewhere. @@ -378,7 +381,6 @@ export function NativeChatResolvedView({ onContextMenuCapture={contextMenu.onContextMenuCapture} className="flex h-full min-h-0 w-full flex-col bg-background focus:outline-none" > -
{viewState.kind === 'loading' ? ( @@ -394,7 +396,7 @@ export function NativeChatResolvedView({ fontScale={fontScale.scale} workingStartedAt={hookWorkingEpoch} showTurnStatus={false} - onLinkClick={nativeChatFileLinkClick} + onLinkClick={onLinkClick} allowFileUriLinks={fileLinkContext !== null} failedDeliveryMessageIds={failedLaunchPromptMessageIds} /> @@ -434,6 +436,7 @@ export function NativeChatResolvedView({ /> )} {contextMenu.menu} +
) } diff --git a/src/renderer/src/components/native-chat/NativeChatStructuredSession.test.tsx b/src/renderer/src/components/native-chat/NativeChatStructuredSession.test.tsx index 2b4a9686aaf..bd8ba9ed7ec 100644 --- a/src/renderer/src/components/native-chat/NativeChatStructuredSession.test.tsx +++ b/src/renderer/src/components/native-chat/NativeChatStructuredSession.test.tsx @@ -203,7 +203,9 @@ describe('NativeChatStructuredSession', () => { ) expect(mocks.messageListProps?.allowFileUriLinks).toBe(true) - expect(mocks.messageListProps?.onLinkClick).toBe(mocks.fileLinkClick) + const event = { preventDefault: vi.fn(), stopPropagation: vi.fn() } + mocks.messageListProps?.onLinkClick?.(event, 'file:///repo/src/a.ts') + expect(mocks.fileLinkClick).toHaveBeenCalledWith(event, 'file:///repo/src/a.ts') }) // Turn status and transcript image previews shipped Codex-first. Every diff --git a/src/renderer/src/components/native-chat/NativeChatStructuredSession.tsx b/src/renderer/src/components/native-chat/NativeChatStructuredSession.tsx index 64f4c7c1253..07aec023d78 100644 --- a/src/renderer/src/components/native-chat/NativeChatStructuredSession.tsx +++ b/src/renderer/src/components/native-chat/NativeChatStructuredSession.tsx @@ -12,11 +12,11 @@ import { NativeChatMessageList } from './NativeChatMessageList' import { NativeChatQuestionCard } from './NativeChatQuestionCard' import { selectNativeChatViewState } from './native-chat-view-state' import { useNativeChatFontScale } from './use-native-chat-font-scale' -import { useNativeChatFileLinkClick } from './use-native-chat-file-link-click' +import { LinkActionPopover } from '@/components/link-actions/LinkActionPopover' +import { useNativeChatLinkActions } from './use-native-chat-link-actions' import { useNativeChatFileLinkContext } from './use-native-chat-file-link-context' import { useStructuredAgentSession } from './use-structured-agent-session' import { translate } from '@/i18n/i18n' -import { NativeChatOrchestrationPausedNotice } from './NativeChatOrchestrationPausedNotice' import { useNativeChatImageRuntimeContext } from './native-chat-image-runtime-context' import { useStructuredNativeChatPaneCommands } from './use-structured-native-chat-pane-commands' import type { NativeChatStructuredViewProps } from './native-chat-view-types' @@ -91,7 +91,11 @@ export function NativeChatStructuredSession( const fontScale = useNativeChatFontScale(viewState.kind === 'ready') const fileLinkContext = useNativeChatFileLinkContext(props.tabId) const imageRuntimeContext = useNativeChatImageRuntimeContext(props.tabId) - const fileLinkClick = useNativeChatFileLinkClick(fileLinkContext) + const { onLinkClick, linkActionRequest, closeLinkActions } = useNativeChatLinkActions( + fileLinkContext, + rootRef, + { sessionId: props.sessionId, isVisible: props.isVisible } + ) const activeStoppingBackgroundTasks = stoppingBackgroundTasks?.sessionId === props.sessionId ? stoppingBackgroundTasks : null const prompt = controller.prompts[0] ?? null @@ -140,11 +144,22 @@ export function NativeChatStructuredSession( optionsSurface: controller.optionSurface, optionSnapshot: controller.optionSnapshot, optionPickerRequest, + sessionCommands: controller.sessionCommands, worktreeId: fileLinkContext?.worktreeId, onError: setComposerError, - runtime: (props.target.kind === 'local' ? 'local' : 'remote') as 'local' | 'remote' + runtime: (props.target.kind === 'local' ? 'local' : 'remote') as 'local' | 'remote', + sessionId: props.sessionId, + runtimeEnvironmentId: + props.target.kind === 'local' ? null : (props.target.environmentId ?? null) }), - [controller, fileLinkContext?.worktreeId, optionPickerRequest, props.agent, props.target.kind] + [ + controller, + fileLinkContext?.worktreeId, + optionPickerRequest, + props.agent, + props.sessionId, + props.target + ] ) return ( @@ -164,7 +179,6 @@ export function NativeChatStructuredSession( onContextMenuCapture={paneCommands.onContextMenuCapture} className="flex h-full min-h-0 w-full flex-col bg-background focus:outline-none" > -
{viewState.kind === 'loading' ? ( @@ -181,8 +195,8 @@ export function NativeChatStructuredSession( workingStartedAt={null} showTurnStatus turnActivity={controller.turnActivity} - onLinkClick={fileLinkClick} - allowFileUriLinks={fileLinkClick !== undefined} + onLinkClick={onLinkClick} + allowFileUriLinks={onLinkClick !== undefined} runtimeContext={imageRuntimeContext} /> )} @@ -343,6 +357,7 @@ export function NativeChatStructuredSession( /> )} {paneCommands.menu} +
) } diff --git a/src/renderer/src/components/native-chat/NativeChatView.tsx b/src/renderer/src/components/native-chat/NativeChatView.tsx index 84c397b8d1d..19ffc82c42f 100644 --- a/src/renderer/src/components/native-chat/NativeChatView.tsx +++ b/src/renderer/src/components/native-chat/NativeChatView.tsx @@ -24,8 +24,7 @@ function NativeChatBridgeView({ ownsTabWideLaunchDraft, onSwitchToTerminal, readTerminalScreen, - contextMenuActions, - orchestrationDispatchStatus + contextMenuActions }: Exclude): React.JSX.Element { const { entry: agentStatusEntry, paneKey } = useNativeChatStatusEntry( terminalTabId, @@ -52,7 +51,6 @@ function NativeChatBridgeView({ onSwitchToTerminal={onSwitchToTerminal} readTerminalScreen={readTerminalScreen} contextMenuActions={contextMenuActions} - orchestrationDispatchStatus={orchestrationDispatchStatus} /> )} diff --git a/src/renderer/src/components/native-chat/StructuredAgentSessionStatusBridge.test.tsx b/src/renderer/src/components/native-chat/StructuredAgentSessionStatusBridge.test.tsx index a8eb22ae7a5..c4376dbf66e 100644 --- a/src/renderer/src/components/native-chat/StructuredAgentSessionStatusBridge.test.tsx +++ b/src/renderer/src/components/native-chat/StructuredAgentSessionStatusBridge.test.tsx @@ -6,15 +6,18 @@ import type { AgentSessionStatusEvent, AgentSessionStatusSummary } from '../../../../shared/agent-session-wire' +import { resolveAttention } from '../sidebar/smart-attention' +import type { AgentStatusEntry } from '../../../../shared/agent-status-types' import type { Tab } from '../../../../shared/tab-types' +import type { AppState } from '@/store/types' import type * as RuntimeRpcClientModule from '@/runtime/runtime-rpc-client' const mocks = vi.hoisted(() => ({ removeAgentStatus: vi.fn(), setAgentStatus: vi.fn(), store: null as null | { - getState: () => Record - setState: (state: Record) => void + getState: () => AppState + setState: (state: Partial & { testRuntimeOwner?: string | null }) => void }, subscribeStatus: vi.fn(), subscribeTranscript: vi.fn(), @@ -23,53 +26,19 @@ const mocks = vi.hoisted(() => ({ })) vi.mock('@/store', async () => { - const { create } = await import('zustand') - const useAppStore = create<{ - agentStatusByPaneKey: Record> - removeAgentStatus: (paneKey: string) => void - setAgentStatus: (...args: unknown[]) => void - testRuntimeOwner: string | null - unifiedTabsByWorktree: Record - }>((set, get) => ({ - agentStatusByPaneKey: {}, - removeAgentStatus: (paneKey) => { - mocks.removeAgentStatus(paneKey) - if (!get().agentStatusByPaneKey[paneKey]) { - return - } - const next = { ...get().agentStatusByPaneKey } - delete next[paneKey] - set({ agentStatusByPaneKey: next }) - }, + const { createTestStore } = await import('@/store/slices/store-test-helpers') + const useAppStore = createTestStore() + const { setAgentStatus, removeAgentStatus } = useAppStore.getState() + useAppStore.setState({ setAgentStatus: (...args) => { mocks.setAgentStatus(...args) - const [paneKey, payload, terminalTitle, , routing, metadata] = args as [ - string, - Record, - string, - unknown, - Record, - Record - ] - set((state) => ({ - agentStatusByPaneKey: { - ...state.agentStatusByPaneKey, - [paneKey]: { - ...payload, - ...routing, - ...metadata, - paneKey, - terminalTitle, - updatedAt: Date.now(), - stateStartedAt: Date.now(), - stateHistory: [] - } - } - })) + setAgentStatus(...args) }, - testRuntimeOwner: null, - unifiedTabsByWorktree: {} - })) + removeAgentStatus: (paneKey) => { + mocks.removeAgentStatus(paneKey) + removeAgentStatus(paneKey) + } + }) mocks.store = useAppStore return { useAppStore } }) @@ -126,7 +95,7 @@ function summary(overrides: Partial = {}): AgentSessi } } -function statuses(): Record[] { +function statuses(): AgentStatusEntry[] { return Object.values(mocks.store?.getState().agentStatusByPaneKey ?? {}) } @@ -218,7 +187,9 @@ describe('StructuredAgentSessionStatusBridge', () => { expect(statuses()).toEqual([expect.objectContaining({ state: 'working' })]) act(() => feed().emit({ type: 'status', session: summary({ status: 'idle', updatedAt: 2 }) })) - expect(statuses()).toEqual([expect.objectContaining({ state: 'done', sessionBoundary: true })]) + expect(statuses()).toEqual([ + expect.objectContaining({ state: 'done', sessionBoundary: false, stateStartedAt: 2 }) + ]) act(() => feed().emit({ type: 'status', session: summary({ status: 'attention', updatedAt: 3 }) }) @@ -226,6 +197,71 @@ describe('StructuredAgentSessionStatusBridge', () => { expect(statuses()).toEqual([expect.objectContaining({ state: 'blocked' })]) }) + it('carries the model, the running tool line, and the last assistant message', async () => { + render() + await waitFor(() => expect(mocks.subscribeStatus).toHaveBeenCalledOnce()) + + act(() => + feed().emit({ + type: 'snapshot', + sessions: [ + summary({ + model: 'gpt-5-codex', + toolName: 'shell', + toolInput: 'pnpm test', + lastAssistantMessage: 'Running the suite now.' + }) + ] + }) + ) + expect(statuses()).toEqual([ + expect.objectContaining({ + model: 'gpt-5-codex', + toolName: 'shell', + toolInput: 'pnpm test', + lastAssistantMessage: 'Running the suite now.' + }) + ]) + + // The tool line describes live work, so a settled turn that omits it must clear it. + act(() => + feed().emit({ + type: 'status', + session: summary({ + status: 'idle', + updatedAt: 2, + model: 'gpt-5-codex', + lastAssistantMessage: 'Suite is green.' + }) + }) + ) + expect(statuses()).toEqual([ + expect.objectContaining({ + state: 'done', + model: 'gpt-5-codex', + lastAssistantMessage: 'Suite is green.' + }) + ]) + expect(statuses()[0]?.toolName).toBeUndefined() + expect(statuses()[0]?.toolInput).toBeUndefined() + + // Only the message moves here, so the row updates only if the guard compares it. + act(() => + feed().emit({ + type: 'status', + session: summary({ + status: 'idle', + updatedAt: 3, + model: 'gpt-5-codex', + lastAssistantMessage: 'Suite is green — 412 passed.' + }) + }) + ) + expect(statuses()).toEqual([ + expect.objectContaining({ lastAssistantMessage: 'Suite is green — 412 passed.' }) + ]) + }) + it('shows no status before a persisted turn', async () => { render() await waitFor(() => expect(mocks.subscribeStatus).toHaveBeenCalledOnce()) @@ -244,8 +280,8 @@ describe('StructuredAgentSessionStatusBridge', () => { const before = mocks.store?.getState().agentStatusByPaneKey act(() => { - for (let updatedAt = 2; updatedAt <= 12; updatedAt += 1) { - feed().emit({ type: 'status', session: summary({ updatedAt }) }) + for (let repeat = 0; repeat < 10; repeat += 1) { + feed().emit({ type: 'status', session: summary() }) } }) @@ -253,6 +289,84 @@ describe('StructuredAgentSessionStatusBridge', () => { expect(mocks.store?.getState().agentStatusByPaneKey).toBe(before) }) + it.each(['claude', 'codex'] as const)( + 'sorts restored %s completions by host time and advances identical turns', + async (agent) => { + const now = Date.now() + mocks.store?.setState({ + unifiedTabsByWorktree: { 'wt-1': [{ ...structuredTab, agentSessionAgent: agent }] } + }) + render() + await waitFor(() => expect(mocks.subscribeStatus).toHaveBeenCalledOnce()) + act(() => + feed().emit({ + type: 'snapshot', + sessions: [summary({ status: 'idle', updatedAt: now - 100 })] + }) + ) + expect(statuses()).toEqual([ + expect.objectContaining({ + state: 'done', + sessionBoundary: false, + stateStartedAt: now - 100, + updatedAt: now - 100 + }) + ]) + act(() => + feed().emit({ type: 'status', session: summary({ status: 'idle', updatedAt: now - 50 }) }) + ) + expect(statuses()).toEqual([ + expect.objectContaining({ stateStartedAt: now - 50, updatedAt: now - 50 }) + ]) + expect( + resolveAttention([{ kind: 'hook', entry: statuses()[0], hasLivePty: false }], now) + ).toEqual({ cls: 2, attentionTimestamp: now - 50 }) + } + ) + + it('preserves the working age when host metadata advances during the same turn', async () => { + render() + await waitFor(() => expect(mocks.subscribeStatus).toHaveBeenCalledOnce()) + act(() => feed().emit({ type: 'status', session: summary({ updatedAt: 100 }) })) + act(() => + feed().emit({ + type: 'status', + session: summary({ updatedAt: 200, providerSession: { ...providerSession, id: 'new-id' } }) + }) + ) + expect(statuses()).toEqual([ + expect.objectContaining({ state: 'working', updatedAt: 200, stateStartedAt: 100 }) + ]) + }) + + it('accepts an authoritative older journal age after a host upgrade reconnect', async () => { + render() + await waitFor(() => expect(mocks.subscribeStatus).toHaveBeenCalledOnce()) + act(() => feed().emit({ type: 'status', session: summary({ updatedAt: 800 }) })) + act(() => + feed().emit({ type: 'snapshot', sessions: [summary({ status: 'idle', updatedAt: 900 })] }) + ) + const paneKey = statuses()[0].paneKey + const history = statuses()[0].stateHistory + const acknowledged = { [paneKey]: 950 } + mocks.store?.setState({ acknowledgedAgentsByPaneKey: acknowledged }) + act(() => + feed().emit({ type: 'snapshot', sessions: [summary({ status: 'idle', updatedAt: 200 })] }) + ) + expect(statuses()).toEqual([ + expect.objectContaining({ state: 'done', updatedAt: 200, stateStartedAt: 200 }) + ]) + const before = mocks.store?.getState().agentStatusByPaneKey + const calls = mocks.setAgentStatus.mock.calls.length + expect(statuses()[0].stateHistory).toBe(history) + expect(mocks.store?.getState().acknowledgedAgentsByPaneKey).toBe(acknowledged) + act(() => + feed().emit({ type: 'snapshot', sessions: [summary({ status: 'idle', updatedAt: 200 })] }) + ) + expect(mocks.store?.getState().agentStatusByPaneKey).toBe(before) + expect(mocks.setAgentStatus).toHaveBeenCalledTimes(calls) + }) + it('drops the status and the feed when the last structured tab closes', async () => { render() await waitFor(() => expect(mocks.subscribeStatus).toHaveBeenCalledOnce()) diff --git a/src/renderer/src/components/native-chat/StructuredAgentSessionStatusBridge.tsx b/src/renderer/src/components/native-chat/StructuredAgentSessionStatusBridge.tsx index d4cb74ab93b..a72f28d7c08 100644 --- a/src/renderer/src/components/native-chat/StructuredAgentSessionStatusBridge.tsx +++ b/src/renderer/src/components/native-chat/StructuredAgentSessionStatusBridge.tsx @@ -75,14 +75,26 @@ function projectStatus(tab: StructuredTab, summary: AgentSessionStatusSummary | : 'done', prompt: summary.latestPrompt, agentType: tab.agentSessionAgent, - sessionBoundary: summary.status === 'idle' + // The host projects these from the journal so the row reads like a hook-reported one: + // the running tool while a turn is live, the agent's last words once it settles. + ...(summary.model ? { model: summary.model } : {}), + ...(summary.toolName ? { toolName: summary.toolName } : {}), + ...(summary.toolInput ? { toolInput: summary.toolInput } : {}), + ...(summary.lastAssistantMessage ? { lastAssistantMessage: summary.lastAssistantMessage } : {}), + sessionBoundary: false } as const const current = store.agentStatusByPaneKey?.[paneKey] if ( current?.state === desired.state && current.prompt === desired.prompt && current.agentType === desired.agentType && + // A row keeps the last model it was told about, so only a reported one can differ. + (summary.model === undefined || current.model === summary.model) && + current.toolName === summary.toolName && + current.toolInput === summary.toolInput && + current.lastAssistantMessage === summary.lastAssistantMessage && current.sessionBoundary === desired.sessionBoundary && + current.updatedAt === summary.updatedAt && current.terminalTitle === tab.label && current.tabId === tab.id && current.worktreeId === tab.worktreeId && @@ -99,7 +111,16 @@ function projectStatus(tab: StructuredTab, summary: AgentSessionStatusSummary | paneKey, desired, tab.label, - undefined, + { + updatedAt: summary.updatedAt, + // This ordered host feed can correct a legacy publication clock after upgrade. + allowOlderTimestamp: true, + stateStartedAt: + desired.state !== 'done' && current?.state === desired.state + ? current.stateStartedAt + : summary.updatedAt, + evidenceObservedAt: Date.now() + }, { tabId: tab.id, worktreeId: tab.worktreeId }, { ...(summary.providerSession ? { providerSession: summary.providerSession } : {}), diff --git a/src/renderer/src/components/native-chat/native-chat-composer-state.test.ts b/src/renderer/src/components/native-chat/native-chat-composer-state.test.ts index df9fa3de5a1..483d29f1ff2 100644 --- a/src/renderer/src/components/native-chat/native-chat-composer-state.test.ts +++ b/src/renderer/src/components/native-chat/native-chat-composer-state.test.ts @@ -16,6 +16,7 @@ import { slashCommandDispatchText, type SlashCommandSuggestion } from './native-chat-composer-state' +import { sessionSlashCommandSuggestions } from '../../../../shared/native-chat-slash-commands' import type { DiscoveredSkill } from '../../../../shared/skills' import { getNativeChatAgentProfile } from '../../../../shared/native-chat-agent-profiles' @@ -236,6 +237,46 @@ describe('native skill and command picker', () => { } }) + it('lets a session report replace the disk scan and enrich the names it knows', () => { + const items = buildNativeChatPickerItems( + [], + [ + skill({ + name: 'ref-oss', + description: 'On disk', + skillFilePath: '/home/ref-oss/SKILL.md', + sourceKind: 'home' + }), + skill({ name: 'stale-on-disk', skillFilePath: '/home/stale/SKILL.md', sourceKind: 'home' }) + ], + '', + '/', + ['dataviz', 'ref-oss'] + ) + // The scanned-but-unreported skill is gone; the reported-but-unscanned one is + // offered without a scope, and sorts after the one the scan located. + expect(items.map((item) => item.name)).toEqual(['ref-oss', 'dataviz']) + expect(items[0]).toMatchObject({ kind: 'skill', description: 'On disk' }) + expect(items[1]).toMatchObject({ kind: 'skill', description: null, sources: [] }) + }) + + it('keeps the disk scan only when a session report is absent', () => { + const items = buildNativeChatPickerItems( + [], + [skill({ name: 'ref-oss', skillFilePath: '/home/ref-oss/SKILL.md' })], + '', + '/', + undefined + ) + expect(items.map((item) => item.name)).toEqual(['ref-oss']) + expect(buildNativeChatPickerItems([], [skill({})], '', '/', [])).toEqual([]) + }) + + it('rejects a session-reported name that is not a safe insertion token', () => { + const items = buildNativeChatPickerItems([], [], '', '/', ['ok', 'two words', 'cle\u200bar']) + expect(items.map((item) => item.name)).toEqual(['ok']) + }) + it('ranks exact, prefix, fuzzy, then description matches within a group', () => { const items = buildNativeChatPickerItems( [], @@ -382,3 +423,31 @@ describe('native skill and command picker', () => { ).toBe('none') }) }) + +it('preserves known skill completion for unclassified session members only', () => { + const commands = sessionSlashCommandSuggestions('claude', [ + { name: 'clear', kind: 'command', kindUnspecified: true }, + { name: 'typescript', kind: 'command', kindUnspecified: true }, + { name: 'project-check', kind: 'command', kindUnspecified: true } + ]) + const diskSkills = [ + skill({ description: 'TypeScript skill' }), + skill({ name: 'not-loaded', skillFilePath: '/not-loaded/SKILL.md' }) + ] + const items = buildNativeChatPickerItems(commands, diskSkills, '', '/', []) + expect(items.map(({ name, kind }) => ({ name, kind }))).toEqual([ + { name: 'clear', kind: 'command' }, + { name: 'project-check', kind: 'command' }, + { name: 'typescript', kind: 'skill' } + ]) + expect(items[2]).toMatchObject({ + description: 'TypeScript skill', + sources: [{ sourceKind: 'repo' }] + }) + const classified = sessionSlashCommandSuggestions('claude', [ + { name: 'typescript', kind: 'command' } + ]) + expect( + buildNativeChatPickerItems(classified, diskSkills, '', '/', []).map(({ kind }) => kind) + ).toEqual(['command']) +}) diff --git a/src/renderer/src/components/native-chat/native-chat-composer-state.ts b/src/renderer/src/components/native-chat/native-chat-composer-state.ts index a14bed089f6..3535a90c04b 100644 --- a/src/renderer/src/components/native-chat/native-chat-composer-state.ts +++ b/src/renderer/src/components/native-chat/native-chat-composer-state.ts @@ -51,11 +51,19 @@ export function deriveComposerAutocomplete( skills: readonly DiscoveredSkill[] = [], profile: NativeChatAgentProfile | null = null, discovery: NativeChatSkillDiscoverySnapshot = { ...EMPTY_DISCOVERY, skills }, - dismissedTriggerKey: string | null = null + dismissedTriggerKey: string | null = null, + sessionSkillNames?: readonly string[] ): ComposerAutocomplete { const before = draft.slice(0, caret) if (before.startsWith('/') && !/\s/.test(before)) { - return deriveSlashAutocomplete(before, agentCommands, profile, discovery, dismissedTriggerKey) + return deriveSlashAutocomplete( + before, + agentCommands, + profile, + discovery, + dismissedTriggerKey, + sessionSkillNames + ) } const mentionMatch = before.match(/(?:^|\s)@(\S*)$/) if (mentionMatch) { @@ -81,7 +89,7 @@ export function deriveComposerAutocomplete( grouped: false, commandsEnabled: false, skillsEnabled: true, - items: buildNativeChatPickerItems([], discovery.skills, query, '$'), + items: buildNativeChatPickerItems([], discovery.skills, query, '$', sessionSkillNames), skillStatus: discovery.status === 'idle' ? 'loading' : discovery.status, ...(discovery.errorKind ? { skillErrorKind: discovery.errorKind } : {}) } @@ -92,7 +100,8 @@ function deriveSlashAutocomplete( agentCommands: readonly SlashCommandSuggestion[], profile: NativeChatAgentProfile | null, discovery: NativeChatSkillDiscoverySnapshot, - dismissedTriggerKey: string | null + dismissedTriggerKey: string | null, + sessionSkillNames: readonly string[] | undefined ): ComposerAutocomplete { const triggerKey = '/:0' if (dismissedTriggerKey === triggerKey) { @@ -106,7 +115,8 @@ function deriveSlashAutocomplete( agentCommands, hasSlashSkills ? discovery.skills : [], query, - '/' + '/', + hasSlashSkills ? sessionSkillNames : [] ) return { mode: 'slash', diff --git a/src/renderer/src/components/native-chat/native-chat-composer-types.ts b/src/renderer/src/components/native-chat/native-chat-composer-types.ts index 9c314f90264..ab3284ebfe3 100644 --- a/src/renderer/src/components/native-chat/native-chat-composer-types.ts +++ b/src/renderer/src/components/native-chat/native-chat-composer-types.ts @@ -1,3 +1,4 @@ +import type { AgentSessionSlashCommand } from '../../../../shared/agent-session-wire' import type { AgentType } from '../../../../shared/agent-status-types' import type { StructuredAgentSessionCommandOutcome } from '../../../../shared/structured-agent-session-composer' import type { @@ -18,9 +19,16 @@ export type NativeChatStructuredComposerTransport = { optionsSurface: SessionOptionsSurface optionSnapshot: SessionOptionDescriptor[] optionPickerRequest?: NativeChatOptionPickerRequest | null + /** The `/` surface the running session reports. Absent keeps the curated + * per-agent catalog, which is what an older host leaves the client with. */ + sessionCommands?: readonly AgentSessionSlashCommand[] worktreeId?: string onError: (message: string | null) => void runtime: 'local' | 'remote' + /** The session behind this composer; a real user send relinquishes orchestration ownership. */ + sessionId: string + /** Owning runtime for that report; null is the local runtime. */ + runtimeEnvironmentId: string | null } export type NativeChatComposerProps = { diff --git a/src/renderer/src/components/native-chat/native-chat-http-link-source-owner.test.ts b/src/renderer/src/components/native-chat/native-chat-http-link-source-owner.test.ts new file mode 100644 index 00000000000..cdd85e37ef4 --- /dev/null +++ b/src/renderer/src/components/native-chat/native-chat-http-link-source-owner.test.ts @@ -0,0 +1,96 @@ +import { afterEach, describe, expect, it, vi } from 'vitest' +import type { AppState } from '@/store/types' +import { + canNativeChatOpenOwnedBrowser, + resolveNativeChatHttpLinkSourceOwner +} from './native-chat-http-link-source-owner' + +const mocks = vi.hoisted(() => ({ + getRuntimeEnvironmentIdForWorktree: vi.fn(), + getConnectionIdFromState: vi.fn(), + canOpenWorkspaceBrowserTabOnRuntime: vi.fn(), + canOpenWorkspaceBrowserTabOnSsh: vi.fn() +})) + +vi.mock('@/lib/worktree-runtime-owner', () => ({ + getRuntimeEnvironmentIdForWorktree: mocks.getRuntimeEnvironmentIdForWorktree +})) +vi.mock('@/lib/connection-owner-resolution', () => ({ + getConnectionIdFromState: mocks.getConnectionIdFromState +})) +vi.mock('@/lib/workspace-browser-tab-open', () => ({ + canOpenWorkspaceBrowserTabOnRuntime: mocks.canOpenWorkspaceBrowserTabOnRuntime, + canOpenWorkspaceBrowserTabOnSsh: mocks.canOpenWorkspaceBrowserTabOnSsh +})) + +const state = {} as AppState + +afterEach(() => { + vi.clearAllMocks() +}) + +describe('resolveNativeChatHttpLinkSourceOwner', () => { + it('prefers the workspace runtime owner', () => { + mocks.getRuntimeEnvironmentIdForWorktree.mockReturnValue('env-1') + + expect(resolveNativeChatHttpLinkSourceOwner(state, 'wt-1')).toEqual({ + kind: 'runtime', + runtimeEnvironmentId: 'env-1' + }) + expect(mocks.getConnectionIdFromState).not.toHaveBeenCalled() + }) + + it('falls back to the SSH connection that owns the workspace', () => { + mocks.getRuntimeEnvironmentIdForWorktree.mockReturnValue(null) + mocks.getConnectionIdFromState.mockReturnValue('ssh-1') + + expect(resolveNativeChatHttpLinkSourceOwner(state, 'wt-1')).toEqual({ + kind: 'ssh', + connectionId: 'ssh-1' + }) + }) + + it('reads a null connection as local', () => { + mocks.getRuntimeEnvironmentIdForWorktree.mockReturnValue(null) + mocks.getConnectionIdFromState.mockReturnValue(null) + + expect(resolveNativeChatHttpLinkSourceOwner(state, 'wt-1')).toEqual({ kind: 'local' }) + }) + + // An unresolved owner must not be mistaken for local: a remote link would then + // open against the wrong host. + it('reports an unresolved owner as unknown', () => { + mocks.getRuntimeEnvironmentIdForWorktree.mockReturnValue(null) + mocks.getConnectionIdFromState.mockReturnValue(undefined) + + expect(resolveNativeChatHttpLinkSourceOwner(state, 'wt-1')).toEqual({ kind: 'unknown' }) + }) +}) + +describe('canNativeChatOpenOwnedBrowser', () => { + it('asks the runtime browser-route check for a runtime owner', () => { + mocks.canOpenWorkspaceBrowserTabOnRuntime.mockReturnValue(true) + + expect( + canNativeChatOpenOwnedBrowser(state, 'wt-1', { + kind: 'runtime', + runtimeEnvironmentId: 'env-1' + }) + ).toBe(true) + expect(mocks.canOpenWorkspaceBrowserTabOnRuntime).toHaveBeenCalledWith(state, 'wt-1', 'env-1') + }) + + it('asks the SSH browser-route check for an SSH owner', () => { + mocks.canOpenWorkspaceBrowserTabOnSsh.mockReturnValue(false) + + expect( + canNativeChatOpenOwnedBrowser(state, 'wt-1', { kind: 'ssh', connectionId: 'ssh-1' }) + ).toBe(false) + expect(mocks.canOpenWorkspaceBrowserTabOnSsh).toHaveBeenCalledWith(state, 'wt-1', 'ssh-1') + }) + + it('never claims an owned browser for local or unknown owners', () => { + expect(canNativeChatOpenOwnedBrowser(state, 'wt-1', { kind: 'local' })).toBe(false) + expect(canNativeChatOpenOwnedBrowser(state, 'wt-1', { kind: 'unknown' })).toBe(false) + }) +}) diff --git a/src/renderer/src/components/native-chat/native-chat-http-link-source-owner.ts b/src/renderer/src/components/native-chat/native-chat-http-link-source-owner.ts new file mode 100644 index 00000000000..8028ff2ca3d --- /dev/null +++ b/src/renderer/src/components/native-chat/native-chat-http-link-source-owner.ts @@ -0,0 +1,40 @@ +import { getConnectionIdFromState } from '@/lib/connection-owner-resolution' +import type { HttpLinkSourceOwner } from '@/lib/http-link-routing' +import { + canOpenWorkspaceBrowserTabOnRuntime, + canOpenWorkspaceBrowserTabOnSsh +} from '@/lib/workspace-browser-tab-open' +import { getRuntimeEnvironmentIdForWorktree } from '@/lib/worktree-runtime-owner' +import type { AppState } from '@/store/types' + +/** The chat transcript has no PTY, so link ownership comes from the session's + * workspace: a runtime id wins, then an SSH connection; an unresolved owner + * stays 'unknown' rather than claiming local. */ +export function resolveNativeChatHttpLinkSourceOwner( + state: AppState, + worktreeId: string +): HttpLinkSourceOwner { + const runtimeEnvironmentId = getRuntimeEnvironmentIdForWorktree(state, worktreeId) + if (runtimeEnvironmentId) { + return { kind: 'runtime', runtimeEnvironmentId } + } + const connectionId = getConnectionIdFromState(state, worktreeId) + if (connectionId === undefined) { + return { kind: 'unknown' } + } + return connectionId === null ? { kind: 'local' } : { kind: 'ssh', connectionId } +} + +export function canNativeChatOpenOwnedBrowser( + state: AppState, + worktreeId: string, + sourceOwner: HttpLinkSourceOwner +): boolean { + if (sourceOwner.kind === 'runtime') { + return canOpenWorkspaceBrowserTabOnRuntime(state, worktreeId, sourceOwner.runtimeEnvironmentId) + } + return ( + sourceOwner.kind === 'ssh' && + canOpenWorkspaceBrowserTabOnSsh(state, worktreeId, sourceOwner.connectionId) + ) +} diff --git a/src/renderer/src/components/native-chat/native-chat-picker-items.ts b/src/renderer/src/components/native-chat/native-chat-picker-items.ts index 6938c2078a4..427e3b74497 100644 --- a/src/renderer/src/components/native-chat/native-chat-picker-items.ts +++ b/src/renderer/src/components/native-chat/native-chat-picker-items.ts @@ -47,13 +47,20 @@ export function buildNativeChatPickerItems( commands: readonly SlashCommandSuggestion[], skills: readonly DiscoveredSkill[], query: string, - prefix: '/' | '$' + prefix: '/' | '$', + sessionSkillNames?: readonly string[] ): NativeChatPickerItem[] { - const mergedSkills = mergeNativeChatSkills(skills) + const unclassifiedNames = new Set( + commands.filter((command) => command.kindUnspecified).map((command) => command.name) + ) + const mergedSkills = mergeNativeChatSkills(skills, sessionSkillNames, unclassifiedNames) const skillNames = new Set(mergedSkills.map((skill) => skill.name)) - const commandNames = new Set(commands.map((command) => command.name)) + const resolvedCommands = commands.filter( + (command) => !(command.kindUnspecified && skillNames.has(command.name)) + ) + const commandNames = new Set(resolvedCommands.map((command) => command.name)) const commandItems = rankItems( - commands.map((command, index) => ({ + resolvedCommands.map((command, index) => ({ item: { kind: 'command' as const, // Why: the name is the dispatch token and the catalog is curated, so @@ -80,7 +87,9 @@ export function buildNativeChatPickerItems( } function mergeNativeChatSkills( - skills: readonly DiscoveredSkill[] + skills: readonly DiscoveredSkill[], + sessionSkillNames: readonly string[] | undefined, + unclassifiedNames: ReadonlySet ): Extract[] { const exactPaths = new Map() for (const skill of skills) { @@ -96,23 +105,43 @@ function mergeNativeChatSkills( } byName.set(safeName, [...(byName.get(safeName) ?? []), { ...skill, name: safeName }]) } - return [...byName.entries()] - .map(([name, namedSkills]) => { - const sorted = [...namedSkills].sort(compareDiscoveredSkills) - return { - kind: 'skill' as const, - id: `skill:${name}`, - name, - description: sorted[0]?.description ? sanitizePickerText(sorted[0].description, 240) : null, - sources: sorted.map((skill) => ({ - sourceKind: skill.sourceKind, - skillFilePath: skill.skillFilePath - })) - } - }) + const discovered = new Map( + [...byName.entries()].map(([name, namedSkills]) => [name, pickerSkill(name, namedSkills)]) + ) + // Why: when the running session reports its own skills, that report is the + // authority on which ones exist — a disk scan cannot see what the session + // actually loaded (plugin roots, setting-source filters), and a scanned root + // the session ignored must not be offered. The scan stays the source of + // description and scope for the names both know about. + const names = + sessionSkillNames !== undefined + ? [ + ...sessionSkillNames.filter(isTokenSafe), + ...[...discovered.keys()].filter((name) => unclassifiedNames.has(name)) + ] + : [...discovered.keys()] + return [...new Set(names)] + .map((name) => discovered.get(name) ?? pickerSkill(name, [])) .sort(comparePickerSkills) } +function pickerSkill( + name: string, + namedSkills: readonly DiscoveredSkill[] +): Extract { + const sorted = [...namedSkills].sort(compareDiscoveredSkills) + return { + kind: 'skill' as const, + id: `skill:${name}`, + name, + description: sorted[0]?.description ? sanitizePickerText(sorted[0].description, 240) : null, + sources: sorted.map((skill) => ({ + sourceKind: skill.sourceKind, + skillFilePath: skill.skillFilePath + })) + } +} + function rankItems( entries: { item: T; stableOrder: number }[], query: string @@ -197,12 +226,21 @@ function compareDiscoveredSkills(a: DiscoveredSkill, b: DiscoveredSkill): number ) } +// A session-reported skill this host could not locate on disk sorts last: it is +// real and invocable, but carries no scope or description to rank on. +const UNLOCATED_SCOPE_PRIORITY = Object.keys(SCOPE_PRIORITY).length + +function skillScopePriority(item: Extract): number { + const sourceKind = item.sources[0]?.sourceKind + return sourceKind === undefined ? UNLOCATED_SCOPE_PRIORITY : SCOPE_PRIORITY[sourceKind] +} + function comparePickerSkills( a: Extract, b: Extract ): number { return ( - SCOPE_PRIORITY[a.sources[0].sourceKind] - SCOPE_PRIORITY[b.sources[0].sourceKind] || + skillScopePriority(a) - skillScopePriority(b) || compareBaseSensitivityLocaleText(a.name, b.name) ) } diff --git a/src/renderer/src/components/native-chat/native-chat-retire-persisted-model.test.ts b/src/renderer/src/components/native-chat/native-chat-retire-persisted-model.test.ts index 2ac8eac3a16..9dc3e403482 100644 --- a/src/renderer/src/components/native-chat/native-chat-retire-persisted-model.test.ts +++ b/src/renderer/src/components/native-chat/native-chat-retire-persisted-model.test.ts @@ -3,7 +3,6 @@ import { renderHook } from '@testing-library/react' import { beforeEach, describe, expect, it, vi } from 'vitest' import type { CatalogModel } from '../../../../shared/agent-session-option-catalog' -import type { PersistedNativeChatSessionOptions } from '../../../../shared/native-chat-session-options' import { clearNativeChatModelEnrichmentForTests, ensureNativeChatModelEnrichment, @@ -11,15 +10,14 @@ import { } from './native-chat-session-option-enrichment' const mocks = vi.hoisted(() => ({ - storeState: { - settings: {} as { nativeChatSessionOptions?: PersistedNativeChatSessionOptions }, - updateSettings: vi.fn() - }, + callRuntimeRpc: vi.fn(), createNativeChatPtySessionOptions: vi.fn(), discoverNativeChatCatalogModels: vi.fn() })) -vi.mock('../../store', () => ({ useAppStore: { getState: () => mocks.storeState } })) +vi.mock('@/runtime/runtime-rpc-client', () => ({ + callRuntimeRpc: mocks.callRuntimeRpc +})) vi.mock('./native-chat-pty-session-options', () => ({ createNativeChatPtySessionOptions: mocks.createNativeChatPtySessionOptions @@ -36,79 +34,63 @@ const { retirePersistedModelMissingFromDiscovery, useNativeChatSessionOptions } const models = (...ids: string[]): CatalogModel[] => ids.map((id) => ({ id, label: id, options: [] })) -function persist(options: PersistedNativeChatSessionOptions): void { - mocks.storeState.settings = { nativeChatSessionOptions: options } -} +const LOCAL_TARGET = { kind: 'local' } as const /** The persisted model becomes `-m ` at every launch site, including ones that * never render the picker, and grok exits fatally on an id it no longer lists. */ describe('retirePersistedModelMissingFromDiscovery', () => { beforeEach(() => { - mocks.storeState.updateSettings.mockReset().mockResolvedValue(undefined) - mocks.storeState.settings = {} + mocks.callRuntimeRpc.mockReset().mockResolvedValue({ ok: true }) }) it('clears a persisted id the authoritative probe no longer lists', async () => { - persist({ grok: { model: 'grok-build', valuesByModel: { 'grok-build': { effort: 'low' } } } }) await retirePersistedModelMissingFromDiscovery('grok', models('grok-4.5')) - expect(mocks.storeState.updateSettings).toHaveBeenCalledWith({ - // Why keep valuesByModel: the option values are still valid if the user - // reselects that model on another host. - nativeChatSessionOptions: { - grok: { valuesByModel: { 'grok-build': { effort: 'low' } } } + expect(mocks.callRuntimeRpc).toHaveBeenCalledWith( + LOCAL_TARGET, + 'settings.mutateNativeChatSessionOptions', + { + type: 'clear-model-if-missing', + agent: 'grok', + availableModelIds: ['grok-4.5'] } - }) + ) }) - it('keeps a persisted id the probe still lists', async () => { - persist({ grok: { model: 'grok-4.5' } }) + it('lets the host keep a concurrently selected model from the available list', async () => { await retirePersistedModelMissingFromDiscovery('grok', models('grok-4.5', 'grok-build')) - expect(mocks.storeState.updateSettings).not.toHaveBeenCalled() + expect(mocks.callRuntimeRpc).toHaveBeenCalledWith( + LOCAL_TARGET, + 'settings.mutateNativeChatSessionOptions', + { + type: 'clear-model-if-missing', + agent: 'grok', + availableModelIds: ['grok-4.5', 'grok-build'] + } + ) }) it('treats an empty list as a failed probe, not an empty account', async () => { - persist({ grok: { model: 'grok-build' } }) await retirePersistedModelMissingFromDiscovery('grok', []) - expect(mocks.storeState.updateSettings).not.toHaveBeenCalled() + expect(mocks.callRuntimeRpc).not.toHaveBeenCalled() }) it('leaves additive agents alone, whose lists extend the seed rather than replace it', async () => { - // Cursor's probe not listing a model is no evidence the model is gone. - persist({ cursor: { model: 'gpt-5.3-codex' } }) await retirePersistedModelMissingFromDiscovery('cursor', models('auto')) - expect(mocks.storeState.updateSettings).not.toHaveBeenCalled() + expect(mocks.callRuntimeRpc).not.toHaveBeenCalled() }) - it('does nothing when no model was ever persisted', async () => { - persist({ grok: { valuesByModel: { 'grok-4.5': { effort: 'high' } } } }) - await retirePersistedModelMissingFromDiscovery('grok', models('grok-4.5')) - expect(mocks.storeState.updateSettings).not.toHaveBeenCalled() - }) - - it('survives settings that were never written', async () => { - mocks.storeState.settings = {} + it('does not depend on a client-local settings snapshot', async () => { await expect( retirePersistedModelMissingFromDiscovery('grok', models('grok-4.5')) ).resolves.toBeUndefined() - expect(mocks.storeState.updateSettings).not.toHaveBeenCalled() + expect(mocks.callRuntimeRpc).toHaveBeenCalledOnce() }) - it('re-reads live settings at apply so a pick landing mid-retirement survives', async () => { - // Regression: retirement captured the settings snapshot before its write was - // queued, so a pick landing in between was clobbered back to the old shape. - persist({ grok: { model: 'grok-build' } }) - const pending = retirePersistedModelMissingFromDiscovery('grok', models('grok-5')) - persist({ grok: { model: 'grok-5' } }) - await pending - expect(mocks.storeState.updateSettings).not.toHaveBeenCalled() - }) - - it('retires only the named agent, leaving other agents’ picks intact', async () => { - persist({ grok: { model: 'grok-build' }, claude: { model: 'opus' } }) - await retirePersistedModelMissingFromDiscovery('grok', models('grok-4.5')) - expect(mocks.storeState.updateSettings).toHaveBeenCalledWith({ - nativeChatSessionOptions: { grok: {}, claude: { model: 'opus' } } - }) + it('swallows a failed best-effort retirement write', async () => { + mocks.callRuntimeRpc.mockRejectedValue(new Error('runtime offline')) + await expect( + retirePersistedModelMissingFromDiscovery('grok', models('grok-4.5')) + ).resolves.toBeUndefined() }) }) @@ -128,8 +110,7 @@ describe('useNativeChatSessionOptions retirement on mount', () => { beforeEach(() => { clearNativeChatModelEnrichmentForTests() - mocks.storeState.updateSettings.mockReset().mockResolvedValue(undefined) - mocks.storeState.settings = {} + mocks.callRuntimeRpc.mockReset().mockResolvedValue({ ok: true }) mocks.discoverNativeChatCatalogModels.mockReset().mockResolvedValue(null) // A stable snapshot reference: useSyncExternalStore re-renders forever otherwise. const emptySnapshot: never[] = [] @@ -143,7 +124,6 @@ describe('useNativeChatSessionOptions retirement on mount', () => { }) it('retires a persisted id against models the probe already cached', async () => { - persist({ grok: { model: 'grok-build' } }) ensureNativeChatModelEnrichment({ agent: 'grok', hostKey: 'local', @@ -154,19 +134,46 @@ describe('useNativeChatSessionOptions retirement on mount', () => { mountPane() await vi.waitFor(() => - expect(mocks.storeState.updateSettings).toHaveBeenCalledWith({ - nativeChatSessionOptions: { grok: {} } - }) + expect(mocks.callRuntimeRpc).toHaveBeenCalledWith( + LOCAL_TARGET, + 'settings.mutateNativeChatSessionOptions', + expect.objectContaining({ type: 'clear-model-if-missing', agent: 'grok' }) + ) ) }) it('leaves the persisted id alone while the probe is still in flight', async () => { - persist({ grok: { model: 'grok-build' } }) mocks.discoverNativeChatCatalogModels.mockReturnValue(new Promise(() => {})) mountPane() await Promise.resolve() - expect(mocks.storeState.updateSettings).not.toHaveBeenCalled() + expect(mocks.callRuntimeRpc).not.toHaveBeenCalled() + }) + + it('keeps PTY picks in the client settings record used by paired launches', async () => { + mountPane() + const persistSelection = mocks.createNativeChatPtySessionOptions.mock.calls[0]?.[0] + ?.persistSelection as + | ((pick: { + modelId: string + optionId: string + value: string + adoptModelAsLaunchDefault: boolean + }) => Promise) + | undefined + + await persistSelection?.({ + modelId: 'grok-4.5', + optionId: 'effort', + value: 'high', + adoptModelAsLaunchDefault: true + }) + + expect(mocks.callRuntimeRpc).toHaveBeenCalledWith( + LOCAL_TARGET, + 'settings.mutateNativeChatSessionOptions', + expect.objectContaining({ type: 'apply-picks', agent: 'grok' }) + ) }) }) diff --git a/src/renderer/src/components/native-chat/native-chat-session-option-settings-write.ts b/src/renderer/src/components/native-chat/native-chat-session-option-settings-write.ts new file mode 100644 index 00000000000..fa4fb3d14a9 --- /dev/null +++ b/src/renderer/src/components/native-chat/native-chat-session-option-settings-write.ts @@ -0,0 +1,15 @@ +import type { NativeChatSessionOptionSettingsMutation } from '../../../../shared/native-chat-session-options' +import { callRuntimeRpc, type RuntimeClientTarget } from '@/runtime/runtime-rpc-client' + +/** + * The executing runtime applies deltas to its latest record. This keeps paired-runtime + * choices on their owner and prevents desktop/mobile writes from replacing one another. + */ +export function enqueueSessionOptionSettingsWrite( + target: RuntimeClientTarget, + mutation: NativeChatSessionOptionSettingsMutation +): Promise { + return callRuntimeRpc(target, 'settings.mutateNativeChatSessionOptions', mutation) + .then(() => undefined) + .catch(() => undefined) +} diff --git a/src/renderer/src/components/native-chat/native-chat-structured-send-composition-clear.test.tsx b/src/renderer/src/components/native-chat/native-chat-structured-send-composition-clear.test.tsx index 216472ac8a6..a6298530b9e 100644 --- a/src/renderer/src/components/native-chat/native-chat-structured-send-composition-clear.test.tsx +++ b/src/renderer/src/components/native-chat/native-chat-structured-send-composition-clear.test.tsx @@ -93,6 +93,8 @@ function transport( optionSnapshot: [], onError: vi.fn(), runtime: 'remote', + sessionId: 'session-test', + runtimeEnvironmentId: null, ...overrides } } diff --git a/src/renderer/src/components/native-chat/native-chat-turn-activity.test.ts b/src/renderer/src/components/native-chat/native-chat-turn-activity.test.ts index f2d30101826..a819e3054a2 100644 --- a/src/renderer/src/components/native-chat/native-chat-turn-activity.test.ts +++ b/src/renderer/src/components/native-chat/native-chat-turn-activity.test.ts @@ -34,6 +34,68 @@ describe('selectStructuredAgentTurnActivity', () => { expect(activity).toEqual({ kind: 'description', text: 'Preparing the answer' }) }) + it('prefers matching ephemeral provider activity over journal-derived status', () => { + const activity = selectStructuredAgentTurnActivity( + [turnStart, item(2, { kind: 'status', text: 'Older journal status' })], + 'turn-1', + { turnId: 'turn-1', text: 'Inspecting the session wire' } + ) + + expect(activity).toEqual({ kind: 'description', text: 'Inspecting the session wire' }) + }) + + it('ignores ephemeral activity from another or settled turn', () => { + const providerActivity = { turnId: 'turn-1', text: 'Inspecting the session wire' } + + expect(selectStructuredAgentTurnActivity([turnStart], 'turn-2', providerActivity)).toBeNull() + expect(selectStructuredAgentTurnActivity([turnStart], null, providerActivity)).toBeNull() + }) + + it.each([ + ['active', 'Still running pnpm test'], + ['most recently settled', 'Running shell pnpm lint now'] + ])('never repeats the %s tool label as provider activity', (_kind, text) => { + const activity = selectStructuredAgentTurnActivity( + [ + turnStart, + item(2, { + kind: 'tool-call', + name: 'shell', + input: { command: 'pnpm test' }, + state: 'running' + }), + item(3, { + kind: 'tool-call', + name: 'shell', + input: { command: 'pnpm lint' }, + state: 'completed' + }) + ], + 'turn-1', + { turnId: 'turn-1', text } + ) + + expect(activity).toBeNull() + }) + + it('does not fall through to a journal status that repeats a recent tool label', () => { + const activity = selectStructuredAgentTurnActivity( + [ + turnStart, + item(2, { + kind: 'tool-call', + name: 'shell', + input: { command: 'pnpm lint' }, + state: 'completed' + }), + item(3, { kind: 'status', text: 'Running pnpm lint' }) + ], + 'turn-1' + ) + + expect(activity).toBeNull() + }) + it('ignores active and settled tools so the tail can use a broad fallback', () => { const activity = selectStructuredAgentTurnActivity( [ diff --git a/src/renderer/src/components/native-chat/native-chat-turn-activity.ts b/src/renderer/src/components/native-chat/native-chat-turn-activity.ts index 37f9fc75015..1444e535a2f 100644 --- a/src/renderer/src/components/native-chat/native-chat-turn-activity.ts +++ b/src/renderer/src/components/native-chat/native-chat-turn-activity.ts @@ -1,5 +1,10 @@ import type { AgentJournalRenderItem } from '../../../../shared/agent-session-journal-types' +import type { AgentSessionTurnActivity } from '../../../../shared/agent-session-wire' import { normalizePromptField } from '../../../../shared/agent-status-field-normalization' +import { + describeActiveToolCall, + formatActiveToolLabel +} from '../../../../shared/native-chat-tool-activity' export type NativeChatTurnActivity = { kind: 'description'; text: string } @@ -12,10 +17,62 @@ function activityLine(text: string): string | null { return latest ? normalizePromptField(latest) || null : null } +function recentToolActivityLabels(items: readonly AgentJournalRenderItem[]): Set { + const labels = new Set() + let foundRunning = false + let foundSettled = false + for (let index = items.length - 1; index >= 0 && (!foundRunning || !foundSettled); index -= 1) { + const body = items[index]?.body + if (body?.kind !== 'tool-call') { + continue + } + const isRunning = body.state === 'running' + if ((isRunning && foundRunning) || (!isRunning && foundSettled)) { + continue + } + const descriptor = describeActiveToolCall({ + type: 'tool-call', + name: body.name, + input: body.input, + state: body.state + }) + const candidates = [ + formatActiveToolLabel(descriptor), + descriptor.preview, + descriptor.preview ? `${descriptor.toolName} ${descriptor.preview}` : descriptor.toolName + ] + for (const candidate of candidates) { + const label = activityLine(candidate)?.toLowerCase() + if (label) { + labels.add(label) + } + } + foundRunning ||= isRunning + foundSettled ||= !isRunning + } + return labels +} + +function repeatsRecentToolLabel(text: string, labels: ReadonlySet): boolean { + const normalized = text.toLowerCase() + for (const label of labels) { + if ( + normalized === label || + normalized.startsWith(`${label} `) || + normalized.endsWith(` ${label}`) || + normalized.includes(` ${label} `) + ) { + return true + } + } + return false +} + /** Prefer provider-authored activity copy; callers provide the broad fallback. */ export function selectStructuredAgentTurnActivity( items: readonly AgentJournalRenderItem[], - turnId: string | null + turnId: string | null, + providerActivity?: AgentSessionTurnActivity | null ): NativeChatTurnActivity | null { if (!turnId) { return null @@ -27,13 +84,20 @@ export function selectStructuredAgentTurnActivity( item.body.turnLifecycle.state === 'running' ) const turnItems = items.slice(Math.max(0, turnStartIndex)) + const toolLabels = recentToolActivityLabels(turnItems) + if (providerActivity?.turnId === turnId) { + const text = activityLine(providerActivity.text) + if (text && !repeatsRecentToolLabel(text, toolLabels)) { + return { kind: 'description', text } + } + } for (let index = turnItems.length - 1; index >= 0; index -= 1) { const body = turnItems[index]?.body if (body?.kind !== 'status' || body.turnLifecycle || body.providerFrame) { continue } const text = activityLine(body.text) - if (text) { + if (text && !repeatsRecentToolLabel(text, toolLabels)) { return { kind: 'description', text } } } diff --git a/src/renderer/src/components/native-chat/native-chat-view-types.ts b/src/renderer/src/components/native-chat/native-chat-view-types.ts index 920bede7028..1a519b5a1a2 100644 --- a/src/renderer/src/components/native-chat/native-chat-view-types.ts +++ b/src/renderer/src/components/native-chat/native-chat-view-types.ts @@ -1,17 +1,10 @@ -import type { - AgentStatusOrchestrationContext, - AgentType -} from '../../../../shared/agent-status-types' +import type { AgentType } from '../../../../shared/agent-status-types' import type { TuiAgent } from '../../../../shared/tui-agent' import type { RuntimeClientTarget } from '@/runtime/runtime-rpc-client' import type { NativeChatSession } from '../../../../shared/native-chat-types' import type { NativeChatContextMenuActions } from './use-native-chat-context-menu' -type NativeChatOrchestrationProps = { - orchestrationDispatchStatus?: AgentStatusOrchestrationContext['dispatchStatus'] -} - -export type NativeChatBridgeViewProps = NativeChatOrchestrationProps & { +export type NativeChatBridgeViewProps = { mode?: 'bridge' /** The terminal tab hosting the agent. paneKey is `${tabId}:${leafId}`. */ terminalTabId: string @@ -34,7 +27,7 @@ export type NativeChatBridgeViewProps = NativeChatOrchestrationProps & { contextMenuActions?: Omit } -export type NativeChatStructuredViewProps = NativeChatOrchestrationProps & { +export type NativeChatStructuredViewProps = { mode: 'structured' tabId: string groupId?: string @@ -45,7 +38,7 @@ export type NativeChatStructuredViewProps = NativeChatOrchestrationProps & { contextMenuActions?: Omit } -export type NativeChatResolvedViewProps = NativeChatOrchestrationProps & { +export type NativeChatResolvedViewProps = { paneKey: string agent: NativeChatSession['agent'] sessionId: string | null diff --git a/src/renderer/src/components/native-chat/native-chat-web-link-actions.test.ts b/src/renderer/src/components/native-chat/native-chat-web-link-actions.test.ts new file mode 100644 index 00000000000..447e8e86831 --- /dev/null +++ b/src/renderer/src/components/native-chat/native-chat-web-link-actions.test.ts @@ -0,0 +1,182 @@ +import { afterEach, describe, expect, it, vi } from 'vitest' +import type { LinkActionRequest } from '@/components/link-actions/link-action-request' +import type * as HttpLinkDestinations from '@/lib/http-link-destinations' +import type { HttpLinkActionDestinations } from '@/lib/http-link-destinations' +import { handleNativeChatWebLink } from './native-chat-web-link-actions' + +const mocks = vi.hoisted(() => ({ openRoutedHttpLink: vi.fn() })) + +vi.mock('@/lib/http-link-destinations', async (importOriginal) => ({ + ...(await importOriginal()), + openRoutedHttpLink: mocks.openRoutedHttpLink +})) + +function stubPlatform(isMac: boolean): void { + vi.stubGlobal('navigator', { userAgent: isMac ? 'Mac OS X' : 'Windows NT 10.0' }) +} + +type ClickInit = { + metaKey?: boolean + ctrlKey?: boolean + shiftKey?: boolean + altKey?: boolean + button?: number +} + +function click(init: ClickInit = {}) { + return { + altKey: false, + ctrlKey: false, + metaKey: false, + shiftKey: false, + button: 0, + clientX: 120, + clientY: 240, + preventDefault: vi.fn(), + ...init + } +} + +function deps( + overrides: { + destinations?: HttpLinkActionDestinations + actionsEnabled?: boolean + } = {} +) { + const requests: LinkActionRequest[] = [] + return { + requests, + deps: { + worktreeId: 'wt-1', + sourceOwner: { kind: 'local' } as const, + destinations: overrides.destinations ?? { primary: 'system', alternate: 'orca' }, + actionsEnabled: overrides.actionsEnabled ?? true, + restoreFocus: vi.fn(), + request: (request: LinkActionRequest) => requests.push(request) + } + } +} + +afterEach(() => { + vi.clearAllMocks() + vi.unstubAllGlobals() +}) + +describe('handleNativeChatWebLink', () => { + it('anchors keyboard activation to the focused link', () => { + stubPlatform(true) + const { deps: d, requests } = deps() + handleNativeChatWebLink( + { + ...click(), + detail: 0, + currentTarget: { + getBoundingClientRect: () => ({ left: 80, bottom: 160 }) as DOMRect + } + }, + 'https://example.com', + d + ) + expect(requests[0]).toMatchObject({ anchorX: 80, anchorY: 160 }) + }) + + it('opens the destination popover on a plain click', () => { + stubPlatform(true) + const { deps: d, requests } = deps() + const event = click() + + expect(handleNativeChatWebLink(event, 'https://example.com/', d)).toBe(true) + expect(event.preventDefault).toHaveBeenCalledOnce() + expect(mocks.openRoutedHttpLink).not.toHaveBeenCalled() + expect(requests).toHaveLength(1) + expect(requests[0]).toMatchObject({ + anchorX: 120, + anchorY: 240, + destination: 'https://example.com/', + kind: 'url' + }) + expect(requests[0]?.primary.label).toBe('System Browser') + expect(requests[0]?.alternate?.label).toBe('Orca Browser') + }) + + it('routes the popover actions to their destinations', () => { + stubPlatform(true) + const { deps: d, requests } = deps() + handleNativeChatWebLink(click(), 'https://example.com/', d) + + void requests[0]?.alternate?.run() + expect(mocks.openRoutedHttpLink).toHaveBeenCalledWith('https://example.com/', { + worktreeId: 'wt-1', + sourceOwner: { kind: 'local' }, + modifierHeld: false, + forceDestination: 'orca' + }) + }) + + it('opens the primary destination directly on a modifier click', () => { + stubPlatform(true) + const { deps: d, requests } = deps() + const event = click({ metaKey: true }) + + expect(handleNativeChatWebLink(event, 'https://example.com/', d)).toBe(true) + expect(requests).toHaveLength(0) + expect(mocks.openRoutedHttpLink).toHaveBeenCalledWith( + 'https://example.com/', + expect.objectContaining({ forceDestination: 'system' }) + ) + }) + + it('opens the alternate destination on a shift+modifier click', () => { + stubPlatform(false) + const { deps: d } = deps() + + expect(handleNativeChatWebLink(click({ ctrlKey: true, shiftKey: true }), 'https://a/', d)).toBe( + true + ) + expect(mocks.openRoutedHttpLink).toHaveBeenCalledWith( + 'https://a/', + expect.objectContaining({ forceDestination: 'orca' }) + ) + }) + + it('falls back to the primary destination when no alternate is offered', () => { + stubPlatform(true) + const { deps: d } = deps({ destinations: { primary: 'system' } }) + + handleNativeChatWebLink(click({ metaKey: true, shiftKey: true }), 'https://a/', d) + expect(mocks.openRoutedHttpLink).toHaveBeenCalledWith( + 'https://a/', + expect.objectContaining({ forceDestination: 'system' }) + ) + }) + + it('opens the link outright on a plain click when link actions are disabled', () => { + stubPlatform(true) + const { deps: d, requests } = deps({ actionsEnabled: false }) + const event = click() + + expect(handleNativeChatWebLink(event, 'https://example.com/', d)).toBe(true) + expect(event.preventDefault).toHaveBeenCalledOnce() + expect(requests).toHaveLength(0) + expect(mocks.openRoutedHttpLink).toHaveBeenCalledWith( + 'https://example.com/', + expect.objectContaining({ forceDestination: 'system' }) + ) + }) + + it.each([ + ['shift-only click', { shiftKey: true }], + ['alt click', { altKey: true }], + ['middle click', { button: 1 }], + ['mac ctrl click', { ctrlKey: true }] + ])('leaves the anchor default for a %s', (_label, init) => { + stubPlatform(true) + const { deps: d, requests } = deps() + const event = click(init) + + expect(handleNativeChatWebLink(event, 'https://example.com/', d)).toBe(false) + expect(event.preventDefault).not.toHaveBeenCalled() + expect(requests).toHaveLength(0) + expect(mocks.openRoutedHttpLink).not.toHaveBeenCalled() + }) +}) diff --git a/src/renderer/src/components/native-chat/native-chat-web-link-actions.ts b/src/renderer/src/components/native-chat/native-chat-web-link-actions.ts new file mode 100644 index 00000000000..54e3f408b32 --- /dev/null +++ b/src/renderer/src/components/native-chat/native-chat-web-link-actions.ts @@ -0,0 +1,77 @@ +import type { LinkActionRequest } from '@/components/link-actions/link-action-request' +// Chat shares the terminal's link-click vocabulary: plain click asks, modifier click opens. +import { + isTerminalLinkActionActivation, + isTerminalLinkDirectActivation +} from '@/components/terminal-pane/terminal-link-activation' +import { + buildHttpLinkActions, + openRoutedHttpLink, + type HttpLinkActionDestinations, + type HttpLinkDestination +} from '@/lib/http-link-destinations' +import type { HttpLinkSourceOwner } from '@/lib/http-link-routing' + +export type NativeChatWebLinkDeps = { + worktreeId: string + sourceOwner: HttpLinkSourceOwner + destinations: HttpLinkActionDestinations + /** Off: a plain click opens the routed destination outright, as it did before actions existed. */ + actionsEnabled: boolean + restoreFocus: () => void + request: (request: LinkActionRequest) => void +} + +type ChatLinkMouseEvent = Pick< + MouseEvent, + 'altKey' | 'clientX' | 'clientY' | 'ctrlKey' | 'metaKey' | 'shiftKey' +> & { + detail?: number + currentTarget?: Pick + button?: number + preventDefault: () => void +} + +/** Returns true when the click was consumed; false leaves the anchor's default. */ +export function handleNativeChatWebLink( + event: ChatLinkMouseEvent, + url: string, + deps: NativeChatWebLinkDeps +): boolean { + const open = (destination: HttpLinkDestination | undefined): void => + openRoutedHttpLink(url, { + worktreeId: deps.worktreeId, + sourceOwner: deps.sourceOwner, + modifierHeld: false, + ...(destination ? { forceDestination: destination } : {}) + }) + + if (isTerminalLinkDirectActivation(event)) { + event.preventDefault() + open( + event.shiftKey + ? (deps.destinations.alternate ?? deps.destinations.primary) + : deps.destinations.primary + ) + return true + } + if (!isTerminalLinkActionActivation(event)) { + return false + } + + event.preventDefault() + if (!deps.actionsEnabled) { + open(deps.destinations.primary) + return true + } + const keyboardAnchor = event.detail === 0 ? event.currentTarget?.getBoundingClientRect() : null + deps.request({ + anchorX: keyboardAnchor?.left ?? event.clientX, + anchorY: keyboardAnchor?.bottom ?? event.clientY, + destination: url, + kind: 'url', + restoreFocus: deps.restoreFocus, + ...buildHttpLinkActions(deps.destinations, open) + }) + return true +} diff --git a/src/renderer/src/components/native-chat/structured-session-takeover-report.test.tsx b/src/renderer/src/components/native-chat/structured-session-takeover-report.test.tsx new file mode 100644 index 00000000000..b44df6c8304 --- /dev/null +++ b/src/renderer/src/components/native-chat/structured-session-takeover-report.test.tsx @@ -0,0 +1,86 @@ +// @vitest-environment happy-dom + +/** + * A user typing into a structured worker's chat pane is a TAKEOVER. + * + * The guard has always existed — `worker_terminal_resources.ownership_state = 'user_owned'` makes + * `worker-release` retain with `user_takeover` — and for a structured worker it was simply never + * armed: `reportWorkerTerminalUserInput` has one call site, on a PTY connection. So a user + * mid-conversation in a chat pane had their session closed and its tab retired, while + * `orchestration-worker-specs.ts` promised "Never closes … user-taken-over terminals". + */ + +import { describe, expect, it, vi } from 'vitest' +import { renderHook } from '@testing-library/react' + +const reportStructuredSessionUserInput = vi.hoisted(() => vi.fn()) +const dispatchStructuredComposerText = vi.hoisted(() => vi.fn()) + +vi.mock('@/lib/worker-terminal-takeover-report', () => ({ + reportStructuredSessionUserInput, + reportWorkerTerminalUserInput: vi.fn() +})) +vi.mock('@/lib/native-chat-telemetry', () => ({ emitNativeChatMessageSent: vi.fn() })) +vi.mock('./native-chat-structured-composer-dispatch', () => ({ + dispatchNativeChatStructuredComposerText: dispatchStructuredComposerText +})) + +import { useNativeChatStructuredComposerSend } from './use-native-chat-structured-composer-send' +import type { NativeChatStructuredComposerTransport } from './native-chat-composer-types' + +function transport(): NativeChatStructuredComposerTransport { + return { + send: vi.fn(() => true), + dispatchCommand: vi.fn(async () => ({ handled: false, accepted: false, error: null })), + optionsSurface: { + getSnapshot: () => [], + setOption: vi.fn(), + invokeAction: vi.fn(), + subscribe: () => () => {} + }, + optionSnapshot: [], + onError: vi.fn(), + runtime: 'local', + sessionId: 'session-1', + runtimeEnvironmentId: null + } +} + +function send(structuredTransport: NativeChatStructuredComposerTransport): (text: string) => void { + const { result } = renderHook(() => + useNativeChatStructuredComposerSend({ + agent: 'claude', + imageAttachments: [], + structuredTransport, + clearImageAttachments: vi.fn(), + clearSkillOrigin: vi.fn(), + setHistory: vi.fn(), + setDraft: vi.fn(), + setCaret: vi.fn() + }) + ) + return result.current +} + +describe('a real user send from a structured chat pane', () => { + it('reports the takeover, addressed by session and never by pane key', async () => { + // By session on purpose: the worker's pane key is a random identity credential held in main, + // and a renderer echoing it back would make it learnable by anyone who can see a chat pane. + reportStructuredSessionUserInput.mockClear() + dispatchStructuredComposerText.mockResolvedValue({ accepted: true, error: null }) + send(transport())('ship it') + await vi.waitFor(() => + expect(reportStructuredSessionUserInput).toHaveBeenCalledWith('session-1', null) + ) + }) + + it('reports nothing when the transport refused the send', async () => { + // A refused send is not a takeover; relinquishing ownership on one would retain every worker + // whose composer merely errored. + reportStructuredSessionUserInput.mockClear() + dispatchStructuredComposerText.mockResolvedValue({ accepted: false, error: 'nope' }) + send(transport())('ship it') + await new Promise((resolve) => setTimeout(resolve, 0)) + expect(reportStructuredSessionUserInput).not.toHaveBeenCalled() + }) +}) diff --git a/src/renderer/src/components/native-chat/use-native-chat-composer-catalog.test.tsx b/src/renderer/src/components/native-chat/use-native-chat-composer-catalog.test.tsx new file mode 100644 index 00000000000..c5e9054894b --- /dev/null +++ b/src/renderer/src/components/native-chat/use-native-chat-composer-catalog.test.tsx @@ -0,0 +1,114 @@ +// @vitest-environment happy-dom +import { renderHook } from '@testing-library/react' +import { describe, expect, it, vi } from 'vitest' +import { buildNativeChatPickerItems } from './native-chat-picker-items' +import { useNativeChatComposerKeyDown } from './use-native-chat-composer-keydown' +import { EMPTY_HISTORY } from './native-chat-composer-state' +import { useNativeChatComposerCatalog } from './use-native-chat-composer-catalog' +import type { NativeChatStructuredComposerTransport } from './native-chat-composer-types' +import { getVerifiedNativeChatCommands } from '../../../../shared/native-chat-agent-profiles' +import { structuredSlashCommands } from '../../../../shared/structured-agent-session-composer' + +function transport(sessionCommands?: NativeChatStructuredComposerTransport['sessionCommands']) { + return { sessionCommands } as NativeChatStructuredComposerTransport +} + +describe('composer catalog authority', () => { + it('keeps PTY and unsupported structured providers on their original catalogs', () => { + const pty = renderHook(() => useNativeChatComposerCatalog('claude')) + expect(pty.result.current.agentCommands).toEqual(getVerifiedNativeChatCommands('claude')) + expect(pty.result.current.sessionSkillNames).toBeUndefined() + const oldHost = renderHook(() => useNativeChatComposerCatalog('claude', transport())) + expect(oldHost.result.current.agentCommands).toEqual(structuredSlashCommands('claude')) + expect(oldHost.result.current.sessionSkillNames).toBeUndefined() + }) + it('respects empty catalogs and command-only catalogs without reviving disk skills', () => { + const { result, rerender } = renderHook( + ({ reported }) => useNativeChatComposerCatalog('claude', transport(reported)), + { + initialProps: { + reported: [] as NonNullable + } + } + ) + expect(result.current).toEqual({ agentCommands: [], sessionSkillNames: [] }) + rerender({ reported: [{ name: 'custom-command', kind: 'command' }] }) + expect(result.current).toEqual({ + agentCommands: [{ name: 'custom-command' }], + sessionSkillNames: [] + }) + }) +}) + +it('Enter completes a known pre-init skill while still dispatching a built-in command', () => { + const reported = [ + { name: 'clear', kind: 'command' as const, kindUnspecified: true as const }, + { name: 'project-skill', kind: 'command' as const, kindUnspecified: true as const } + ] + const complete = vi.fn(), + dispatch = vi.fn() + const { result, rerender } = renderHook( + ({ activeSuggestion }) => { + const catalog = useNativeChatComposerCatalog('claude', transport(reported)) + const items = buildNativeChatPickerItems( + catalog.agentCommands, + [ + { + id: 'project-skill', + name: 'project-skill', + description: 'Project skill', + providers: ['claude'], + sourceKind: 'repo', + sourceLabel: 'Project', + rootPath: '/project/.claude/skills', + directoryPath: '/project/.claude/skills/project-skill', + skillFilePath: '/project/.claude/skills/project-skill/SKILL.md', + installed: true, + updatedAt: null + } + ], + '', + '/', + catalog.sessionSkillNames + ) + return useNativeChatComposerKeyDown({ + autocomplete: { + mode: 'slash', + query: '', + items, + triggerKey: '/', + prefix: '/', + grouped: true, + commandsEnabled: true, + skillsEnabled: true, + skillStatus: 'ready' + }, + activeSuggestion, + draft: '/', + history: EMPTY_HISTORY, + isComposing: () => false, + completePickerItem: complete, + dispatchPickerCommand: dispatch, + dismissPicker: vi.fn(), + interrupt: vi.fn(), + send: vi.fn(), + setActiveSuggestion: vi.fn(), + setDraft: vi.fn(), + setCaret: vi.fn(), + setHistory: vi.fn() + }) + }, + { initialProps: { activeSuggestion: 1 } } + ) + const enter = { key: 'Enter', nativeEvent: {}, preventDefault: vi.fn() } as unknown as Parameters< + typeof result.current + >[0] + result.current(enter) + expect(complete).toHaveBeenCalledWith( + expect.objectContaining({ name: 'project-skill', kind: 'skill' }) + ) + expect(dispatch).not.toHaveBeenCalled() + rerender({ activeSuggestion: 0 }) + result.current(enter) + expect(dispatch).toHaveBeenCalledWith(expect.objectContaining({ name: 'clear', kind: 'command' })) +}) diff --git a/src/renderer/src/components/native-chat/use-native-chat-composer-catalog.ts b/src/renderer/src/components/native-chat/use-native-chat-composer-catalog.ts new file mode 100644 index 00000000000..e9a7b84e09a --- /dev/null +++ b/src/renderer/src/components/native-chat/use-native-chat-composer-catalog.ts @@ -0,0 +1,44 @@ +import { useMemo } from 'react' +import type { AgentType } from '../../../../shared/agent-status-types' +import { getVerifiedNativeChatCommands } from '../../../../shared/native-chat-agent-profiles' +import { + sessionReportedSkillNames, + sessionSlashCommandSuggestions, + type SlashCommandSuggestion +} from '../../../../shared/native-chat-slash-commands' +import { structuredSlashCommands } from '../../../../shared/structured-agent-session-composer' +import type { NativeChatStructuredComposerTransport } from './native-chat-composer-types' + +export type NativeChatComposerCatalog = { + agentCommands: readonly SlashCommandSuggestion[] + sessionSkillNames: readonly string[] | undefined +} + +/** + * What the `/` menu offers. A structured session reports the surface it actually + * loaded — the only list that includes this repo's own commands and the skills + * that reach the session through plugin roots — so it wins whenever it is + * present. The curated per-agent catalog remains the answer for the PTY lane and + * for a host that predates the report. + */ +export function useNativeChatComposerCatalog( + agent: AgentType, + structuredTransport?: NativeChatStructuredComposerTransport +): NativeChatComposerCatalog { + const structured = Boolean(structuredTransport) + const reported = structuredTransport?.sessionCommands + const agentCommands = useMemo( + () => + !structured + ? getVerifiedNativeChatCommands(agent) + : reported !== undefined + ? sessionSlashCommandSuggestions(agent, reported) + : structuredSlashCommands(agent), + [agent, reported, structured] + ) + const sessionSkillNames = useMemo( + () => (reported !== undefined ? sessionReportedSkillNames(reported) : undefined), + [reported] + ) + return { agentCommands, sessionSkillNames } +} diff --git a/src/renderer/src/components/native-chat/use-native-chat-link-actions.test.tsx b/src/renderer/src/components/native-chat/use-native-chat-link-actions.test.tsx new file mode 100644 index 00000000000..5037c175cbc --- /dev/null +++ b/src/renderer/src/components/native-chat/use-native-chat-link-actions.test.tsx @@ -0,0 +1,184 @@ +// @vitest-environment happy-dom +import type { ReactNode } from 'react' +import { useRef } from 'react' +import { cleanup, fireEvent, render, screen } from '@testing-library/react' +import { afterEach, describe, expect, it, vi } from 'vitest' +import { LinkActionPopover } from '@/components/link-actions/LinkActionPopover' +import CommentMarkdown from '@/components/sidebar/CommentMarkdown' +import { useNativeChatLinkActions } from './use-native-chat-link-actions' + +const mocks = vi.hoisted(() => ({ + openHttpLink: vi.fn(), + openFileLink: vi.fn(), + settings: { openLinksInApp: true, terminalLinkActionPopoverEnabled: true } as { + openLinksInApp?: boolean + terminalLinkActionPopoverEnabled?: boolean + } +})) + +vi.mock('@/lib/http-link-routing', () => ({ openHttpLink: mocks.openHttpLink })) + +vi.mock('./native-chat-http-link-source-owner', () => ({ + resolveNativeChatHttpLinkSourceOwner: () => ({ kind: 'local' }), + canNativeChatOpenOwnedBrowser: () => false +})) + +vi.mock('./use-native-chat-file-link-click', () => ({ + useNativeChatFileLinkClick: (context: unknown) => (context ? mocks.openFileLink : undefined) +})) + +vi.mock('@/store', () => ({ + useAppStore: Object.assign( + (selector: (state: Record) => unknown) => + selector({ openSettingsPage: vi.fn(), openSettingsTarget: vi.fn() }), + { getState: () => ({ settings: mocks.settings }) } + ) +})) + +vi.mock('@/components/ui/tooltip', () => ({ + Tooltip: ({ children }: { children: ReactNode }) => children, + TooltipTrigger: ({ children }: { children: ReactNode }) => children, + TooltipContent: ({ children }: { children: ReactNode }) => {children} +})) + +vi.mock('@/components/ui/popover', () => ({ + Popover: ({ children, open }: { children: ReactNode; open: boolean }) => + open ?
{children}
: null, + PopoverAnchor: () => null, + PopoverContent: ({ children }: { children: ReactNode }) =>
{children}
+})) + +const context = { worktreeId: 'wt-1', worktreePath: '/repo', runtimeEnvironmentId: null } + +function Transcript({ + markdown, + sessionId = 'session-1', + isVisible = true, + linkContext = context +}: { + markdown: string + sessionId?: string + isVisible?: boolean + linkContext?: typeof context | null +}): React.JSX.Element { + const rootRef = useRef(null) + const { onLinkClick, linkActionRequest, closeLinkActions } = useNativeChatLinkActions( + linkContext, + rootRef, + { sessionId, isVisible } + ) + return ( +
+ + +
+ ) +} + +afterEach(() => { + cleanup() + vi.clearAllMocks() + mocks.settings = { openLinksInApp: true, terminalLinkActionPopoverEnabled: true } +}) + +describe('native chat transcript links', () => { + it.each(['hidden', 'session', 'workspace', 'no-context'] as const)( + 'dismisses a request when the transcript is %s', + async (change) => { + const markdown = '[link](https://example.com)' + const { rerender } = render() + fireEvent.click(await screen.findByRole('link', { name: 'link' })) + expect(screen.getByText('System Browser')).toBeTruthy() + rerender( + + ) + expect(screen.queryByText('System Browser')).toBeNull() + rerender() + expect(screen.queryByText('System Browser')).toBeNull() + expect(mocks.openHttpLink).not.toHaveBeenCalled() + } + ) + + it('offers both destinations when a rendered http link is clicked', async () => { + render() + + fireEvent.click(await screen.findByRole('link', { name: 'the PR' })) + + expect(screen.getByText('https://github.com/o/r/pull/1')).toBeTruthy() + expect(screen.getByText('Orca Browser')).toBeTruthy() + expect(screen.getByText('System Browser')).toBeTruthy() + expect(mocks.openHttpLink).not.toHaveBeenCalled() + }) + + it('keeps mailto links on the anchor default', async () => { + const { container } = render() + const anchorDefault = vi.fn((event: Event) => { + expect(event.defaultPrevented).toBe(false) + event.preventDefault() + }) + container.addEventListener('click', anchorDefault) + fireEvent.click(await screen.findByRole('link', { name: 'email' })) + expect(anchorDefault).toHaveBeenCalledOnce() + expect(mocks.openHttpLink).not.toHaveBeenCalled() + expect(mocks.openFileLink).not.toHaveBeenCalled() + expect(screen.queryByText('System Browser')).toBeNull() + }) + + it('restores focus to the clicked transcript link', async () => { + render() + const anchor = await screen.findByRole('link', { name: 'link' }) + fireEvent.click(anchor) + fireEvent.click(screen.getByText('System Browser')) + expect(document.activeElement).toBe(anchor) + }) + + it('routes the chosen destination through the shared link opener', async () => { + render() + fireEvent.click(await screen.findByRole('link', { name: 'the PR' })) + + fireEvent.click(screen.getByText('System Browser')) + + expect(mocks.openHttpLink).toHaveBeenCalledWith( + 'https://github.com/o/r/pull/1', + expect.objectContaining({ forceSystemBrowser: true, worktreeId: 'wt-1' }) + ) + }) + + it('opens the routed destination outright when link actions are off', async () => { + mocks.settings = { openLinksInApp: true, terminalLinkActionPopoverEnabled: false } + render() + + fireEvent.click(await screen.findByRole('link', { name: 'the PR' })) + + expect(screen.queryByText('Orca Browser')).toBeNull() + expect(mocks.openHttpLink).toHaveBeenCalledWith( + 'https://github.com/o/r/pull/1', + expect.objectContaining({ forceInApp: true }) + ) + }) + + it('leaves file links on the existing native chat opener', async () => { + render() + + fireEvent.click(await screen.findByRole('link', { name: 'the file' })) + + expect(mocks.openFileLink).toHaveBeenCalledOnce() + expect(screen.queryByText('System Browser')).toBeNull() + }) +}) diff --git a/src/renderer/src/components/native-chat/use-native-chat-link-actions.ts b/src/renderer/src/components/native-chat/use-native-chat-link-actions.ts new file mode 100644 index 00000000000..cf0b5e9f5ee --- /dev/null +++ b/src/renderer/src/components/native-chat/use-native-chat-link-actions.ts @@ -0,0 +1,87 @@ +import { useCallback, useState, type RefObject } from 'react' +import { + closeLinkActionRequest, + type LinkActionRequest +} from '@/components/link-actions/link-action-request' +import { httpLinkActionDestinationsFor } from '@/lib/http-link-destinations' +import type { CommentMarkdownLinkClickHandler } from '@/components/sidebar/CommentMarkdown' +import { routeNativeChatHref } from '../../../../shared/native-chat-href-routing' +import { useAppStore } from '../../store' +import type { NativeChatFileLinkContext } from './native-chat-file-link' +import { + canNativeChatOpenOwnedBrowser, + resolveNativeChatHttpLinkSourceOwner +} from './native-chat-http-link-source-owner' +import { handleNativeChatWebLink } from './native-chat-web-link-actions' +import { useNativeChatFileLinkClick } from './use-native-chat-file-link-click' + +export type NativeChatLinkActions = { + onLinkClick: CommentMarkdownLinkClickHandler | undefined + linkActionRequest: LinkActionRequest | null + closeLinkActions: (dismissed?: LinkActionRequest) => void +} + +/** Transcript links: file targets open in Orca, http(s) targets offer the same + * destination popover the terminal shows. */ +export function useNativeChatLinkActions( + context: NativeChatFileLinkContext | null, + rootRef: RefObject, + scope: { sessionId: string | null; isVisible: boolean } +): NativeChatLinkActions { + const openFileLink = useNativeChatFileLinkClick(context) + const [linkActionRequest, setLinkActionRequest] = useState(null) + const scopeKey = JSON.stringify([ + context?.worktreeId, + context?.runtimeEnvironmentId, + scope.sessionId + ]) + const [previousScopeKey, setPreviousScopeKey] = useState(scopeKey) + if (previousScopeKey !== scopeKey || (!scope.isVisible && linkActionRequest !== null)) { + setPreviousScopeKey(scopeKey) + setLinkActionRequest(null) + } + const closeLinkActions = useCallback((dismissed?: LinkActionRequest) => { + setLinkActionRequest((current) => closeLinkActionRequest(current, dismissed)) + }, []) + + const onLinkClick = useCallback( + (event, href) => { + if (!context) { + return + } + const route = routeNativeChatHref(href) + if (route.kind === 'file') { + openFileLink?.(event, href) + return + } + // mailto: and other schemes keep the anchor's default handling. + if (route.kind !== 'web' || !/^https?:/i.test(route.url)) { + return + } + // Read at click time: settings and workspace ownership must not re-render the transcript. + const state = useAppStore.getState() + const sourceOwner = resolveNativeChatHttpLinkSourceOwner(state, context.worktreeId) + const anchor = event.currentTarget + handleNativeChatWebLink(event, route.url, { + worktreeId: context.worktreeId, + sourceOwner, + destinations: httpLinkActionDestinationsFor( + state.settings, + sourceOwner, + canNativeChatOpenOwnedBrowser(state, context.worktreeId, sourceOwner) + ), + actionsEnabled: state.settings?.terminalLinkActionPopoverEnabled !== false, + restoreFocus: () => + (anchor.isConnected ? anchor : rootRef.current)?.focus({ preventScroll: true }), + request: setLinkActionRequest + }) + }, + [context, openFileLink, rootRef] + ) + + return { + onLinkClick: context ? onLinkClick : undefined, + linkActionRequest, + closeLinkActions + } +} diff --git a/src/renderer/src/components/native-chat/use-native-chat-picker-state.ts b/src/renderer/src/components/native-chat/use-native-chat-picker-state.ts index d66228356a2..189b4f16cfa 100644 --- a/src/renderer/src/components/native-chat/use-native-chat-picker-state.ts +++ b/src/renderer/src/components/native-chat/use-native-chat-picker-state.ts @@ -46,6 +46,8 @@ export function useNativeChatPickerState(args: { draft: string caret: number agentCommands: readonly SlashCommandSuggestion[] + /** Skill names the running session reports; undefined keeps the host disk scan. */ + sessionSkillNames?: readonly string[] textareaRef: RefObject setDraft: (value: string) => void setCaret: Dispatch> @@ -58,6 +60,7 @@ export function useNativeChatPickerState(args: { draft, caret, agentCommands, + sessionSkillNames, textareaRef, setDraft, setCaret, @@ -86,9 +89,19 @@ export function useNativeChatPickerState(args: { discovery.skills, profile, discovery, - dismissed?.context === dismissalContext ? dismissed.triggerKey : null + dismissed?.context === dismissalContext ? dismissed.triggerKey : null, + sessionSkillNames ), - [agentCommands, caret, dismissalContext, dismissed, discovery, draft, profile] + [ + agentCommands, + caret, + dismissalContext, + dismissed, + discovery, + draft, + profile, + sessionSkillNames + ] ) useEffect(() => { @@ -152,7 +165,9 @@ export function useNativeChatPickerState(args: { agentCommands, discovery.skills, profile, - discovery + discovery, + null, + sessionSkillNames ) if ( (next.mode !== 'slash' && next.mode !== 'skill') || @@ -161,7 +176,7 @@ export function useNativeChatPickerState(args: { setDismissed(null) } }, - [agentCommands, dismissalContext, dismissed, discovery, draft, profile] + [agentCommands, dismissalContext, dismissed, discovery, draft, profile, sessionSkillNames] ) const classifySend = useCallback( diff --git a/src/renderer/src/components/native-chat/use-native-chat-session-options.ts b/src/renderer/src/components/native-chat/use-native-chat-session-options.ts index 54237b05628..4e4d9a58c4c 100644 --- a/src/renderer/src/components/native-chat/use-native-chat-session-options.ts +++ b/src/renderer/src/components/native-chat/use-native-chat-session-options.ts @@ -4,15 +4,7 @@ import { getAgentSessionOptionCatalog, type CatalogModel } from '../../../../shared/agent-session-option-catalog' -import { - clearNativeChatSessionOptionModel, - updateNativeChatSessionOptionDefaults -} from '../../../../shared/native-chat-session-option-defaults' -import type { - PersistedNativeChatSessionOptions, - SessionOptionDescriptor -} from '../../../../shared/native-chat-session-options' -import { useAppStore } from '../../store' +import type { SessionOptionDescriptor } from '../../../../shared/native-chat-session-options' import { createNativeChatPtySessionOptions, type NativeChatPtySessionOptionsSurface @@ -28,35 +20,12 @@ import { resolveNativeChatModelDiscoveryContext } from './native-chat-session-option-discovery' import { readClaudeSessionOptionsFromTerminalScreen } from './claude-terminal-session-options' +import { enqueueSessionOptionSettingsWrite } from './native-chat-session-option-settings-write' const EMPTY_SNAPSHOT: SessionOptionDescriptor[] = [] const subscribeEmpty = (): (() => void) => () => {} const getEmptySnapshot = (): SessionOptionDescriptor[] => EMPTY_SNAPSHOT - -/** - * Why: every nativeChatSessionOptions writer — a pick from any pane, a probe - * retirement — serializes on this one chain and re-reads live settings at apply - * time. updateSettings shallow-merges the whole object, so an interleaved write - * from a snapshot captured earlier would silently clobber a concurrent pick. - * The update runs against the settled base and may return null to skip writing. - */ -let settingsWrite: Promise = Promise.resolve() -function enqueueSessionOptionSettingsWrite( - update: ( - base: PersistedNativeChatSessionOptions | undefined - ) => PersistedNativeChatSessionOptions | null -): Promise { - const write = settingsWrite - .catch(() => undefined) - .then(() => { - const next = update(useAppStore.getState().settings?.nativeChatSessionOptions) - return next - ? useAppStore.getState().updateSettings({ nativeChatSessionOptions: next }) - : undefined - }) - settingsWrite = write - return write -} +const CLIENT_SETTINGS_TARGET = { kind: 'local' } as const /** * Why: the picker drops a retired model, but the persisted default is what launches @@ -75,11 +44,10 @@ export async function retirePersistedModelMissingFromDiscovery( if (models.length === 0) { return } - await enqueueSessionOptionSettingsWrite((persisted) => { - const modelId = persisted?.[agent]?.model - return typeof modelId === 'string' && modelId && !models.some((model) => model.id === modelId) - ? clearNativeChatSessionOptionModel(persisted, agent) - : null + await enqueueSessionOptionSettingsWrite(CLIENT_SETTINGS_TARGET, { + type: 'clear-model-if-missing', + agent, + availableModelIds: models.map((model) => model.id) }) } @@ -133,16 +101,12 @@ export function useNativeChatSessionOptions(args: { dispatchCommand, onAgentPicker, persistSelection: ({ modelId, optionId, value, adoptModelAsLaunchDefault }) => - enqueueSessionOptionSettingsWrite((persisted) => - updateNativeChatSessionOptionDefaults({ - persisted, - agent, - modelId, - optionId, - value, - adoptModelAsLaunchDefault - }) - ) + // Paired PTY launches still assemble their launch preferences from client settings. + enqueueSessionOptionSettingsWrite(CLIENT_SETTINGS_TARGET, { + type: 'apply-picks', + agent, + picks: [{ modelId, optionId, value, adoptModelAsLaunchDefault }] + }) }) }, [ agent, diff --git a/src/renderer/src/components/native-chat/use-native-chat-structured-composer-send.ts b/src/renderer/src/components/native-chat/use-native-chat-structured-composer-send.ts index e871d65c62c..c3ccc09a19e 100644 --- a/src/renderer/src/components/native-chat/use-native-chat-structured-composer-send.ts +++ b/src/renderer/src/components/native-chat/use-native-chat-structured-composer-send.ts @@ -1,5 +1,6 @@ import { useCallback } from 'react' import { emitNativeChatMessageSent } from '@/lib/native-chat-telemetry' +import { reportStructuredSessionUserInput } from '@/lib/worker-terminal-takeover-report' import { isStructuredAgentSessionComposerCommand } from '../../../../shared/structured-agent-session-composer' import type { AgentType } from '../../../../shared/agent-status-types' import { dispatchNativeChatStructuredComposerText } from './native-chat-structured-composer-dispatch' @@ -49,6 +50,13 @@ export function useNativeChatStructuredComposerSend({ return } emitNativeChatMessageSent({ agent, runtime: structuredTransport.runtime }) + // A real user send is a takeover, exactly as typing into a worker's pane is. Only past + // `accepted`, and only from this hook: the outbox dispatcher retries and would re-fire, + // and orchestration's own pointer nudges never reach the composer at all. + reportStructuredSessionUserInput( + structuredTransport.sessionId, + structuredTransport.runtimeEnvironmentId + ) setHistory((previous) => pushHistory(previous, text)) setDraft('') setCaret(0) diff --git a/src/renderer/src/components/native-chat/use-structured-agent-session.test.tsx b/src/renderer/src/components/native-chat/use-structured-agent-session.test.tsx index 9a42ccb6da6..abf185cb913 100644 --- a/src/renderer/src/components/native-chat/use-structured-agent-session.test.tsx +++ b/src/renderer/src/components/native-chat/use-structured-agent-session.test.tsx @@ -3,17 +3,27 @@ import { act, renderHook, waitFor } from '@testing-library/react' import { beforeEach, describe, expect, it, vi } from 'vitest' -const mocks = vi.hoisted(() => ({ call: vi.fn(), operationId: vi.fn() })) +const mocks = vi.hoisted(() => ({ + call: vi.fn(), + operationId: vi.fn(), + enqueueSettingsWrite: vi.fn() +})) let fence = 3 +let sessionCommands: { name: string; kind: 'command' | 'skill' }[] | undefined vi.mock('@/runtime/structured-agent-session-client', () => ({ callStructuredAgentSession: mocks.call })) +vi.mock('./native-chat-session-option-settings-write', () => ({ + enqueueSessionOptionSettingsWrite: mocks.enqueueSettingsWrite +})) + vi.mock('./use-structured-agent-session-read', () => ({ useStructuredAgentSessionRead: () => ({ state: { fence, + commands: sessionCommands, items: [], submissions: [], status: 'ready', @@ -37,8 +47,26 @@ vi.mock('./use-structured-agent-session-outbox', () => ({ }) })) +import { + applyNativeChatSessionOptionSettingsMutation, + resolveStructuredLaunchSeedOptions +} from '../../../../shared/native-chat-session-option-defaults' +import type { PersistedNativeChatSessionOptions } from '../../../../shared/native-chat-session-options' import { useStructuredAgentSession } from './use-structured-agent-session' +/** Replay every host mutation in order, exactly as the runtime does. */ +function seededByNextLaunch(): Record | undefined { + let persisted: PersistedNativeChatSessionOptions | undefined + for (const [, mutation] of mocks.enqueueSettingsWrite.mock.calls) { + persisted = + applyNativeChatSessionOptionSettingsMutation( + persisted, + mutation as Parameters[1] + ) ?? persisted + } + return resolveStructuredLaunchSeedOptions(persisted, 'codex') +} + const LOCAL_TARGET = { kind: 'local' } as const const OPTIONS = { @@ -332,4 +360,166 @@ describe('useStructuredAgentSession options', () => { taskId: 'task-2' }) }) + + it('remembers a model pick so the next launch seeds the pair the provider settled on', async () => { + mocks.call.mockImplementation((_target, method) => + method === 'agentSession.options' + ? Promise.resolve(OPTIONS) + : Promise.resolve({ + ok: true, + value: { + key: 'model', + value: 'gpt-fast', + options: { model: 'gpt-fast', effort: 'low' } + } + }) + ) + const { result } = renderHook(() => + useStructuredAgentSession({ + sessionId: 'session-1', + target: LOCAL_TARGET, + agent: 'codex', + isVisible: true + }) + ) + await waitFor(() => expect(result.current.optionSnapshot).toHaveLength(2)) + + await act(async () => { + expect(await result.current.setStructuredOption('model', 'gpt-fast')).toBe(true) + }) + + expect(seededByNextLaunch()).toEqual({ model: 'gpt-fast', effort: 'low' }) + expect(mocks.enqueueSettingsWrite).toHaveBeenCalledWith(LOCAL_TARGET, { + type: 'apply-picks', + agent: 'codex', + picks: [ + { modelId: 'gpt-fast', optionId: 'model', value: 'gpt-fast' }, + { modelId: 'gpt-fast', optionId: 'effort', value: 'low' } + ] + }) + }) + + it('writes through the session runtime target', async () => { + const remoteTarget = { kind: 'environment', environmentId: 'remote-1' } as const + mocks.call.mockImplementation((_target, method) => + method === 'agentSession.options' + ? Promise.resolve(OPTIONS) + : Promise.resolve({ + ok: true, + value: { key: 'effort', value: 'high', options: { effort: 'high' } } + }) + ) + const { result } = renderHook(() => + useStructuredAgentSession({ + sessionId: 'session-1', + target: remoteTarget, + agent: 'codex', + isVisible: true + }) + ) + await waitFor(() => expect(result.current.optionSnapshot).toHaveLength(2)) + + await act(async () => { + expect(await result.current.setStructuredOption('effort', 'high')).toBe(true) + }) + + expect(mocks.enqueueSettingsWrite).toHaveBeenCalledWith( + remoteTarget, + expect.objectContaining({ type: 'apply-picks', agent: 'codex' }) + ) + }) + + it('pins the model an effort-only pick was made against', async () => { + mocks.call.mockImplementation((_target, method) => + method === 'agentSession.options' + ? Promise.resolve(OPTIONS) + : Promise.resolve({ + ok: true, + value: { key: 'effort', value: 'high', options: { effort: 'high' } } + }) + ) + const { result } = renderHook(() => + useStructuredAgentSession({ + sessionId: 'session-1', + target: LOCAL_TARGET, + agent: 'codex', + isVisible: true + }) + ) + await waitFor(() => expect(result.current.optionSnapshot).toHaveLength(2)) + + await act(async () => { + expect(await result.current.setStructuredOption('effort', 'high')).toBe(true) + }) + + // Without the model the launch resolves nothing, so the remembered effort would be dead. + expect(seededByNextLaunch()).toEqual({ model: 'gpt-live', effort: 'high' }) + }) + + it('remembers nothing when the provider refuses the pick', async () => { + mocks.call.mockImplementation((_target, method) => + method === 'agentSession.options' + ? Promise.resolve(OPTIONS) + : Promise.reject(new Error('provider rejected option')) + ) + const { result } = renderHook(() => + useStructuredAgentSession({ + sessionId: 'session-1', + target: LOCAL_TARGET, + agent: 'codex', + isVisible: true + }) + ) + await waitFor(() => expect(result.current.optionSnapshot).toHaveLength(2)) + + await act(async () => { + expect(await result.current.setStructuredOption('model', 'gpt-fast')).toBe(false) + }) + + expect(mocks.enqueueSettingsWrite).not.toHaveBeenCalled() + }) +}) + +describe('session command catalog stream', () => { + beforeEach(() => { + vi.clearAllMocks() + fence = 3 + sessionCommands = undefined + mocks.call.mockResolvedValue(OPTIONS) + }) + + const args = { sessionId: 'one', target: LOCAL_TARGET, agent: 'claude' as const, isVisible: true } + const commands = [{ name: 'plugin:review', kind: 'skill' as const }] + + it('uses owner-scoped catalog state without a separate command RPC or stale cache', () => { + sessionCommands = commands + const { result, rerender } = renderHook((props) => useStructuredAgentSession(props), { + initialProps: args + }) + expect(result.current.sessionCommands).toEqual(commands) + sessionCommands = undefined + rerender({ ...args, sessionId: 'two' }) + expect(result.current.sessionCommands).toBeUndefined() + sessionCommands = [] + rerender({ ...args, sessionId: 'two' }) + expect(result.current.sessionCommands).toEqual([]) + expect( + mocks.call.mock.calls.filter(([, method]) => method === 'agentSession.commands') + ).toHaveLength(0) + }) + + it('adopts idle catalog updates and does no command reads on repeated transcript renders', () => { + sessionCommands = commands + const { result, rerender } = renderHook(() => useStructuredAgentSession(args)) + expect(result.current.sessionCommands).toEqual(commands) + for (let index = 0; index < 30; index += 1) { + rerender() + } + sessionCommands = [] + rerender() + expect(result.current.sessionCommands).toEqual([]) + expect( + mocks.call.mock.calls.filter(([, method]) => method === 'agentSession.commands') + ).toHaveLength(0) + }) }) diff --git a/src/renderer/src/components/native-chat/use-structured-agent-session.ts b/src/renderer/src/components/native-chat/use-structured-agent-session.ts index b0bab73669c..7cba19940ea 100644 --- a/src/renderer/src/components/native-chat/use-structured-agent-session.ts +++ b/src/renderer/src/components/native-chat/use-structured-agent-session.ts @@ -16,6 +16,7 @@ import { canSetStructuredAgentSessionOption, commitStructuredAgentSessionOptionValues, createStructuredAgentSessionOptionState, + structuredAgentSessionOptionPicks, structuredAgentSessionOptionSnapshot } from '../../../../shared/structured-agent-session-options' import { activeStructuredAgentSessionTurnId } from '../../../../shared/structured-agent-session-projection' @@ -29,6 +30,7 @@ import { useStructuredAgentSessionHold } from './use-structured-agent-session-ho import { useStructuredAgentSessionRead } from './use-structured-agent-session-read' import { projectStructuredAgentSessionMessages } from './structured-agent-session-message-projection' import { selectStructuredAgentTurnActivity } from './native-chat-turn-activity' +import { enqueueSessionOptionSettingsWrite } from './native-chat-session-option-settings-write' export type StructuredPromptItem = AgentJournalRenderItem & { body: Extract @@ -142,8 +144,8 @@ export function useStructuredAgentSession(args: { // rather than leaving the last write unconfirmed for the life of the session. const turnId = activeStructuredAgentSessionTurnId(state.items) const turnActivity = useMemo( - () => selectStructuredAgentTurnActivity(state.items, turnId), - [state.items, turnId] + () => selectStructuredAgentTurnActivity(state.items, turnId, state.activity), + [state.activity, state.items, turnId] ) const isMonitoringBackgroundTasks = turnId === null && state.backgroundTasks?.state === 'monitoring' @@ -192,11 +194,20 @@ export function useStructuredAgentSession(args: { { key: id, value } ) if (result && activeOptionRecordRef.current === targetRecord) { + const committed = result.options ?? { [id]: value } setOptionState((current) => current.record === targetRecord - ? commitStructuredAgentSessionOptionValues(current, result.options ?? { [id]: value }) + ? commitStructuredAgentSessionOptionValues(current, committed) : current ) + const picks = structuredAgentSessionOptionPicks(optionState, committed) + if (picks.length > 0) { + void enqueueSessionOptionSettingsWrite(target, { + type: 'apply-picks', + agent, + picks + }) + } } return Boolean(result) } finally { @@ -207,7 +218,7 @@ export function useStructuredAgentSession(args: { ) } }, - [mutate, optionState] + [agent, mutate, optionState, target] ) const setOption = useCallback( async (id: string, value: string | boolean) => { @@ -270,6 +281,7 @@ export function useStructuredAgentSession(args: { ), optionSnapshot, optionSurface, + sessionCommands: state.commands ?? undefined, setStructuredOption } } diff --git a/src/renderer/src/components/ports/WorkspacePortScanner.test.tsx b/src/renderer/src/components/ports/WorkspacePortScanner.test.tsx index 0ae53fee931..7741368d025 100644 --- a/src/renderer/src/components/ports/WorkspacePortScanner.test.tsx +++ b/src/renderer/src/components/ports/WorkspacePortScanner.test.tsx @@ -602,3 +602,113 @@ describe('WorkspacePortScanner', () => { expect(getPublishedRemoteWorktreePorts()).toBeUndefined() }) }) + +describe('advertised URL refresh bursts', () => { + async function mountLocalScanner(): Promise<() => void> { + useAppStore.setState({ settings: getDefaultSettings('/tmp/orca-workspaces') }) + await act(async () => { + root?.render() + await flushPromises() + }) + localScan.mockClear() + return vi.mocked(window.api.workspacePorts.onAdvertisedUrlChanged).mock + .calls[0][0] as () => void + } + + it('coalesces sequential URL changes into one immediate scan and one settled scan', async () => { + const changed = await mountLocalScanner() + for (let index = 0; index < 5; index++) { + await act(async () => { + changed() + await flushPromises() + await vi.advanceTimersByTimeAsync(100) + }) + } + expect(localScan).toHaveBeenCalledTimes(1) + await act(async () => { + await vi.advanceTimersByTimeAsync(1_000) + }) + expect(localScan).toHaveBeenCalledTimes(2) + await act(async () => { + changed() + await flushPromises() + }) + expect(localScan).toHaveBeenCalledTimes(3) + }) + + it('cancels the settled scan on unmount', async () => { + const changed = await mountLocalScanner() + await act(async () => { + changed() + await flushPromises() + }) + act(() => root?.unmount()) + root = null + await vi.advanceTimersByTimeAsync(2_000) + expect(localScan).toHaveBeenCalledTimes(1) + }) + + it('skips the settled scan while hidden and accepts the next visible URL change', async () => { + let visibility: DocumentVisibilityState = 'visible' + const restore = overrideDocumentVisibilityState(() => visibility) + try { + const changed = await mountLocalScanner() + await act(async () => { + changed() + await flushPromises() + }) + visibility = 'hidden' + await act(async () => { + await vi.advanceTimersByTimeAsync(2_000) + }) + expect(localScan).toHaveBeenCalledTimes(1) + visibility = 'visible' + await act(async () => { + changed() + await flushPromises() + }) + expect(localScan).toHaveBeenCalledTimes(2) + } finally { + restore() + } + }) +}) + +it('releases the URL burst when its leading scan finishes while hidden', async () => { + let visibility: DocumentVisibilityState = 'visible' + const restore = overrideDocumentVisibilityState(() => visibility) + try { + useAppStore.setState({ settings: getDefaultSettings('/tmp/orca-workspaces') }) + await act(async () => { + root?.render() + await flushPromises() + }) + const changed = vi.mocked(window.api.workspacePorts.onAdvertisedUrlChanged).mock + .calls[0][0] as () => void + let finish!: (scan: WorkspacePortScanResult) => void + localScan.mockClear() + localScan.mockImplementationOnce( + () => + new Promise((resolve) => { + finish = resolve + }) + ) + await act(async () => { + changed() + await flushPromises() + }) + visibility = 'hidden' + await act(async () => { + finish(emptyScan) + await flushPromises() + }) + visibility = 'visible' + await act(async () => { + changed() + await flushPromises() + }) + expect(localScan).toHaveBeenCalledTimes(2) + } finally { + restore() + } +}) diff --git a/src/renderer/src/components/ports/WorkspacePortScanner.tsx b/src/renderer/src/components/ports/WorkspacePortScanner.tsx index 2b12bd86cd8..f8f2d11ee40 100644 --- a/src/renderer/src/components/ports/WorkspacePortScanner.tsx +++ b/src/renderer/src/components/ports/WorkspacePortScanner.tsx @@ -280,6 +280,7 @@ export function WorkspacePortScanner({ enabled = true }: { enabled?: boolean }): return } + let burstRefresh: Promise | null = null let eventSequence = 0 let disposed = false let retryTimer: ReturnType | null = null @@ -296,15 +297,24 @@ export function WorkspacePortScanner({ enabled = true }: { enabled?: boolean }): const sequence = eventSequence clearRetryTimer() if (!isWindowVisible()) { + burstRefresh = null return } - void refresh({ force: true, targets: [runtimeTarget] }).finally(() => { - if (disposed || sequence !== eventSequence || !isWindowVisible()) { + // Keep the leading scan through the quiet window so sequential events share it too. + burstRefresh ??= refresh({ force: true, targets: [runtimeTarget] }) + void burstRefresh.finally(() => { + if (disposed || sequence !== eventSequence) { + return + } + if (!isWindowVisible()) { + burstRefresh = null return } // Why: some dev servers print their URL just before the listener is // visible to lsof/netstat. One quiet settle scan catches that startup race. retryTimer = setTimeout(() => { + retryTimer = null + burstRefresh = null if (disposed || sequence !== eventSequence || !isWindowVisible()) { return } diff --git a/src/renderer/src/components/right-sidebar/source-control/listing/branch-section.tsx b/src/renderer/src/components/right-sidebar/source-control/listing/branch-section.tsx index 3f2ab4723f7..b96d756a5ef 100644 --- a/src/renderer/src/components/right-sidebar/source-control/listing/branch-section.tsx +++ b/src/renderer/src/components/right-sidebar/source-control/listing/branch-section.tsx @@ -91,8 +91,8 @@ export function SourceControlBranchSection({ -
{actions}
+ {conflictCount > 0 && ( + + {conflictCount}{' '} + {translate('auto.components.right.sidebar.SourceControl.413a3ba113', 'conflict')} + {conflictCount === 1 ? '' : 's'} + + )} +
+ +
{actions}
) diff --git a/src/renderer/src/components/right-sidebar/source-control/listing/uncommitted-sections.tsx b/src/renderer/src/components/right-sidebar/source-control/listing/uncommitted-sections.tsx index 36822be6e5d..c8591534055 100644 --- a/src/renderer/src/components/right-sidebar/source-control/listing/uncommitted-sections.tsx +++ b/src/renderer/src/components/right-sidebar/source-control/listing/uncommitted-sections.tsx @@ -113,8 +113,7 @@ export function SourceControlUncommittedSections(props: { onToggle={() => props.toggleSection(id)} actions={ <> - {/* Why: bulk actions are hover-only, but forced visible on no-hover pointers (touch/SSH; see AGENTS.md "SSH Use Case"). One wrapper so focusing any action reveals all three (else keyboard tabs into an invisible stop). */} -
+
{canRevertAll && ( entry.conflictStatus === 'unresolved') - ? 'h-6 px-1.5 text-[10px] text-muted-foreground hover:text-foreground' - : 'h-auto px-1.5 py-0.5 text-xs text-muted-foreground hover:text-foreground' - } + size="xs" + className="px-1.5 text-muted-foreground hover:text-foreground" onClick={(event) => { event.stopPropagation() props.onViewSection(sectionViewAction) diff --git a/src/renderer/src/components/settings/AccountsPane.tsx b/src/renderer/src/components/settings/AccountsPane.tsx index 82e2b1c3683..eb19fc54306 100644 --- a/src/renderer/src/components/settings/AccountsPane.tsx +++ b/src/renderer/src/components/settings/AccountsPane.tsx @@ -80,6 +80,8 @@ export function AccountsPane({ const runtimeEnvironments = useAppStore((s) => s.runtimeEnvironments) const recordedOpenCodeSettingEditsRef = useRef>(new Set()) const [miniMaxCookieDraft, setMiniMaxCookieDraft] = useState('') + const [miniMaxApiKeyDraft, setMiniMaxApiKeyDraft] = useState('') + const [miniMaxApiKeyConfigured, setMiniMaxApiKeyConfigured] = useState(false) const [miniMaxConfigured, setMiniMaxConfigured] = useState(false) const [miniMaxCredentialBusy, setMiniMaxCredentialBusy] = useState(false) const localAccountRuntime = getSelectedAccountRuntime( @@ -222,18 +224,23 @@ export function AccountsPane({ const refreshMiniMaxCredentialStatus = async (): Promise => { try { const status = await window.api.minimaxCredentials.getStatus() - setMiniMaxConfigured(status.configured) + setMiniMaxConfigured(status.cookieConfigured) + setMiniMaxApiKeyConfigured(status.apiKeyConfigured) } catch (error) { console.error('Failed to load MiniMax credential status:', error) } } - const { saveMiniMaxCookie, clearMiniMaxCookie } = createMiniMaxCredentialActions({ - miniMaxCookieDraft, - setMiniMaxCookieDraft, - setMiniMaxConfigured, - setMiniMaxCredentialBusy, - recordFeatureInteraction - }) + const { saveMiniMaxCookie, clearMiniMaxCookie, saveMiniMaxApiKey, clearMiniMaxApiKey } = + createMiniMaxCredentialActions({ + miniMaxCookieDraft, + setMiniMaxCookieDraft, + miniMaxApiKeyDraft, + setMiniMaxApiKeyDraft, + setMiniMaxApiKeyConfigured, + setMiniMaxConfigured, + setMiniMaxCredentialBusy, + recordFeatureInteraction + }) useEffect(() => { void refreshMiniMaxCredentialStatus() @@ -335,6 +342,11 @@ export function AccountsPane({ runCodexAccountAction, recordOpenCodeSettingEdit, miniMaxRateLimits, + miniMaxApiKeyDraft, + setMiniMaxApiKeyDraft, + miniMaxApiKeyConfigured, + saveMiniMaxApiKey, + clearMiniMaxApiKey, miniMaxCookieDraft, setMiniMaxCookieDraft, miniMaxConfigured, diff --git a/src/renderer/src/components/settings/BrowserTerminalLinkActionsSetting.tsx b/src/renderer/src/components/settings/BrowserTerminalLinkActionsSetting.tsx index b658052db81..3335c456ebd 100644 --- a/src/renderer/src/components/settings/BrowserTerminalLinkActionsSetting.tsx +++ b/src/renderer/src/components/settings/BrowserTerminalLinkActionsSetting.tsx @@ -19,7 +19,7 @@ export function BrowserTerminalLinkActionsSetting({ }: BrowserTerminalLinkActionsSettingProps): React.JSX.Element { const title = translate( 'auto.components.settings.BrowserTerminalLinkActionsSetting.title', - 'Show terminal link actions' + 'Show link actions' ) const description = getTerminalLinkActionsDescription({ isMac }) diff --git a/src/renderer/src/components/settings/GrokAccountsSection.test.tsx b/src/renderer/src/components/settings/GrokAccountsSection.test.tsx index cd418597719..09e9372283b 100644 --- a/src/renderer/src/components/settings/GrokAccountsSection.test.tsx +++ b/src/renderer/src/components/settings/GrokAccountsSection.test.tsx @@ -8,7 +8,8 @@ import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' const mocks = vi.hoisted(() => ({ getStatus: vi.fn(), - refreshGrokRateLimits: vi.fn() + refreshGrokRateLimits: vi.fn(), + grokUsage: vi.fn<() => unknown>(() => null) })) vi.mock('@/lib/agent-catalog', () => ({ @@ -29,7 +30,8 @@ vi.mock('../../store', () => ({ useAppStore: (selector: (state: Record) => unknown) => selector({ refreshGrokRateLimits: mocks.refreshGrokRateLimits, - rateLimits: { grok: null } + settingsSearchQuery: '', + rateLimits: { grok: mocks.grokUsage() } }) })) @@ -45,6 +47,7 @@ describe('GrokAccountsSection', () => { error: null }) mocks.refreshGrokRateLimits.mockResolvedValue(undefined) + mocks.grokUsage.mockReturnValue(null) Object.defineProperty(window, 'api', { configurable: true, value: { grokAccounts: { getStatus: mocks.getStatus } } @@ -66,4 +69,23 @@ describe('GrokAccountsSection', () => { ).toBeInTheDocument() expect(screen.queryByText(/grok login/i)).not.toBeInTheDocument() }) + + // Why: #15740 — an unreported percentage must be stated, never shown as 0%. + it('shows why usage is unknown instead of hiding the row', async () => { + mocks.grokUsage.mockReturnValue({ + provider: 'grok', + session: null, + weekly: null, + updatedAt: Date.now(), + error: 'Grok did not report a usage percentage for this account', + status: 'unavailable' + }) + + render() + + expect( + await screen.findByText('Grok did not report a usage percentage for this account') + ).toBeInTheDocument() + expect(screen.queryByText('0%')).not.toBeInTheDocument() + }) }) diff --git a/src/renderer/src/components/settings/GrokAccountsSection.tsx b/src/renderer/src/components/settings/GrokAccountsSection.tsx index 8df194dbd2d..b4f6a64157f 100644 --- a/src/renderer/src/components/settings/GrokAccountsSection.tsx +++ b/src/renderer/src/components/settings/GrokAccountsSection.tsx @@ -56,6 +56,12 @@ export function GrokAccountsSection(): React.JSX.Element { // monthly included usage instead of hiding the usage row entirely. const usageIsWeekly = Boolean(grokUsage?.weekly) const usageWindow = grokUsage?.weekly ?? grokUsage?.monthly ?? null + // Why: hiding the row entirely left signed-in users with no explanation when + // Grok reports no percentage — never let unknown usage read as healthy (#15740). + const unavailableReason = + signedIn && !usageWindow && grokUsage?.status === 'unavailable' + ? (grokUsage.error ?? null) + : null return (
@@ -198,6 +204,17 @@ export function GrokAccountsSection(): React.JSX.Element { ) : null}
+ ) : unavailableReason ? ( + +

{unavailableReason}

+
) : null} ) diff --git a/src/renderer/src/components/settings/MobileSettingsPane.tsx b/src/renderer/src/components/settings/MobileSettingsPane.tsx index ff8de2b16e5..2dd10a006f4 100644 --- a/src/renderer/src/components/settings/MobileSettingsPane.tsx +++ b/src/renderer/src/components/settings/MobileSettingsPane.tsx @@ -13,7 +13,7 @@ export { getMobileSettingsPaneSearchEntries } const ORCA_IOS_APP_STORE_URL = 'https://apps.apple.com/app/orca-ide/id6766130217' const ORCA_ANDROID_APK_URL = - 'https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.47/app-release.apk' + 'https://github.com/stablyai/orca/releases/download/mobile-android-v0.0.48/app-release.apk' export function MobileSettingsPane(): React.JSX.Element { const showMobileButton = useAppStore((s) => s.settings?.showMobileButton !== false) diff --git a/src/renderer/src/components/settings/accounts-pane-minimax-actions.ts b/src/renderer/src/components/settings/accounts-pane-minimax-actions.ts index a670ef3ff3b..0b5c220d50b 100644 --- a/src/renderer/src/components/settings/accounts-pane-minimax-actions.ts +++ b/src/renderer/src/components/settings/accounts-pane-minimax-actions.ts @@ -4,6 +4,9 @@ import { toast } from 'sonner' import { translate } from '@/i18n/i18n' type MiniMaxCredentialActionContext = { + miniMaxApiKeyDraft: string + setMiniMaxApiKeyDraft: Dispatch> + setMiniMaxApiKeyConfigured: Dispatch> miniMaxCookieDraft: string setMiniMaxCookieDraft: Dispatch> setMiniMaxConfigured: Dispatch> @@ -12,10 +15,15 @@ type MiniMaxCredentialActionContext = { } export function createMiniMaxCredentialActions(context: MiniMaxCredentialActionContext): { + saveMiniMaxApiKey: () => Promise + clearMiniMaxApiKey: () => Promise saveMiniMaxCookie: () => Promise clearMiniMaxCookie: () => Promise } { const { + miniMaxApiKeyDraft, + setMiniMaxApiKeyDraft, + setMiniMaxApiKeyConfigured, miniMaxCookieDraft, setMiniMaxCookieDraft, setMiniMaxConfigured, @@ -32,7 +40,7 @@ export function createMiniMaxCredentialActions(context: MiniMaxCredentialActionC setMiniMaxCredentialBusy(true) try { const status = await window.api.minimaxCredentials.saveCookie(miniMaxCookieDraft.trim()) - if (!status.configured) { + if (!status.cookieConfigured) { throw new Error( translate( 'auto.components.settings.AccountsPane.8e6f0cb1d8', @@ -40,7 +48,7 @@ export function createMiniMaxCredentialActions(context: MiniMaxCredentialActionC ) ) } - setMiniMaxConfigured(status.configured) + setMiniMaxConfigured(status.cookieConfigured) setMiniMaxCookieDraft('') recordFeatureInteraction('usage-tracking') toast.success( @@ -52,7 +60,7 @@ export function createMiniMaxCredentialActions(context: MiniMaxCredentialActionC 'auto.components.settings.AccountsPane.b43e761fe5', 'MiniMax cookie update failed.' ), - { description: String((error as Error)?.message ?? error) } + { description: error instanceof Error ? error.message : String(error) } ) } finally { setMiniMaxCredentialBusy(false) @@ -63,7 +71,7 @@ export function createMiniMaxCredentialActions(context: MiniMaxCredentialActionC setMiniMaxCredentialBusy(true) try { const status = await window.api.minimaxCredentials.clearCookie() - setMiniMaxConfigured(status.configured) + setMiniMaxConfigured(status.cookieConfigured) setMiniMaxCookieDraft('') recordFeatureInteraction('usage-tracking') } catch (error) { @@ -72,12 +80,72 @@ export function createMiniMaxCredentialActions(context: MiniMaxCredentialActionC 'auto.components.settings.AccountsPane.b43e761fe5', 'MiniMax cookie update failed.' ), - { description: String((error as Error)?.message ?? error) } + { description: error instanceof Error ? error.message : String(error) } ) } finally { setMiniMaxCredentialBusy(false) } } - return { saveMiniMaxCookie, clearMiniMaxCookie } + const saveMiniMaxApiKey = async (): Promise => { + if (!miniMaxApiKeyDraft.trim()) { + toast.error( + translate( + 'auto.components.settings.AccountsPane.d6f1b9b6a2', + 'MiniMax API key is required.' + ) + ) + return + } + setMiniMaxCredentialBusy(true) + try { + const status = await window.api.minimaxCredentials.saveApiKey(miniMaxApiKeyDraft.trim()) + if (!status.apiKeyConfigured) { + throw new Error( + translate( + 'auto.components.settings.AccountsPane.7c5d8a4e1b', + 'MiniMax API key was not saved.' + ) + ) + } + setMiniMaxApiKeyConfigured(status.apiKeyConfigured) + setMiniMaxApiKeyDraft('') + recordFeatureInteraction('usage-tracking') + toast.success( + translate('auto.components.settings.AccountsPane.4d2c7b9e83', 'MiniMax API key saved.') + ) + } catch (error) { + toast.error( + translate( + 'auto.components.settings.AccountsPane.b43e761fe5', + 'MiniMax credential update failed.' + ), + { description: error instanceof Error ? error.message : String(error) } + ) + } finally { + setMiniMaxCredentialBusy(false) + } + } + + const clearMiniMaxApiKey = async (): Promise => { + setMiniMaxCredentialBusy(true) + try { + const status = await window.api.minimaxCredentials.clearApiKey() + setMiniMaxApiKeyConfigured(status.apiKeyConfigured) + setMiniMaxApiKeyDraft('') + recordFeatureInteraction('usage-tracking') + } catch (error) { + toast.error( + translate( + 'auto.components.settings.AccountsPane.b43e761fe5', + 'MiniMax credential update failed.' + ), + { description: error instanceof Error ? error.message : String(error) } + ) + } finally { + setMiniMaxCredentialBusy(false) + } + } + + return { saveMiniMaxCookie, clearMiniMaxCookie, saveMiniMaxApiKey, clearMiniMaxApiKey } } diff --git a/src/renderer/src/components/settings/accounts-pane-minimax-credentials.tsx b/src/renderer/src/components/settings/accounts-pane-minimax-credentials.tsx new file mode 100644 index 00000000000..9fa2b3c35d0 --- /dev/null +++ b/src/renderer/src/components/settings/accounts-pane-minimax-credentials.tsx @@ -0,0 +1,275 @@ +import { HelpCircle, Loader2, Lock, LockOpen } from 'lucide-react' +import { useNow } from '../../hooks/use-now' +import { translate } from '@/i18n/i18n' +import { formatUiRelativeTime } from '@/i18n/relative-time-format' +import { Badge } from '../ui/badge' +import { Button } from '../ui/button' +import { Input } from '../ui/input' +import { Label } from '../ui/label' +import { Popover, PopoverContent, PopoverTrigger } from '../ui/popover' +import { SearchableSetting } from './SearchableSetting' +import type { AccountsPaneSectionModel } from './accounts-pane-types' + +function formatMiniMaxRelativeRefresh(updatedAt: number, now: number): string { + const diffMs = Math.max(0, now - updatedAt) + if (diffMs < 60_000) { + return translate('auto.components.settings.AccountsPane.3a30aaf526', 'just now') + } + return formatUiRelativeTime(-diffMs) +} + +function MiniMaxCookieHelpPopover({ consoleUrl }: { consoleUrl: string }): React.JSX.Element { + const steps = [ + translate( + 'auto.components.settings.AccountsPane.openSelectedConsole', + 'Open {{url}} in your browser and sign in.', + { url: consoleUrl } + ), + translate('auto.components.settings.AccountsPane.24560fe830', 'Open DevTools.'), + translate( + 'auto.components.settings.AccountsPane.4cab0fa42d', + 'Go to the Network tab and enable Preserve log.' + ), + translate('auto.components.settings.AccountsPane.bee4e63e1c', 'Reload the page.'), + translate( + 'auto.components.settings.AccountsPane.87f814af6f', + 'Filter for remains and select the coding_plan/remains request.' + ), + translate( + 'auto.components.settings.AccountsPane.435df0ee51', + 'Under Request Headers, copy the Cookie value.' + ), + translate('auto.components.settings.AccountsPane.7492fb3bba', 'Paste it here and click Save.') + ] + return ( +
+
+

+ {translate('auto.components.settings.AccountsPane.9fec52de4b', 'How to copy the cookie')} +

+

+ {translate( + 'auto.components.settings.AccountsPane.cookieSelectedEndpoint', + 'Stored locally and sent to the selected MiniMax endpoint for usage refreshes.' + )} +

+
+
    + {steps.map((step) => ( +
  1. {step}
  2. + ))} +
+
+ ) +} + +export function MiniMaxCredentials({ + model, + consoleUrl +}: { + model: AccountsPaneSectionModel + consoleUrl: string +}): React.JSX.Element { + const { + miniMaxCookieDraft, + setMiniMaxCookieDraft, + miniMaxConfigured, + miniMaxCredentialBusy, + miniMaxRateLimits, + saveMiniMaxCookie, + clearMiniMaxCookie, + miniMaxApiKeyDraft, + setMiniMaxApiKeyDraft, + miniMaxApiKeyConfigured, + saveMiniMaxApiKey, + clearMiniMaxApiKey + } = model + const now = useNow(60_000) + return ( + <> + +
+
+ + + {miniMaxConfigured ? : } + {miniMaxConfigured + ? translate('auto.components.settings.AccountsPane.73ea15f24b', 'Saved') + : translate('auto.components.settings.AccountsPane.23afe8f226', 'Not saved')} + +
+ + + + + + + + +
+
+ setMiniMaxCookieDraft(e.target.value)} + placeholder={translate( + 'auto.components.settings.AccountsPane.b8a4f21c3e', + 'Paste the Cookie header from DevTools' + )} + spellCheck={false} + className="flex-1 text-xs" + /> + + {miniMaxConfigured ? ( + + ) : null} +
+

+ {translate( + 'auto.components.settings.AccountsPane.copySelectedConsoleCookie', + 'Open the selected console, sign in, then copy the Cookie request header from DevTools (Network → any remains request → Cookie).' + )} +

+ {miniMaxConfigured && + miniMaxRateLimits?.status === 'ok' && + miniMaxRateLimits.error === null ? ( +

+ {translate( + 'auto.components.settings.AccountsPane.53f7b8c7a2', + 'Last refresh: {{value0}}', + { + value0: formatMiniMaxRelativeRefresh(miniMaxRateLimits.updatedAt, now) + } + )} +

+ ) : null} +

+ {translate( + 'auto.components.settings.AccountsPane.31d24a4e87', + 'Cookie expires when you sign out in the browser.' + )} +

+
+ + +
+
+ + + {miniMaxApiKeyConfigured ? ( + + ) : ( + + )} + {miniMaxApiKeyConfigured + ? translate('auto.components.settings.AccountsPane.73ea15f24b', 'Saved') + : translate('auto.components.settings.AccountsPane.23afe8f226', 'Not saved')} + +
+
+
+ setMiniMaxApiKeyDraft(e.target.value)} + placeholder={translate( + 'auto.components.settings.AccountsPane.4f2c8a7e1b', + 'Paste your MiniMax API key' + )} + spellCheck={false} + className="flex-1 text-xs" + /> + + {miniMaxApiKeyConfigured ? ( + + ) : null} +
+

+ {translate( + 'auto.components.settings.AccountsPane.apiKeyInstructions', + 'Copy the API key from your MiniMax console → API keys. A saved API key takes priority over the cookie; use Forget key to switch back to the cookie.' + )} +

+
+ + ) +} diff --git a/src/renderer/src/components/settings/accounts-pane-minimax-section.tsx b/src/renderer/src/components/settings/accounts-pane-minimax-section.tsx index ce72bc0a16d..8a13b1e4f87 100644 --- a/src/renderer/src/components/settings/accounts-pane-minimax-section.tsx +++ b/src/renderer/src/components/settings/accounts-pane-minimax-section.tsx @@ -1,83 +1,36 @@ -import { ExternalLink, HelpCircle, Loader2, Lock, LockOpen, ShieldCheck } from 'lucide-react' +import { ExternalLink, ShieldCheck } from 'lucide-react' import { translate } from '@/i18n/i18n' -import { formatUiRelativeTime } from '@/i18n/relative-time-format' import { cn } from '@/lib/utils' -import { Badge } from '../ui/badge' -import { Button } from '../ui/button' -import { Input } from '../ui/input' import { Label } from '../ui/label' -import { Popover, PopoverContent, PopoverTrigger } from '../ui/popover' import { MiniMaxIcon } from '../status-bar/icons' import { SearchableSetting } from './SearchableSetting' import type { AccountsPaneSectionModel } from './accounts-pane-types' import { DebouncedSettingsTextInput } from './DebouncedSettingsTextInput' -const MINIMAX_CONSOLE_URL = 'https://platform.minimax.io/console/usage' - -function formatMiniMaxRelativeRefresh(updatedAt: number, now: number): string { - const diffMs = Math.max(0, now - updatedAt) - if (diffMs < 60_000) { - return translate('auto.components.settings.AccountsPane.3a30aaf526', 'just now') - } - return formatUiRelativeTime(-diffMs) -} - -function MiniMaxCookieHelpPopover(): React.JSX.Element { - const steps = [ - translate( - 'auto.components.settings.AccountsPane.f5d8d2a6a1', - 'Open platform.minimax.io/console/usage in your browser and sign in.' - ), - translate('auto.components.settings.AccountsPane.24560fe830', 'Open DevTools.'), - translate( - 'auto.components.settings.AccountsPane.4cab0fa42d', - 'Go to the Network tab and enable Preserve log.' - ), - translate('auto.components.settings.AccountsPane.bee4e63e1c', 'Reload the page.'), - translate( - 'auto.components.settings.AccountsPane.87f814af6f', - 'Filter for remains and select the coding_plan/remains request.' - ), - translate( - 'auto.components.settings.AccountsPane.435df0ee51', - 'Under Request Headers, copy the Cookie value.' - ), - translate('auto.components.settings.AccountsPane.7492fb3bba', 'Paste it here and click Save.') - ] - return ( -
-
-

- {translate('auto.components.settings.AccountsPane.9fec52de4b', 'How to copy the cookie')} -

-

- {translate( - 'auto.components.settings.AccountsPane.4e32e030b2', - 'Stored locally. Orca sends it only to platform.minimax.io for usage refreshes.' - )} -

-
-
    - {steps.map((step) => ( -
  1. {step}
  2. - ))} -
-
- ) -} +import { MiniMaxCredentials } from './accounts-pane-minimax-credentials' +import { Select, SelectContent, SelectItem, SelectTrigger, SelectValue } from '../ui/select' export function renderMiniMaxAccountsSection(model: AccountsPaneSectionModel): React.JSX.Element { const { - clearMiniMaxCookie, miniMaxConfigured, - miniMaxCookieDraft, + miniMaxApiKeyConfigured, miniMaxCredentialBusy, - miniMaxRateLimits, - saveMiniMaxCookie, - setMiniMaxCookieDraft, settings, - updateSettings + updateSettings, + recordFeatureInteraction } = model + const consoleUrl = + settings.minimaxEndpoint === 'cn' + ? 'https://platform.minimaxi.com/console/usage' + : 'https://platform.minimax.io/console/usage' + const configured = miniMaxConfigured || miniMaxApiKeyConfigured + const handleMiniMaxEndpointChange = (value: string): void => { + if ((value !== 'overseas' && value !== 'cn') || value === settings.minimaxEndpoint) { + return + } + recordFeatureInteraction('usage-tracking') + void updateSettings({ minimaxEndpoint: value }) + } return (
@@ -88,13 +41,13 @@ export function renderMiniMaxAccountsSection(model: AccountsPaneSectionModel): R

{translate( - 'auto.components.settings.AccountsPane.15e831350e', - 'Configure MiniMax usage tracking from platform.minimax.io.' + 'auto.components.settings.AccountsPane.usageTracking', + 'Configure MiniMax usage tracking for your account.' )}

- {miniMaxConfigured + {configured ? translate('auto.components.settings.AccountsPane.0b8c1c7e02', 'Stored locally') - : translate('auto.components.settings.AccountsPane.1fd1b1b6b4', 'Cookie not set')} + : translate( + 'auto.components.settings.AccountsPane.credentialsNotSet', + 'Credentials not set' + )}

{translate( - 'auto.components.settings.AccountsPane.5e08b0fe57', - 'Stored locally and sent only to platform.minimax.io for usage refreshes.' + 'auto.components.settings.AccountsPane.selectedEndpointStorage', + 'Stored locally and sent to the selected MiniMax endpoint for usage refreshes.' )}

-
-
- - - {miniMaxConfigured ? : } - {miniMaxConfigured - ? translate('auto.components.settings.AccountsPane.73ea15f24b', 'Saved') - : translate('auto.components.settings.AccountsPane.23afe8f226', 'Not saved')} - -
- - - - - - - - -
-
- setMiniMaxCookieDraft(e.target.value)} - placeholder={translate( - 'auto.components.settings.AccountsPane.b8a4f21c3e', - 'Paste the Cookie header from DevTools' - )} - spellCheck={false} - className="flex-1 text-xs" - /> - - {miniMaxConfigured ? ( - - ) : null} -
-

- {translate( - 'auto.components.settings.AccountsPane.79418c782a', - 'Open platform.minimax.io/console/usage in your browser, sign in, then copy the Cookie request header from DevTools (Network → any remains request → Cookie).' - )} -

- {miniMaxConfigured && - miniMaxRateLimits?.status === 'ok' && - miniMaxRateLimits.error === null ? ( -

- {translate( - 'auto.components.settings.AccountsPane.53f7b8c7a2', - 'Last refresh: {{value0}}', + + + +

diff --git a/src/renderer/src/components/settings/accounts-pane-types.ts b/src/renderer/src/components/settings/accounts-pane-types.ts index c4ea87bf4bf..2799cfd6509 100644 --- a/src/renderer/src/components/settings/accounts-pane-types.ts +++ b/src/renderer/src/components/settings/accounts-pane-types.ts @@ -105,6 +105,11 @@ export type AccountsPaneSectionModel = { runCodexAccountAction: CodexAccountActionRunner recordOpenCodeSettingEdit: (field: 'cookie' | 'workspaceId') => void miniMaxRateLimits: ProviderRateLimits | null + miniMaxApiKeyDraft: string + setMiniMaxApiKeyDraft: Dispatch> + miniMaxApiKeyConfigured: boolean + saveMiniMaxApiKey: () => Promise + clearMiniMaxApiKey: () => Promise miniMaxCookieDraft: string setMiniMaxCookieDraft: Dispatch> miniMaxConfigured: boolean diff --git a/src/renderer/src/components/settings/accounts-search.test.ts b/src/renderer/src/components/settings/accounts-search.test.ts index 7958045b8e1..7e18f59722e 100644 --- a/src/renderer/src/components/settings/accounts-search.test.ts +++ b/src/renderer/src/components/settings/accounts-search.test.ts @@ -23,8 +23,8 @@ describe('getAccountsMiniMaxSearchEntries', () => { expect(entries).toHaveLength(1) const [entry] = entries expect(entry.title).toBe('MiniMax Usage') - expect(entry.description).toContain('platform.minimax.io') expect(entry.description.toLowerCase()).toContain('cookie') + expect(entry.description.toLowerCase()).toContain('api key') }) it('exposes the keywords that drive the Settings search index', () => { diff --git a/src/renderer/src/components/settings/accounts-search.ts b/src/renderer/src/components/settings/accounts-search.ts index 6ff07366bc9..efcf16c7d69 100644 --- a/src/renderer/src/components/settings/accounts-search.ts +++ b/src/renderer/src/components/settings/accounts-search.ts @@ -176,12 +176,16 @@ export const getAccountsMiniMaxSearchEntries = createLocalizedCatalog(() => [ title: translate('auto.components.settings.accounts.search.733f9e2a93', 'MiniMax Usage'), description: translate( 'auto.components.settings.accounts.search.f8374c3151', - 'Paste your platform.minimax.io session cookie for local rate-limit fetching.' + 'Configure MiniMax usage tracking. Pick the overseas or China endpoint, then paste a session cookie or save an API key that works on either host.' ), keywords: [ ...translateSearchKeyword('auto.components.settings.accounts.search.d16378a88f', 'minimax'), ...translateSearchKeyword('auto.components.settings.accounts.search.61f7d1fcbe', 'cookie'), ...translateSearchKeyword('auto.components.settings.accounts.search.9c4e40cf6b', 'session'), + ...translateSearchKeyword('auto.components.settings.accounts.search.b2c4e7f1a8', 'endpoint'), + ...translateSearchKeyword('auto.components.settings.accounts.search.3a9b6d2c4e', 'api key'), + ...translateSearchKeyword('auto.components.settings.accounts.search.5d8f1a3b7c', 'china'), + ...translateSearchKeyword('auto.components.settings.accounts.search.7e2a4b8c1d', 'overseas'), ...translateSearchKeyword( 'auto.components.settings.accounts.search.e949b08ffb', 'rate limit' diff --git a/src/renderer/src/components/settings/browser-link-routing-copy.ts b/src/renderer/src/components/settings/browser-link-routing-copy.ts index bdf005749e2..688887ac810 100644 --- a/src/renderer/src/components/settings/browser-link-routing-copy.ts +++ b/src/renderer/src/components/settings/browser-link-routing-copy.ts @@ -7,7 +7,7 @@ export function getBrowserLinkRoutingShortcutLabel(platform: { isMac: boolean }) export function getTerminalLinkActionsDescription(platform: { isMac: boolean }): string { return translate( 'auto.components.settings.BrowserTerminalLinkActionsSetting.description', - 'Show available actions when you click a terminal link. Turn this off to require {{modifier}}-click.', + 'Show available actions when you click a link in the terminal or a chat transcript. Turn this off to require {{modifier}}-click in the terminal.', { modifier: platform.isMac ? '⌘' : 'Ctrl' } ) } diff --git a/src/renderer/src/components/settings/browser-search.test.ts b/src/renderer/src/components/settings/browser-search.test.ts index 2acc2c0aed4..f6fa3ebb18a 100644 --- a/src/renderer/src/components/settings/browser-search.test.ts +++ b/src/renderer/src/components/settings/browser-search.test.ts @@ -61,7 +61,7 @@ describe('browser settings search copy', () => { expect(linkRoutingEntry?.keywords).not.toContain('cmd') const terminalActionsEntry = getBrowserPaneSearchEntries({ isMac: false }).find( - (entry) => entry.title === 'Show terminal link actions' + (entry) => entry.title === 'Show link actions' ) expect(terminalActionsEntry?.description).toContain('Ctrl-click') expect(terminalActionsEntry?.description).not.toContain('Cmd/Ctrl') @@ -102,7 +102,7 @@ describe('browser link routing modifier copy', () => { 'Default Zoom', 'Link Routing', 'Hold Shift to open in Orca', - 'Show terminal link actions', + 'Show link actions', 'Localhost Worktree Labels', 'Session & Cookies', 'Remote server workspaces', diff --git a/src/renderer/src/components/settings/browser-search.ts b/src/renderer/src/components/settings/browser-search.ts index 2903ab9f062..0a99a37de74 100644 --- a/src/renderer/src/components/settings/browser-search.ts +++ b/src/renderer/src/components/settings/browser-search.ts @@ -34,6 +34,10 @@ export function getTerminalLinkActionSearchKeywords(platform: BrowserShortcutPla 'auto.components.settings.browser.search.terminalLinkActions.terminal', 'terminal' ), + ...translateSearchKeyword( + 'auto.components.settings.browser.search.terminalLinkActions.chat', + 'chat' + ), ...translateSearchKeyword( 'auto.components.settings.browser.search.terminalLinkActions.click', 'click' @@ -182,7 +186,7 @@ export function getBrowserPaneSearchEntries( { title: translate( 'auto.components.settings.BrowserTerminalLinkActionsSetting.title', - 'Show terminal link actions' + 'Show link actions' ), description: getTerminalLinkActionsDescription(platform), keywords: getTerminalLinkActionSearchKeywords(platform) diff --git a/src/renderer/src/components/sidebar/delete-worktree-toast.ts b/src/renderer/src/components/sidebar/delete-worktree-toast.ts index 1c013da0da6..946e99eadee 100644 --- a/src/renderer/src/components/sidebar/delete-worktree-toast.ts +++ b/src/renderer/src/components/sidebar/delete-worktree-toast.ts @@ -74,6 +74,23 @@ export function getDeleteWorktreeToastCopy( isDestructive: false } } + if (forceDeleteReason === 'running-agent-session') { + return { + title: translate( + 'auto.components.sidebar.delete.worktree.toast.1d0fa5c0a5', + 'Failed to delete workspace {{value0}}', + { value0: worktreeName } + ), + // Why this is not the "could not confirm" wording: Orca watched these sessions stay + // attached, so there is no doubt to waive — Force Delete ends a conversation that is + // running right now, and any work it holds goes with it. + description: translate( + 'auto.components.sidebar.delete.worktree.toast.runningAgentSession', + 'This workspace still has running agent sessions, so Orca stopped before deleting any files. Force Delete will close them and discard any work they hold.' + ), + isDestructive: false + } + } if (forceDeleteReason === 'missing-registration') { return { title: translate( diff --git a/src/renderer/src/components/sidebar/folder-workspace-composer-submit.ts b/src/renderer/src/components/sidebar/folder-workspace-composer-submit.ts index b6f1a1654f3..ca685dded64 100644 --- a/src/renderer/src/components/sidebar/folder-workspace-composer-submit.ts +++ b/src/renderer/src/components/sidebar/folder-workspace-composer-submit.ts @@ -182,9 +182,7 @@ export async function submitFolderWorkspaceCreate({ linkedTask: toFolderWorkspaceLinkedTask(linkedWorkItem), ...(linkedTaskSourceContext ? { linkedTaskSourceContext } : {}), ...(quickAgent ? { createdWithAgent: quickAgent } : {}), - ...(pendingFirstAgentMessageRename && !structuredLaunch - ? { pendingFirstAgentMessageRename: true } - : {}) + ...(pendingFirstAgentMessageRename ? { pendingFirstAgentMessageRename: true } : {}) }) if (!workspace) { return false diff --git a/src/renderer/src/components/stats/GrokUsagePane.test.tsx b/src/renderer/src/components/stats/GrokUsagePane.test.tsx index d1ef3700f83..42e8e6577ab 100644 --- a/src/renderer/src/components/stats/GrokUsagePane.test.tsx +++ b/src/renderer/src/components/stats/GrokUsagePane.test.tsx @@ -38,6 +38,7 @@ const mockStoreState = { status: 'ok' }, minimaxCookieConfigured: false, + minimaxApiKeyConfigured: false, grokAuthConfigured: true, claudeTarget: { runtime: 'host', wslDistro: null }, codexTarget: { runtime: 'host', wslDistro: null }, diff --git a/src/renderer/src/components/status-bar/status-bar-provider-visibility.test.ts b/src/renderer/src/components/status-bar/status-bar-provider-visibility.test.ts index e833bce7fa4..41a933a850b 100644 --- a/src/renderer/src/components/status-bar/status-bar-provider-visibility.test.ts +++ b/src/renderer/src/components/status-bar/status-bar-provider-visibility.test.ts @@ -73,6 +73,7 @@ function usageSettings(overrides: Partial = {}): UsagePro geminiCliOAuthEnabled: false, antigravityUsageConfigured: false, minimaxCookieConfigured: false, + minimaxApiKeyConfigured: false, grokAuthConfigured: false, ...overrides } @@ -127,6 +128,7 @@ describe('hasUsageProviderSettings', () => { false ) expect(hasUsageProviderSettings(usageSettings({ minimaxCookieConfigured: true }))).toBe(true) + expect(hasUsageProviderSettings(usageSettings({ minimaxApiKeyConfigured: true }))).toBe(true) expect(hasUsageProviderSettings(usageSettings({ grokAuthConfigured: true }))).toBe(true) }) @@ -196,6 +198,24 @@ describe('hasUsageProviderSettingsForProvider', () => { expect(hasUsageProviderSettingsForProvider('minimax', null)).toBe(false) }) + it('treats minimaxApiKeyConfigured as a parallel durable signal for MiniMax', () => { + // Why: CN endpoint users can configure MiniMax with an API key only. The + // visibility check must accept either credential so the status bar stays + // visible while the snapshot is still pending. + expect( + hasUsageProviderSettingsForProvider( + 'minimax', + usageSettings({ minimaxApiKeyConfigured: true }) + ) + ).toBe(true) + expect( + hasUsageProviderSettingsForProvider( + 'minimax', + usageSettings({ minimaxApiKeyConfigured: false, minimaxCookieConfigured: false }) + ) + ).toBe(false) + }) + it('treats grokAuthConfigured as the durable signal for Grok', () => { expect( hasUsageProviderSettingsForProvider('grok', usageSettings({ grokAuthConfigured: true })) diff --git a/src/renderer/src/components/status-bar/status-bar-provider-visibility.ts b/src/renderer/src/components/status-bar/status-bar-provider-visibility.ts index f1afd97f5f4..19258592f1f 100644 --- a/src/renderer/src/components/status-bar/status-bar-provider-visibility.ts +++ b/src/renderer/src/components/status-bar/status-bar-provider-visibility.ts @@ -16,6 +16,7 @@ export type UsageProviderSettings = Pick< antigravityUsageConfigured: boolean // Why: MiniMax/Grok sign-in live on disk, not in settings; main sets these each poll. minimaxCookieConfigured: boolean + minimaxApiKeyConfigured: boolean grokAuthConfigured: boolean } @@ -77,6 +78,7 @@ export function hasUsageProviderSettings( // Antigravity's durable signal requires geminiCliOAuthEnabled, so it is // already covered by the gemini term above. settings?.minimaxCookieConfigured === true || + settings?.minimaxApiKeyConfigured === true || settings?.grokAuthConfigured === true ) } @@ -107,7 +109,7 @@ export function hasUsageProviderSettingsForProvider( return settings.antigravityUsageConfigured === true && settings.geminiCliOAuthEnabled === true } if (providerId === 'minimax') { - return settings.minimaxCookieConfigured === true + return settings.minimaxCookieConfigured === true || settings.minimaxApiKeyConfigured === true } if (providerId === 'grok') { return settings.grokAuthConfigured === true diff --git a/src/renderer/src/components/status-bar/use-status-bar-controller.ts b/src/renderer/src/components/status-bar/use-status-bar-controller.ts index 0bbbb6d1071..33db1976b0d 100644 --- a/src/renderer/src/components/status-bar/use-status-bar-controller.ts +++ b/src/renderer/src/components/status-bar/use-status-bar-controller.ts @@ -112,6 +112,7 @@ export function useStatusBarController(floatingTerminalOpen: boolean) { ...settings, antigravityUsageConfigured, minimaxCookieConfigured: rateLimits.minimaxCookieConfigured, + minimaxApiKeyConfigured: rateLimits.minimaxApiKeyConfigured, grokAuthConfigured: rateLimits.grokAuthConfigured } const visibleClaude = getVisibleUsageProvider('claude', claude, usageSettings) diff --git a/src/renderer/src/components/tab-bar/EditorFileTab.test.tsx b/src/renderer/src/components/tab-bar/EditorFileTab.test.tsx index 6eb614550c9..edc15a53f3c 100644 --- a/src/renderer/src/components/tab-bar/EditorFileTab.test.tsx +++ b/src/renderer/src/components/tab-bar/EditorFileTab.test.tsx @@ -345,6 +345,15 @@ function findMenuItemByText(node: unknown, label: string): ReactElementLike { return item } +/** Picks Rename, then fires the close-autofocus that actually opens the input. */ +function selectRenameFromMenu(node: unknown): void { + ;(findMenuItemByText(node, 'Rename').props.onSelect as () => void)() + const content = findElementsByType(node, 'DropdownMenuContent')[0]! + ;(content.props.onCloseAutoFocus as (event: { preventDefault: () => void }) => void)({ + preventDefault: vi.fn() + }) +} + function findSpanByText(node: unknown, label: string): ReactElementLike { const span = findElementsByType(node, 'span').find( (candidate) => @@ -401,7 +410,7 @@ describe('EditorFileTab rename menu', () => { // isUntitled; the tab menu must let users rename the screenshot-style // "untitled-N.md" files directly. expect(renameItem.props.disabled).toBe(false) - ;(renameItem.props.onSelect as () => void)() + selectRenameFromMenu(firstRender) const secondRender = expandNode((await renderEditorFileTab(file, onActivate)).element) const inputs = findElementsByType(secondRender, 'input') @@ -425,9 +434,8 @@ describe('EditorFileTab rename menu', () => { it('ignores IME composition Enter before renaming the editor file tab', async () => { const file = baseFile() const firstRender = expandNode((await renderEditorFileTab(file)).element) - const renameItem = findMenuItemByText(firstRender, 'Rename') - ;(renameItem.props.onSelect as () => void)() + selectRenameFromMenu(firstRender) const secondRender = expandNode((await renderEditorFileTab(file)).element) const input = findElementsByType(secondRender, 'input')[0] @@ -457,9 +465,8 @@ describe('EditorFileTab rename menu', () => { it('does not re-commit when unmounting the rename input emits multiple blur events', async () => { const file = baseFile() const firstRender = expandNode((await renderEditorFileTab(file)).element) - const renameItem = findMenuItemByText(firstRender, 'Rename') - ;(renameItem.props.onSelect as () => void)() + selectRenameFromMenu(firstRender) const secondRender = expandNode((await renderEditorFileTab(file)).element) const input = findElementsByType(secondRender, 'input')[0] diff --git a/src/renderer/src/components/tab-bar/EditorFileTab.tsx b/src/renderer/src/components/tab-bar/EditorFileTab.tsx index 78d17b8b929..064ac1fdb0e 100644 --- a/src/renderer/src/components/tab-bar/EditorFileTab.tsx +++ b/src/renderer/src/components/tab-bar/EditorFileTab.tsx @@ -171,8 +171,8 @@ export default function EditorFileTab({ if (!input) { return } - // Why: Radix closes the context menu after onSelect; defer focus so its - // teardown cannot steal focus back or blur-commit the newly mounted input. + // Why: the tab re-lays out around the input; focus on the next frame so + // that swap has settled before selecting text. renameFocusFrameRef.current = requestAnimationFrame(() => { renameFocusFrameRef.current = null if (renameInputRef.current !== input) { diff --git a/src/renderer/src/components/tab-bar/EditorFileTabContextMenu.test.tsx b/src/renderer/src/components/tab-bar/EditorFileTabContextMenu.test.tsx index 00b2d2a210c..812079563af 100644 --- a/src/renderer/src/components/tab-bar/EditorFileTabContextMenu.test.tsx +++ b/src/renderer/src/components/tab-bar/EditorFileTabContextMenu.test.tsx @@ -202,7 +202,9 @@ function extractText(node: unknown): string { return el.props && 'children' in el.props ? extractText(el.props.children) : '' } -async function renderMenu(): Promise { +async function renderMenu( + overrides: { onActivate?: () => void; onOpenRenameInput?: () => void } = {} +): Promise { const module = await import('./EditorFileTabContextMenu') return module.EditorFileTabContextMenu({ open: true, @@ -238,7 +240,8 @@ async function renderMenu(): Promise { onCloseAll: vi.fn(), onCloseToRight: vi.fn(), onCloseToLeft: vi.fn(), - onOpenMarkdownPreview: vi.fn() + onOpenMarkdownPreview: vi.fn(), + ...overrides }) } @@ -266,6 +269,27 @@ describe('EditorFileTabContextMenu close-all shortcut', () => { vi.unstubAllGlobals() }) + it('opens rename only after menu close releases focus and consumes the request once', async () => { + const onActivate = vi.fn() + const onOpenRenameInput = vi.fn() + const tree = expandNode(await renderMenu({ onActivate, onOpenRenameInput })) + const rename = findElementsByType(tree, 'DropdownMenuItem').find((item) => + extractText(item.props.children).includes('Rename') + )! + const content = findElementsByType(tree, 'DropdownMenuContent')[0]! + ;(rename.props.onSelect as () => void)() + expect(onActivate).not.toHaveBeenCalled() + expect(onOpenRenameInput).not.toHaveBeenCalled() + const preventDefault = vi.fn() + const close = content.props.onCloseAutoFocus as (event: { preventDefault: () => void }) => void + close({ preventDefault }) + expect(preventDefault).toHaveBeenCalledTimes(1) + expect(onActivate).toHaveBeenCalledTimes(1) + expect(onOpenRenameInput).toHaveBeenCalledTimes(1) + close({ preventDefault }) + expect(onOpenRenameInput).toHaveBeenCalledTimes(1) + }) + it('renders assigned shortcuts next to Rename, Close, and Close All Editor Tabs', async () => { const tree = expandNode(await renderMenu()) const menuItems = findElementsByType(tree, 'DropdownMenuItem') diff --git a/src/renderer/src/components/tab-bar/EditorFileTabContextMenu.tsx b/src/renderer/src/components/tab-bar/EditorFileTabContextMenu.tsx index 32e29595738..1813265573f 100644 --- a/src/renderer/src/components/tab-bar/EditorFileTabContextMenu.tsx +++ b/src/renderer/src/components/tab-bar/EditorFileTabContextMenu.tsx @@ -126,6 +126,10 @@ export function EditorFileTabContextMenu({ } skipMenuFocusRestoreRef.current = false event.preventDefault() + // Why: opening the input in onSelect lets the still-closing menu reclaim + // focus, and the resulting blur commits the rename away before the user types. + onActivate() + onOpenRenameInput() }} > { skipMenuFocusRestoreRef.current = true - onActivate() - onOpenRenameInput() }} > diff --git a/src/renderer/src/components/tab-bar/TabBarCreateEntry.tab-results.test.tsx b/src/renderer/src/components/tab-bar/TabBarCreateEntry.tab-results.test.tsx index d73156d533a..85009b38f25 100644 --- a/src/renderer/src/components/tab-bar/TabBarCreateEntry.tab-results.test.tsx +++ b/src/renderer/src/components/tab-bar/TabBarCreateEntry.tab-results.test.tsx @@ -11,6 +11,7 @@ import type { OpenTabSearchEntries } from './open-tab-search-entries' import type { TabAgentLaunchOption } from './tab-agent-launch-options' import type { TabCreateMenuOption } from './tab-create-menu-options' import type { TabEntryOption } from './tab-create-entry-action' +import { encodePaletteIdentity } from '@/lib/palette-match/palette-ranking' // Why: the real entry-action module pulls in runtime IPC + the app store; these // tests only need a controllable option list beneath the tab rows. @@ -132,11 +133,15 @@ import TabBarCreateEntry from './TabBarCreateEntry' ;(globalThis as { IS_REACT_ACT_ENVIRONMENT?: boolean }).IS_REACT_ACT_ENVIRONMENT = true +function openWorkspaceTabId(tabId: string): string { + return encodePaletteIdentity(['workspace-tab', 'local', 'wt', tabId]) +} + function terminalResult(overrides: Partial = {}): OpenTabSearchResult { return { executionHostId: 'local', source: 'workspace', - id: 'open-tab:workspace:tab-1', + id: openWorkspaceTabId('tab-1'), title: 'Add tab search and jump in worktree', matchedText: null, worktreeId: 'wt', @@ -264,7 +269,11 @@ describe('TabBarCreateEntry tab results', () => { it('shows the matched text rather than the shared label when tabs share a title (AE2)', () => { tabSearchMock.resultsByQuery['fix the flaky'] = [ terminalResult({ title: 'Claude Code', matchedText: 'fix the flaky retry test' }), - terminalResult({ id: 'open-tab:workspace:tab-2', tabId: 'tab-2', title: 'Claude Code' }) + terminalResult({ + id: openWorkspaceTabId('tab-2'), + tabId: 'tab-2', + title: 'Claude Code' + }) ] renderEntry() @@ -415,7 +424,11 @@ describe('TabBarCreateEntry tab results', () => { activationMocks.workspace.mockReturnValue({ status: 'failed', reason: 'missing-tab' }) tabSearchMock.resultsByQuery['add tab'] = [ terminalResult(), - terminalResult({ id: 'open-tab:workspace:tab-2', tabId: 'tab-2', title: 'second tab' }) + terminalResult({ + id: openWorkspaceTabId('tab-2'), + tabId: 'tab-2', + title: 'second tab' + }) ] const onDidOpenEntry = vi.fn() const onOpenEntry = vi.fn().mockResolvedValue(undefined) diff --git a/src/renderer/src/components/tab-bar/TabBarCreateEntry.tsx b/src/renderer/src/components/tab-bar/TabBarCreateEntry.tsx index 5c39d07bf84..8998fa5cd32 100644 --- a/src/renderer/src/components/tab-bar/TabBarCreateEntry.tsx +++ b/src/renderer/src/components/tab-bar/TabBarCreateEntry.tsx @@ -88,7 +88,8 @@ function TabBarCreateEntrySession({ const tabResults = useTabCreateEntrySearchResults({ enabled: menuOpen && !terminalQueryMode, query, - worktreeId + worktreeId, + retainedResultId: pinnedOptionId }) const shouldResolveAbsolutePaths = menuOpen && !terminalQueryMode && isTabEntryAbsolutePathLike(query.trim()) diff --git a/src/renderer/src/components/tab-bar/TabBarCreateEntryRow.tsx b/src/renderer/src/components/tab-bar/TabBarCreateEntryRow.tsx index 280574c7a62..f93e2ad2190 100644 --- a/src/renderer/src/components/tab-bar/TabBarCreateEntryRow.tsx +++ b/src/renderer/src/components/tab-bar/TabBarCreateEntryRow.tsx @@ -184,7 +184,9 @@ function getActionPresentation( } if (option.kind === 'tab') { return { - detail: option.option.matchedText ?? option.option.title, + detail: option.option.matchedTexts?.length + ? option.option.matchedTexts.join(' · ') + : (option.option.matchedText ?? option.option.title), icon: getOpenTabIcon(option.option), label: translate('auto.components.tab.bar.TabBarCreateEntry.8f0a1c4d92', 'Switch to tab'), showDetail: true diff --git a/src/renderer/src/components/tab-bar/open-tab-search-entries.ts b/src/renderer/src/components/tab-bar/open-tab-search-entries.ts index 4361e392ce4..dd02eb998ee 100644 --- a/src/renderer/src/components/tab-bar/open-tab-search-entries.ts +++ b/src/renderer/src/components/tab-bar/open-tab-search-entries.ts @@ -14,12 +14,12 @@ import { type SearchableWorkspaceTab } from '@/lib/workspace-tab-palette-search' import type { AppState } from '@/store/types' -import { getIndexedAllWorktrees } from '@/store/worktree-repo-index' import { getRepoExecutionHostId, getWorktreeExecutionHostId, type ExecutionHostId } from '../../../../shared/execution-host' +import { getPaletteOwnershipWorktreeIds } from '@/lib/unified-tab-host-ownership' export type OpenTabSearchEntries = { workspaceTabs: readonly SearchableWorkspaceTab[] @@ -40,14 +40,15 @@ export type OpenTabSearchEntryState = Pick< | 'activeWorktreeId' | 'browserPagesByWorkspace' | 'browserTabsByWorktree' + | 'folderWorkspaces' | 'groupsByWorktree' | 'openFiles' | 'tabsByWorktree' | 'unifiedTabsByWorktree' + | 'worktreesByRepo' > & { executionHostId: ExecutionHostId generatedTitlesEnabled: boolean - ownershipWorktrees: readonly Pick[] repo: Pick | null worktree: Worktree } @@ -100,14 +101,15 @@ export function selectOpenTabSearchEntryState( browserPagesByWorkspace: state.browserPagesByWorkspace, browserTabsByWorktree: state.browserTabsByWorktree, executionHostId, + folderWorkspaces: state.folderWorkspaces, generatedTitlesEnabled: state.settings?.tabAutoGenerateTitle === true, groupsByWorktree: state.groupsByWorktree, openFiles: state.openFiles, - ownershipWorktrees: getIndexedAllWorktrees(state.worktreesByRepo), repo, tabsByWorktree: state.tabsByWorktree, unifiedTabsByWorktree: state.unifiedTabsByWorktree, - worktree + worktree, + worktreesByRepo: state.worktreesByRepo } } @@ -133,7 +135,7 @@ export function buildOpenTabSearchEntries( const worktrees = [scopedWorktree] const scope = { worktrees, - ownershipWorktrees: state.ownershipWorktrees, + ownershipWorktrees: getPaletteOwnershipWorktreeIds(state), repoMap: new Map(repo ? [[repo.id, repo]] : []), worktreeOrder: new Map([[worktree.id, 0]]) } diff --git a/src/renderer/src/components/tab-bar/open-tab-search.test.ts b/src/renderer/src/components/tab-bar/open-tab-search.test.ts index 665b5a460de..d304c37a519 100644 --- a/src/renderer/src/components/tab-bar/open-tab-search.test.ts +++ b/src/renderer/src/components/tab-bar/open-tab-search.test.ts @@ -21,6 +21,7 @@ import { type OpenTabSearchInput, type OpenTabSearchResult } from './open-tab-search' +import { createPaletteSearchContext } from '@/lib/palette-match/palette-ranking' const worktree: Worktree = { id: 'wt-1', @@ -71,7 +72,8 @@ function makeWorkspaceTab({ occupantAgent = null, tabSortIndex = 0, groupSortIndex = 0, - isCurrentTab = false + isCurrentTab = false, + createdAt = 0 }: { id: string title: string @@ -83,10 +85,13 @@ function makeWorkspaceTab({ tabSortIndex?: number groupSortIndex?: number isCurrentTab?: boolean + createdAt?: number }): SearchableWorkspaceTab { const searchTexts = secondarySearchTexts ?? (secondaryText ? [secondaryText] : []) + const tab = makeTab(id, contentType) as SearchableWorkspaceTab['tab'] + tab.createdAt = createdAt return { - tab: makeTab(id, contentType) as SearchableWorkspaceTab['tab'], + tab, worktree, repoName: REPO_NAME, worktreeSortIndex: 0, @@ -218,7 +223,62 @@ function search(input: Partial & { query: string }): OpenTab }) } +function readableId(result: OpenTabSearchResult): string { + return `open-tab:${result.source}:${result.source === 'browser' ? result.pageId : result.tabId}` +} + describe('searchOpenTabs ranking', () => { + it('uses the shared Atlas order before applying the four-row cap', () => { + const now = 100 * 24 * 60 * 60 * 1000 + const age = (milliseconds: number): number => now - milliseconds + const workspaceTabs = [ + makeWorkspaceTab({ + id: 'old-prefix-2d', + title: 'atlas-follow-up-draft-2026-09-01.md', + createdAt: age(2 * 24 * 60 * 60 * 1000) + }), + makeWorkspaceTab({ + id: 'old-prefix-3d', + title: 'atlas-meeting-todo.md', + createdAt: age(3 * 24 * 60 * 60 * 1000) + }), + makeWorkspaceTab({ + id: 'recent-title', + title: 'Clarify Atlas action items', + createdAt: age(30_000) + }), + makeWorkspaceTab({ + id: 'recent-path', + title: 'questions-and-answers.md', + secondaryText: 'notes/atlas/questions.md', + createdAt: age(30 * 60 * 1000) + }), + makeWorkspaceTab({ + id: 'older-path', + title: 'worklog.md', + secondaryText: 'notes/atlas/worklog.md', + createdAt: age(9 * 60 * 60 * 1000) + }), + makeWorkspaceTab({ + id: 'older-title', + title: 'Advance Atlas security review', + createdAt: age(19 * 60 * 60 * 1000) + }) + ] + const results = search({ + query: 'atlas', + context: createPaletteSearchContext(now), + workspaceTabs + }) + + expect(results.map((result) => (result.source === 'workspace' ? result.tabId : ''))).toEqual([ + 'recent-title', + 'older-title', + 'old-prefix-2d', + 'old-prefix-3d' + ]) + }) + it('ranks a title-prefix match above a title-substring match from another source', () => { const results = search({ query: 'zebra', @@ -226,13 +286,10 @@ describe('searchOpenTabs ranking', () => { browserPages: [makeBrowserPage({ id: 'page-1', title: 'Zebra release notes' })] }) - expect(results.map((result) => result.id)).toEqual([ - 'open-tab:browser:page-1', - 'open-tab:workspace:tab-1' - ]) + expect(results.map(readableId)).toEqual(['open-tab:browser:page-1', 'open-tab:workspace:tab-1']) }) - it('ranks any title match above any secondary match', () => { + it('ranks a primary word match above a comparable secondary word match', () => { const results = search({ query: 'zebra', workspaceTabs: [ @@ -246,13 +303,13 @@ describe('searchOpenTabs ranking', () => { simulatorTabs: [makeSimulatorTab({ id: 'sim-1', label: 'Trailing zebra' })] }) - expect(results.map((result) => result.id)).toEqual([ + expect(results.map(readableId)).toEqual([ 'open-tab:simulator:sim-1', 'open-tab:workspace:tab-secondary' ]) }) - // Both land in the secondary tier, so match rank has to beat tab position: the + // Both use secondary coverage, so match rank has to beat tab position: the // agent tab sits earlier in the group and would win a position-only tie-break. it('ranks a path match above an agent-snippet match on tabs in the same group', () => { const results = search({ @@ -274,13 +331,13 @@ describe('searchOpenTabs ranking', () => { ] }) - expect(results.map((result) => result.id)).toEqual([ + expect(results.map(readableId)).toEqual([ 'open-tab:workspace:tab-path', 'open-tab:workspace:tab-agent' ]) }) - it('breaks tier ties on source order, then on engine score', () => { + it('breaks semantic and activity ties on source order, then engine score', () => { const results = search({ query: 'zebra', workspaceTabs: [ @@ -291,7 +348,7 @@ describe('searchOpenTabs ranking', () => { simulatorTabs: [makeSimulatorTab({ id: 'sim-1', label: 'Zebra emulator' })] }) - expect(results.map((result) => result.id)).toEqual([ + expect(results.map(readableId)).toEqual([ 'open-tab:workspace:tab-early', 'open-tab:workspace:tab-late', 'open-tab:browser:page-1', @@ -312,13 +369,50 @@ describe('searchOpenTabs ranking', () => { }) expect(results).toHaveLength(OPEN_TAB_SEARCH_RESULT_LIMIT) - expect(results.map((result) => result.id)).toEqual([ + expect(results.map(readableId)).toEqual([ 'open-tab:workspace:tab-0', 'open-tab:workspace:tab-1', 'open-tab:workspace:tab-2', 'open-tab:workspace:tab-3' ]) }) + + it('reserves one capped slot for a retained eligible result', () => { + const input = { + query: 'zebra', + workspaceTabs: [0, 1, 2, 3, 4].map((index) => + makeWorkspaceTab({ id: `tab-${index}`, title: `Zebra ${index}`, tabSortIndex: index }) + ) + } + const uncappedSelection = searchOpenTabs({ + browserPages: [], + simulatorTabs: [], + ...input + })[3] + input.workspaceTabs[4].tab.createdAt = Date.now() + const retained = searchOpenTabs({ + browserPages: [], + simulatorTabs: [], + ...input, + retainedResultId: uncappedSelection.id + }) + + expect(retained).toHaveLength(OPEN_TAB_SEARCH_RESULT_LIMIT) + expect(retained.some((result) => result.id === uncappedSelection.id)).toBe(true) + }) + + it('ranks an exact browser destination above a workspace typo', () => { + const results = search({ + query: 'zebra', + workspaceTabs: [makeWorkspaceTab({ id: 'tab-typo', title: 'zebrb' })], + browserPages: [makeBrowserPage({ id: 'page-exact', title: 'Notes', url: 'zebra' })] + }) + + expect(results.map(readableId)).toEqual([ + 'open-tab:browser:page-exact', + 'open-tab:workspace:tab-typo' + ]) + }) }) describe('searchOpenTabs filtering', () => { @@ -334,7 +428,7 @@ describe('searchOpenTabs filtering', () => { ] }) - expect(results.map((result) => result.id)).toEqual([ + expect(results.map(readableId)).toEqual([ 'open-tab:workspace:tab-1', 'open-tab:browser:page-1', 'open-tab:simulator:sim-1' @@ -372,12 +466,49 @@ describe('searchOpenTabs filtering', () => { ] }) - expect(results.map((result) => result.id)).toEqual([ - 'open-tab:workspace:tab-1', - 'open-tab:browser:page-1' + expect(results.map(readableId)).toEqual(['open-tab:workspace:tab-1', 'open-tab:browser:page-1']) + }) + + it('keeps branch matches while excluding worktree and repository fields', () => { + expect( + search({ + query: 'main', + workspaceTabs: [makeWorkspaceTab({ id: 'tab-1', title: 'Notes' })] + }).map(readableId) + ).toEqual(['open-tab:workspace:tab-1']) + }) + + it('uses an admissible title proof when the unrestricted match prefers the worktree', () => { + const entry = makeWorkspaceTab({ id: 'tab-1', title: 'atlaz' }) + entry.document = buildPaletteTabDocument({ + id: 'tab-1', + title: 'atlaz', + secondaryTexts: [], + worktreeName: 'atlas', + branch: BRANCH_NAME, + repoName: REPO_NAME + }) + + expect(search({ query: 'atlas', workspaceTabs: [entry] }).map(readableId)).toEqual([ + 'open-tab:workspace:tab-1' ]) }) + it('does not create a snippet fallback when only excluded structured fields match', () => { + expect( + search({ + query: 'aurora', + workspaceTabs: [ + makeWorkspaceTab({ + id: 'tab-1', + title: 'Notes', + agentSnippets: ['aurora agent notes'] + }) + ] + }) + ).toEqual([]) + }) + // Both tokens land on the "ios simulator" alias, so the row fills no title or // secondary range — the inverse test would drop it. it('keeps a simulator alias match that spans two keywords', () => { @@ -386,7 +517,7 @@ describe('searchOpenTabs filtering', () => { simulatorTabs: [makeSimulatorTab({ id: 'sim-1', label: 'Pixel 8' })] }) - expect(results.map((result) => result.id)).toEqual(['open-tab:simulator:sim-1']) + expect(results.map(readableId)).toEqual(['open-tab:simulator:sim-1']) }) }) @@ -433,6 +564,51 @@ describe('searchOpenTabs result fields', () => { }) }) + it('keeps editor paths scoped to their host and worktree when tab ids repeat', () => { + const local = makeWorkspaceTab({ + id: 'same-tab', + title: 'Atlas', + contentType: 'editor', + secondaryText: 'local/atlas.ts' + }) + const remote = makeWorkspaceTab({ + id: 'same-tab', + title: 'Atlas', + contentType: 'editor', + secondaryText: 'remote/atlas.ts' + }) + remote.worktree = { ...worktree, hostId: 'ssh:remote' } + remote.tab = { ...remote.tab, executionHostId: 'ssh:remote' } + const sibling = makeWorkspaceTab({ + id: 'same-tab', + title: 'Atlas', + contentType: 'editor', + secondaryText: 'sibling/atlas.ts' + }) + sibling.worktree = { ...worktree, id: 'wt-2' } + sibling.tab = { ...sibling.tab, worktreeId: 'wt-2' } + + expect(search({ query: 'Atlas', workspaceTabs: [local, remote, sibling] })).toEqual( + expect.arrayContaining([ + expect.objectContaining({ + executionHostId: 'local', + worktreeId: 'wt-1', + relativePath: 'local/atlas.ts' + }), + expect.objectContaining({ + executionHostId: 'ssh:remote', + worktreeId: 'wt-1', + relativePath: 'remote/atlas.ts' + }), + expect.objectContaining({ + executionHostId: 'local', + worktreeId: 'wt-2', + relativePath: 'sibling/atlas.ts' + }) + ]) + ) + }) + it('copies a confident occupant agent onto workspace results', () => { const results = search({ query: 'grok', diff --git a/src/renderer/src/components/tab-bar/open-tab-search.ts b/src/renderer/src/components/tab-bar/open-tab-search.ts index 05aa6126872..b4d58b04b29 100644 --- a/src/renderer/src/components/tab-bar/open-tab-search.ts +++ b/src/renderer/src/components/tab-bar/open-tab-search.ts @@ -1,11 +1,16 @@ // Merges the three Cmd+J open-tab engines into one ranked list for the new-tab // omnibox. Pure: no store, no React. +import { capPaletteSection } from '../cmd-j/palette-section-render-cap' import { isClipboardTextByteLengthOverLimit } from '../../../../shared/clipboard-text' +import type { PaletteDocumentRank } from '@/lib/palette-match/palette-document' import { - comparePaletteDocumentRank, - type PaletteDocumentRank -} from '@/lib/palette-match/palette-document' + comparePaletteEntityRanks, + createPaletteSearchContext, + encodePaletteIdentity, + type PaletteActivityRank, + type PaletteSearchContext +} from '@/lib/palette-match/palette-ranking' import { LOCAL_EXECUTION_HOST_ID, type ExecutionHostId } from '../../../../shared/execution-host' import { searchBrowserPages, @@ -17,6 +22,7 @@ import { type SearchableSimulatorTab, type SimulatorPaletteSearchResult } from '@/lib/simulator-palette-search' +import { getUnifiedTabPaletteExecutionHostId } from '@/lib/unified-tab-host-ownership' import type { TuiAgent } from '../../../../shared/tui-agent' import { searchWorkspaceTabs, @@ -39,6 +45,7 @@ type OpenTabSearchResultBase = { title: string /** Engine secondary text when the match came from a secondary field. */ matchedText: string | null + matchedTexts?: readonly string[] worktreeId: string } @@ -72,14 +79,16 @@ export type OpenTabSearchInput = { browserPages: readonly SearchableBrowserPage[] simulatorTabs: readonly SearchableSimulatorTab[] query: string + context?: PaletteSearchContext + retainedResultId?: string | null } type RankedResult = { result: OpenTabSearchResult - tier: number - sourceRank: number - matchRank: PaletteDocumentRank | null - score: number + matchRank: PaletteDocumentRank + activity: PaletteActivityRank + position: readonly [number, number] + identity: string } const SOURCE_RANK: Record = { @@ -88,13 +97,6 @@ const SOURCE_RANK: Record = { simulator: 2 } -const TITLE_PREFIX_TIER = 0 -const TITLE_SUBSTRING_TIER = 1 -// Why one tier for every secondary match: path and agent-snippet matches share -// `secondaryRanges`, so splitting on offset would outrank the engine's own match -// rank, which is compared explicitly below. See the plan's tiering decision. -const SECONDARY_TIER = 2 - function isOpenTabSearchQueryTooLarge( query: string, maxBytes = OPEN_TAB_SEARCH_QUERY_MAX_BYTES @@ -107,21 +109,6 @@ type EngineResult = | BrowserPaletteSearchResult | SimulatorPaletteSearchResult -// Why the positive signal rather than "no title and no secondary range": the -// simulator alias branch and the browser workspace-label branch are real matches -// that carry neither range, and would be dropped by the inverse test. -function isNameOnlyMatch(result: EngineResult): boolean { - return result.worktreeRanges.length > 0 || result.repoRanges.length > 0 -} - -function getTier(result: EngineResult): number { - const titleRange = result.titleRanges[0] - if (!titleRange) { - return SECONDARY_TIER - } - return titleRange.start === 0 ? TITLE_PREFIX_TIER : TITLE_SUBSTRING_TIER -} - function getMatchedText(result: EngineResult): string | null { return result.secondaryRanges.length > 0 ? result.secondaryText : null } @@ -138,16 +125,15 @@ function getEditorRelativePath(entry: SearchableWorkspaceTab | undefined): strin } function baseResult( - source: OpenTabSearchSource, - id: string, result: EngineResult, executionHostId: ExecutionHostId ): OpenTabSearchResultBase { return { executionHostId, - id: `open-tab:${source}:${id}`, + id: result.paletteIdentity, title: result.title, matchedText: getMatchedText(result), + matchedTexts: result.secondaryMatches.map((match) => match.text).filter(Boolean), worktreeId: result.worktreeId } } @@ -157,84 +143,129 @@ function rank( results: readonly TEngine[], toResult: (result: TEngine) => OpenTabSearchResult ): RankedResult[] { - return results - .filter((result) => !isNameOnlyMatch(result)) - .map((result) => ({ - tier: getTier(result), - sourceRank: SOURCE_RANK[source], - matchRank: result.rank, - score: result.score, - result: toResult(result) - })) + return results.flatMap((result) => { + if (!result.rank) { + return [] + } + const converted = toResult(result) + return [ + { + matchRank: result.rank, + activity: result.activity, + position: [SOURCE_RANK[source], result.score], + result: converted, + identity: converted.id + } + ] + }) } -export function searchOpenTabs({ +export function searchOpenTabCandidates({ workspaceTabs, browserPages, simulatorTabs, - query + query, + context: suppliedContext }: OpenTabSearchInput): OpenTabSearchResult[] { const trimmed = query.trim() if (!trimmed || isOpenTabSearchQueryTooLarge(query)) { return [] } - // Single-worktree builders stamp one host on every entry; resolve once. - const executionHostId = - workspaceTabs[0]?.worktree.hostId ?? - browserPages[0]?.worktree.hostId ?? - simulatorTabs[0]?.worktree.hostId ?? - LOCAL_EXECUTION_HOST_ID + const context = suppliedContext ?? createPaletteSearchContext(Date.now()) // Why map workspace only: editor relativePath is read from the searchable entry. - const workspaceEntriesByTabId = new Map(workspaceTabs.map((entry) => [entry.tab.id, entry])) + const workspaceEntriesByIdentity = new Map( + workspaceTabs.map((entry) => [ + encodePaletteIdentity([ + getUnifiedTabPaletteExecutionHostId(entry.tab, entry.worktree) ?? LOCAL_EXECUTION_HOST_ID, + entry.worktree.id, + entry.tab.id + ]), + entry + ]) + ) return [ // Why no isCurrentTab filter: Cmd+J lists the tab you are on, and hiding it // made the omnibox look broken when you searched for the tab on screen. - ...rank('workspace', searchWorkspaceTabs([...workspaceTabs], trimmed), (result) => ({ - ...baseResult('workspace', result.tabId, result, executionHostId), - source: 'workspace', - contentType: result.contentType, - tabId: result.tabId, - entityId: result.entityId, - groupId: result.groupId, - relativePath: getEditorRelativePath(workspaceEntriesByTabId.get(result.tabId)), - occupantAgent: result.occupantAgent - })), - ...rank('browser', searchBrowserPages([...browserPages], trimmed), (result) => ({ - ...baseResult('browser', result.pageId, result, executionHostId), - source: 'browser', - contentType: 'browser', - pageId: result.pageId, - workspaceId: result.workspaceId, - url: result.url, - faviconUrl: result.faviconUrl - })), - ...rank('simulator', searchSimulatorTabs([...simulatorTabs], trimmed), (result) => ({ - ...baseResult('simulator', result.tabId, result, executionHostId), - source: 'simulator', - contentType: 'simulator', - tabId: result.tabId, - groupId: result.groupId - })) + ...rank( + 'workspace', + searchWorkspaceTabs([...workspaceTabs], trimmed, { context, fieldMode: 'omnibox' }), + (result) => ({ + ...baseResult(result, result.executionHostId ?? LOCAL_EXECUTION_HOST_ID), + source: 'workspace', + contentType: result.contentType, + tabId: result.tabId, + entityId: result.entityId, + groupId: result.groupId, + relativePath: getEditorRelativePath( + workspaceEntriesByIdentity.get( + encodePaletteIdentity([ + result.executionHostId ?? LOCAL_EXECUTION_HOST_ID, + result.worktreeId, + result.tabId + ]) + ) + ), + occupantAgent: result.occupantAgent + }) + ), + ...rank( + 'browser', + searchBrowserPages([...browserPages], trimmed, { context, fieldMode: 'omnibox' }), + (result) => ({ + ...baseResult(result, result.executionHostId ?? LOCAL_EXECUTION_HOST_ID), + source: 'browser', + contentType: 'browser', + pageId: result.pageId, + workspaceId: result.workspaceId, + url: result.url, + faviconUrl: result.faviconUrl + }) + ), + ...rank( + 'simulator', + searchSimulatorTabs([...simulatorTabs], trimmed, { context, fieldMode: 'omnibox' }), + (result) => ({ + ...baseResult(result, result.executionHostId ?? LOCAL_EXECUTION_HOST_ID), + source: 'simulator', + contentType: 'simulator', + tabId: result.tabId, + groupId: result.groupId + }) + ) ] .sort((a, b) => { - if (a.tier !== b.tier) { - return a.tier - b.tier - } - if (a.sourceRank !== b.sourceRank) { - return a.sourceRank - b.sourceRank - } - // Why before position: `score` is position-only now, so without this an - // agent-snippet fallback in an earlier tab would outrank a real path match. - if (a.matchRank && b.matchRank) { - const byMatch = comparePaletteDocumentRank(a.matchRank, b.matchRank) - if (byMatch !== 0) { - return byMatch + return comparePaletteEntityRanks( + { + rank: a.matchRank, + activity: a.activity, + position: a.position, + identity: a.identity + }, + { + rank: b.matchRank, + activity: b.activity, + position: b.position, + identity: b.identity } - } - return a.score - b.score + ) }) - .slice(0, OPEN_TAB_SEARCH_RESULT_LIMIT) .map((ranked) => ranked.result) } + +export function searchOpenTabs(input: OpenTabSearchInput): OpenTabSearchResult[] { + return capOpenTabSearchCandidates(searchOpenTabCandidates(input), input.retainedResultId) +} + +export function capOpenTabSearchCandidates( + candidates: readonly OpenTabSearchResult[], + retainedResultId?: string | null +): OpenTabSearchResult[] { + const capped = capPaletteSection( + candidates, + OPEN_TAB_SEARCH_RESULT_LIMIT, + (result) => result.id === retainedResultId + ) + return [...capped.visible] +} diff --git a/src/renderer/src/components/tab-bar/use-open-tab-search.test.ts b/src/renderer/src/components/tab-bar/use-open-tab-search.test.ts index c5a647ad96e..56bb65b8a09 100644 --- a/src/renderer/src/components/tab-bar/use-open-tab-search.test.ts +++ b/src/renderer/src/components/tab-bar/use-open-tab-search.test.ts @@ -1,7 +1,7 @@ // @vitest-environment happy-dom import { act, renderHook } from '@testing-library/react' -import { beforeEach, describe, expect, it } from 'vitest' +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' import type { BrowserPage, BrowserWorkspace } from '../../../../shared/browser-workspace-types' import type { Repo } from '../../../../shared/repo-types' import type { Tab, TabContentType, TabGroup } from '../../../../shared/tab-types' @@ -13,6 +13,8 @@ import { useOpenTabSearch } from './use-open-tab-search' const initialAppState = useAppStore.getInitialState() +afterEach(() => vi.restoreAllMocks()) + function makeWorktree(id: string, displayName: string): Worktree { return { id, @@ -387,6 +389,33 @@ describe('useOpenTabSearch', () => { expect(result.current.results.map((entry) => entry.title)).toEqual(['zebra epsilon']) }) + it('uses a fresh shared clock when the tab snapshot changes', () => { + const clock = vi.spyOn(Date, 'now').mockReturnValue(1_000) + const { result } = renderSearch() + + clock.mockReturnValue(2_000) + const state = useAppStore.getState() + act(() => { + useAppStore.setState({ + unifiedTabsByWorktree: { + ...state.unifiedTabsByWorktree, + 'wt-1': (state.unifiedTabsByWorktree['wt-1'] ?? []).map((tab) => + tab.id === 'tab-a' + ? { ...tab, lastFocusedAt: 1_800 } + : tab.id === 'tab-b' + ? { ...tab, lastFocusedAt: 1_900 } + : tab + ) + } + }) + }) + + expect(result.current.results.slice(0, 2).map((entry) => entry.title)).toEqual([ + 'zebra beta', + 'zebra alpha' + ]) + }) + it('reflects the generated-titles setting in matched titles', () => { seedStore({ tabsByWorktree: { diff --git a/src/renderer/src/components/tab-bar/use-open-tab-search.ts b/src/renderer/src/components/tab-bar/use-open-tab-search.ts index 693e6941e1d..2baf1387779 100644 --- a/src/renderer/src/components/tab-bar/use-open-tab-search.ts +++ b/src/renderer/src/components/tab-bar/use-open-tab-search.ts @@ -9,7 +9,12 @@ import { selectOpenTabSearchEntryState, type OpenTabSearchEntries } from './open-tab-search-entries' -import { searchOpenTabs, type OpenTabSearchResult } from './open-tab-search' +import { + capOpenTabSearchCandidates, + searchOpenTabCandidates, + type OpenTabSearchResult +} from './open-tab-search' +import { usePaletteSearchEvaluationContext } from '@/hooks/use-palette-search-evaluation-context' const EMPTY_RESULTS: OpenTabSearchResult[] = [] @@ -17,6 +22,8 @@ export type UseOpenTabSearchOptions = { enabled: boolean query: string worktreeId: string + /** Keyboard-selected result's `id`; keep it inside the display cap while it still matches. */ + retainedResultId?: string | null } export type OpenTabSearchSnapshot = { @@ -30,7 +37,8 @@ export type OpenTabSearchSnapshot = { export function useOpenTabSearch({ enabled, query, - worktreeId + worktreeId, + retainedResultId }: UseOpenTabSearchOptions): OpenTabSearchSnapshot { // Why null while disabled: a closed menu stays stable across store churn. const state = useAppStore( @@ -48,13 +56,29 @@ export function useOpenTabSearch({ [agentState, state] ) const deferredQuery = useDeferredValue(query) + const evaluationSnapshot = useMemo( + () => ({ deferredQuery, enabled, entries }), + [deferredQuery, enabled, entries] + ) + const context = usePaletteSearchEvaluationContext(evaluationSnapshot) + const candidates = useMemo( + () => + entries + ? searchOpenTabCandidates({ + ...entries, + query: deferredQuery, + context + }) + : EMPTY_RESULTS, + [context, deferredQuery, entries] + ) return useMemo( () => ({ query: deferredQuery, entries, - results: entries ? searchOpenTabs({ ...entries, query: deferredQuery }) : EMPTY_RESULTS + results: capOpenTabSearchCandidates(candidates, retainedResultId) }), - [deferredQuery, entries] + [candidates, deferredQuery, entries, retainedResultId] ) } diff --git a/src/renderer/src/components/tab-bar/use-tab-create-entry-search-results.ts b/src/renderer/src/components/tab-bar/use-tab-create-entry-search-results.ts index 72cb38a50d0..a4496a107ee 100644 --- a/src/renderer/src/components/tab-bar/use-tab-create-entry-search-results.ts +++ b/src/renderer/src/components/tab-bar/use-tab-create-entry-search-results.ts @@ -6,16 +6,19 @@ import type { OpenTabSearchResult } from './open-tab-search' export function useTabCreateEntrySearchResults({ enabled, query, - worktreeId + worktreeId, + retainedResultId }: { enabled: boolean query: string worktreeId: string + retainedResultId?: string | null }): readonly OpenTabSearchResult[] { const tabSearch = useOpenTabSearch({ enabled, query: enabled ? query : '', - worktreeId + worktreeId, + retainedResultId }) // Why retain instead of clearing: emptying deferred rows flashes the list on // every keystroke. Retention re-checks each row against the live query, so diff --git a/src/renderer/src/components/task-page-github-status-actions.ts b/src/renderer/src/components/task-page-github-status-actions.ts index 3722678a310..9a47c6a31a2 100644 --- a/src/renderer/src/components/task-page-github-status-actions.ts +++ b/src/renderer/src/components/task-page-github-status-actions.ts @@ -82,7 +82,7 @@ export function getTaskPageGitHubDuplicateTargetErrorMessage( } export function getTaskPageGitHubDuplicateCandidates( - items: GitHubWorkItem[], + items: readonly GitHubWorkItem[], currentIssueNumber: number, query: string ): GitHubWorkItem[] { diff --git a/src/renderer/src/components/task-page/github/StatusCell.tsx b/src/renderer/src/components/task-page/github/StatusCell.tsx index 38151d5e24c..1884d4fd663 100644 --- a/src/renderer/src/components/task-page/github/StatusCell.tsx +++ b/src/renderer/src/components/task-page/github/StatusCell.tsx @@ -31,6 +31,8 @@ import { cn } from '@/lib/utils' import { CircleDot, ChevronDown, Copy, CheckCircle2, Ban, ChevronRight } from 'lucide-react' import type { TaskPageGitHubWorkItemMutationRunner } from '../../task-page-linear-jira-list-model' import { TaskPageGitHubDuplicatePicker } from './DuplicatePicker' +import { useGitHubDuplicateIssueCandidates } from '@/components/github/github-duplicate-issue-candidates' + export function GHStatusCell({ item, repo, @@ -50,27 +52,7 @@ export function GHStatusCell({ const [duplicatePickerOpen, setDuplicatePickerOpen] = useState(false) const [duplicateSearch, setDuplicateSearch] = useState('') const [duplicateError, setDuplicateError] = useState(null) - const duplicateIssueCandidates = useAppStore( - useShallow((s) => { - if (!duplicatePickerOpen) { - return [] - } - const deduped = new Map() - for (const entry of Object.values(s.workItemsCache)) { - for (const candidate of entry.data ?? []) { - if ( - candidate.type === 'issue' && - candidate.repoId === item.repoId && - candidate.number !== item.number && - !deduped.has(candidate.number) - ) { - deduped.set(candidate.number, candidate) - } - } - } - return Array.from(deduped.values()).sort((a, b) => b.number - a.number) - }) - ) + const duplicateIssueCandidates = useGitHubDuplicateIssueCandidates(item, duplicatePickerOpen) const repoOwnerSettings = useAppStore( useShallow((s) => getSettingsForRepoRuntimeOwner(s, repo?.id ?? null)) ) diff --git a/src/renderer/src/components/terminal-pane/TerminalPaneNativeChatPortal.tsx b/src/renderer/src/components/terminal-pane/TerminalPaneNativeChatPortal.tsx index da2a3f02133..426221aa464 100644 --- a/src/renderer/src/components/terminal-pane/TerminalPaneNativeChatPortal.tsx +++ b/src/renderer/src/components/terminal-pane/TerminalPaneNativeChatPortal.tsx @@ -17,7 +17,6 @@ export function TerminalPaneNativeChatPortal({ chatPaneOwnsTabWideLaunchDraft, chatPanePtyId, chatPaneResolvedAgent, - chatPaneDispatchStatus, contextMenu, effectiveChatViewMode, expandedPaneId, @@ -77,7 +76,6 @@ export function TerminalPaneNativeChatPortal({ isVisible={isRendererVisible} target={structuredChatTarget} contextMenuActions={contextMenuActions} - orchestrationDispatchStatus={chatPaneDispatchStatus} /> ) : ( )}
, diff --git a/src/renderer/src/components/terminal-pane/TerminalPaneSurface.tsx b/src/renderer/src/components/terminal-pane/TerminalPaneSurface.tsx index dc346ce95b4..4df42a74de2 100644 --- a/src/renderer/src/components/terminal-pane/TerminalPaneSurface.tsx +++ b/src/renderer/src/components/terminal-pane/TerminalPaneSurface.tsx @@ -9,7 +9,7 @@ import TerminalPaneHeaderOverlay from './TerminalPaneHeaderOverlay' import { isPaneOwnerUnverifiedError, TerminalErrorToast } from './TerminalErrorToast' import { requestTerminalPaneRecovery } from './terminal-pane-recovery' import { TerminalSessionStateSaveFailureDialog } from './TerminalSessionStateSaveFailureDialog' -import { TerminalLinkActionPopover } from './TerminalLinkActionPopover' +import { LinkActionPopover } from '@/components/link-actions/LinkActionPopover' import { TerminalAgentSessionForkDialog } from './TerminalAgentSessionForkDialog' import { SessionRestoredBannerPortals } from './SessionRestoredBannerPortals' import { handleInternalTerminalFileDrop } from './terminal-drop-handler' @@ -262,10 +262,7 @@ export function TerminalPaneSurface({ canCopyAgentSessionId={menuAgentSessionId !== null} onCopyAgentSessionId={() => void contextMenu.onCopyAgentSessionId()} /> - + {quickCommandEditorOpen ? ( { vi.stubGlobal('HTMLElement', FakeHTMLElement) vi.stubGlobal('document', { activeElement, + querySelectorAll: vi.fn(() => []), querySelector: vi.fn((selector: string) => selector === '[data-tab-rename-input="true"]' && renameInputMounted ? (new FakeHTMLElement({ tagName: 'INPUT' }) as unknown as Element) diff --git a/src/renderer/src/components/terminal-pane/pane-helpers.ts b/src/renderer/src/components/terminal-pane/pane-helpers.ts index 84715b241ca..94e4d96a162 100644 --- a/src/renderer/src/components/terminal-pane/pane-helpers.ts +++ b/src/renderer/src/components/terminal-pane/pane-helpers.ts @@ -1,4 +1,5 @@ import type { PaneManager } from '@/lib/pane-manager/pane-manager' +import { focusPanePreservingOverlays } from '@/lib/pane-manager/pane-overlay-focus' export function fitPanes(manager: PaneManager): void { manager.fitAllPanes() @@ -16,7 +17,9 @@ export function focusActivePane(manager: PaneManager): void { } const panes = manager.getPanes() const activePane = manager.getActivePane() ?? panes[0] - activePane?.terminal.focus() + if (activePane) { + focusPanePreservingOverlays(activePane) + } } export function fitAndFocusPanes(manager: PaneManager): void { diff --git a/src/renderer/src/components/terminal-pane/parked-terminal-byte-watcher.test.ts b/src/renderer/src/components/terminal-pane/parked-terminal-byte-watcher.test.ts index 1a9653d15c8..edc1e574fff 100644 --- a/src/renderer/src/components/terminal-pane/parked-terminal-byte-watcher.test.ts +++ b/src/renderer/src/components/terminal-pane/parked-terminal-byte-watcher.test.ts @@ -305,7 +305,7 @@ describe('startParkedTerminalByteWatcher', () => { dispose() }) - it('skips completion dispatch when tracking is fully disabled, keeping the cache timer', async () => { + it('keeps mobile completion detection active when desktop notifications and attention are off', async () => { mockStoreState.settings = { ...mockStoreState.settings, experimentalTerminalAttention: false, @@ -318,7 +318,10 @@ describe('startParkedTerminalByteWatcher', () => { flushSideEffects() vi.advanceTimersByTime(NOTIFICATION_GRACE_MS * 4) - expect(dispatchTerminalNotification).not.toHaveBeenCalled() + expect(dispatchTerminalNotification).toHaveBeenCalledWith( + WORKTREE_ID, + expect.objectContaining({ source: 'agent-task-complete', suppressOsNotification: true }) + ) expect(mockStoreState.setCacheTimerStartedAt).toHaveBeenLastCalledWith( PANE_KEY, expect.any(Number) diff --git a/src/renderer/src/components/terminal-pane/terminal-layout-overlay-focus.test.tsx b/src/renderer/src/components/terminal-pane/terminal-layout-overlay-focus.test.tsx new file mode 100644 index 00000000000..d8684dd5df1 --- /dev/null +++ b/src/renderer/src/components/terminal-pane/terminal-layout-overlay-focus.test.tsx @@ -0,0 +1,97 @@ +// @vitest-environment happy-dom +import { afterEach, describe, expect, it, vi } from 'vitest' +import type { PaneManager } from '@/lib/pane-manager/pane-manager' +import { fitAndFocusPanes } from './pane-helpers' + +function createLayoutFixture() { + const textarea = document.createElement('textarea') + textarea.className = 'xterm-helper-textarea' + document.body.append(textarea) + textarea.focus() + const terminal = { focus: vi.fn(() => textarea.focus()) } + const manager = { + fitAllPanes: vi.fn(), + getActivePane: () => ({ terminal }), + getPanes: () => [{ terminal }] + } as unknown as PaneManager + return { manager, terminal, textarea } +} + +function mountOverlay(role: string) { + const overlay = document.createElement('div') + overlay.setAttribute('role', role) + overlay.tabIndex = -1 + vi.spyOn(overlay, 'getClientRects').mockReturnValue([ + new DOMRect(0, 0, 100, 100) + ] as unknown as DOMRectList) + document.body.append(overlay) + return overlay +} + +afterEach(() => { + document.body.replaceChildren() + vi.restoreAllMocks() +}) + +describe('terminal layout preserves overlay focus', () => { + it.each(['menu', 'dialog', 'alertdialog', 'listbox'])( + 'does not blur an open %s during a queued fit', + (role) => { + const { manager, terminal } = createLayoutFixture() + const overlay = mountOverlay(role) + overlay.focus() + const blurred = vi.fn() + overlay.addEventListener('blur', blurred) + + fitAndFocusPanes(manager) + + expect(manager.fitAllPanes).toHaveBeenCalledOnce() + expect(terminal.focus).not.toHaveBeenCalled() + expect(document.activeElement).toBe(overlay) + expect(blurred).not.toHaveBeenCalled() + } + ) + + it('leaves a mounted menu time to acquire focus', () => { + const { manager, terminal, textarea } = createLayoutFixture() + textarea.blur() + mountOverlay('menu') + + fitAndFocusPanes(manager) + + expect(terminal.focus).not.toHaveBeenCalled() + expect(document.activeElement).toBe(document.body) + }) + + it('allows focus after the menu closes', () => { + const { manager, terminal, textarea } = createLayoutFixture() + const overlay = mountOverlay('menu') + overlay.focus() + overlay.remove() + + fitAndFocusPanes(manager) + + expect(terminal.focus).toHaveBeenCalledOnce() + expect(document.activeElement).toBe(textarea) + }) + + it('hands focus back to a menu that is animating closed', () => { + const { manager, terminal, textarea } = createLayoutFixture() + mountOverlay('menu').setAttribute('data-state', 'closed') + + fitAndFocusPanes(manager) + + expect(terminal.focus).toHaveBeenCalledOnce() + expect(document.activeElement).toBe(textarea) + }) + + it('does not treat the workspace sidebar as a focus-owning overlay', () => { + const { manager, terminal } = createLayoutFixture() + const sidebar = mountOverlay('listbox') + sidebar.setAttribute('data-worktree-sidebar', '') + + fitAndFocusPanes(manager) + + expect(terminal.focus).toHaveBeenCalledOnce() + }) +}) diff --git a/src/renderer/src/components/terminal-pane/terminal-link-action-request.ts b/src/renderer/src/components/terminal-pane/terminal-link-action-request.ts index 108a670939b..833d05ffa4d 100644 --- a/src/renderer/src/components/terminal-pane/terminal-link-action-request.ts +++ b/src/renderer/src/components/terminal-pane/terminal-link-action-request.ts @@ -1,24 +1,17 @@ import type { TerminalLinkPointerGesture } from './terminal-link-pointer-gesture' import { isTerminalLinkActionActivation } from './terminal-link-activation' +import { + closeLinkActionRequest, + type LinkAction, + type LinkActionKind, + type LinkActionRequest +} from '@/components/link-actions/link-action-request' -export type TerminalLinkActionKind = 'url' | 'file' | 'workspace' | 'terminal' | 'task' +export type TerminalLinkActionKind = LinkActionKind -export type TerminalLinkAction = { - external?: boolean - label: string - run: () => void | Promise -} +export type TerminalLinkAction = LinkAction -export type TerminalLinkActionRequest = { - paneId: number - anchorX: number - anchorY: number - destination: string - kind: TerminalLinkActionKind - primary: TerminalLinkAction - alternate?: TerminalLinkAction - focusTerminal: () => void -} +export type TerminalLinkActionRequest = LinkActionRequest & { paneId: number } export type TerminalLinkActionRequester = (request: TerminalLinkActionRequest) => void @@ -34,7 +27,7 @@ export function closeTerminalLinkActionRequest( current: TerminalLinkActionRequest | null, dismissed?: TerminalLinkActionRequest ): TerminalLinkActionRequest | null { - return dismissed && current !== dismissed ? current : null + return closeLinkActionRequest(current, dismissed) } type LinkActionDetails = Pick< @@ -65,7 +58,7 @@ export function requestTerminalLinkAction( paneId: context.paneId, anchorX: event.clientX, anchorY: event.clientY, - focusTerminal: context.focusTerminal + restoreFocus: context.focusTerminal }) return true } diff --git a/src/renderer/src/components/terminal-pane/terminal-link-open-hints.test.ts b/src/renderer/src/components/terminal-pane/terminal-link-open-hints.test.ts index 466f95b6a67..f19573ff786 100644 --- a/src/renderer/src/components/terminal-pane/terminal-link-open-hints.test.ts +++ b/src/renderer/src/components/terminal-pane/terminal-link-open-hints.test.ts @@ -1,9 +1,5 @@ import { afterEach, describe, expect, it, vi } from 'vitest' -import { - getTerminalUrlOpenHint, - terminalHttpLinkActionDestinationsFor, - terminalUrlOpenHintOptionsFor -} from './terminal-link-open-hints' +import { getTerminalUrlOpenHint, terminalUrlOpenHintOptionsFor } from './terminal-link-open-hints' function stubPlatform(isMac: boolean): void { vi.stubGlobal('navigator', { userAgent: isMac ? 'Mac OS X' : 'Windows NT 10.0' }) @@ -169,37 +165,3 @@ describe('terminalUrlOpenHintOptionsFor', () => { expect(options.modifierInverts).toBe(true) }) }) - -describe('terminalHttpLinkActionDestinationsFor', () => { - it.each([ - ['local', { kind: 'local' } as const, false], - ['capable runtime', { kind: 'runtime', runtimeEnvironmentId: 'env-1' } as const, true], - ['eligible SSH', { kind: 'ssh', connectionId: 'ssh-1' } as const, true] - ])( - 'offers both destinations for a %s owner and follows the preference', - (_label, owner, canOpen) => { - expect( - terminalHttpLinkActionDestinationsFor({ openLinksInApp: true }, owner, canOpen) - ).toEqual({ - primary: 'orca', - alternate: 'system' - }) - expect( - terminalHttpLinkActionDestinationsFor({ openLinksInApp: false }, owner, canOpen) - ).toEqual({ - primary: 'system', - alternate: 'orca' - }) - } - ) - - it.each([ - ['incapable runtime', { kind: 'runtime', runtimeEnvironmentId: 'env-1' } as const], - ['ineligible SSH', { kind: 'ssh', connectionId: 'ssh-1' } as const], - ['unknown owner', { kind: 'unknown' } as const] - ])('offers only the system browser for an %s', (_label, owner) => { - expect(terminalHttpLinkActionDestinationsFor({ openLinksInApp: true }, owner, false)).toEqual({ - primary: 'system' - }) - }) -}) diff --git a/src/renderer/src/components/terminal-pane/terminal-link-open-hints.ts b/src/renderer/src/components/terminal-pane/terminal-link-open-hints.ts index d0646a9db1d..1af33f4f402 100644 --- a/src/renderer/src/components/terminal-pane/terminal-link-open-hints.ts +++ b/src/renderer/src/components/terminal-pane/terminal-link-open-hints.ts @@ -1,5 +1,5 @@ +import { canSourceOwnerOpenInOrca } from '@/lib/http-link-destinations' import type { HttpLinkSourceOwner } from '@/lib/http-link-routing' -import type { TerminalHttpLinkActionDestinations } from './terminal-url-link-hit-testing' export function isMacPlatform(): boolean { return navigator.userAgent.includes('Mac') @@ -37,30 +37,6 @@ export type TerminalUrlOpenHintOptions = { showActions?: boolean } -function canSourceOwnerOpenInOrca( - sourceOwner: HttpLinkSourceOwner, - canOpenOwnedBrowser: boolean -): boolean { - return ( - sourceOwner.kind === 'local' || - ((sourceOwner.kind === 'runtime' || sourceOwner.kind === 'ssh') && canOpenOwnedBrowser) - ) -} - -export function terminalHttpLinkActionDestinationsFor( - settings: { openLinksInApp?: boolean } | null | undefined, - sourceOwner: HttpLinkSourceOwner, - canOpenOwnedBrowser: boolean -): TerminalHttpLinkActionDestinations { - const canOpenInOrca = canSourceOwnerOpenInOrca(sourceOwner, canOpenOwnedBrowser) - if (!canOpenInOrca) { - return { primary: 'system' } - } - return settings?.openLinksInApp === true - ? { primary: 'orca', alternate: 'system' } - : { primary: 'system', alternate: 'orca' } -} - // Why: remote owners advertise Orca only when their existing browser route is eligible. export function terminalUrlOpenHintOptionsFor( settings: diff --git a/src/renderer/src/components/terminal-pane/terminal-pane-hook-order-parity.test.ts b/src/renderer/src/components/terminal-pane/terminal-pane-hook-order-parity.test.ts index a6b96168047..80150075f7a 100644 --- a/src/renderer/src/components/terminal-pane/terminal-pane-hook-order-parity.test.ts +++ b/src/renderer/src/components/terminal-pane/terminal-pane-hook-order-parity.test.ts @@ -14,8 +14,10 @@ const TERMINAL_PANE_HOOK_SOURCE_PATTERN = // Restoring the terminal/chat switcher added four `useCallback`s -- three in // chat-state (can-toggle, toggle-for-leaf, toggle-active) and the context-menu // toggle in projection (208 hooks, still 8 useMemo). +// Then chat-state's orchestration dispatch-status subscription went with the +// paused notice that read it (207 hooks, still 8 useMemo). const PRE_REFACTOR_HOOK_ORDER_SHA256 = - '983ad067c9feca82c5435eb1b865674344489c368ec2007dc7bb40c81aef037c' + '2bbb42427b61e3722114ac37c407230cb7daffbf9b899090c7a635f15731ccad' const sourceFiles = readdirSync(__dirname) .filter((name) => TERMINAL_PANE_HOOK_SOURCE_PATTERN.test(name)) @@ -80,7 +82,7 @@ function readFlattenedHookOrder(): string[] { describe('TerminalPane refactor hook parity', () => { it('preserves the recursively flattened render hook order', () => { const hooks = readFlattenedHookOrder() - expect(hooks).toHaveLength(208) + expect(hooks).toHaveLength(207) expect(hooks.filter((hook) => hook === 'useMemo')).toHaveLength(8) expect(createHash('sha256').update(hooks.join('\n')).digest('hex')).toBe( PRE_REFACTOR_HOOK_ORDER_SHA256 diff --git a/src/renderer/src/components/terminal-pane/terminal-pane-manager-options.ts b/src/renderer/src/components/terminal-pane/terminal-pane-manager-options.ts index 128b715a3c5..d42e41dc76c 100644 --- a/src/renderer/src/components/terminal-pane/terminal-pane-manager-options.ts +++ b/src/renderer/src/components/terminal-pane/terminal-pane-manager-options.ts @@ -1,4 +1,5 @@ import type { IDisposable } from '@xterm/xterm' +import { FLOATING_TERMINAL_WORKTREE_ID } from '../../../../shared/constants' import type { PaneManagerOptions } from '@/lib/pane-manager/pane-manager' import { useAppStore } from '@/store' import { resolveTerminalLigaturesEnabled } from '../../../../shared/terminal-ligatures' @@ -176,6 +177,8 @@ export function createTerminalPaneManagerOptions( formatLinkTooltip: (paneId, url, hint) => formatTerminalUrlTooltip(url, hint, context.getHttpLinkSourceOwnerForPane(paneId)), initialRenderingSuspended: !isVisibleRef.current, + // Reopening the floating panel must rebuild silently corrupted glyph atlases. + retainHiddenWebgl: worktreeId !== FLOATING_TERMINAL_WORKTREE_ID, terminalGpuAcceleration: settingsRef.current?.terminalGpuAcceleration ?? 'auto', debugLabel: `tab:${tabId}/wt:${worktreeId}` } diff --git a/src/renderer/src/components/terminal-pane/terminal-pane-mount-preparation.ts b/src/renderer/src/components/terminal-pane/terminal-pane-mount-preparation.ts index d64ab5dfd7b..56c9825212d 100644 --- a/src/renderer/src/components/terminal-pane/terminal-pane-mount-preparation.ts +++ b/src/renderer/src/components/terminal-pane/terminal-pane-mount-preparation.ts @@ -1,6 +1,7 @@ import type { PaneManager } from '@/lib/pane-manager/pane-manager' import { useAppStore } from '@/store' import { getConnectionId } from '@/lib/connection-context' +import { httpLinkActionDestinationsFor } from '@/lib/http-link-destinations' import { canOpenWorkspaceBrowserTabOnRuntime, canOpenWorkspaceBrowserTabOnSsh @@ -8,7 +9,6 @@ import { import { resolvePaneWslDistro } from './terminal-pane-wsl-distro' import { resolveTerminalHttpLinkSourceOwner } from './terminal-http-link-source-owner' import { - terminalHttpLinkActionDestinationsFor, getTerminalFileOpenHint, getTerminalUrlOpenHint, terminalUrlOpenHintOptionsFor @@ -124,7 +124,7 @@ export function prepareTerminalPaneMount( ) } const getHttpLinkActionDestinations = (paneId: number): TerminalHttpLinkActionDestinations => - terminalHttpLinkActionDestinationsFor( + httpLinkActionDestinationsFor( deps.settingsRef.current, getHttpLinkSourceOwnerForPane(paneId), canOpenOwnedBrowserForPane(paneId) diff --git a/src/renderer/src/components/terminal-pane/terminal-pane-store-subscription-budget.test.tsx b/src/renderer/src/components/terminal-pane/terminal-pane-store-subscription-budget.test.tsx index 2418a5da6d3..a43afded2f8 100644 --- a/src/renderer/src/components/terminal-pane/terminal-pane-store-subscription-budget.test.tsx +++ b/src/renderer/src/components/terminal-pane/terminal-pane-store-subscription-budget.test.tsx @@ -4,8 +4,9 @@ * synchronously on every publication, so the per-pane subscription count is a * direct multiplier on agent-status burn (docs/reference/renderer-agent-status-performance.md). * - * On `main` one mounted pane opened 49 listeners; 32 of them selected values that - * can never change — 28 store actions and 4 duplicate reads of one unified tab. + * On `main` one mounted pane opened 49 listeners; 33 of them earned nothing — 28 + * store actions and 4 duplicate reads of one unified tab, all of which can never + * change, plus a dispatch-status read left behind by the notice that consumed it. */ import { act, createRef, type ReactNode } from 'react' import { createRoot, type Root } from 'react-dom/client' @@ -25,7 +26,7 @@ import { * visit per store publication for every retained tab in the app — read the doc * above before you do. */ -const TERMINAL_PANE_LISTENER_BUDGET = 17 +const TERMINAL_PANE_LISTENER_BUDGET = 16 /** What the same mount cost before the stable-action and unified-tab folds. */ const PRE_FOLD_LISTENERS_PER_PANE = 49 @@ -103,8 +104,8 @@ describe('TerminalPane store subscription budget', () => { expect(perPane).toBe(TERMINAL_PANE_LISTENER_BUDGET) expect(perPane).toBeLessThan(PRE_FOLD_LISTENERS_PER_PANE) - // 28 stable actions plus four duplicate unified-tab reads. - expect(PRE_FOLD_LISTENERS_PER_PANE - perPane).toBe(TERMINAL_PANE_STORE_ACTION_KEYS.length + 4) + // 28 stable actions, four duplicate unified-tab reads, one dead dispatch-status read. + expect(PRE_FOLD_LISTENERS_PER_PANE - perPane).toBe(TERMINAL_PANE_STORE_ACTION_KEYS.length + 5) unmount() expect(listenerCount()).toBe(baseline) diff --git a/src/renderer/src/components/terminal-pane/terminal-url-link-hit-testing.ts b/src/renderer/src/components/terminal-pane/terminal-url-link-hit-testing.ts index df223a4bbd1..0139df29e16 100644 --- a/src/renderer/src/components/terminal-pane/terminal-url-link-hit-testing.ts +++ b/src/renderer/src/components/terminal-pane/terminal-url-link-hit-testing.ts @@ -1,5 +1,5 @@ import type { IBufferLine, IBufferRange, IDisposable, Terminal } from '@xterm/xterm' -import { openHttpLink, type HttpLinkSourceOwner } from '@/lib/http-link-routing' +import type { HttpLinkSourceOwner } from '@/lib/http-link-routing' import { buildEdgeWrappedHttpLogicalLineCandidates } from './edge-wrapped-terminal-http-links' import { buildHardWrappedHttpLogicalLineCandidates } from './hard-wrapped-terminal-http-links' import { dedupeLogicalLines } from './terminal-file-link-hit-testing' @@ -12,7 +12,13 @@ import { getTerminalBufferPositionForMouseEvent } from './terminal-mouse-buffer- import { extractTerminalHttpLinks } from './terminal-http-url-extraction' import { buildWrappedLogicalLine, rangeForParsedFileLink } from './wrapped-terminal-link-ranges' import { isTerminalLinkifierHoverActive } from '@/lib/pane-manager/terminal-linkifier-hover-reset' -import { translate } from '@/i18n/i18n' +import { + buildHttpLinkActions, + openRoutedHttpLink, + type HttpLinkActionDestinations, + type HttpLinkDestination, + type HttpLinkRoutingPreferenceRequester +} from '@/lib/http-link-destinations' import { isTerminalOwnedLinkGesture } from './terminal-link-activation' import { requestTerminalLinkAction, @@ -46,16 +52,11 @@ export type HttpLinkClickFallbackBinding = IDisposable & { ptyMouseSuppression: TerminalLinkPtyMouseSuppression } -export type TerminalHttpLinkDestination = 'orca' | 'system' +export type TerminalHttpLinkDestination = HttpLinkDestination -export type TerminalHttpLinkActionDestinations = { - primary: TerminalHttpLinkDestination - alternate?: TerminalHttpLinkDestination -} +export type TerminalHttpLinkActionDestinations = HttpLinkActionDestinations -export type TerminalLinkRoutingPreferenceRequester = ( - url: string -) => boolean | Promise | null | undefined +export type TerminalLinkRoutingPreferenceRequester = HttpLinkRoutingPreferenceRequester function isDesktopHttpLinkFallbackActivation(event: MouseEvent): boolean { if (event.defaultPrevented || event.button !== 0) { @@ -74,7 +75,7 @@ export function handleTerminalHttpLink( const forceDestination = event?.shiftKey ? (deps.actionDestinations?.alternate ?? deps.actionDestinations?.primary) : deps.actionDestinations?.primary - openTerminalHttpLink(url, { + openRoutedHttpLink(url, { ...deps, modifierHeld: forceDestination ? false : Boolean(event?.shiftKey), forceDestination @@ -82,51 +83,12 @@ export function handleTerminalHttpLink( return true } - const actionDestinations = deps.actionDestinations - const primaryDestination = actionDestinations?.primary - const labelForDestination = (destination: TerminalHttpLinkDestination): string => - destination === 'orca' - ? translate( - 'auto.components.terminal.pane.TerminalLinkActionPopover.orcaBrowser', - 'Orca Browser' - ) - : translate( - 'auto.components.terminal.pane.TerminalLinkActionPopover.systemBrowser', - 'System Browser' - ) - return requestTerminalLinkAction(event, deps.linkActionContext, { destination: deps.actionDestination ?? url, kind: 'url', - primary: { - external: primaryDestination === 'system', - label: primaryDestination - ? labelForDestination(primaryDestination) - : translate( - 'auto.components.terminal.pane.TerminalLinkActionPopover.openLink', - 'Open link' - ), - run: () => - openTerminalHttpLink(url, { - ...deps, - modifierHeld: false, - forceDestination: primaryDestination - }) - }, - ...(actionDestinations?.alternate - ? { - alternate: { - external: actionDestinations.alternate === 'system', - label: labelForDestination(actionDestinations.alternate), - run: () => - openTerminalHttpLink(url, { - ...deps, - modifierHeld: false, - forceDestination: actionDestinations.alternate - }) - } - } - : {}) + ...buildHttpLinkActions(deps.actionDestinations, (destination) => + openRoutedHttpLink(url, { ...deps, modifierHeld: false, forceDestination: destination }) + ) }) } @@ -224,7 +186,7 @@ export function openHttpLinkAtBufferPosition( if (!url) { return false } - openTerminalHttpLink(url, deps) + openRoutedHttpLink(url, deps) return true } @@ -271,58 +233,3 @@ function rangeContainsBufferPosition( const current = position.y * terminalColumns + position.x return lower <= current && current <= upper } - -export function openTerminalHttpLink(url: string, deps: UrlLinkHitTestDeps): void { - // Why: pane ownership beats the global active runtime for both local and remote routes. - const sourceOwner = deps.sourceOwner ?? { kind: 'local' } - if (deps.forceDestination) { - openHttpLink(url, { - allowRemoteInApp: true, - worktreeId: deps.worktreeId, - forceInApp: deps.forceDestination === 'orca', - forceSystemBrowser: deps.forceDestination === 'system', - sourceOwner - }) - return - } - if (deps.modifierHeld) { - // Why: the modifier states a destination outright, so it also skips the - // one-time routing prompt; openHttpLink resolves which destination it means. - openHttpLink(url, { - allowRemoteInApp: true, - worktreeId: deps.worktreeId, - modifierHeld: true, - sourceOwner - }) - return - } - - // Why: remote panes use the persisted routing preference and never prompt the viewing client. - const preferenceDecision = - sourceOwner.kind === 'local' ? deps.requestOpenLinksInAppPreference?.(url) : null - if (preferenceDecision === null || preferenceDecision === undefined) { - openHttpLink(url, { allowRemoteInApp: true, worktreeId: deps.worktreeId, sourceOwner }) - return - } - - // Why: the first terminal link click may need an async preference dialog. - // Suppress the browser's default link handling first, then route after the - // persisted choice is available. - void Promise.resolve(preferenceDecision) - .then((openInOrca) => { - openHttpLink(url, { - allowRemoteInApp: true, - worktreeId: deps.worktreeId, - forceSystemBrowser: !openInOrca, - sourceOwner - }) - }) - .catch(() => { - openHttpLink(url, { - allowRemoteInApp: true, - worktreeId: deps.worktreeId, - forceSystemBrowser: true, - sourceOwner - }) - }) -} diff --git a/src/renderer/src/components/terminal-pane/use-notification-dispatch.test.ts b/src/renderer/src/components/terminal-pane/use-notification-dispatch.test.ts index 1267b986827..529edda1466 100644 --- a/src/renderer/src/components/terminal-pane/use-notification-dispatch.test.ts +++ b/src/renderer/src/components/terminal-pane/use-notification-dispatch.test.ts @@ -341,7 +341,7 @@ describe('dispatchTerminalNotification', () => { expect(mockState.markAgentCompletionPaneUnread).toHaveBeenCalledWith(paneKey) }) - it('can mark terminal attention without dispatching an OS notification', () => { + it('offers attention-only completion to main for independent mobile delivery', () => { dispatchTerminalNotification('wt-primary', { source: 'agent-task-complete', terminalTitle: 'codex', @@ -352,7 +352,7 @@ describe('dispatchTerminalNotification', () => { expect(mockState.markWorktreeUnread).toHaveBeenCalledWith('wt-primary') expect(mockState.markTerminalTabUnread).toHaveBeenCalledWith('tab-1') expect(mockState.markTerminalPaneUnread).toHaveBeenCalledWith(paneKey) - expect(window.api.notifications.dispatch).not.toHaveBeenCalled() + expect(window.api.notifications.dispatch).toHaveBeenCalled() }) it('does not mark the visible focused pane unread', () => { diff --git a/src/renderer/src/components/terminal-pane/use-notification-dispatch.ts b/src/renderer/src/components/terminal-pane/use-notification-dispatch.ts index 483ba4c792e..2a13fe01f5b 100644 --- a/src/renderer/src/components/terminal-pane/use-notification-dispatch.ts +++ b/src/renderer/src/components/terminal-pane/use-notification-dispatch.ts @@ -174,9 +174,7 @@ export function dispatchTerminalNotification( } } - if (event.suppressOsNotification) { - return - } + // Desktop settings are applied in main after independent mobile delivery. // Why: prefer worktree.repoId over string-parsing the worktreeId. The // `${repoId}::${path}` format is an implementation detail of id diff --git a/src/renderer/src/components/terminal-pane/use-parked-terminal-watcher-synchronization.ts b/src/renderer/src/components/terminal-pane/use-parked-terminal-watcher-synchronization.ts index d582177b3a3..7c7e501f68c 100644 --- a/src/renderer/src/components/terminal-pane/use-parked-terminal-watcher-synchronization.ts +++ b/src/renderer/src/components/terminal-pane/use-parked-terminal-watcher-synchronization.ts @@ -122,11 +122,16 @@ function selectWatcherReconciliationStoreInputs( state: AppState, terminalTabs: readonly TerminalTab[] ): WatcherReconciliationStoreInputs { - return terminalTabs.flatMap((tab) => [ - state.ptyIdsByTabId[tab.id] ?? EMPTY_PTY_IDS, - state.terminalLayoutsByTabId[tab.id] ?? null, - Object.keys(state.runtimePaneTitlesByTabId[tab.id] ?? {}).join(',') - ]) + return terminalTabs.flatMap((tab) => { + // Why not `?? {}`: this runs per tab on every store write, and the fallback + // object was allocated only to be thrown away — Object.keys({}).join(',') is ''. + const paneTitles = state.runtimePaneTitlesByTabId[tab.id] + return [ + state.ptyIdsByTabId[tab.id] ?? EMPTY_PTY_IDS, + state.terminalLayoutsByTabId[tab.id] ?? null, + paneTitles ? Object.keys(paneTitles).join(',') : '' + ] + }) } export function useParkedTerminalWatcherSynchronization(args: { diff --git a/src/renderer/src/components/terminal-pane/use-terminal-pane-chat-state.ts b/src/renderer/src/components/terminal-pane/use-terminal-pane-chat-state.ts index 2fd75f3f093..24a4e580532 100644 --- a/src/renderer/src/components/terminal-pane/use-terminal-pane-chat-state.ts +++ b/src/renderer/src/components/terminal-pane/use-terminal-pane-chat-state.ts @@ -5,7 +5,6 @@ import { useAppStore } from '../../store' import { getCachedTerminalTabForWorktree } from './terminal-tab-lookup' import { selectTerminalTabAgentTypesByLeaf } from './terminal-tab-agent-type-index' import { collectLeafIdsInOrder, EMPTY_LAYOUT } from './layout-serialization' -import { makePaneKey } from '../../../../shared/stable-pane-id' import { sanitizeTerminalLayoutPaneTitles } from '@/lib/terminal-pane-title-sanitization' import { resolveNativeChatLeafTitleAgent } from './native-chat-leaf-title-agent' import { useTerminalPaneStoreActions } from './use-terminal-pane-store-actions' @@ -56,11 +55,6 @@ export function useTerminalPaneChatState(controller: TerminalPaneTitleController ) const nativeChatEnabled = useAppStore((store) => store.settings?.experimentalNativeChat === true) const effectiveChatViewMode = nativeChatEnabled && isChatViewMode - const chatPaneDispatchStatus = useAppStore((store) => - chatLeafId - ? store.agentStatusByPaneKey[makePaneKey(tabId, chatLeafId)]?.orchestration?.dispatchStatus - : undefined - ) const runtimePaneTitlesByPaneId = useAppStore( useShallow((store) => store.runtimePaneTitlesByTabId[tabId] ?? {}) ) @@ -280,7 +274,6 @@ export function useTerminalPaneChatState(controller: TerminalPaneTitleController structuredSessionId, nativeChatEnabled, effectiveChatViewMode, - chatPaneDispatchStatus, unifiedTabLabel, runtimePaneTitlesByPaneId, tabAgentTypeByLeaf, diff --git a/src/renderer/src/components/terminal-pane/use-terminal-pane-projection.ts b/src/renderer/src/components/terminal-pane/use-terminal-pane-projection.ts index 7ef23834873..36f59ebfb2e 100644 --- a/src/renderer/src/components/terminal-pane/use-terminal-pane-projection.ts +++ b/src/renderer/src/components/terminal-pane/use-terminal-pane-projection.ts @@ -25,7 +25,6 @@ export function useTerminalPaneProjection(controller: TerminalPaneMobileControll applyNativeChatLeafRoute, canToggleChatForLeaf, chatLeafId, - chatPaneDispatchStatus, contextMenu, contextMenuLeafId, effectiveChatViewMode, @@ -214,7 +213,6 @@ export function useTerminalPaneProjection(controller: TerminalPaneMobileControll structuredChatAgent, structuredChatTarget, structuredSessionId, - chatPaneDispatchStatus, chatPaneOwnsTabWideLaunchDraft, activePaneIsChatLeaf, resolveAgentForLeaf, diff --git a/src/renderer/src/components/terminal/initial-terminal-structured-launch.test.tsx b/src/renderer/src/components/terminal/initial-terminal-structured-launch.test.tsx new file mode 100644 index 00000000000..7a76434ce2f --- /dev/null +++ b/src/renderer/src/components/terminal/initial-terminal-structured-launch.test.tsx @@ -0,0 +1,83 @@ +// @vitest-environment happy-dom +import { act } from 'react' +import { createRoot, type Root } from 'react-dom/client' +import { afterEach, describe, expect, it, vi } from 'vitest' +import { useTerminalWatcherEffects } from '../use-terminal-watcher-effects' +import type { TerminalColdActivationController } from '../terminal-cold-activation' + +const mocks = vi.hoisted(() => ({ + gate: vi.fn(), + launchStatus: vi.fn((_worktreeId: string, _provider: string): string => 'idle'), + createTab: vi.fn() +})) +vi.mock('@/store', () => ({ + useAppStore: Object.assign(() => 'none', { + getState: () => ({ activeWorktreeId: 'wt-1' }) + }) +})) +vi.mock('@/lib/worktree-agent-activation-gate', () => ({ + gateWorktreeAgentActivation: mocks.gate +})) +vi.mock('@/lib/structured-agent-session-launch', () => ({ + getStructuredAgentLaunchStatus: mocks.launchStatus +})) +vi.mock('@/lib/resume-sleeping-agent-session', () => ({ + resumeSleepingAgentSessionsForWorktree: vi.fn() +})) +vi.mock('@/lib/workspace-terminal-host-authority', () => ({ + createWorkspaceTerminalHostAuthoritySelector: () => () => 'none' +})) +vi.mock('../terminal-pane/terminal-parked-tab-watchers', () => ({ + pruneParkedTerminalWatchers: vi.fn(), + terminalWatcherLiveWorkspaceIds: () => new Set(), + syncParkedTerminalTabWatchersForWorkspaces: vi.fn(), + disposeAllParkedTerminalWatchers: vi.fn() +})) + +;(globalThis as { IS_REACT_ACT_ENVIRONMENT?: boolean }).IS_REACT_ACT_ENVIRONMENT = true +let root: Root | undefined +afterEach(async () => { + await act(async () => root?.unmount()) + vi.clearAllMocks() +}) + +function Watcher(): null { + useTerminalWatcherEffects({ + activeWorktreeId: 'wt-1', + workspaceSessionReady: true, + terminalStartupRestorationReady: true, + workspaceSurfaceIds: [], + tabsByWorktree: {}, + createTab: mocks.createTab, + reconcileWorktreeTabModel: () => ({ renderableTabCount: 0 }) + } as unknown as TerminalColdActivationController) + return null +} + +describe('passive terminal seeding during native chat creation', () => { + it.each([ + ['claude', 'pending', 0], + ['codex', 'pending', 0], + ['claude', 'unknown', 0], + ['codex', 'unknown', 0], + ['claude', 'idle', 1] + ] as const)('handles %s launch status %s', async (agent, status, expectedTabs) => { + let finishGate!: (outcome: 'empty') => void + mocks.gate.mockReturnValue( + new Promise((resolve) => { + finishGate = resolve + }) + ) + mocks.launchStatus.mockReturnValue('idle') + root = createRoot(document.createElement('div')) + await act(async () => root?.render()) + + // A create starts after the inventory probe but before its empty result returns. + mocks.launchStatus.mockImplementation((_worktreeId, provider) => + provider === agent ? status : 'idle' + ) + await act(async () => finishGate('empty')) + + expect(mocks.createTab).toHaveBeenCalledTimes(expectedTabs) + }) +}) diff --git a/src/renderer/src/components/use-terminal-watcher-effects.ts b/src/renderer/src/components/use-terminal-watcher-effects.ts index 9e82b821c31..3c6a89fb323 100644 --- a/src/renderer/src/components/use-terminal-watcher-effects.ts +++ b/src/renderer/src/components/use-terminal-watcher-effects.ts @@ -13,6 +13,8 @@ import { useAppStore } from '@/store' import { gateWorktreeAgentActivation } from '@/lib/worktree-agent-activation-gate' import { resumeSleepingAgentSessionsForWorktree } from '@/lib/resume-sleeping-agent-session' import { createWorkspaceTerminalHostAuthoritySelector } from '@/lib/workspace-terminal-host-authority' +import { getStructuredAgentLaunchStatus } from '@/lib/structured-agent-session-launch' +import { AGENT_SESSION_PROVIDER_HANDLE_PROVIDERS } from '../../../shared/agent-session-provider-handle' import type { TerminalColdActivationController } from './terminal-cold-activation' export function useTerminalWatcherEffects(controller: TerminalColdActivationController): void { @@ -159,6 +161,14 @@ export function useTerminalWatcherEffects(controller: TerminalColdActivationCont ) { return } + // A pending or unanswered chat create owns the surface even before its tab is published. + if ( + AGENT_SESSION_PROVIDER_HANDLE_PROVIDERS.some( + (agent) => getStructuredAgentLaunchStatus(activeWorktreeId, agent) !== 'idle' + ) + ) { + return + } // Why: the activation gate reconciles durable/live agent state first; only an actually empty, never-visited workspace receives a default shell. const { renderableTabCount } = reconcileWorktreeTabModel(activeWorktreeId) if (shouldAutoCreateInitialTerminal(renderableTabCount, activeWorktreeHasTerminalState)) { diff --git a/src/renderer/src/components/use-worktree-jump-palette-browser-ownership.test.ts b/src/renderer/src/components/use-worktree-jump-palette-browser-ownership.test.ts new file mode 100644 index 00000000000..c8aea613f03 --- /dev/null +++ b/src/renderer/src/components/use-worktree-jump-palette-browser-ownership.test.ts @@ -0,0 +1,82 @@ +// @vitest-environment happy-dom + +import { cleanup, renderHook } from '@testing-library/react' +import { afterEach, expect, it } from 'vitest' +import { useAppStore } from '@/store' +import type { BrowserPage, BrowserWorkspace } from '../../../shared/browser-workspace-types' +import type { Tab } from '../../../shared/tab-types' +import { makeUnifiedTab, makeWorktree } from './worktree-jump-palette-test-fixtures' +import { useWorktreeJumpPaletteOpenTabs } from './use-worktree-jump-palette-open-tabs' + +afterEach(cleanup) + +it('keeps same-id browser results on their owner with recency, and follows ownership changes', () => { + const worktrees = [ + makeWorktree('same-id', 'Local workspace', { hostId: 'local' }), + makeWorktree('same-id', 'Remote workspace', { hostId: 'runtime:paired' }) + ] + const page: BrowserPage = { + id: 'page', + workspaceId: 'browser', + worktreeId: 'same-id', + url: 'https://example.test/docs', + title: 'Browser proof', + loading: false, + faviconUrl: null, + canGoBack: false, + canGoForward: false, + loadError: null, + createdAt: 1 + } + const workspace: BrowserWorkspace = { + ...page, + id: 'browser', + activePageId: page.id, + pageIds: [page.id] + } + const tab: Tab = { + ...makeUnifiedTab('tab', 'same-id', 'browser', 'Browser proof'), + contentType: 'browser', + executionHostId: 'runtime:paired', + lastFocusedAt: 5_000 + } + type PaletteInput = Parameters[0] + const input: Partial = { + ...useAppStore.getInitialState(), + // The store holds {key, result}; the hook takes the unwrapped result. + workspacePortScan: null, + paletteStatusInputsActive: true, + allWorktrees: worktrees, + browserSortedWorktrees: worktrees, + repoMap: new Map(), + repoByHostIdentity: new Map(), + worktreeOrder: new Map(), + worktreeMatches: [], + hasQuery: true, + deferredQuery: 'Browser proof', + browserTabsByWorktree: { 'same-id': [workspace] }, + browserPagesByWorkspace: { browser: [page] }, + unifiedTabsByWorktree: { 'same-id': [tab] } + } + const { result, rerender } = renderHook( + (props: Partial) => useWorktreeJumpPaletteOpenTabs(props as PaletteInput), + { initialProps: input } + ) + // lastActiveAt rides the same map: without it every browser row sorts as never-focused. + const owners = () => + result.current.browserItems.map(({ result: entry }) => [ + entry.pageId, + entry.executionHostId, + entry.lastActiveAt + ]) + + expect(owners()).toEqual([['page', 'runtime:paired', 5_000]]) + + rerender({ + ...input, + unifiedTabsByWorktree: { + 'same-id': [{ ...tab, executionHostId: 'local' }] + } + }) + expect(owners()).toEqual([['page', 'local', 5_000]]) +}) diff --git a/src/renderer/src/components/use-worktree-jump-palette-controller.ts b/src/renderer/src/components/use-worktree-jump-palette-controller.ts index 0c167dfb5f8..97d3391e011 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-controller.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-controller.ts @@ -13,6 +13,9 @@ import { useWorktreeJumpPaletteSelectionActions } from './use-worktree-jump-pale import { useWorktreeJumpPaletteCreateAction } from './use-worktree-jump-palette-create-action' import { useWorktreeJumpPaletteTaskUrl } from './use-worktree-jump-palette-task-url' import { useWorkspaceEmojiShortcodeInput } from '@/components/workspace-emoji/useWorkspaceEmojiShortcodeInput' +import { usePaletteSearchEvaluationContext } from '@/hooks/use-palette-search-evaluation-context' +import type { WorktreePaletteRequestGuard } from '@/lib/worktree-palette-create-action' +import { useMemo } from 'react' export function useWorktreeJumpPaletteController({ visible, @@ -25,6 +28,34 @@ export function useWorktreeJumpPaletteController({ }) { const storeState = useWorktreeJumpPaletteStoreState({ visible, lingering }) const localState = useWorktreeJumpPaletteLocalState({ createLookupGuard, visible }) + const paletteEvaluationSnapshot = useMemo( + () => ({ + query: localState.paletteSearchQuery, + agentStatus: storeState.agentStatusByPaneKey, + worktrees: storeState.allWorktrees, + browserPages: storeState.browserPagesByWorkspace, + browserWorkspaces: storeState.browserTabsByWorktree, + openFiles: storeState.openFiles, + retainedAgents: storeState.retainedAgentsByPaneKey, + sleepingAgents: storeState.sleepingAgentSessionsByPaneKey, + unifiedTabs: storeState.unifiedTabsByWorktree, + visible + }), + [ + localState.paletteSearchQuery, + storeState.agentStatusByPaneKey, + storeState.allWorktrees, + storeState.browserPagesByWorkspace, + storeState.browserTabsByWorktree, + storeState.openFiles, + storeState.retainedAgentsByPaneKey, + storeState.sleepingAgentSessionsByPaneKey, + storeState.unifiedTabsByWorktree, + visible + ] + ) + const paletteSearchContext = usePaletteSearchEvaluationContext(paletteEvaluationSnapshot) + const evaluation = { paletteSearchContext } const taskUrl = useWorktreeJumpPaletteTaskUrl({ visible, createWorktreeName: localState.createWorktreeName, @@ -35,13 +66,15 @@ export function useWorktreeJumpPaletteController({ const worktrees = useWorktreeJumpPaletteWorktrees({ ...storeState, ...localState, - ...filter + ...filter, + ...evaluation }) const openTabs = useWorktreeJumpPaletteOpenTabs({ ...storeState, ...localState, ...filter, - ...worktrees + ...worktrees, + ...evaluation }) const recentTabs = useWorktreeJumpPaletteRecentTabs({ ...storeState, @@ -130,10 +163,10 @@ export function useWorktreeJumpPaletteController({ ...listEntries, ...selectionLifecycle, ...selectionActions, + paletteNowMs: worktrees.hasQuery ? paletteSearchContext.nowMs : storeState.paletteNowMs, emojiInput, ...createAction } } export type WorktreeJumpPaletteController = ReturnType -import type { WorktreePaletteRequestGuard } from '@/lib/worktree-palette-create-action' diff --git a/src/renderer/src/components/use-worktree-jump-palette-filter.ts b/src/renderer/src/components/use-worktree-jump-palette-filter.ts index ec238cff920..c9eefe0ed7a 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-filter.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-filter.ts @@ -1,11 +1,10 @@ -import { useEffect, useMemo } from 'react' +import { useMemo } from 'react' import { buildSidebarHostOptions } from '@/components/sidebar/sidebar-host-options' import { getProjectGroupExecutionHostIdForRows } from '@/components/sidebar/worktree-list/listing/host-filtering' import { buildPaletteFilterModel } from '@/components/cmd-j/palette-filter-options' import { buildPaletteFilterPredicate, - isPaletteFilterActive, - reconcilePaletteFilter + isPaletteFilterActive } from '@/components/cmd-j/palette-filter' import { getRepoHostIdentity } from '@/store/slices/repo-host-identity' import { getHostDisplayLabelOverrides } from '../../../shared/host-setting-overrides' @@ -26,7 +25,7 @@ type WorktreeJumpPaletteFilterInput = Pick< | 'projectHostSetups' | 'projectGroups' > & - Pick + Pick export function useWorktreeJumpPaletteFilter({ repos, @@ -39,8 +38,7 @@ export function useWorktreeJumpPaletteFilter({ projects, projectHostSetups, projectGroups, - rawFilter, - setRawFilter + filter }: WorktreeJumpPaletteFilterInput) { const repoMap = useMemo(() => new Map(repos.map((repo) => [repo.id, repo])), [repos]) const repoByHostIdentity = useMemo( @@ -83,14 +81,6 @@ export function useWorktreeJumpPaletteFilter({ }), [allWorktrees, defaultHostId, hostOptions, projectHostSetups, projects, repos] ) - const filter = useMemo( - () => reconcilePaletteFilter(rawFilter, filterModel), - [rawFilter, filterModel] - ) - useEffect(() => { - setRawFilter((current) => reconcilePaletteFilter(current, filterModel)) - // oxlint-disable-next-line react-hooks/exhaustive-deps -- local-state setter identity is stable across extraction. - }, [filterModel]) const filterActive = isPaletteFilterActive(filter) const hostFilterActive = filter.hostIds.length > 0 const filterPredicate = useMemo( @@ -115,7 +105,6 @@ export function useWorktreeJumpPaletteFilter({ canCreateWorktree, defaultHostId, filterModel, - filter, filterActive, hostFilterActive, filterPredicate, diff --git a/src/renderer/src/components/use-worktree-jump-palette-local-state.ts b/src/renderer/src/components/use-worktree-jump-palette-local-state.ts index 5bfd941c8fd..5115634e053 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-local-state.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-local-state.ts @@ -1,6 +1,11 @@ import { useDeferredValue, useMemo, useRef, useState } from 'react' +import { useShallow } from 'zustand/react/shallow' import type { WorktreePaletteRequestGuard } from '@/lib/worktree-palette-create-action' -import { EMPTY_PALETTE_FILTER, type PaletteFilterState } from '@/components/cmd-j/palette-filter' +import { + buildPaletteFilterFromSidebarScope, + type PaletteFilterState +} from '@/components/cmd-j/palette-filter' +import { useAppStore } from '@/store' import { parseCmdJTaskSourceUrl } from '@/lib/worktree-palette-task-url-match' import { getWorktreePaletteCreateActionState } from '@/lib/worktree-palette-create-action' import type { CmdJActiveGroupSnapshot } from '@/components/cmd-j/quick-action-context' @@ -14,6 +19,13 @@ export function useWorktreeJumpPaletteLocalState({ createLookupGuard: WorktreePaletteRequestGuard visible: boolean }) { + const sidebarScope = useAppStore( + useShallow((state) => ({ + filterRepoIds: state.filterRepoIds, + visibleWorkspaceHostIds: state.visibleWorkspaceHostIds, + workspaceHostScope: state.workspaceHostScope + })) + ) const [query, setQuery] = useState('') const deferredQuery = useDeferredValue(query) const liveQueryRef = useRef(query) @@ -34,7 +46,9 @@ export function useWorktreeJumpPaletteLocalState({ // Create is armed by an explicit keyboard/pointer move, except for task URLs. const selectionMovedByUserRef = useRef(false) const digitShortcutItemsRef = useRef([]) - const [rawFilter, setRawFilter] = useState(EMPTY_PALETTE_FILTER) + const [filter, setFilter] = useState(() => + buildPaletteFilterFromSidebarScope(sidebarScope) + ) const [dialogElement, setDialogElement] = useState(null) const previousWorktreeIdRef = useRef(null) const previousActiveTabTypeRef = useRef('terminal') @@ -51,15 +65,18 @@ export function useWorktreeJumpPaletteLocalState({ const preserveCreateLookupOnCloseRef = useRef(false) const [expandedSectionCaps, setExpandedSectionCaps] = useState>({}) - // Reset expansion after a new query or a fresh open without adding an extra effect render. + // Reset expansion and seed each open before the palette paints. const [previousQuery, setPreviousQuery] = useState(query) const [previousVisible, setPreviousVisible] = useState(visible) - if (previousQuery !== query || previousVisible !== visible) { + const visibilityChanged = previousVisible !== visible + if (previousQuery !== query || visibilityChanged) { setPreviousQuery(query) setPreviousVisible(visible) setExpandedSectionCaps({}) + if (visibilityChanged && visible) { + setFilter(buildPaletteFilterFromSidebarScope(sidebarScope)) + } } - return { query, setQuery, @@ -75,8 +92,8 @@ export function useWorktreeJumpPaletteLocalState({ autoSelectedItemIdRef, selectionMovedByUserRef, digitShortcutItemsRef, - rawFilter, - setRawFilter, + filter, + setFilter, dialogElement, setDialogElement, previousWorktreeIdRef, @@ -94,8 +111,7 @@ export function useWorktreeJumpPaletteLocalState({ createLookupGuard, preserveCreateLookupOnCloseRef, expandedSectionCaps, - setExpandedSectionCaps, - previousVisible + setExpandedSectionCaps } } diff --git a/src/renderer/src/components/use-worktree-jump-palette-open-tabs.ts b/src/renderer/src/components/use-worktree-jump-palette-open-tabs.ts index 4175906bac5..d6c62711670 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-open-tabs.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-open-tabs.ts @@ -12,7 +12,7 @@ import { type SearchableWorkspaceTab } from '@/lib/workspace-tab-palette-search' import { comparePaletteRankedItems } from '@/lib/cmd-j-section-leadership' -import { getWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' +import { getPaletteWorktreeIdentity } from '@/lib/palette-repo-resolution' import type { BrowserPaletteItem, OpenTabPaletteItem, @@ -24,6 +24,16 @@ import type { WorktreeJumpPaletteFilter } from './use-worktree-jump-palette-filt import type { WorktreeJumpPaletteLocalState } from './use-worktree-jump-palette-local-state' import type { WorktreeJumpPaletteStoreState } from './use-worktree-jump-palette-store-state' import type { WorktreeJumpPaletteWorktrees } from './use-worktree-jump-palette-worktrees' +import { + encodePaletteIdentity, + type PaletteSearchContext +} from '@/lib/palette-match/palette-ranking' +import { + buildBrowserPaletteItems, + buildOpenTabPaletteItems, + buildSimulatorPaletteItems, + buildWorkspaceTabPaletteItems +} from './worktree-jump-palette-open-tab-items' const EMPTY_BROWSER_PAGE_ENTRIES: SearchableBrowserPage[] = [] const EMPTY_SIMULATOR_TAB_ENTRIES: SearchableSimulatorTab[] = [] @@ -32,7 +42,9 @@ const EMPTY_WORKSPACE_TAB_ENTRIES: SearchableWorkspaceTab[] = [] type WorktreeJumpPaletteOpenTabsInput = WorktreeJumpPaletteStoreState & WorktreeJumpPaletteWorktrees & Pick & - Pick + Pick & { + paletteSearchContext: PaletteSearchContext + } export function useWorktreeJumpPaletteOpenTabs({ paletteStatusInputsActive, @@ -64,6 +76,7 @@ export function useWorktreeJumpPaletteOpenTabs({ terminalLayoutsByTabId, paneForegroundAgentByPaneKey, deferredQuery, + paletteSearchContext, hasQuery, worktreeMatches, resolveWorktree @@ -83,7 +96,8 @@ export function useWorktreeJumpPaletteOpenTabs({ activeBrowserTabId, activeWorktreeId, activeWorkspaceExecutionHostId, - activeTabType + activeTabType, + unifiedTabsByWorktree }) }, [ paletteStatusInputsActive, @@ -97,11 +111,15 @@ export function useWorktreeJumpPaletteOpenTabs({ browserSortedWorktrees, repoByHostIdentity, repoMap, + unifiedTabsByWorktree, worktreeOrder ]) const browserMatches = useMemo( - () => searchBrowserPages(browserPageEntries, deferredQuery.trim()), - [browserPageEntries, deferredQuery] + () => + searchBrowserPages(browserPageEntries, deferredQuery.trim(), { + context: paletteSearchContext + }), + [browserPageEntries, deferredQuery, paletteSearchContext] ) const simulatorTabEntries = useMemo(() => { if (!paletteStatusInputsActive) { @@ -135,8 +153,11 @@ export function useWorktreeJumpPaletteOpenTabs({ worktreeOrder ]) const simulatorMatches = useMemo( - () => searchSimulatorTabs(simulatorTabEntries, deferredQuery.trim()), - [simulatorTabEntries, deferredQuery] + () => + searchSimulatorTabs(simulatorTabEntries, deferredQuery.trim(), { + context: paletteSearchContext + }), + [simulatorTabEntries, deferredQuery, paletteSearchContext] ) const workspaceTabEntries = useMemo(() => { if (!paletteStatusInputsActive) { @@ -196,15 +217,23 @@ export function useWorktreeJumpPaletteOpenTabs({ worktreeOrder ]) const workspaceTabMatches = useMemo( - () => searchWorkspaceTabs(workspaceTabEntries, deferredQuery.trim()), - [workspaceTabEntries, deferredQuery] + () => + searchWorkspaceTabs(workspaceTabEntries, deferredQuery.trim(), { + context: paletteSearchContext + }), + [workspaceTabEntries, deferredQuery, paletteSearchContext] ) const worktreeItems = useMemo(() => { const items = worktreeMatches .map((match) => { const worktree = resolveWorktree(match.worktreeId, match.worktreeHostId) return worktree - ? { id: `worktree:${worktree.id}`, type: 'worktree' as const, match, worktree } + ? { + id: encodePaletteIdentity(['worktree', getPaletteWorktreeIdentity(worktree)]), + type: 'worktree' as const, + match, + worktree + } : null }) .filter((item): item is WorktreePaletteItem => item !== null) @@ -212,69 +241,41 @@ export function useWorktreeJumpPaletteOpenTabs({ return items } const orderByIdentity = new Map( - items.map((item, index) => [getWorktreeHostIdentity(item.worktree), index]) + items.map((item, index) => [getPaletteWorktreeIdentity(item.worktree), index]) ) return items.sort((left, right) => comparePaletteRankedItems( { rank: left.match.rank, - order: orderByIdentity.get(getWorktreeHostIdentity(left.worktree)) ?? 0, - id: left.id + order: orderByIdentity.get(getPaletteWorktreeIdentity(left.worktree)) ?? 0, + identity: left.id, + activity: left.match.activity }, { rank: right.match.rank, - order: orderByIdentity.get(getWorktreeHostIdentity(right.worktree)) ?? 0, - id: right.id + order: orderByIdentity.get(getPaletteWorktreeIdentity(right.worktree)) ?? 0, + identity: right.id, + activity: right.match.activity } ) ) }, [hasQuery, resolveWorktree, worktreeMatches]) const browserItems = useMemo( - () => - browserMatches.map((result) => ({ - id: `browser-page:${result.pageId}`, - type: 'browser-page' as const, - result - })), + () => buildBrowserPaletteItems(browserMatches), [browserMatches] ) const simulatorItems = useMemo( - () => - simulatorMatches.map((result) => ({ - id: `simulator-tab:${result.tabId}`, - type: 'simulator-tab' as const, - result - })), + () => buildSimulatorPaletteItems(simulatorMatches), [simulatorMatches] ) const workspaceTabItems = useMemo( - () => - workspaceTabMatches.map((result) => ({ - id: `workspace-tab:${result.tabId}`, - type: 'workspace-tab' as const, - result - })), + () => buildWorkspaceTabPaletteItems(workspaceTabMatches), [workspaceTabMatches] ) - const openTabItems = useMemo(() => { - const items = [...browserItems, ...simulatorItems, ...workspaceTabItems] - return items.sort((left, right) => - comparePaletteRankedItems( - { - rank: left.result.rank, - order: left.result.score, - id: left.id, - lastActiveAt: left.result.lastActiveAt ?? undefined - }, - { - rank: right.result.rank, - order: right.result.score, - id: right.id, - lastActiveAt: right.result.lastActiveAt ?? undefined - } - ) - ) - }, [browserItems, simulatorItems, workspaceTabItems]) + const openTabItems = useMemo( + () => buildOpenTabPaletteItems({ browserItems, simulatorItems, workspaceTabItems }), + [browserItems, simulatorItems, workspaceTabItems] + ) return { browserPageEntries, diff --git a/src/renderer/src/components/use-worktree-jump-palette-recent-tabs.ts b/src/renderer/src/components/use-worktree-jump-palette-recent-tabs.ts index 62e82f0328d..129114ab4d1 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-recent-tabs.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-recent-tabs.ts @@ -4,7 +4,6 @@ import { type TabPaneInputSources } from '@/components/sidebar/smart-attention' import { - buildFocusedGroupTabRecency, orderRecentWorkspaceTabs, type RecentWorkspaceTabRow } from '@/lib/recent-workspace-tab-rows' @@ -15,49 +14,32 @@ import { type PaletteItem } from './worktree-jump-palette-model' import { shouldIncludeOpenTabInRecentSection } from './worktree-jump-palette-recent-inclusion' -import type { WorktreeJumpPaletteFilter } from './use-worktree-jump-palette-filter' import type { WorktreeJumpPaletteLocalState } from './use-worktree-jump-palette-local-state' import type { WorktreeJumpPaletteOpenTabs } from './use-worktree-jump-palette-open-tabs' import type { WorktreeJumpPaletteStoreState } from './use-worktree-jump-palette-store-state' import type { WorktreeJumpPaletteWorktrees } from './use-worktree-jump-palette-worktrees' +import { + getPaletteWorktreeExecutionHostId, + getPaletteWorktreeIdentity +} from '@/lib/palette-repo-resolution' +import { encodePaletteIdentity } from '@/lib/palette-match/palette-ranking' type WorktreeJumpPaletteRecentTabsInput = WorktreeJumpPaletteStoreState & WorktreeJumpPaletteOpenTabs & Pick & - Pick & - Pick + Pick< + WorktreeJumpPaletteLocalState, + 'query' | 'filter' | 'autoSelectedItemIdRef' | 'setSelectedItemId' + > -function getRecentTabOccurrenceBase(item: OpenTabRecentRow['item']): string { - if (item.type === 'browser-page') { - const result = item.result - return JSON.stringify([ - item.type, - item.id, - result.executionHostId ?? '', - result.worktreeId, - result.workspaceId, - result.pageId - ]) - } - if (item.type === 'simulator-tab') { - const result = item.result - return JSON.stringify([ - item.type, - item.id, - result.executionHostId ?? '', - result.worktreeId, - result.tabId - ]) - } - const result = item.result - return JSON.stringify([ - item.type, - item.id, - result.executionHostId ?? '', - result.worktreeId, - result.tabId, - result.entityId - ]) +type RecentTabOrderSnapshot = { + order: readonly string[] + attentionReady: boolean +} + +const EMPTY_RECENT_TAB_SNAPSHOT: RecentTabOrderSnapshot = { + order: EMPTY_RECENT_TAB_ORDER, + attentionReady: false } export function useWorktreeJumpPaletteRecentTabs({ @@ -68,36 +50,54 @@ export function useWorktreeJumpPaletteRecentTabs({ runtimePaneTitlesByTabId, terminalLayoutsByTabId, openTabItems, + workspaceTabEntries, + simulatorTabEntries, + browserPageEntries, resolveWorktree, unreadTerminalTabs, unreadAgentCompletionPanes, visible, hasQuery, query, - filterActive, - lastVisitedAtByWorktreeId, - activeGroupIdByWorktree, - groupsByWorktree, + filter, autoSelectedItemIdRef, setSelectedItemId }: WorktreeJumpPaletteRecentTabsInput) { + const tabFocusTimes = useMemo(() => { + const times = new Map() + for (const entry of [...workspaceTabEntries, ...simulatorTabEntries]) { + times.set( + encodePaletteIdentity(['tab', getPaletteWorktreeIdentity(entry.worktree), entry.tab.id]), + entry.tab.lastFocusedAt + ) + } + for (const entry of browserPageEntries) { + times.set( + encodePaletteIdentity(['page', getPaletteWorktreeIdentity(entry.worktree), entry.page.id]), + entry.lastFocusedAt + ) + } + return times + }, [workspaceTabEntries, simulatorTabEntries, browserPageEntries]) const occurrenceIds = useMemo(() => { const counts = new Map() return openTabItems.map((item) => { - const base = getRecentTabOccurrenceBase(item) + const base = item.id const ordinal = counts.get(base) ?? 0 counts.set(base, ordinal + 1) return `recent-tab:${base}:${ordinal}` }) }, [openTabItems]) - const terminalTabsById = useMemo(() => { - const byId = new Map() - for (const tabs of Object.values(tabsByWorktree)) { + const terminalTabsByWorktree = useMemo(() => { + const byWorktree = new Map>() + for (const [worktreeId, tabs] of Object.entries(tabsByWorktree)) { + const byId = new Map() for (const tab of tabs ?? []) { - byId.set(tab.id, tab) + byId.set(tab.id, byId.has(tab.id) ? null : tab) } + byWorktree.set(worktreeId, byId) } - return byId + return byWorktree }, [tabsByWorktree]) const recentTabPaneSources = useMemo( () => ({ @@ -133,18 +133,25 @@ export function useWorktreeJumpPaletteRecentTabs({ id: item.id, occurrenceId, worktreeId: worktree.id, - worktreeHostId: worktree.hostId, + worktreeHostId: getPaletteWorktreeExecutionHostId(worktree), + lastFocusedAt: tabFocusTimes.get( + encodePaletteIdentity([ + item.type === 'browser-page' ? 'page' : 'tab', + getPaletteWorktreeIdentity(worktree), + item.type === 'browser-page' ? item.result.pageId : item.result.tabId + ]) + ), unifiedTabId: item.type === 'browser-page' ? null : item.result.tabId, terminalTab: item.type === 'workspace-tab' && item.result.contentType === 'terminal' - ? (terminalTabsById.get(item.result.entityId) ?? null) + ? (terminalTabsByWorktree.get(worktree.id)?.get(item.result.entityId) ?? null) : null, worktreeLastActivityAt: worktree.lastActivityAt } }) } return entries - }, [occurrenceIds, openTabItems, resolveWorktree, terminalTabsById]) + }, [occurrenceIds, openTabItems, resolveWorktree, terminalTabsByWorktree, tabFocusTimes]) const recentTabRowByItem = useMemo( () => new Map(openTabRecentRows.map(({ item, row }) => [item, row])), [openTabRecentRows] @@ -169,9 +176,10 @@ export function useWorktreeJumpPaletteRecentTabs({ } return rows }, [openTabRecentRows, recentTabPaneSources, unreadAgentCompletionPanes, unreadTerminalTabs]) - const [recentTabOrder, setRecentTabOrder] = useState(EMPTY_RECENT_TAB_ORDER) - const recentTabOrderCapturedRef = useRef(false) - const recentTabOrderAttentionReadyRef = useRef(false) + const [recentTabSnapshot, setRecentTabSnapshot] = useState(EMPTY_RECENT_TAB_SNAPSHOT) + // Why: recent rows are already narrowed by the filter, so a filter change mid-open must + // re-capture — a frozen order would otherwise hide rows a cleared chip brought back. + const capturedFilterRef = useRef(filter) const recentOrderAttentionIncomplete = useMemo(() => { for (const { item, worktree, row } of openTabRecentRows) { if ( @@ -188,49 +196,42 @@ export function useWorktreeJumpPaletteRecentTabs({ }, [openTabRecentRows]) useLayoutEffect(() => { if (!visible) { - recentTabOrderCapturedRef.current = false - recentTabOrderAttentionReadyRef.current = false autoSelectedItemIdRef.current = null - setRecentTabOrder(EMPTY_RECENT_TAB_ORDER) + setRecentTabSnapshot(EMPTY_RECENT_TAB_SNAPSHOT) return } - if (hasQuery || query.length > 0 || filterActive) { + if (hasQuery || query.length > 0) { return } + const filterChanged = capturedFilterRef.current !== filter + if (filterChanged) { + capturedFilterRef.current = filter + } if ( - recentTabOrderCapturedRef.current && - (recentTabOrderAttentionReadyRef.current || recentOrderAttentionIncomplete) + !filterChanged && + recentTabSnapshot.order.length > 0 && + (recentTabSnapshot.attentionReady || recentOrderAttentionIncomplete) ) { return } const order = orderRecentWorkspaceTabs({ - rows: recentTabRows, - paneSources: recentTabPaneSources, - now: Date.now(), - lastVisitedAtByWorktreeId, - focusedGroupTabRecency: buildFocusedGroupTabRecency(activeGroupIdByWorktree, groupsByWorktree) + rows: recentTabRows }) if (order.length === 0) { - recentTabOrderCapturedRef.current = false - recentTabOrderAttentionReadyRef.current = false - setRecentTabOrder(EMPTY_RECENT_TAB_ORDER) + setRecentTabSnapshot(EMPTY_RECENT_TAB_SNAPSHOT) return } - recentTabOrderCapturedRef.current = true - recentTabOrderAttentionReadyRef.current = !recentOrderAttentionIncomplete - setRecentTabOrder(order) + setRecentTabSnapshot({ order, attentionReady: !recentOrderAttentionIncomplete }) setSelectedItemId((current) => current === '' || current === autoSelectedItemIdRef.current ? '' : current ) // oxlint-disable-next-line react-hooks/exhaustive-deps -- controller refs and setters preserve their original stable identities. }, [ - activeGroupIdByWorktree, - filterActive, - groupsByWorktree, + filter, hasQuery, - lastVisitedAtByWorktreeId, query.length, recentOrderAttentionIncomplete, + recentTabSnapshot, recentTabPaneSources, recentTabRows, visible @@ -239,8 +240,10 @@ export function useWorktreeJumpPaletteRecentTabs({ const itemByOccurrenceId = new Map( openTabRecentRows.map(({ occurrenceId, item }) => [occurrenceId, item]) ) - return recentTabOrder.flatMap((occurrenceId) => itemByOccurrenceId.get(occurrenceId) ?? []) - }, [openTabRecentRows, recentTabOrder]) + return recentTabSnapshot.order.flatMap( + (occurrenceId) => itemByOccurrenceId.get(occurrenceId) ?? [] + ) + }, [openTabRecentRows, recentTabSnapshot.order]) return { recentTabPaneSources, recentTabRowByItem, recentTabItems, openTabRecentRows } } diff --git a/src/renderer/src/components/use-worktree-jump-palette-sections.ts b/src/renderer/src/components/use-worktree-jump-palette-sections.ts index abb91d1ac2f..42555b2dda0 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-sections.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-sections.ts @@ -38,7 +38,11 @@ type WorktreeJumpPaletteSectionsInput = WorktreeJumpPaletteOpenTabs & Pick & Pick< WorktreeJumpPaletteLocalState, - 'createWorktreeName' | 'showCreateAction' | 'expandedSectionCaps' | 'setExpandedSectionCaps' + | 'createWorktreeName' + | 'showCreateAction' + | 'expandedSectionCaps' + | 'setExpandedSectionCaps' + | 'selectedItemId' > export function useWorktreeJumpPaletteSections({ @@ -51,19 +55,20 @@ export function useWorktreeJumpPaletteSections({ createWorktreeName, showCreateAction, expandedSectionCaps, - setExpandedSectionCaps + setExpandedSectionCaps, + selectedItemId }: WorktreeJumpPaletteSectionsInput) { const openTabsLeadSections = useMemo(() => { if (!hasQuery) { return true } return shouldOpenTabsLeadPaletteSections({ - bestWorktreeQualityRank: worktreeItems[0] - ? bestPaletteQualityRank([worktreeItems[0].match.qualityClass]) - : NO_PALETTE_QUALITY_RANK, - bestOpenTabQualityRank: openTabItems[0] - ? bestPaletteQualityRank([openTabItems[0].result.qualityClass]) - : NO_PALETTE_QUALITY_RANK + bestWorktreeQualityRank: bestPaletteQualityRank( + worktreeItems.map((item) => item.match.qualityClass) + ), + bestOpenTabQualityRank: bestPaletteQualityRank( + openTabItems.map((item) => item.result.qualityClass) + ) }) }, [hasQuery, openTabItems, worktreeItems]) @@ -72,11 +77,11 @@ export function useWorktreeJumpPaletteSections({ return false } const bestEntityQualityRank = Math.min( - worktreeItems[0] - ? bestPaletteQualityRank([worktreeItems[0].match.qualityClass]) + worktreeItems.length + ? bestPaletteQualityRank(worktreeItems.map((item) => item.match.qualityClass)) : NO_PALETTE_QUALITY_RANK, - openTabItems[0] - ? bestPaletteQualityRank([openTabItems[0].result.qualityClass]) + openTabItems.length + ? bestPaletteQualityRank(openTabItems.map((item) => item.result.qualityClass)) : NO_PALETTE_QUALITY_RANK ) return shouldIntentSectionLeadPaletteSections({ @@ -98,15 +103,35 @@ export function useWorktreeJumpPaletteSections({ [setExpandedSectionCaps] ) + const openTabsCap = PALETTE_SECTION_RENDER_CAP + (expandedSectionCaps['open-tabs'] ?? 0) + const typedWorktreeCap = PALETTE_SECTION_RENDER_CAP + (expandedSectionCaps.worktrees ?? 0) + const openTabIndexById = useMemo( + () => new Map(openTabItems.map((item, index) => [item.id, index])), + [openTabItems] + ) + const worktreeIndexById = useMemo( + () => new Map(worktreeItems.map((item, index) => [item.id, index])), + [worktreeItems] + ) + const retainedOpenTabId = + hasQuery && (openTabIndexById.get(selectedItemId ?? '') ?? -1) >= openTabsCap + ? selectedItemId + : null + const retainedWorktreeId = + hasQuery && (worktreeIndexById.get(selectedItemId ?? '') ?? -1) >= typedWorktreeCap + ? selectedItemId + : null + const paletteSections = useMemo(() => { - const openTabsCap = PALETTE_SECTION_RENDER_CAP + (expandedSectionCaps['open-tabs'] ?? 0) + const retainOpenTab = (item: { id: string }): boolean => item.id === retainedOpenTabId + const retainWorktree = (item: { id: string }): boolean => item.id === retainedWorktreeId // Why: "See more" drops the above-the-fold trim outright instead of stepping 20 at a time, so one // click reveals the whole recent history the shared render cap allows. const recentTabsCap = expandedSectionCaps['open-tabs'] ? openTabsCap : EMPTY_QUERY_RECENT_TAB_CAP const openTabs = hasQuery - ? capPaletteSection(openTabItems, openTabsCap) + ? capPaletteSection(openTabItems, openTabsCap, retainOpenTab) : capPaletteSection(recentTabItems, recentTabsCap) const baseWorktreeCap = hasQuery ? Infinity @@ -115,10 +140,10 @@ export function useWorktreeJumpPaletteSections({ Math.max(1, EMPTY_QUERY_ROW_BUDGET - openTabs.visible.length) ) const worktreeCap = hasQuery - ? PALETTE_SECTION_RENDER_CAP + (expandedSectionCaps.worktrees ?? 0) + ? typedWorktreeCap : baseWorktreeCap + (expandedSectionCaps.worktrees ?? 0) const worktrees = hasQuery - ? capPaletteSection(worktreeItems, worktreeCap) + ? capPaletteSection(worktreeItems, worktreeCap, retainWorktree) : { visible: worktreeItems.slice(0, worktreeCap), overflowCount: Math.max(0, worktreeItems.length - worktreeCap) @@ -141,7 +166,9 @@ export function useWorktreeJumpPaletteSections({ TYPED_QUERY_LEADING_PREVIEW + (expandedSectionCaps[openTabsLeadSections ? 'open-tabs' : 'worktrees'] ?? 0), leadingHardCap: openTabsLeadSections ? openTabsCap : worktreeCap, - trailingHardCap: openTabsLeadSections ? worktreeCap : openTabsCap + trailingHardCap: openTabsLeadSections ? worktreeCap : openTabsCap, + leadingRetain: openTabsLeadSections ? retainOpenTab : retainWorktree, + trailingRetain: openTabsLeadSections ? retainWorktree : retainOpenTab }) : null return { @@ -162,8 +189,12 @@ export function useWorktreeJumpPaletteSections({ middleItems, openTabItems, openTabsLeadSections, + openTabsCap, projectTargetItems, recentTabItems, + retainedOpenTabId, + retainedWorktreeId, + typedWorktreeCap, worktreeItems ]) diff --git a/src/renderer/src/components/use-worktree-jump-palette-selection-actions.ts b/src/renderer/src/components/use-worktree-jump-palette-selection-actions.ts index 77294e63931..e0da6c158fe 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-selection-actions.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-selection-actions.ts @@ -14,6 +14,7 @@ import { getUnavailableQuickActionMessage } from './use-worktree-jump-palette-qu import type { SettingsNavTarget } from '@/lib/settings-navigation-types' import type { Worktree } from '../../../shared/worktree/types' import { useAppStore } from '@/store' +import { getPaletteWorktreeExecutionHostId } from '@/lib/palette-repo-resolution' import { translate } from '@/i18n/i18n' import type { PaletteItem } from './worktree-jump-palette-model' import type { WorktreeJumpPaletteLocalState } from './use-worktree-jump-palette-local-state' @@ -56,7 +57,8 @@ export function useWorktreeJumpPaletteSelectionActions({ }: WorktreeJumpPaletteSelectionActionsInput) { const handleSelectWorktree = useCallback( (worktree: Worktree) => { - const current = useAppStore.getState().getKnownWorktreeById(worktree.id, worktree.hostId) + const executionHostId = getPaletteWorktreeExecutionHostId(worktree) + const current = useAppStore.getState().getKnownWorktreeById(worktree.id, executionHostId) if (!current) { toast.error( translate('auto.components.WorktreeJumpPalette.2c38630a01', 'Workspace no longer exists') @@ -65,7 +67,7 @@ export function useWorktreeJumpPaletteSelectionActions({ } const activation = activateAndRevealWorktree( worktree.id, - worktree.hostId ? { executionHostId: worktree.hostId } : {} + executionHostId ? { executionHostId } : {} ) recordFeatureInteraction('cmd-j-workspace-open') skipRestoreFocusRef.current = true diff --git a/src/renderer/src/components/use-worktree-jump-palette-selection-lifecycle.ts b/src/renderer/src/components/use-worktree-jump-palette-selection-lifecycle.ts index 8906add43e2..5fd2159f8ca 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-selection-lifecycle.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-selection-lifecycle.ts @@ -4,7 +4,6 @@ import { queueBrowserFocusRequest } from '@/components/browser-pane/host-guest/browser-focus' import { captureCmdJActiveGroupSnapshot } from '@/components/cmd-j/quick-action-context' -import { EMPTY_PALETTE_FILTER } from '@/components/cmd-j/palette-filter' import { resolvePaletteFocusRestoreTarget } from '@/components/cmd-j/palette-focus-restore-target' import { CREATE_WORKTREE_ITEM_ID, @@ -56,7 +55,7 @@ export function useWorktreeJumpPaletteSelectionLifecycle({ latestQueryRef, setQuery, setSelectedItemId, - setRawFilter, + setExpandedSectionCaps, selectionMovedByUserRef, taskSourceUrl, listRef, @@ -81,10 +80,8 @@ export function useWorktreeJumpPaletteSelectionLifecycle({ if (visible && !wasVisibleRef.current) { recordFeatureInteraction('cmd-j') createLookupGuard.invalidate() - activeGroupSnapshotRef.current = captureCmdJActiveGroupSnapshot( - useAppStore.getState(), - activeWorktreeId - ) + const appState = useAppStore.getState() + activeGroupSnapshotRef.current = captureCmdJActiveGroupSnapshot(appState, activeWorktreeId) previousWorktreeIdRef.current = activeWorktreeId previousActiveTabTypeRef.current = activeTabType previousBrowserPageIdRef.current = @@ -107,11 +104,12 @@ export function useWorktreeJumpPaletteSelectionLifecycle({ latestQueryRef.current = '' setQuery('') setSelectedItemId('') + setExpandedSectionCaps({}) selectionMovedByUserRef.current = false - setRawFilter(EMPTY_PALETTE_FILTER) listRef.current?.scrollTo(0, 0) } if (!visible && wasVisibleRef.current) { + setExpandedSectionCaps({}) if (preserveCreateLookupOnCloseRef.current) { preserveCreateLookupOnCloseRef.current = false } else { @@ -171,6 +169,7 @@ export function useWorktreeJumpPaletteSelectionLifecycle({ latestQueryRef.current = nextQuery setQuery(nextQuery) setSelectedItemId('') + setExpandedSectionCaps({}) listRef.current?.scrollTo(0, 0) }, // oxlint-disable-next-line react-hooks/exhaustive-deps -- controller refs and setters preserve their original stable identities. diff --git a/src/renderer/src/components/use-worktree-jump-palette-store-state.ts b/src/renderer/src/components/use-worktree-jump-palette-store-state.ts index c10778abaa4..21b0bc67f42 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-store-state.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-store-state.ts @@ -2,9 +2,9 @@ import { useMemo } from 'react' import { useTranslation } from 'react-i18next' import { useShallow } from 'zustand/react/shallow' import { useAppStore } from '@/store' -import { useAllWorktrees } from '@/store/selectors' import { usePluginCommands } from '@/store/plugin-panels' import { useSettingsNavigationMetadata } from '@/hooks/useSettingsNavigationMetadata' +import { dedupePaletteWorktrees } from '@/lib/palette-repo-resolution' import { selectPaletteIndexStatusSnapshot, selectPaletteStatusInputs @@ -29,7 +29,10 @@ export function useWorktreeJumpPaletteStoreState({ const recordFeatureInteraction = useAppStore((state) => state.recordFeatureInteraction) const revealSidebarRow = useAppStore((state) => state.revealSidebarRow) const worktreesByRepo = useAppStore((state) => state.worktreesByRepo) - const allWorktrees = useAllWorktrees() + const allWorktrees = useMemo( + () => dedupePaletteWorktrees(Object.values(worktreesByRepo).flat()), + [worktreesByRepo] + ) const repos = useAppStore((state) => state.repos) const projectGroups = useAppStore((state) => state.projectGroups) const projects = useAppStore((state) => state.projects) diff --git a/src/renderer/src/components/use-worktree-jump-palette-worktrees.ts b/src/renderer/src/components/use-worktree-jump-palette-worktrees.ts index 5f5392cabcb..e07dcd3a668 100644 --- a/src/renderer/src/components/use-worktree-jump-palette-worktrees.ts +++ b/src/renderer/src/components/use-worktree-jump-palette-worktrees.ts @@ -27,16 +27,20 @@ import type { WorktreeJumpPaletteLocalState } from './use-worktree-jump-palette- import type { WorktreeJumpPaletteStoreState } from './use-worktree-jump-palette-store-state' import { buildWorktreeJumpPaletteDocumentIndex } from './worktree-jump-palette-document-index' import { buildWorktreeJumpPaletteWorktreeMaps } from './worktree-jump-palette-worktree-maps' +import type { PaletteSearchContext } from '@/lib/palette-match/palette-ranking' type WorktreeJumpPaletteWorktreesInput = WorktreeJumpPaletteStoreState & Pick< WorktreeJumpPaletteFilter, 'filterPredicate' | 'repoMap' | 'repoByHostIdentity' | 'hostOptions' | 'hostFilterActive' > & - Pick + Pick & { + paletteSearchContext: PaletteSearchContext + } export function useWorktreeJumpPaletteWorktrees({ paletteSearchQuery, + paletteSearchContext, repos, worktreesByRepo, agentStatusByPaneKey, @@ -266,11 +270,13 @@ export function useWorktreeJumpPaletteWorktrees({ documents: worktreeDocuments, repoMap, repoMapByHostIdentity: repoByHostIdentity, - checksReviewByWorktree + checksReviewByWorktree, + context: paletteSearchContext }), [ checksReviewByWorktree, paletteSearchQuery, + paletteSearchContext, repoByHostIdentity, repoMap, sortedWorktrees, diff --git a/src/renderer/src/components/worktree-jump-palette-browser-simulator-rows.tsx b/src/renderer/src/components/worktree-jump-palette-browser-simulator-rows.tsx index 3c88e3daf0c..95220a25f2e 100644 --- a/src/renderer/src/components/worktree-jump-palette-browser-simulator-rows.tsx +++ b/src/renderer/src/components/worktree-jump-palette-browser-simulator-rows.tsx @@ -66,6 +66,7 @@ export function WorktreeJumpPaletteSimulatorRow({ titleRanges={result.titleRanges} secondaryText={result.secondaryText} secondaryRanges={result.secondaryRanges} + secondaryMatches={result.secondaryMatches} worktreeName={result.worktreeName} worktreeRanges={result.worktreeRanges} sessionAge={simulatorSessionAge} @@ -87,6 +88,11 @@ export function WorktreeJumpPaletteSimulatorRow({ } /> + {result.typeAliasMatches.length ? ( + + {result.typeAliasMatches.map((match) => match.text).join(', ')} + + ) : null}
@@ -158,6 +164,7 @@ export function WorktreeJumpPaletteBrowserRow({ titleRanges={result.titleRanges} secondaryText={result.secondaryText} secondaryRanges={result.secondaryRanges} + secondaryMatches={result.secondaryMatches} worktreeName={result.worktreeName} worktreeRanges={result.worktreeRanges} sessionAge={browserSessionAge} diff --git a/src/renderer/src/components/worktree-jump-palette-document-index.ts b/src/renderer/src/components/worktree-jump-palette-document-index.ts index 922aea88c70..1dc70ec5efc 100644 --- a/src/renderer/src/components/worktree-jump-palette-document-index.ts +++ b/src/renderer/src/components/worktree-jump-palette-document-index.ts @@ -2,14 +2,16 @@ import { getPaletteHostBadge } from '@/components/cmd-j/palette-host-badge' import type { SidebarHostOption } from '@/components/sidebar/sidebar-host-options' import { getWorkspacePortsByWorktreeId } from '@/lib/workspace-port-groups' import { buildWorktreePaletteDocuments } from '@/lib/worktree-palette-document' -import { resolvePaletteRepoForWorktree } from '@/lib/palette-repo-resolution' +import { + getPaletteWorktreeIdentity, + resolvePaletteRepoForWorktree +} from '@/lib/palette-repo-resolution' import type { PaletteDocument } from '@/lib/palette-match/palette-document' import type { AppState } from '@/store/types' import type { Repo } from '../../../shared/repo-types' import type { Worktree } from '../../../shared/worktree/types' import type { WorkspacePortScanResult } from '../../../shared/workspace-ports' import type { HostedReviewInfo } from '../../../shared/hosted-review' -import { getWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' export function buildWorktreeJumpPaletteDocumentIndex({ worktrees, @@ -37,7 +39,7 @@ export function buildWorktreeJumpPaletteDocumentIndex({ const repo = resolvePaletteRepoForWorktree(worktree, repoMap, repoByHostIdentity) const badge = getPaletteHostBadge(repo, hostOptions, hostFilterActive) if (badge) { - hostLabelByWorktreeId.set(getWorktreeHostIdentity(worktree), badge.label) + hostLabelByWorktreeId.set(getPaletteWorktreeIdentity(worktree), badge.label) } } return buildWorktreePaletteDocuments( diff --git a/src/renderer/src/components/worktree-jump-palette-interleaved-sections.test.tsx b/src/renderer/src/components/worktree-jump-palette-interleaved-sections.test.tsx index f52bf725a5a..2d45ee26bcd 100644 --- a/src/renderer/src/components/worktree-jump-palette-interleaved-sections.test.tsx +++ b/src/renderer/src/components/worktree-jump-palette-interleaved-sections.test.tsx @@ -10,6 +10,7 @@ import type { TerminalTab } from '../../../shared/terminal-tab-types' import type { Worktree } from '../../../shared/worktree/types' import { useAppStore } from '@/store' import type { AppState } from '@/store/types' +import { encodePaletteIdentity } from '@/lib/palette-match/palette-ranking' import { layoutMultiPrimaryPaletteSections, orderMultiPrimaryPaletteItems @@ -124,6 +125,21 @@ let testRoot: Root let testContainer: HTMLDivElement let setCommandQuery: ((next: string) => void) | null = null +const WORKSPACE_TAB_ITEM_PREFIX = encodePaletteIdentity(['workspace-tab']) +const WORKTREE_ITEM_PREFIX = encodePaletteIdentity(['worktree']) + +function workspaceTabItemId(worktreeId: string, tabId: string): string { + return encodePaletteIdentity(['workspace-tab', '', worktreeId, tabId]) +} + +function isWorkspaceTabItemId(id: string): boolean { + return id.startsWith(WORKSPACE_TAB_ITEM_PREFIX) +} + +function isWorktreeItemId(id: string): boolean { + return id.startsWith(WORKTREE_ITEM_PREFIX) +} + function makeRepo(): Repo { return { id: 'repo-1', @@ -306,7 +322,7 @@ function getPrimaryRowsBySectionHeader(): { header: string; rowId: string }[] { )) { const rowId = node.dataset.commandItem if (rowId) { - if (rowId.startsWith('workspace-tab:') || rowId.startsWith('worktree:')) { + if (isWorkspaceTabItemId(rowId) || isWorktreeItemId(rowId)) { pairs.push({ header, rowId }) } continue @@ -347,10 +363,10 @@ describe('WorktreeJumpPalette interleaved primary sections', () => { const rows = getPrimaryRowsBySectionHeader() // Why the counts: both remainders must still render, just under a re-emitted header. - expect(rows.filter((row) => row.rowId.startsWith('workspace-tab:'))).toHaveLength(8) - expect(rows.filter((row) => row.rowId.startsWith('worktree:'))).toHaveLength(5) + expect(rows.filter((row) => isWorkspaceTabItemId(row.rowId))).toHaveLength(8) + expect(rows.filter((row) => isWorktreeItemId(row.rowId))).toHaveLength(5) for (const { header, rowId } of rows) { - expect(header).toBe(rowId.startsWith('workspace-tab:') ? 'Open Tabs' : 'Worktrees') + expect(header).toBe(isWorkspaceTabItemId(rowId) ? 'Open Tabs' : 'Worktrees') } }) @@ -366,10 +382,10 @@ describe('WorktreeJumpPalette interleaved primary sections', () => { testContainer.querySelectorAll('[data-command-item]') ) .map((el) => el.dataset.commandItem!) - .filter((id) => id.startsWith('workspace-tab:') || id.startsWith('worktree:')) + .filter((id) => isWorkspaceTabItemId(id) || isWorktreeItemId(id)) - const tabIds = renderedIds.filter((id) => id.startsWith('workspace-tab:')) - const worktreeIds = renderedIds.filter((id) => id.startsWith('worktree:')) + const tabIds = renderedIds.filter(isWorkspaceTabItemId) + const worktreeIds = renderedIds.filter(isWorktreeItemId) const layout = layoutMultiPrimaryPaletteSections({ leadingItems: tabIds, @@ -402,7 +418,7 @@ describe('WorktreeJumpPalette interleaved primary sections', () => { await flushEffects() const rows = getPrimaryRowsBySectionHeader() - expect(rows).toEqual([{ header: 'Open Tabs', rowId: 'workspace-tab:tab-0' }]) + expect(rows).toEqual([{ header: 'Open Tabs', rowId: workspaceTabItemId('wt-tabs', 'tab-0') }]) expect(testContainer.textContent).toContain('Open Tabs') expect(testContainer.textContent).not.toContain('Worktrees') }) @@ -425,12 +441,14 @@ describe('WorktreeJumpPalette interleaved primary sections', () => { activeGroupIdByWorktree: { 'wt-tabs': 'group-wt-tabs' } }) - const row = testContainer.querySelector('[data-command-item="workspace-tab:tab-0"]') + const row = testContainer.querySelector( + `[data-command-item="${workspaceTabItemId('wt-tabs', 'tab-0')}"]` + ) expect(row).not.toBeNull() const title = row?.querySelector('[data-slot="palette-open-tab-title"]') const worktree = row?.querySelector('[data-slot="palette-open-tab-worktree"]') expect(title?.textContent).toBe(longTitle) - expect(title?.classList.contains('flex-1')).toBe(true) + expect(title?.classList.contains('flex-auto')).toBe(true) expect(worktree?.textContent).toBe('user-support') expect(worktree?.compareDocumentPosition(title ?? document.createElement('span'))).toBe( Node.DOCUMENT_POSITION_PRECEDING @@ -454,7 +472,9 @@ describe('WorktreeJumpPalette interleaved primary sections', () => { activeGroupIdByWorktree: { 'wt-tabs': 'group-wt-tabs' } }) - const row = testContainer.querySelector('[data-command-item="workspace-tab:tab-0"]') + const row = testContainer.querySelector( + `[data-command-item="${workspaceTabItemId('wt-tabs', 'tab-0')}"]` + ) expect(row).not.toBeNull() const worktree = row?.querySelector('[data-slot="palette-open-tab-worktree"]') expect(worktree?.textContent).toBe('main') @@ -577,13 +597,15 @@ describe('WorktreeJumpPalette interleaved primary sections', () => { await flushEffects() // After expanding by 20: 30 worktrees are rendered, 5 more - const renderedItems = testContainer.querySelectorAll('[data-command-item^="worktree:"]') + const renderedItems = testContainer.querySelectorAll( + `[data-command-item^="${WORKTREE_ITEM_PREFIX}"]` + ) expect(renderedItems).toHaveLength(30) expect(testContainer.textContent).toContain('5 more') const firstRevealedItemId = Array.from(testContainer.querySelectorAll('[cmdk-item]'))[ seeMoreIndex ]?.getAttribute('data-value') - expect(firstRevealedItemId).toMatch(/^worktree:/) + expect(firstRevealedItemId).toMatch(new RegExp(`^${WORKTREE_ITEM_PREFIX}`)) expect(firstRevealedItemId).not.toBe(initialItemIds[0]) expect( testContainer @@ -601,7 +623,9 @@ describe('WorktreeJumpPalette interleaved primary sections', () => { }) await flushEffects() - const renderedItemsAll = testContainer.querySelectorAll('[data-command-item^="worktree:"]') + const renderedItemsAll = testContainer.querySelectorAll( + `[data-command-item^="${WORKTREE_ITEM_PREFIX}"]` + ) expect(renderedItemsAll).toHaveLength(35) expect(testContainer.textContent).not.toContain('more') }) diff --git a/src/renderer/src/components/worktree-jump-palette-open-tab-items.ts b/src/renderer/src/components/worktree-jump-palette-open-tab-items.ts new file mode 100644 index 00000000000..dff7ce1ded6 --- /dev/null +++ b/src/renderer/src/components/worktree-jump-palette-open-tab-items.ts @@ -0,0 +1,67 @@ +import { comparePaletteRankedItems } from '@/lib/cmd-j-section-leadership' +import type { BrowserPaletteSearchResult } from '@/lib/browser-palette-search' +import type { SimulatorPaletteSearchResult } from '@/lib/simulator-palette-search' +import type { WorkspaceTabPaletteSearchResult } from '@/lib/workspace-tab-palette-search' +import type { + BrowserPaletteItem, + OpenTabPaletteItem, + SimulatorPaletteItem, + WorkspaceTabPaletteItem +} from './worktree-jump-palette-model' + +export function buildBrowserPaletteItems( + results: readonly BrowserPaletteSearchResult[] +): BrowserPaletteItem[] { + return results.map((result) => ({ + id: result.paletteIdentity, + type: 'browser-page', + result + })) +} + +export function buildSimulatorPaletteItems( + results: readonly SimulatorPaletteSearchResult[] +): SimulatorPaletteItem[] { + return results.map((result) => ({ + id: result.paletteIdentity, + type: 'simulator-tab', + result + })) +} + +export function buildWorkspaceTabPaletteItems( + results: readonly WorkspaceTabPaletteSearchResult[] +): WorkspaceTabPaletteItem[] { + return results.map((result) => ({ + id: result.paletteIdentity, + type: 'workspace-tab', + result + })) +} + +export function buildOpenTabPaletteItems({ + browserItems, + simulatorItems, + workspaceTabItems +}: { + browserItems: readonly BrowserPaletteItem[] + simulatorItems: readonly SimulatorPaletteItem[] + workspaceTabItems: readonly WorkspaceTabPaletteItem[] +}): OpenTabPaletteItem[] { + return [...browserItems, ...simulatorItems, ...workspaceTabItems].sort((left, right) => + comparePaletteRankedItems( + { + rank: left.result.rank, + order: left.result.score, + identity: left.id, + activity: left.result.activity + }, + { + rank: right.result.rank, + order: right.result.score, + identity: right.id, + activity: right.result.activity + } + ) + ) +} diff --git a/src/renderer/src/components/worktree-jump-palette-primitives.test.tsx b/src/renderer/src/components/worktree-jump-palette-primitives.test.tsx new file mode 100644 index 00000000000..a54a48329be --- /dev/null +++ b/src/renderer/src/components/worktree-jump-palette-primitives.test.tsx @@ -0,0 +1,47 @@ +// @vitest-environment happy-dom + +import { cleanup, render, type RenderResult, screen } from '@testing-library/react' +import { afterEach, expect, it } from 'vitest' +import { TooltipProvider } from '@/components/ui/tooltip' +import { PaletteOpenTabPrimaryLine } from './worktree-jump-palette-primitives' + +afterEach(() => cleanup()) + +function renderPrimaryLine( + secondaryMatches: readonly { text: string; ranges: readonly never[] }[] +): RenderResult { + return render( + + + + ) +} + +it('exposes the extra secondary matches through the row text, not the tab order', () => { + const { container } = renderPrimaryLine([ + { text: 'src/app.ts', ranges: [] }, + { text: 'src/deep/nested.ts', ranges: [] }, + { text: 'docs/readme.md', ranges: [] } + ]) + + const extraMatches = container.querySelector('[data-slot="palette-open-tab-extra-matches"]') + expect(extraMatches?.textContent).toBe('src/deep/nested.ts, docs/readme.md') + + const badge = screen.getByText('+2') + expect(badge.getAttribute('aria-hidden')).toBe('true') + expect(badge.tabIndex).toBe(-1) +}) + +it('renders no badge when every secondary match is already shown', () => { + renderPrimaryLine([{ text: 'src/app.ts', ranges: [] }]) + + expect(screen.queryByText(/^\+\d+$/)).toBeNull() +}) diff --git a/src/renderer/src/components/worktree-jump-palette-primitives.tsx b/src/renderer/src/components/worktree-jump-palette-primitives.tsx index 497828bb8b9..4162a56cb6e 100644 --- a/src/renderer/src/components/worktree-jump-palette-primitives.tsx +++ b/src/renderer/src/components/worktree-jump-palette-primitives.tsx @@ -1,5 +1,4 @@ -import { useLayoutEffect, useRef, useState } from 'react' -import type React from 'react' +import React, { useLayoutEffect, useRef, useState } from 'react' import { ShortcutKeyCombo } from '@/components/ShortcutKeyCombo' import { translate } from '@/i18n/i18n' import type { PaletteHostBadge } from '@/components/cmd-j/palette-host-badge' @@ -8,6 +7,8 @@ import { Tooltip, TooltipContent, TooltipTrigger } from '@/components/ui/tooltip import type { Worktree } from '../../../shared/worktree/types' import { resolveWorktreeBranchLabel } from '@/lib/worktree-default-display-name' +const NO_SECONDARY_MATCHES: readonly { text: string; ranges: readonly MatchRange[] }[] = [] + export function PaletteRowShortcutBadge({ index, modifierKeys @@ -30,10 +31,12 @@ export function PaletteRowShortcutBadge({ export function HighlightedText({ text, - matchRanges + matchRanges, + highlightClassName = 'font-semibold text-foreground' }: { text: string matchRanges?: readonly MatchRange[] | null + highlightClassName?: string }): React.JSX.Element { const ranges = (matchRanges ?? []).filter( (range) => range.start < range.end && range.start < text.length @@ -51,7 +54,7 @@ export function HighlightedText({ } if (end > start) { parts.push( - + {text.slice(start, end)} ) @@ -69,6 +72,7 @@ export function PaletteOpenTabPrimaryLine({ titleRanges, secondaryText, secondaryRanges, + secondaryMatches = NO_SECONDARY_MATCHES, worktreeName, worktreeRanges, sessionAge, @@ -78,6 +82,7 @@ export function PaletteOpenTabPrimaryLine({ titleRanges: readonly MatchRange[] secondaryText: string secondaryRanges: readonly MatchRange[] + secondaryMatches?: readonly { text: string; ranges: readonly MatchRange[] }[] worktreeName: string worktreeRanges: readonly MatchRange[] sessionAge?: string @@ -85,12 +90,15 @@ export function PaletteOpenTabPrimaryLine({ }): React.JSX.Element { const showSecondary = secondaryText.trim().length > 0 const showWorktree = worktreeName.trim().length > 0 + const additionalSecondaryMatches = secondaryMatches.filter( + (match) => match.text && match.text !== secondaryText + ) return (
@@ -115,6 +123,37 @@ export function PaletteOpenTabPrimaryLine({ ) : null} + {additionalSecondaryMatches.length ? ( + <> + {/* Tab selects the palette filter, so the badge stays out of the tab order and + reads its matches through the row's own accessible name instead. */} + + {additionalSecondaryMatches.map((match) => match.text).join(', ')} + + + + + +{additionalSecondaryMatches.length} + + + + {additionalSecondaryMatches.map((match) => ( +
+ +
+ ))} +
+
+ + ) : null} {showWorktree ? ( <> · diff --git a/src/renderer/src/components/worktree-jump-palette-surface.tsx b/src/renderer/src/components/worktree-jump-palette-surface.tsx index 40013f97f9d..09c0e3443b9 100644 --- a/src/renderer/src/components/worktree-jump-palette-surface.tsx +++ b/src/renderer/src/components/worktree-jump-palette-surface.tsx @@ -70,7 +70,7 @@ export function WorktreeJumpPaletteSurface({ @@ -93,7 +93,7 @@ export function WorktreeJumpPaletteSurface({ } /> + {result.typeAliasMatches.length ? ( + + {result.typeAliasMatches.map((match) => match.text).join(', ')} + + ) : null}
diff --git a/src/renderer/src/components/worktree-jump-palette-worktree-maps.ts b/src/renderer/src/components/worktree-jump-palette-worktree-maps.ts index e069ca29bc4..335b0167865 100644 --- a/src/renderer/src/components/worktree-jump-palette-worktree-maps.ts +++ b/src/renderer/src/components/worktree-jump-palette-worktree-maps.ts @@ -1,5 +1,5 @@ -import { getWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' import type { Worktree } from '../../../shared/worktree/types' +import { getPaletteWorktreeIdentity } from '@/lib/palette-repo-resolution' export function buildWorktreeJumpPaletteWorktreeMaps(worktrees: readonly Worktree[]): { worktreeMap: Map @@ -8,13 +8,13 @@ export function buildWorktreeJumpPaletteWorktreeMaps(worktrees: readonly Worktre const worktreeMap = new Map() for (const worktree of worktrees) { // Keep a host-qualified map for consumers that only have an identity key. - worktreeMap.set(getWorktreeHostIdentity(worktree), worktree) + worktreeMap.set(getPaletteWorktreeIdentity(worktree), worktree) if (!worktreeMap.has(worktree.id)) { worktreeMap.set(worktree.id, worktree) } } const worktreeOrder = new Map( - worktrees.map((worktree, index) => [getWorktreeHostIdentity(worktree), index]) + worktrees.map((worktree, index) => [getPaletteWorktreeIdentity(worktree), index]) ) return { worktreeMap, worktreeOrder } } diff --git a/src/renderer/src/components/worktree-jump-palette-worktree-row.tsx b/src/renderer/src/components/worktree-jump-palette-worktree-row.tsx index 479625fbe86..80610a700ca 100644 --- a/src/renderer/src/components/worktree-jump-palette-worktree-row.tsx +++ b/src/renderer/src/components/worktree-jump-palette-worktree-row.tsx @@ -51,7 +51,10 @@ export function WorktreeJumpPaletteWorktreeRow({ activeWorktreeId, controller.activeWorkspaceExecutionHostId ) - const sessionAge = formatPaletteSessionAge(worktree.lastActivityAt, controller.paletteNowMs) + const sessionAge = formatPaletteSessionAge( + controller.hasQuery ? entry.match.lastActiveAt : worktree.lastActivityAt, + controller.paletteNowMs + ) const sshConnectionId = repo?.connectionId && !isRuntimeOwnedSshTargetId(repo.connectionId) ? repo.connectionId : null const sshStatus = sshConnectionId diff --git a/src/renderer/src/hooks/composer-state/full-creation-execution.ts b/src/renderer/src/hooks/composer-state/full-creation-execution.ts index 7cb5f88f17d..f199ca66f0c 100644 --- a/src/renderer/src/hooks/composer-state/full-creation-execution.ts +++ b/src/renderer/src/hooks/composer-state/full-creation-execution.ts @@ -175,7 +175,7 @@ export function useFullCreationExecution(input: FullCreationExecutionInput) { smartGitHubResolution.kind === 'none' ? (linkedGitLabMR ?? undefined) : undefined, smartGitHubResolution.kind === 'none' ? (linkedGitLabIssue ?? undefined) : undefined, effectiveBackendStartup, - structuredLaunch ? false : pendingFirstAgentMessageRename, + pendingFirstAgentMessageRename, undefined, linkedLinearIssueWorkspaceId, linkedLinearIssueOrganizationUrlKey, diff --git a/src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge.ts b/src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge.ts index 1fd3eed2039..a297c751a29 100644 --- a/src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge.ts +++ b/src/renderer/src/hooks/ipc-events/terminal-request-ipc-bridge.ts @@ -3,6 +3,7 @@ import { getConnectionIdFromState } from '@/lib/connection-context' import { initialAgentTabViewModeProps } from '@/lib/native-chat-initial-view-mode' import { isNativeChatTranscriptLocalReadable } from '@/lib/native-chat-transcript-readability' import { resolveTerminalWorktreeRoute } from '@/lib/terminal-worktree-route' +import { insertUnifiedTabAfterAnchor } from '@/lib/unified-tab-anchor-insertion' import { translate } from '@/i18n/i18n' import { useAppStore } from '../../store' import { @@ -83,30 +84,11 @@ export function registerTerminalRequestIpcBridge(unsubs: (() => void)[]): void { requestBackgroundTerminalWorktreeMount({ worktreeId, tabIds: [tab.id] }) } if (data.afterTabId) { - const createdUnifiedTab = useAppStore + const createdUnifiedTabId = useAppStore .getState() - .unifiedTabsByWorktree[worktreeId]?.find((item) => item.entityId === tab.id) - const anchorUnifiedTab = useAppStore - .getState() - .unifiedTabsByWorktree[worktreeId]?.find((item) => item.id === data.afterTabId) - if ( - createdUnifiedTab && - anchorUnifiedTab && - createdUnifiedTab.groupId === anchorUnifiedTab.groupId - ) { - const group = useAppStore - .getState() - .groupsByWorktree[worktreeId]?.find((item) => item.id === createdUnifiedTab.groupId) - const order = (group?.tabOrder ?? []).filter((id) => id !== createdUnifiedTab.id) - const anchorIndex = order.indexOf(anchorUnifiedTab.id) - order.splice( - anchorIndex === -1 ? order.length : anchorIndex + 1, - 0, - createdUnifiedTab.id - ) - useAppStore.getState().reorderUnifiedTabs(createdUnifiedTab.groupId, order, { - recordInteraction: false - }) + .unifiedTabsByWorktree[worktreeId]?.find((item) => item.entityId === tab.id)?.id + if (createdUnifiedTabId) { + insertUnifiedTabAfterAnchor(worktreeId, createdUnifiedTabId, data.afterTabId) } } if (shouldActivate) { diff --git a/src/renderer/src/hooks/use-palette-search-evaluation-context.test.ts b/src/renderer/src/hooks/use-palette-search-evaluation-context.test.ts new file mode 100644 index 00000000000..984397a9393 --- /dev/null +++ b/src/renderer/src/hooks/use-palette-search-evaluation-context.test.ts @@ -0,0 +1,34 @@ +// @vitest-environment happy-dom + +import { renderHook } from '@testing-library/react' +import { afterEach, describe, expect, it, vi } from 'vitest' +import { usePaletteSearchEvaluationContext } from './use-palette-search-evaluation-context' + +afterEach(() => vi.restoreAllMocks()) + +describe('usePaletteSearchEvaluationContext', () => { + it('captures one clock per snapshot without committing a stale ranking pass', () => { + const clock = vi.spyOn(Date, 'now').mockReturnValue(1_000) + const evaluations: number[] = [] + const snapshot = { query: 'atlas' } + const { result, rerender } = renderHook( + ({ snapshot }) => { + const context = usePaletteSearchEvaluationContext(snapshot) + evaluations.push(context.nowMs) + return context + }, + { initialProps: { snapshot } } + ) + expect(evaluations).toEqual([1_000]) + const initial = result.current + + clock.mockReturnValue(2_000) + rerender({ snapshot }) + expect(result.current).toBe(initial) + + evaluations.length = 0 + rerender({ snapshot: { query: 'atlas notes' } }) + expect(evaluations).toEqual([2_000]) + expect(result.current).not.toBe(initial) + }) +}) diff --git a/src/renderer/src/hooks/use-palette-search-evaluation-context.ts b/src/renderer/src/hooks/use-palette-search-evaluation-context.ts new file mode 100644 index 00000000000..5ad43ca0bc5 --- /dev/null +++ b/src/renderer/src/hooks/use-palette-search-evaluation-context.ts @@ -0,0 +1,14 @@ +import { useMemo } from 'react' +import { + createPaletteSearchContext, + type PaletteSearchContext +} from '@/lib/palette-match/palette-ranking' + +/** One clock for every source participating in the current search snapshot. */ +export function usePaletteSearchEvaluationContext(snapshot: unknown): PaletteSearchContext { + return useMemo(() => { + void snapshot + // oxlint-disable-next-line react/purity -- Each changed snapshot starts one synchronous evaluation clock. + return createPaletteSearchContext(Date.now()) + }, [snapshot]) +} diff --git a/src/renderer/src/i18n/en-runtime-required.json b/src/renderer/src/i18n/en-runtime-required.json index af43a08f37a..64902d783ac 100644 --- a/src/renderer/src/i18n/en-runtime-required.json +++ b/src/renderer/src/i18n/en-runtime-required.json @@ -226,10 +226,6 @@ "AutomationEditorDialogHeader": { "4c8e1a72b9": "A recurring agent task" }, - "AutomationListSortHeader": { - "sortedAscending": "{{value0}}, sorted ascending", - "sortedDescending": "{{value0}}, sorted descending" - }, "AutomationRunHistory": { "fdb3caa8fb": "known" }, @@ -1053,13 +1049,20 @@ }, "settings": { "AccountsPane": { + "15e831350e": "Configure MiniMax usage tracking from platform.minimax.io.", + "1fd1b1b6b4": "Cookie not set", "3455cf43fa": "Claude login.", "350b2a1aa7": "Use your current", + "4e32e030b2": "Stored locally. Orca sends it only to platform.minimax.io for usage refreshes.", "566d9a99ab": "_token=…; minimax_group_id_v2=…", + "5e08b0fe57": "Stored locally and sent only to platform.minimax.io for usage refreshes.", + "79418c782a": "Open platform.minimax.io/console/usage in your browser, sign in, then copy the Cookie request header from DevTools (Network → any remains request → Cookie).", "9107406589": "Could not load Claude accounts.", "b10cb4f696": "adding", "b11078a9c2": "wsl", - "b8c2905c2b": "Could not load Codex accounts." + "b43e761fe5": "MiniMax cookie update failed.", + "b8c2905c2b": "Could not load Codex accounts.", + "f5d8d2a6a1": "Open platform.minimax.io/console/usage in your browser and sign in." }, "AdvancedNetworkSettingsSection": { "d93c7cd531": "Configure app-level network routing.", @@ -1525,63 +1528,6 @@ "7c3bb36706": "remove", "e2b0ee267f": "stale" }, - "accounts": { - "search": { - "02c438bc7b": "expired", - "042885c07c": "out of date", - "06662af91e": "account", - "0b4d948eb5": "wsl", - "35b461d817": "sign in", - "421c6be25e": "id", - "488a7e9206": "linux", - "593720c17f": "location", - "5b3f18ef4a": "switch", - "61f7d1fcbe": "cookie", - "70d1b8def5": "codex", - "7118d2f908": "credentials", - "77e32a2ad3": "reauthenticate", - "7e67d7d1b6": "wrk", - "8630464352": "cli", - "86edc96bc9": "status bar", - "8b06729e0f": "active", - "8dcbef1856": "opencode", - "933deaf732": "oauth", - "9c4e40cf6b": "session", - "9f70aa706c": "provider", - "a9f3d7b5c8": "login", - "b0a4e8c6d9": "oauth", - "b7c2cee442": "experimental", - "bdbd1e668e": "windows", - "be8b621bdc": "workspace", - "c1b5f9d7e0": "xai", - "c759741d77": "quota", - "d2c6a0e8f1": "grok", - "e02c136ad0": "auth", - "e14049e1a8": "claude", - "e8e1ff3887": "gemini", - "e949b08ffb": "rate limit", - "f2d666a886": "optional" - } - }, - "advanced": { - "search": { - "2b4d26d11e": "networking", - "4383251647": "vpn", - "48a1c8f534": "http", - "4b4ae4345a": "http2", - "4d44352eea": "network", - "621233008b": "http/1.1", - "6576fce4d2": "troubleshooting", - "65bf6af262": "compatibility", - "79e0947e95": "support", - "a0f71bd909": "http/2", - "a7002e1ac4": "updater", - "e04e9db503": "advanced", - "e61ed8ab33": "updates", - "f8ff125ebe": "http1", - "f98a60af11": "proxy" - } - }, "agent-awake-copy": { "95d3031db2": "Keeps this computer and display awake while agents are working. Lid-close behavior follows this device's power settings.", "a42f6fbdd8": "Keeps this computer and display awake while agents are working. Orca also asks this device to stay awake when the lid is closed, subject to its power policy.", @@ -1600,7 +1546,6 @@ }, "agents": { "search": { - "2814401339": "installed", "042c551bc5": "config", "0d1c334987": "lid", "0d752916f8": "hooks", @@ -1618,16 +1563,12 @@ "66b6b82eb4": "awake", "6956646a1e": "title", "6984d4291a": "status", - "719f53350c": "path", - "77c02fa3c3": "windows", - "839e82c81f": "detect", "845ad9128a": "power", "848dcae8d3": "generated", "8599603496": "done", "87fffe6c20": "show", "8a17fd6026": "stable", "966890236d": "name", - "96ba2373b6": "agent", "a6d594c17d": "install", "a79d266f71": "session", "afbf35be68": "stable session", @@ -1637,8 +1578,6 @@ "c1317fe641": "restore", "c64059f50d": "prompt", "cbdd7f3b9e": "Choose whether installed agents are detected on this device or in WSL.", - "d2952dfd74": "location", - "d608654c03": "wsl", "d8f3a8b8a0": "default", "dbc8aca6b0": "sleep", "e2b7c0dcd7": "github", @@ -1646,104 +1585,9 @@ "ef804b7337": "Agent Location", "f2932bf22b": "detected", "f412abbba5": "claude", - "f622b8eb2a": "linux", "ff8de8a2ad": "display" } }, - "appearance": { - "search": { - "006e67b279": "ports", - "00a028f25f": "usage", - "08c86bf58e": "gitignore", - "0952091186": "scale", - "0c83659f48": "shortcut", - "0d5a74b606": "tasks", - "1f2880a9d5": "orca", - "24094af355": "font", - "25e51b62ee": "rate limit", - "262fe1d24f": "dark", - "2804a920ad": "gemini", - "2cfb3420c0": "app icon", - "2ee4810f38": "github", - "2f12e1aa3a": "ui", - "35565867cb": "moonshot", - "36e006efc1": "app", - "3a9b69d734": "system", - "3ae5de6101": "zoom", - "40e5c3c285": "kimi", - "4355f18ac6": "memory", - "43cfba3b95": "server", - "44d873fd18": "light", - "468448bba4": "watercolor", - "46d21eef62": "localhost", - "4c920ab2d1": "schedule", - "4ddbde4999": "cpu", - "5095258df2": "interface", - "51b0ccd6a2": "google", - "51f957ce39": "name", - "58f4e22fa2": "automation", - "5bff6a2ef0": "sidebar", - "5e5b8878bf": "phone", - "648eeada79": "hide", - "651f35b2c6": "switcher", - "6b846424cc": "linear", - "6cf5f54ce1": "button", - "6ecad74eb3": "ssh", - "74618577c7": "mobile", - "839fb1e3ed": "toolbox", - "896eb53fd4": "status bar", - "8b36fb3f64": "typography", - "8dfd676c28": "codex", - "90bdc043ea": "disk", - "96b4fb0064": "terminal", - "97957e374e": "openai", - "9c4d5f0894": "manager", - "9f2df826ac": "ignored", - "a0e09aed9c": "typeface", - "a278406ed5": "remote", - "a895d0f938": "brand", - "a9d56852eb": "opencode", - "ac79fe4a04": "show", - "afbb6a3767": "tokens", - "antigravityKeyword": "antigravity", - "b186f3cefb": "automations", - "bce3ac317a": "git", - "bed343b03e": "titlebar", - "c1bca1885a": "file explorer", - "c5b9f8d1e3": "xai", - "c690a15849": "resource", - "c9fe3a7876": "claude", - "cb1cc62cf8": "space", - "d16378a88f": "minimax", - "d18b54ca90": "dock", - "d6c0a9e2f4": "grok", - "d77537b580": "opencode-go", - "d9e7cef86f": "cookie", - "dc02c8759d": "workspace", - "de586def95": "subscription", - "dea0a9a665": "anthropic", - "e5bc35d59e": "window", - "edbf0f63a0": "cost", - "f4997e0f8a": "connection", - "f586abfa35": "blue", - "fab91464dd": "ide", - "fe192b060e": "host", - "language": { - "i18n": "i18n", - "locale": "locale", - "translation": "translation" - }, - "workspaceCardLayout": { - "cardLayout": "card layout", - "compact": "compact", - "compactDisplay": "compact display", - "detailed": "detailed", - "workspaceCards": "workspace cards", - "workspaceOptions": "workspace options", - "worktreeCards": "worktree cards" - } - } - }, "artifacts": { "account": "Orca account", "connected": "Connected", @@ -1756,21 +1600,7 @@ "rename": { "branch": { "search": { - "0971762141": "kebab-case", - "10485c4fc5": "command", - "3ef3cbe98c": "agent", - "40d21f2efc": "prompt", - "427f2cd1eb": "Auto-Rename Branch", - "50139297e6": "built-in prompt", - "502aa57681": "instructions", - "55a1860e47": "rename", - "7803423877": "auto", - "7adefcdd94": "template", - "9319bd9827": "branch", - "a482f6a423": "slug", - "ed677944cc": "worktree", - "f0acf64301": "creature name", - "f41833025e": "generate" + "427f2cd1eb": "Auto-Rename Branch" } } } @@ -1782,142 +1612,6 @@ } } }, - "browser": { - "search": { - "0732ebe6fb": "private", - "0bb34eacc9": "query", - "0dbb1eaf4e": "homepage", - "16bd69cd82": "search", - "1c1e097985": "arc", - "1f8153acfb": "duckduckgo", - "29193a51d5": "cookies", - "291f480a5e": "home", - "2d2d995c58": "browser", - "2e7f951773": "import", - "3538b3aaeb": "token", - "3910a41f32": "auth", - "44d14df30d": "preview", - "4596a52cf7": "landing", - "483a0eb5e0": "new tab", - "4a98ed195f": "zoom", - "4fda4fb066": "url", - "5164c47e31": "blank", - "533a253deb": "edge", - "5448f4097b": "default", - "54f4ea55f7": "scale", - "66dd641a47": "session", - "68d1db8929": "markdown", - "726f2a8556": "page zoom", - "72b4b89970": "engine", - "72c58f7792": "webview", - "7539f6336c": "profile", - "75a0d435b7": "chrome", - "82ba1c80ea": "localhost", - "854ef6ce83": "login", - "8a489aab8d": "google", - "8b8ed06e4b": "omnibox", - "8dd4805991": "file", - "90425d313c": "shift", - "95944898e0": "percentage", - "a7a07d5415": "editor", - "ad40e75d13": "bing", - "bea27bac4b": "links", - "e1c2a57f07": "kagi", - "linkRoutingModifier": { - "invert": "invert", - "modifier": "modifier", - "opposite": "opposite", - "routing": "routing" - }, - "terminalLinkActions": { - "actions": "actions", - "click": "click", - "disable": "disable", - "menu": "menu", - "popover": "popover", - "terminal": "terminal" - } - }, - "use": { - "search": { - "02837ee497": "session", - "034c5e8d7f": "enable", - "088e7a9012": "chrome", - "20c1323d1e": "computer use", - "22fb801af8": "chrome profile", - "2e1b09897b": "edge", - "30c74aaa1f": "path", - "3f4c559deb": "arc profile", - "3ffafc9b95": "command", - "48557f639c": "login", - "59968bb9b4": "authenticated browser", - "62e2a790c0": "existing session", - "63a66da648": "system browser", - "6ea88e5206": "npx", - "7e0dcb257a": "shell", - "85fab5e12c": "cli", - "96ce3d2de2": "auth", - "9d97446873": "agent", - "a2d489263e": "skill", - "a57c2172dc": "agent-browser", - "ab349a2dd0": "arc", - "ba4eb53b72": "browser use", - "cee44fb442": "automation", - "d5ad1f7aad": "import", - "d5afa54d21": "edge profile", - "e56c7b55c9": "setup", - "e5a784bc54": "install", - "f5b8fdddf5": "orca-cli", - "fb8178824f": "cookies", - "ff05cbc344": "orca" - } - } - }, - "commit": { - "message": { - "ai": { - "search": { - "3766941527": "agent", - "0f29331fed": "arguments", - "110be48b81": "pull request", - "127d512e75": "commit", - "181cdb0637": "open", - "37c65bbb44": "fix", - "402f101af8": "prompt", - "53e8504fb2": "ci", - "542e1a00a7": "codex", - "57c851a68c": "cli", - "61117e57f3": "args", - "7e264b926b": "draft", - "82109d627d": "source control", - "8e0bcc5d99": "model", - "8e9cc598d7": "generate", - "93e5210da8": "message", - "b261c88609": "pr", - "b7d50da4d8": "template", - "c33cb1b982": "ai", - "c46e665f7e": "checks", - "d22a6459e4": "conflicts", - "d32936bb2a": "branch", - "ee14a9e9f7": "enabled", - "f121bec167": "claude", - "f4731b22bf": "command" - } - } - } - }, - "computer": { - "use": { - "search": { - "26c1290d83": "screen recording", - "6e88da3508": "skill", - "798be54d7e": "automation", - "82f01c2d2c": "accessibility", - "e27f8bafbf": "screenshot", - "fefb452f5b": "computer use" - } - } - }, "computerUseSkillRuntime": { "thisDevice": "This device" }, @@ -1925,303 +1619,56 @@ "permissionsRequired_one": "1 permission required before agents can operate app windows.", "permissionsRequired_other": "{{value0}} permissions required before agents can operate app windows." }, - "developer": { - "permissions": { - "search": { - "00e954319e": "whisper", - "0a467b750e": "screenshot", - "0c13b249e3": "tcc", - "11653d3f42": "mdns", - "1e6e27b202": "ffmpeg", - "2270ccff3f": "privacy", - "259b829b84": "camera", - "3e0131e45d": "icloud", - "4438f81bfa": "documents", - "5610022e1e": "automation", - "6c82846f66": "device", - "6db4fca386": "macos", - "78a10b826f": "bonjour", - "7f145a3984": "window", - "87620e6416": "lan", - "a0c19119fb": "downloads", - "a765112513": "video", - "a98aa11a9c": "permissions", - "af122938a3": "voice", - "b192432ef0": "audio", - "c4a4a02ea4": "usb", - "ce07159ff5": "desktop", - "e3fbc48083": "bluetooth", - "ed7c12bdb4": "microphone", - "f061f08b7b": "sox", - "fa3239cd42": "local network" - } - } - }, "experimental": { "search": { - "01567f19ca": "attention", - "051203d37c": "pet", - "0d24759f14": "experimental", "10b52f79c1": "worktrees", "244a0ecd3d": "activity", - "268e99d957": "highlight", - "2a33975d72": "mascot", "3021571c30": "shared", "3028f0bd3a": "link", "44c7f209d5": "node_modules", "4ad605f222": "env", "4d63251595": "Threaded left-sidebar feed for agent completions and blocking states.", - "5f067ba0f9": "agent", "603d29ed74": "Automatically materialize configured files or folders into newly created worktrees using APFS clone-copy on macOS when possible, otherwise symlinks.", - "65df471ab2": "animated", - "7695fd30e9": "notification", "78c2a8dc74": "Shared paths on worktrees", - "791fefc0b0": "corner", - "7b79081695": "unread", - "8facf10138": "bell", "92a9357d1f": "agents view", - "9af7a518db": "character", - "9bb3bd5098": "terminal", - "9f5609bfb8": "overlay", - "agentDashboard": { - "dashboard": "dashboard" - }, - "agentHibernation": { - "agent": "agent", - "agents": "agents", - "minutes": "minutes", - "sleep": "sleep", - "terminal": "terminal" - }, - "b54cea709b": "sidekick", "bff1ff7768": "symlinks", "c387565812": "symlink", "ca5d1f3f46": "timeline", "ccc5548ac5": "Agents View", "d01b3882ba": "notifications", "d23ae13990": "worktree", - "edc49480a1": "pane", "f082788cfe": "links", - "f10d307468": "completion", "fa72e71f05": "agents", - "fe5688b761": "sidebar", - "nativeChat": { - "grok": "grok" - }, - "newWorktreeCardStyle": { - "card": "card", - "cards": "cards", - "menu": "menu", - "metadata": "metadata", - "status": "status", - "worktree": "worktree", - "worktrees": "worktrees" - } - } - }, - "floating": { - "workspace": { - "search": { - "156ffeee08": "note", - "2b5efa55c9": "global", - "49db74a92d": "browser", - "52db6e3baf": "notes", - "6410fe83d8": "terminal", - "884e5e6132": "markdown", - "94f4d013c8": "status bar", - "a38bfc3f77": "quick panel", - "a452146574": "toggle button", - "ebeedb2f6a": "quick terminal" - } + "fe5688b761": "sidebar" } }, "general": { "search": { - "06ea5a69a6": "github", - "0a00691c06": "shell command", - "0a02059549": "file tree", - "0a5fa65926": "inline", - "0cb3d94f00": "cursor", - "0efc9d96ad": "prompt", - "12ecc640a8": "mru", - "146728ac2c": "delay", - "19baae651b": "sidebar", - "1ff67ba40c": "notes", - "20b711ac9e": "proxy", - "22572e99c1": "annotations", - "233f7e2f37": "side-by-side", "27d9b996ba": "codex", - "2a254b725e": "tab", - "2b463f0bf9": "view", - "2f42852568": "tree", - "3462308bd3": "tokens", - "3566fce83f": "localhost", - "3a73054565": "bypass", - "3b5733573e": "diff", "3c30fe2d51": "gemini", - "3ca5ab78a5": "code", "41c2f9a025": "default", - "4469b6fa4e": "save", - "4dd5684836": "review", - "54ba13831a": "recent", - "585beac3f8": "ttl", - "5a9df5566f": "open menu", "5baf51c4d9": "open claude", "5d9ba08673": "copilot", "5fdf1dc2d1": "omp", - "6382fe9724": "npx", - "660528b048": "cost", - "68d03d9980": "vscode", - "6c2ce8457c": "file explorer", - "750420dd9a": "control", - "7887a2c262": "folder", - "7baf524b04": "workspace", - "7e9b556873": "skip", - "7edf4f69e2": "automation", "8436ff6f8e": "Proxy Bypass Rules", - "84c67d0108": "delete", - "86f54575c7": "autosave", "882c4896fd": "opencode", - "88d3df9ce9": "terminal", "8ea37a05bc": "agent", - "8f03d44672": "http_proxy", - "8fb00fcd05": "launcher", - "91a46caafc": "no_proxy", - "924a660a78": "cli", - "939b80f5fd": "timer", - "93f6ec5e70": "directory", - "95b63edde7": "claude", - "973ed6bfbf": "combined diff", "9b0bc30160": "pi", - "9bde064915": "subfolder", - "9c72990db8": "minimap", "9da6c875e5": "dock", - "9e86ccd05c": "version", - "9f8558233a": "confirm", - "a0014961ae": "scroll", "aea7d2cccb": "openclaude", - "b2601a778c": "cache", - "b2799ba622": "milliseconds", - "b65665703a": "support", - "b8093e9a93": "open in", - "b9096a44cf": "https_proxy", - "baa263d6d8": "agents", - "bda108e66c": "skill", - "bdfb6dc21b": "like", - "be24c7cd67": "split", "c29f23ab57": "HTTP Proxy", - "c56cb6f1c2": "network", "c61b14be7c": "grok", - "c9d8c1ce66": "release notes", - "c9d9636f24": "finder", - "ca812803ea": "recent tab order", - "ca86dd6e27": "dialog", - "d05f629d2c": "markdown", "db11502270": "Default Agent", - "dbeb1f348e": "command", - "df10666259": "worktree", - "e1ee631696": "editor", "e2da948f59": "Pre-select an AI coding agent in the new-workspace composer.", - "e3919429c0": "overview", "e3b1d42f95": "Proxy URL for Orca network requests and local terminal children.", - "e49e739a59": "download", - "e4fb4516d0": "star", "e55d62dfa4": "launchpad", - "e6b01c8e30": "feedback", "eb8946b2c9": "Hosts that should bypass the configured HTTP proxy.", - "ebf8f056b5": "zed", - "ec5049e510": "nested", - "f472e97440": "aider", - "f89a94773c": "update", - "f8f0ac213a": "sequential", - "fb4f338a3d": "path", - "fb84767421": "switch", - "fe62b3f09f": "ctrl" - } - }, - "git": { - "search": { - "035134fcd9": "worktree", - "0849b571fe": "up to date", - "0c75583ca9": "safely", - "16f53f7323": "gh", - "1d2fae1fa2": "git username", - "28192e3a63": "master", - "40f9b815fd": "api budget", - "4808f065b3": "gitlab", - "564942ffc5": "origin/main", - "65b69d9f80": "graphql", - "6ee3cfff02": "git diff", - "769ddd7f81": "custom", - "ab0e22c9f6": "refresh local main", - "b7e52124c7": "rate limit", - "bae91effdd": "fresh base", - "branchUpstream": "branch upstream", - "c41e345153": "behind main", - "changesFirst": "changes first", - "committedChanges": "committed changes", - "compareBase": "compare base", - "currentBranch": "current branch", - "d088806071": "github", - "d9f70d51a0": "stale main", - "de06e9d105": "base ref", - "defaultBranch": "default branch", - "defaultCompareBase": "default compare base", - "e3e9adde59": "main", - "ead733645f": "glab", - "f83c8937c4": "branch naming", - "gitChanges": "git changes", - "groupOrder": "group order", - "localChanges": "local changes", - "originMaster": "origin/master", - "repositoryDefault": "repository default", - "sourceControl": "source control", - "stagedFirst": "staged first", - "untrackedFirst": "untracked first", - "upstream": "upstream" - } - }, - "input": { - "search": { - "26c83b06c5": "linux", - "31ba58c8ae": "middle click", - "5fb84ba77f": "middle mouse", - "7059cfb00a": "clipboard", - "71905435dd": "x11", - "886597d6b3": "macos", - "b51d47ceb7": "input", - "c4440c3986": "paste", - "de51e18ee9": "primary selection", - "e25165320e": "editing", - "e5cd0e7a46": "selection" + "f472e97440": "aider" } }, "integrations": { "search": { - "03a7b275be": "ado", - "129fc59aa8": "gitea", - "20540996ef": "credentials", - "2ec2bd328c": "api token", - "33180e8c10": "self-hosted", - "371ee914d2": "merge request", - "3c3d3d8ffa": "connect", - "41ccade05c": "gh", - "50d20817f7": "bitbucket", - "581844769a": "mr", - "7319e3015b": "linear", - "7345b7c3e6": "atlassian", - "8c568d761c": "pull request", - "a626990bd2": "disconnect", - "af5ae87847": "access token", - "b38b5d27f1": "azure devops", - "b40cbe5de4": "glab", - "b79c21bd42": "github", - "b939695c69": "gitlab", - "c450244ad7": "integration", - "c97d58a0f3": "Bitbucket Cloud authentication via API token environment variables.", - "e1263dd748": "jira", - "ed63380247": "azure repos", - "faa0b5a0d9": "api key" + "c97d58a0f3": "Bitbucket Cloud authentication via API token environment variables." } }, "jira": { @@ -2255,268 +1702,19 @@ } }, "mobile": { - "emulator": { - "search": { - "04c5f5d901": "device", - "1ad6fb6230": "default device", - "1dc8c52ffa": "default iphone", - "25159de808": "mobile emulator", - "25d7bfbcd4": "udid", - "2bb2e09225": "mobile skill", - "2d67f708ce": "simulator", - "3211e7acf9": "xcrun", - "42bfab45d8": "availability", - "49727355a3": "iphone", - "64494f03c3": "emulator attach", - "6b6407dc1f": "emulator", - "6f728f1456": "emulator tap", - "7650063d17": "simctl", - "7c5a8a2bee": "xcode", - "84e5706975": "serve-sim", - "8ef0f08d36": "runtime", - "9353854ff3": "orca emulator", - "ab4814f3c5": "default simulator", - "ac0a985873": "emulator skill", - "b8ddd13195": "agent emulator", - "bbe4267416": "emulator type", - "bec7231663": "ipad", - "c5eca29310": "ios simulator", - "d4b7833894": "orca cli", - "ec3c4043fd": "default ipad", - "f8b871d655": "agent cli" - } - }, - "pane": { - "search": { - "126afc5dbd": "remote", - "16bff559a0": "tailnet", - "1802188b5d": "wifi", - "1f70d63998": "ip", - "2128a21096": "scan", - "356c31d6dc": "width", - "3a5e31e84b": "leave", - "3c1807a81a": "qr", - "4a0c826f3d": "code", - "5e8fda4d7f": "paired", - "6cd2bfdb0e": "restore", - "6db86f445f": "mobile", - "70f505f3c3": "lan", - "7b37c2e557": "network", - "7d01f93ec0": "connected", - "8015fd9523": "hold", - "82783d9b71": "devices", - "87711f4b8f": "vpn", - "905c65a308": "revoke", - "9e16be01d6": "background", - "a023683767": "interface", - "aa3f736042": "resize", - "ad08035c5f": "phone", - "b34ad5b3a7": "terminal", - "c690e3ee38": "tailscale", - "d0c89bc4a9": "overlay", - "dbccde3a60": "close", - "dd6e671aa9": "address", - "e518cbd61c": "pair", - "fadcbfdd99": "fit" - } - }, "settings": { "search": { - "0b7e585cb9": "scan", - "59b1d75fd1": "code", - "5d5af8e041": "iphone", - "6bfa001752": "apk", - "7e801801ac": "remote", - "87816d1c59": "qr", - "8d4ba0ef09": "beta", - "a7eececc1d": "android", - "b730ff7049": "experimental", - "cf2c93b479": "pair", - "e4f4daea0e": "relay", - "f213400800": "mobile", - "f4ed142753": "phone" + "b730ff7049": "experimental" } } }, - "notifications": { - "search": { - "079c29aeb5": "flac", - "193e1f107c": "task", - "3014ad1b8f": "ding", - "4ada6bfde9": "filtering", - "51ae2183e1": "desktop", - "5362074f19": "mp3", - "57e34a31cd": "wav", - "5f7472d3fb": "complete", - "6e08f78315": "audio", - "6ecb8418cb": "m4a", - "722face52f": "aac", - "72539aede4": "system", - "7fa07e9600": "agent", - "a2ab73b325": "attention", - "a4c3b29a3c": "focused", - "aa288005c3": "test", - "adbc3a0fcf": "native", - "ae0487f8fd": "bell", - "c638ae989d": "terminal", - "ca8faa40d7": "notifications", - "d16ae23645": "ogg", - "d58b64dddf": "volume", - "dc7d7c07cd": "sound", - "dd9d3e5f0f": "idle", - "ecdeff4993": "loudness", - "ef86a782cc": "bong", - "fa60d8e4ab": "suppress" - } - }, - "orchestration": { - "search": { - "08c65b12a2": "examples", - "13ba5c6cbd": "agents", - "21c28ccdf7": "coordinator", - "32c5098e7b": "claude", - "741dfc03fa": "worker", - "7ad948b714": "task", - "91fc8ab7e5": "coordination", - "9a5ebdca31": "messaging", - "a7f76b4ca7": "orchestration", - "c766a01978": "handoff", - "ca54c69806": "DAG", - "d86705ba77": "multi-agent", - "eee028ae14": "dispatch", - "f278fd04db": "codex", - "f5d39af41e": "child agents" - } - }, - "privacy": { - "search": { - "3922051573": "data", - "058550f6bc": "do_not_track", - "10124159f1": "privacy", - "1686c07fee": "support", - "27a27b2f63": "opt out", - "2b5a5c312f": "posthog", - "4104f6f0f3": "analytics", - "4d4bb76bf4": "opt in", - "5854a5c752": "ci", - "664f1a8984": "continuous integration", - "69637f4dc4": "orca_telemetry_disabled", - "77d3180def": "telemetry", - "79c319948b": "usage", - "83a6cd79b3": "do not track", - "94e04427f6": "env", - "b021b9cb81": "anonymous", - "c0494ff48a": "diagnostics", - "d8191ae5ca": "environment variable", - "e8bc614a18": "disable", - "ead1deded2": "share" - } - }, "providerAccountScope": { "localMac": "Local Mac" }, - "quick": { - "commands": { - "search": { - "0073cf8ce9": "terminal", - "0b78c4a165": "launch", - "1c5bdcd0f2": "repository", - "236d4cfac8": "quick", - "2d8aff42be": "run", - "3c316e6ef8": "yarn", - "89d2a9ad9f": "repo", - "8bf43c2dad": "global", - "a26ecdb77b": "snippet", - "b86c727100": "npm", - "b949a7c0a0": "pnpm", - "cfffa6cdb6": "commands", - "d07d130849": "shortcut", - "f58b92a48f": "project", - "fecb031823": "command" - } - } - }, "repository": { "search": { - "0432d2fb7c": "local", - "095fca94fe": "preset", - "0a3a582794": "env", - "130d76dc16": "rename", - "16dc7a4637": "model context protocol", - "19f58d6d89": "advanced", - "1d90a6cfbb": "both", - "1e73e840ff": "emoji", - "1ff4f12c0c": "directory", - "2011a6a4f2": "github issue command", - "26f42fe773": ".cursor/mcp.json", - "27733eb6c1": "favicon", - "3067595d82": "delete", - "343f0a508c": "mcp", - "3c180a251c": "link", - "4733ec2395": "../worktrees", - "491b05d6e6": "setup command", - "4b9a18a56d": "monorepo", - "4c17787d7b": "archive", - "4e2529722c": "directories", - "4f3c0230c2": "sparse", - "5590388dfa": "setup", - "58d8bca414": "relative", - "5e9445bbfd": "authoritative", - "5ff7fe1ade": "pull request", - "603c68b68c": "orca.yaml", - "6438a94c63": "project icon", - "6469de5368": "project", - "66b584bd6c": "issue command", - "6b80f7d3c8": "local settings scripts", - "6d8de2f090": "hex", - "7e228fc439": "symlinks", - "8068d8d0f1": "pr", - "80c490b012": "ask", - "84da7fa2d7": "node_modules", - "8655e3387b": "hooks", - "8d045419b1": "color", - "917dce844a": "branch name", - "92af66c7ce": "project name", - "9811f3d152": "branch", - "9cad92fe77": "orca.yaml hooks", - "9dc60d7f6d": "github", - "9f5ae26ccd": "presets", - "a1a4c51d58": "archive command", - "a31b43a7f8": "setup script", - "a325a89dff": "workspace path", - "a47f51127e": "source control", - "a69c5cbe90": "run by default", - "aa42616e3d": "checkout", - "apfs": "apfs", "availableHosts": "Available Hosts", "availableHostsDescription": "Hosts where this project is set up.", - "b2546efab5": "repository icon", - "bc7e504b8e": ".orca/issue-command", - "bf460fded8": "yaml", - "c06adcf136": "symlink", - "c1075178cf": "badge", - "c5e8bdbcbb": "skip by default", - "cb4b4de666": "avatar", - "cc876ca5f2": "repository", - "cd73b976d7": "repository name", - "cfad7ce5f3": "ai", - "clone": "clone", - "copy": "copy", - "d73fb47b45": ".claude/mcp.json", - "db11b337c4": ".claude.json", - "e760e3fae7": ".mcp.json", - "ec70364df2": "workflow", - "ed269fad69": "command source", - "eec39b3de6": "commit message", - "f1c53f2820": "worktree", - "f1e1bfa89f": "source", - "f3e6dee5fe": "worktree path", - "f41cef5083": "base ref", - "f9d84b7971": "setup run policy", - "fa3131f223": "model", - "fbfd2386e8": "archive script", - "fcb8fa8144": "shared", - "fff8834983": "prompt", "host": "host", "remote": "remote", "ssh": "ssh", @@ -2526,20 +1724,8 @@ "runtime": { "environments": { "search": { - "09568ccc65": "server", - "104f4d7dbd": "pairing", - "2bd988d041": "pairing code", "3517fb2ec0": "Active Server", - "45501ff2c3": "cloud", - "4575341c77": "Choose local desktop, add a saved remote Orca server, or generate a pairing URL.", - "5cd7dca3b8": "remote", - "772e3b4753": "vm", - "81444c4102": "pairing url", - "c6e5a03aa0": "dev box", - "d198440ce3": "runtime", - "d760866285": "client", - "ebd5369acf": "environment", - "f1575f1e09": "web client" + "4575341c77": "Choose local desktop, add a saved remote Orca server, or generate a pairing URL." } } }, @@ -2551,215 +1737,23 @@ "refreshLinks": "Refresh", "showButton": "Show Skills button" }, - "shortcuts": { - "search": { - "0ecba9aa5f": "keyboard", - "0ecfc47434": "conflict", - "0f8cb15582": "agent", - "4811a8264a": "terminal first", - "7e3fc707aa": "terminal", - "7f1b38f59a": "tui", - "afda131738": "orca first", - "ca6a0c2df7": "shortcut", - "f1adebbe8c": "shell" - } - }, "ssh": { "search": { - "00d1fda01a": "new", - "09395490af": "target", - "237b391f7c": "connection", - "2cd40ba0d0": "hosts", - "3b12e064a4": "import", - "5220501141": "config", "62826efbe9": "Add a new remote SSH target.", - "74c6d90d78": "Manage remote SSH targets.", - "7efd17e816": "ssh", - "8cb870b109": "test", - "8fb1cc87cc": "host", - "d41f296f64": "ping", - "d4bcd497c7": "remote", - "f7b6383aec": "add", - "f9493b80c0": "server" - } - }, - "tasks": { - "search": { - "11f001cdd4": "gitlab", - "2ec54bee51": "tasks", - "3d81c26d78": "source", - "412ec3c702": "linear", - "44083ae418": "display", - "5430396e11": "jira", - "58cda6f9c0": "hide", - "604d8e4089": "atlassian", - "apiKey": "api key", - "c10ac2125e": "github", - "cf0e3e0c2f": "provider", - "connect": "connect", - "setup": "setup", - "skill": "skill" + "74c6d90d78": "Manage remote SSH targets." } }, "terminal": { - "clipboard": { - "search": { - "043b32faa1": "ssh", - "10d73e22d3": "clipboard", - "2061d8db1a": "neovim", - "4043e294d2": "gnome", - "5fb3512e8c": "paste", - "5ffcd13c90": "tmux", - "62d1208b90": "osc 52", - "64533e30cc": "nvim", - "664789b73a": "auto", - "737cef6de1": "x11", - "797fdfe4ca": "select", - "9dfc125cd3": "osc52", - "9fda309db9": "fzf", - "a38508c419": "copy", - "c38c18be15": "selection", - "cf83ac3dbd": "linux", - "d106f44fb4": "remote", - "e87c6d776d": "automatic" - } - }, - "search": { - "015c82349f": "block", - "0838b3717b": "window", - "0a05629060": "recover", - "103cdb862f": "typography", - "10f9fb6fea": "settings", - "11fd3fbcf2": "ansi", - "18ce996647": "vertical", - "1ab57a0fbd": "mac", - "1abcf4d7de": "linux", - "20ce287cc6": "weight", - "24f7977756": "japanese", - "25f606d9e5": "blink", - "2ade3ea490": "config", - "33031c1465": "text size", - "34fe1af39d": "typing", - "35c2311a33": "jetbrains mono", - "38f1b4f4cb": "key", - "3982d88725": "history", - "411229c636": "light", - "4529806908": "setup", - "456da64d4d": "clear", - "46d99ef4bb": "opacity", - "4b4e80d850": "acceleration", - "4ba8623632": "palette", - "4cec42dbf7": "intl", - "4ed3e239a8": "boundary", - "4f7f8f28ca": "transparency", - "54a9b3725b": "horizontal", - "56fff3d113": "memory", - "674b7c8436": "color", - "6892fb1019": "restart", - "6b659fff2a": "script", - "6c2f9f05c8": "vibrancy", - "6c4c85ba43": "dimming", - "6cddc858ba": "webgl", - "6ded6297fe": "iosevka", - "6eaf7ee0e4": "cursor", - "71eb45e293": "blur", - "7286cd2566": "word", - "7341e3d00e": "line height", - "7718d70356": "preview", - "781f49d942": "divider", - "7a48c7715b": "workspace", - "7ab424c4d3": "ligature", - "7ace5beec9": "meta", - "7d924d870d": "graphics", - "7db59c4738": "alpha", - "7f7640c29e": "fira code", - "846a7a1204": "pane", - "88561b3499": "frozen", - "920573d65b": "kill all", - "98059d0944": "backslash", - "983d45cf4c": "compose", - "9c35f56625": "yen", - "9f2dda133c": "pty", - "a16224d16a": "calt", - "a3e5297c10": "kill", - "a6e9dcc829": "bar", - "a8d2784214": "manage", - "abaa24752d": "keyboard", - "afc8d5f790": "ligatures", - "affb14efd4": "selection", - "b0bb76ae6b": "font", - "b2f52cb96c": "spacing", - "b37edfc65a": "option", - "b3b94cfcb5": "international", - "b495dc6a9f": "jis", - "b5116e7b12": "follows", - "b872de3926": "location", - "bc7ae1f7c0": "rendering", - "c047f398cc": "launch", - "c4427dc5ff": "alt", - "cde233f5da": "scrollback", - "d1fa00a9cb": "hover", - "d2a366c7f9": "double-click", - "d4aeafac10": "separator", - "d4daf4f612": "unfreeze", - "d5e6c7fab1": "font features", - "d802a578bf": "sessions", - "d8bd6182b8": "override", - "d8d6f7a3c5": "macos", - "da864e6cec": "light mode", - "db82cb13b0": "gpu", - "dd4f6cb541": "german", - "de7bc1d5f5": "split", - "e3aeea308e": "cascadia code", - "e8baf0d12c": "padding", - "ea364ce6e4": "mouse", - "ee611ae238": "hide", - "eefd1d8332": "underline", - "f036794286": "active", - "f25d948664": "margin", - "f35400f7e8": "daemon", - "f44643328e": "tab", - "f5d1e3d472": "focus", - "f637a7dee9": "thickness", - "f6dd9ff606": "background", - "f785374072": "dark", - "fae142a354": "readline", - "fd6c24313d": "new", - "fffa9ab980": "renderer", - "fffdff40a7": "buffer", - "rows": "rows", - "theme_target": { - "keyword_editing": "editing", - "keyword_target": "target" - } - }, "windows": { "search": { "02c772582a": "linux", - "04994f6929": "default", - "07ec155fb6": "bash.exe", - "12519edb5d": "command prompt", "1f402b3651": "WSL Distribution", - "28ff08ed35": "windows", "2b4a340ce0": "distribution", - "2d99cd91be": "powershell", - "4af2f7526e": "version", - "4d09141a42": "context menu", "4ee2579c32": "ubuntu", "5074ad8b5f": "distro", - "591912177b": "git bash", - "5a2db98d23": "bash", - "6cd20b9e64": "cmd", "6e3adf4cba": "wsl", - "768613e483": "powershell 7", - "7c7056940a": "shell", "978457945b": "Choose which WSL distribution new WSL terminals and local agent scans use.", - "d414022016": "pwsh", - "d57f870938": "advanced", - "e55186fe2b": "right click", - "e7d2793b03": "terminal", - "fc564eadaf": "debian", - "fcfa53920b": "paste" + "fc564eadaf": "debian" } } }, @@ -2789,31 +1783,6 @@ "imported_other": "Imported {{value0}} themes", "over_limit_one": "Importing these themes would exceed the {{value0}} custom terminal theme limit. Deselect 1 new theme and try again.", "over_limit_other": "Importing these themes would exceed the {{value0}} custom terminal theme limit. Deselect {{value1}} new themes and try again." - }, - "voice": { - "pane": { - "search": { - "04c25a6fb0": "openai", - "064a9bd94a": "hold", - "080202facb": "model", - "089d31a45b": "dictation", - "10d45a9fce": "stt", - "2d206de105": "api key", - "322d457a0d": "transcription", - "3d8b853963": "speech", - "6fa48bcd41": "toggle", - "7640ed9848": "voice", - "931b1a9e53": "push to talk", - "b9dee49cd7": "download", - "d86f5600da": "mode", - "e360027a65": "microphone", - "f6e0dfa61c": "cloud", - "micAirpods": "airpods", - "micDefault": "system default", - "micDevice": "device", - "micInput": "input" - } - } } }, "shared": { @@ -3311,9 +2280,6 @@ "e3ff145b98": "Split Left", "f7c3d7d5af": "Split Right" }, - "QuickLaunchButton": { - "ec2adf093e": "Launch {{value0}} in a new terminal" - }, "SortableTabContextMenu": { "0ce4bae39d": "Split Left", "21132389e9": "Split Right", @@ -3673,13 +2639,8 @@ "thinking": "Thinking", "toggleDetails": "Toggle turn details", "workedFor": "Worked for {{value0}}", - "working": "Working…", "workingFor": "Working for {{value0}}" }, - "toggle": { - "showChat": "Show chat view", - "showTerminal": "Show terminal" - }, "tool": { "countN": "{{value0}} tool calls", "countOne": "1 tool call", @@ -3696,11 +2657,13 @@ "usedOneSummary": "Used 1 tool" } }, - "tab": { - "bar": { - "SortableTabContextMenu": { - "switchToChatView": "Switch to chat view", - "switchToTerminalView": "Switch to terminal view" + "onboarding": { + "integrations": { + "capabilities": { + "browseIssues": "Browse GitHub issues and pull requests in the Tasks view without leaving Orca", + "managePullRequests": "Read, comment on, and merge pull requests without leaving Orca", + "reviewStatus": "See issue state, review status, and CI checks on every worktree", + "startWorkspaceFromIssue": "Start a workspace from any GitHub issue or pull request, prefilled with its title and context" } } }, @@ -3732,16 +2695,6 @@ "readyOneOne": "1 workspace found, with 1 cleanup suggestion." } } - }, - "onboarding": { - "integrations": { - "capabilities": { - "browseIssues": "Browse GitHub issues and pull requests in the Tasks view without leaving Orca", - "managePullRequests": "Read, comment on, and merge pull requests without leaving Orca", - "reviewStatus": "See issue state, review status, and CI checks on every worktree", - "startWorkspaceFromIssue": "Start a workspace from any GitHub issue or pull request, prefilled with its title and context" - } - } } }, "dashboard": { @@ -3792,15 +2745,6 @@ } }, "settings": { - "appearance": { - "language": { - "chinese": "中文(简体)", - "english": "English", - "japanese": "日本語", - "korean": "한국어", - "spanish": "Español" - } - }, "browser": { "clientHostedRemote": { "description": "Render remote workspace pages on this desktop; network traffic still goes through the remote host. Applies to new pages only.", diff --git a/src/renderer/src/i18n/locales/en.json b/src/renderer/src/i18n/locales/en.json index 1f880f15eb9..2b9bbc9ef24 100644 --- a/src/renderer/src/i18n/locales/en.json +++ b/src/renderer/src/i18n/locales/en.json @@ -5780,7 +5780,8 @@ "locked": "This workspace is locked by Git. Run git worktree unlock from its repository, then retry deletion.", "lockedReason": "This workspace is locked by Git. Git reported: {{value0}}. Run git worktree unlock from its repository, then retry deletion.", "unstoppedPty": "Orca could not confirm every terminal in this workspace has exited, so it stopped before deleting any files. Use Force Delete to remove it anyway.", - "unstoppedPtyLive": "This workspace still has running terminals, so Orca stopped before deleting any files. Force Delete will kill them and discard any uncommitted work they hold." + "unstoppedPtyLive": "This workspace still has running terminals, so Orca stopped before deleting any files. Force Delete will kill them and discard any uncommitted work they hold.", + "runningAgentSession": "This workspace still has running agent sessions, so Orca stopped before deleting any files. Force Delete will close them and discard any work they hold." } } }, @@ -6414,7 +6415,6 @@ "8d61637a77": "MiniMax cookie saved.", "b43e761fe5": "MiniMax cookie update failed.", "5d63bbfbec": "MiniMax", - "15e831350e": "Configure MiniMax usage tracking from platform.minimax.io.", "21d6eb141e": "MiniMax Session Cookie", "33bba5ad83": "Paste your MiniMax session cookie for local rate-limit fetching.", "73ea15f24b": "Saved", @@ -6422,7 +6422,6 @@ "566d9a99ab": "_token=…; minimax_group_id_v2=…", "f38b9cc4bd": "Replace", "590a3130f9": "Save", - "79418c782a": "Open platform.minimax.io/console/usage in your browser, sign in, then copy the Cookie request header from DevTools (Network → any remains request → Cookie).", "9dd50d3f75": "Advanced", "174fb408f9": "Leave these defaults alone unless MiniMax usage refresh points at the wrong workspace or model.", "bf160bb6c0": "Group ID override", @@ -6433,14 +6432,11 @@ "3c92b0d31c": "general", "0d8e77bc40": "Open console", "0b8c1c7e02": "Stored locally", - "1fd1b1b6b4": "Cookie not set", - "5e08b0fe57": "Stored locally and sent only to platform.minimax.io for usage refreshes.", "43d7a45b97": "How to copy", "b8a4f21c3e": "Paste the Cookie header from DevTools", "53f7b8c7a2": "Last refresh: {{value0}}", "31d24a4e87": "Cookie expires when you sign out in the browser.", "3a30aaf526": "just now", - "f5d8d2a6a1": "Open platform.minimax.io/console/usage in your browser and sign in.", "24560fe830": "Open DevTools.", "4cab0fa42d": "Go to the Network tab and enable Preserve log.", "bee4e63e1c": "Reload the page.", @@ -6448,7 +6444,6 @@ "435df0ee51": "Under Request Headers, copy the Cookie value.", "7492fb3bba": "Paste it here and click Save.", "9fec52de4b": "How to copy the cookie", - "4e32e030b2": "Stored locally. Orca sends it only to platform.minimax.io for usage refreshes.", "remoteServerFallback": "the remote server", "loadAccountsFailed": "Could not load provider accounts.", "remoteScopeAccounts": "Showing accounts managed by {{value0}}. Add or re-authenticate accounts on that server.", @@ -6463,7 +6458,31 @@ "codexConfigSyncMissingSource": "Codex is still using the settings it last synced because {{value0}} is missing. Restore that file to resume syncing.", "codexConfigSyncBlankSource": "Codex is still using the settings it last synced because {{value0}} is empty. That is expected while a synced folder finishes downloading.", "codexConfigSyncManagedHomeUnavailable": "Orca could not read this account’s Codex files just now, so settings may not be syncing. This usually clears on its own — antivirus or a backup tool briefly locks them.", - "codexConfigSyncUnreadableSource": "Codex is still using the settings it last synced because {{value0}} could not be read. Check that file's permissions." + "codexConfigSyncUnreadableSource": "Codex is still using the settings it last synced because {{value0}} could not be read. Check that file's permissions.", + "d6f1b9b6a2": "MiniMax API key is required.", + "7c5d8a4e1b": "MiniMax API key was not saved.", + "4d2c7b9e83": "MiniMax API key saved.", + "f8a4b9d210": "MiniMax endpoint", + "0b3a9f6c2e": "Pick the host that matches your account. Both overseas (platform.minimax.io) and China (platform.minimaxi.com) accept either a session cookie or an API key.", + "83b6a1f7c4": "MiniMax API key", + "4f2c8a7e1b": "Paste your MiniMax API key", + "a7b1e3c5d2": "Forget key", + "usageTracking": "Configure MiniMax usage tracking for your account.", + "credentialsNotSet": "Credentials not set", + "selectedEndpointStorage": "Stored locally and sent to the selected MiniMax endpoint for usage refreshes.", + "endpointOverseas": "Overseas (platform.minimax.io)", + "endpointChina": "China (platform.minimaxi.com)", + "openSelectedConsole": "Open {{url}} in your browser and sign in.", + "cookieSelectedEndpoint": "Stored locally and sent to the selected MiniMax endpoint for usage refreshes.", + "copySelectedConsoleCookie": "Open the selected console, sign in, then copy the Cookie request header from DevTools (Network → any remains request → Cookie).", + "apiKeySelectedEndpoint": "Paste the API key from your MiniMax console → API keys. Stored locally and sent to the selected MiniMax endpoint for usage refreshes. The API key takes priority over the cookie.", + "apiKeyInstructions": "Copy the API key from your MiniMax console → API keys. A saved API key takes priority over the cookie; use Forget key to switch back to the cookie.", + "15e831350e": "Configure MiniMax usage tracking from platform.minimax.io.", + "1fd1b1b6b4": "Cookie not set", + "4e32e030b2": "Stored locally. Orca sends it only to platform.minimax.io for usage refreshes.", + "5e08b0fe57": "Stored locally and sent only to platform.minimax.io for usage refreshes.", + "79418c782a": "Open platform.minimax.io/console/usage in your browser, sign in, then copy the Cookie request header from DevTools (Network → any remains request → Cookie).", + "f5d8d2a6a1": "Open platform.minimax.io/console/usage in your browser and sign in." }, "AdvancedPane": { "40b29e0bf3": "Restart", @@ -8830,14 +8849,18 @@ "b84a5b0c8a": "Choose whether provider accounts are inspected and added on this device or in WSL.", "d09fb5ca92": "Account Location", "733f9e2a93": "MiniMax Usage", - "f8374c3151": "Paste your platform.minimax.io session cookie for local rate-limit fetching.", + "f8374c3151": "Configure MiniMax usage tracking. Pick the overseas or China endpoint, then paste a session cookie or save an API key that works on either host.", "f4a8c2e1b7": "Grok (xAI) Usage", "e3b7d1f9a2": "OAuth sign-in via Grok CLI (grok login) for weekly credit usage.", "d2c6a0e8f1": "grok", "c1b5f9d7e0": "xai", "b0a4e8c6d9": "oauth", "a9f3d7b5c8": "login", - "d16378a88f": "minimax" + "d16378a88f": "minimax", + "b2c4e7f1a8": "endpoint", + "3a9b6d2c4e": "api key", + "5d8f1a3b7c": "china", + "7e2a4b8c1d": "overseas" } }, "advanced": { @@ -9148,6 +9171,7 @@ }, "terminalLinkActions": { "terminal": "terminal", + "chat": "chat", "click": "click", "actions": "actions", "popover": "popover", @@ -11028,7 +11052,9 @@ "e6dadc1e2b": "Monthly usage", "75e396bf42": "Included monthly usage for Grok unified-billing accounts.", "b36fa2c908": "Signed in. Orca reads the Grok CLI session stored on disk.", - "f08c41de73": "Session expired — run grok on the computer running Orca and wait for it to start. If prompted, complete sign-in, then click Refresh usage. No chat message is needed." + "f08c41de73": "Session expired — run grok on the computer running Orca and wait for it to start. If prompted, complete sign-in, then click Refresh usage. No chat message is needed.", + "0bb18642b7": "Usage", + "a8f4139350": "Grok reported no usage percentage for this account." }, "AppearanceWindowSidebarSection": { "usagePercentageDisplayUsed": "Used", @@ -11176,8 +11202,8 @@ "descriptionOrca": "Links open in your system browser. When enabled, {{chord}}+click opens one in Orca's built-in browser instead." }, "BrowserTerminalLinkActionsSetting": { - "title": "Show terminal link actions", - "description": "Show available actions when you click a terminal link. Turn this off to require {{modifier}}-click." + "title": "Show link actions", + "description": "Show available actions when you click a link in the terminal or a chat transcript. Turn this off to require {{modifier}}-click in the terminal." }, "PluginConsentDialog": { "workerTrust": "Background worker — runs its own process", @@ -17076,11 +17102,6 @@ "deny": "Deny" }, "launchPromptNotDelivered": "Not delivered — check the terminal", - "orchestrationPaused": { - "label": "Orchestration paused", - "message": "Structured Chat blocks terminal prompts and sends. Orchestration messages remain queued; switch to Terminal, then check the Orca inbox with", - "command": "orca orchestration check" - }, "structuredSessionCloseFailed": "Could not close this chat session", "structuredSessionLaunchFailed": "Could not open {{value0}} chat", "structuredSessionLaunchPending": "Starting {{value0}} chat…", diff --git a/src/renderer/src/i18n/locales/zh.json b/src/renderer/src/i18n/locales/zh.json index 959296cd869..17f60476de0 100644 --- a/src/renderer/src/i18n/locales/zh.json +++ b/src/renderer/src/i18n/locales/zh.json @@ -5451,7 +5451,6 @@ "8d61637a77": "MiniMax Cookie 已保存。", "b43e761fe5": "MiniMax Cookie 更新失败。", "5d63bbfbec": "MiniMax", - "15e831350e": "从 platform.minimax.io 配置 MiniMax 使用量跟踪。", "21d6eb141e": "MiniMax 会话 Cookie", "33bba5ad83": "粘贴 MiniMax 会话 Cookie 以在本地获取速率限制。", "73ea15f24b": "已保存", @@ -5459,7 +5458,6 @@ "566d9a99ab": "_token=…; minimax_group_id_v2=…", "f38b9cc4bd": "替换", "590a3130f9": "保存", - "79418c782a": "在浏览器中打开 platform.minimax.io/console/usage 并登录,然后从 DevTools(网络 → 任一 remains 请求 → Cookie)复制 Cookie 请求头。", "9dd50d3f75": "高级", "174fb408f9": "除非 MiniMax 使用量刷新指向了错误的工作区或模型,否则请保持这些默认值。", "bf160bb6c0": "Group ID 覆盖", @@ -5470,14 +5468,11 @@ "3c92b0d31c": "general", "0d8e77bc40": "打开控制台", "0b8c1c7e02": "已存储在本地", - "1fd1b1b6b4": "未设置 Cookie", - "5e08b0fe57": "存储在本地,仅发送到 platform.minimax.io 以刷新使用量。", "43d7a45b97": "如何复制", "b8a4f21c3e": "粘贴来自 DevTools 的 Cookie 请求头", "53f7b8c7a2": "上次刷新: {{value0}}", "31d24a4e87": "在浏览器中退出登录后,Cookie 将过期。", "3a30aaf526": "刚刚", - "f5d8d2a6a1": "在浏览器中打开 platform.minimax.io/console/usage 并登录。", "24560fe830": "打开 DevTools。", "4cab0fa42d": "转到“网络”选项卡并启用“保留日志”。", "bee4e63e1c": "重新加载页面。", @@ -5485,7 +5480,6 @@ "435df0ee51": "在请求头中,复制 Cookie 的值。", "7492fb3bba": "在此粘贴并点击保存。", "9fec52de4b": "如何复制 Cookie", - "4e32e030b2": "存储在本地。Orca 仅将其发送到 platform.minimax.io 以刷新使用量。", "remoteServerFallback": "远程服务器", "loadAccountsFailed": "无法加载提供商账户。", "remoteScopeAccounts": "正在显示由 {{value0}} 管理的账户。请在该服务器上添加或重新验证账户。", @@ -5500,7 +5494,31 @@ "codexConfigSyncMissingSource": "由于缺少 {{value0}},Codex 仍在使用上次同步的设置。请恢复该文件以恢复同步。", "codexConfigSyncBlankSource": "由于 {{value0}} 为空,Codex 仍在使用上次同步的设置。同步文件夹完成下载前出现这种情况是正常的。", "codexConfigSyncManagedHomeUnavailable": "Orca 暂时无法读取此账户的 Codex 文件,因此设置可能尚未同步。这通常会自行恢复——防病毒软件或备份工具可能只是短暂锁定了这些文件。", - "codexConfigSyncUnreadableSource": "由于无法读取 {{value0}},Codex 仍在使用上次同步的设置。请检查该文件的权限。" + "codexConfigSyncUnreadableSource": "由于无法读取 {{value0}},Codex 仍在使用上次同步的设置。请检查该文件的权限。", + "d6f1b9b6a2": "请填写 MiniMax API 密钥。", + "7c5d8a4e1b": "MiniMax API 密钥未保存。", + "4d2c7b9e83": "MiniMax API 密钥已保存。", + "f8a4b9d210": "MiniMax 端点", + "0b3a9f6c2e": "选择与你的账户匹配的端点。海外 (platform.minimax.io) 和中国 (platform.minimaxi.com) 均支持会话 Cookie 或 API 密钥。", + "83b6a1f7c4": "MiniMax API 密钥", + "4f2c8a7e1b": "粘贴你的 MiniMax API 密钥", + "a7b1e3c5d2": "忘记密钥", + "usageTracking": "为你的账户配置 MiniMax 用量跟踪。", + "credentialsNotSet": "尚未设置凭据", + "selectedEndpointStorage": "保存在本地,并发送到所选的 MiniMax 端点以刷新用量。", + "openSelectedConsole": "在浏览器中打开 {{url}} 并登录。", + "cookieSelectedEndpoint": "保存在本地,并发送到所选的 MiniMax 端点以刷新用量。", + "copySelectedConsoleCookie": "打开所选控制台并登录,然后从开发者工具复制 Cookie 请求标头(Network → 任意 remains 请求 → Cookie)。", + "apiKeySelectedEndpoint": "粘贴 MiniMax 控制台 → API 密钥中的密钥。保存在本地,并发送到所选的 MiniMax 端点以刷新用量。API 密钥优先于 Cookie。", + "apiKeyInstructions": "从 MiniMax 控制台 → API 密钥中复制密钥。已保存的 API 密钥优先于 Cookie;使用“忘记密钥”切换回 Cookie。", + "endpointOverseas": "海外 (platform.minimax.io)", + "endpointChina": "中国 (platform.minimaxi.com)", + "15e831350e": "从 platform.minimax.io 配置 MiniMax 使用量跟踪。", + "1fd1b1b6b4": "未设置 Cookie", + "4e32e030b2": "存储在本地。Orca 仅将其发送到 platform.minimax.io 以刷新使用量。", + "5e08b0fe57": "存储在本地,仅发送到 platform.minimax.io 以刷新使用量。", + "79418c782a": "在浏览器中打开 platform.minimax.io/console/usage 并登录,然后从 DevTools(网络 → 任一 remains 请求 → Cookie)复制 Cookie 请求头。", + "f5d8d2a6a1": "在浏览器中打开 platform.minimax.io/console/usage 并登录。" }, "AdvancedPane": { "40b29e0bf3": "重新启动", @@ -7738,13 +7756,17 @@ "b84a5b0c8a": "选择是否在此设备上或 WSL 中检查和添加提供商账户。", "d09fb5ca92": "账户位置", "733f9e2a93": "MiniMax 使用情况", - "f8374c3151": "粘贴 platform.minimax.io 会话 Cookie 以在本地获取速率限制。", + "f8374c3151": "配置 MiniMax 用量跟踪。选择海外或中国端点,然后粘贴会话 Cookie 或保存适用于该端点的 API 密钥。", "f4a8c2e1b7": "Grok (xAI) 使用情况", "e3b7d1f9a2": "通过 Grok CLI(grok login)OAuth 登录以查看每周额度使用量。", "d2c6a0e8f1": "grok", "c1b5f9d7e0": "xai", "b0a4e8c6d9": "OAuth", - "a9f3d7b5c8": "登录" + "a9f3d7b5c8": "登录", + "b2c4e7f1a8": "端点", + "3a9b6d2c4e": "API 密钥", + "5d8f1a3b7c": "中国", + "7e2a4b8c1d": "海外" } }, "advanced": { diff --git a/src/renderer/src/lib/agent-launch-routing.ts b/src/renderer/src/lib/agent-launch-routing.ts index 6f6eeb14e7b..17cb95a43d0 100644 --- a/src/renderer/src/lib/agent-launch-routing.ts +++ b/src/renderer/src/lib/agent-launch-routing.ts @@ -1,17 +1,21 @@ import type { GlobalSettings } from '../../../shared/global-settings-types' import type { ProjectExecutionRuntimeResolution } from '../../../shared/project-execution-runtime' -import { isAgentSessionHandleProvider } from '../../../shared/agent-session-provider-handle' -import { STRUCTURED_AGENT_SESSION_RUNTIME_CAPABILITY } from '../../../shared/protocol-version' -import type { TuiAgent } from '../../../shared/tui-agent' import { - getTuiAgentDefaultArgs, - getTuiAgentDefaultEnv -} from '../../../shared/tui-agent-launch-defaults' + prefersStructuredNativeChatByDefault, + resolveStructuredNativeChatSupport +} from '../../../shared/structured-native-chat-launch-route' +import type { TuiAgent } from '../../../shared/tui-agent' import { decideInitialAgentTabViewMode, type NativeChatLaunchPromptDelivery } from '@/lib/native-chat-initial-view-mode' +export { + hasExplicitTuiAgentArgs, + hasExplicitTuiLaunchCustomization, + hasSemanticallyNonEmptyAgentArgs +} from '../../../shared/tui-agent-launch-customization' + export type AgentLaunchRoute = 'structured-native-chat' | 'legacy-native-chat' | 'terminal-tui' export type AgentLaunchRoutingInput = { @@ -37,39 +41,6 @@ export type AgentLaunchRoutingInput = { initialSessionOptions?: Readonly> } -export function hasExplicitTuiLaunchCustomization( - settings: - | Pick - | null - | undefined, - agent: TuiAgent -): boolean { - const configuredArgs = settings?.agentDefaultArgs?.[agent] - const configuredEnv = settings?.agentDefaultEnv?.[agent] - const defaultEnv = getTuiAgentDefaultEnv(agent) - const envIsCustomized = - configuredEnv !== undefined && - (Object.keys(configuredEnv).length !== Object.keys(defaultEnv).length || - Object.entries(configuredEnv).some(([key, value]) => defaultEnv[key] !== value)) - return ( - Boolean(settings?.agentCmdOverrides?.[agent]?.trim()) || - hasExplicitTuiAgentArgs(agent, configuredArgs) || - envIsCustomized - ) -} - -export function hasSemanticallyNonEmptyAgentArgs(value: string | null | undefined): boolean { - return Boolean(value?.trim()) -} - -export function hasExplicitTuiAgentArgs( - agent: TuiAgent, - value: string | null | undefined -): boolean { - const trimmed = value?.trim() ?? '' - return trimmed.length > 0 && trimmed !== getTuiAgentDefaultArgs(agent).trim() -} - export function resolveAgentLaunchRoute(input: AgentLaunchRoutingInput): AgentLaunchRoute { const initialViewMode = decideInitialAgentTabViewMode({ experimentalNativeChat: input.settings?.experimentalNativeChat, @@ -82,25 +53,19 @@ export function resolveAgentLaunchRoute(input: AgentLaunchRoutingInput): AgentLa if (initialViewMode !== 'chat') { return 'terminal-tui' } - if (input.settings?.experimentalStructuredNativeChat !== true) { + if (!prefersStructuredNativeChatByDefault(input.settings)) { return 'legacy-native-chat' } - - const projectRuntime = input.projectRuntime - const runtimeRefused = - projectRuntime?.status === 'repair-required' || projectRuntime?.runtime.kind === 'wsl' - const structuredSupported = - isAgentSessionHandleProvider(input.agent) && - input.promptDelivery !== 'draft' && - input.workspaceKind !== 'floating' && - input.requiresTuiLaunchCustomization !== true && - input.executionHostId === 'local' && - // Codex's Windows refusal is deliberate and settled elsewhere, so it stays a client-side - // answer. Claude's is measured by the executing host at create time (agentSession.createSupport) - // because only that host knows whether it can read a provider child's start time. - (input.agent !== 'codex' || input.platform !== 'win32') && - !runtimeRefused && - input.hostCapabilities.includes(STRUCTURED_AGENT_SESSION_RUNTIME_CAPABILITY) - - return structuredSupported ? 'structured-native-chat' : 'legacy-native-chat' + return resolveStructuredNativeChatSupport({ + agent: input.agent, + executionHostId: input.executionHostId, + platform: input.platform, + hostCapabilities: input.hostCapabilities, + workspaceKind: input.workspaceKind, + projectRuntime: input.projectRuntime, + isDraftPrompt: input.promptDelivery === 'draft', + requiresTuiLaunchCustomization: input.requiresTuiLaunchCustomization + }).supported + ? 'structured-native-chat' + : 'legacy-native-chat' } diff --git a/src/renderer/src/lib/browser-page-palette-activation.test.ts b/src/renderer/src/lib/browser-page-palette-activation.test.ts index bf5e94961b4..62b99146a60 100644 --- a/src/renderer/src/lib/browser-page-palette-activation.test.ts +++ b/src/renderer/src/lib/browser-page-palette-activation.test.ts @@ -195,6 +195,47 @@ describe('activateBrowserPagePaletteResult', () => { }) }) + it('rejects colliding child ids before mutating either host', () => { + seedStore({ + worktreesByRepo: { + 'repo-1': [makeWorktree({ hostId: 'ssh:host-1' })], + 'repo-2': [makeWorktree({ repoId: 'repo-2', hostId: 'ssh:host-2' })] + }, + unifiedTabsByWorktree: { + 'wt-1': [ + makeBrowserTab({ executionHostId: 'ssh:host-1', groupId: 'group-host-1' }), + makeBrowserTab({ executionHostId: 'ssh:host-2', groupId: 'group-host-2' }) + ] + }, + groupsByWorktree: { + 'wt-1': [makeGroup({ id: 'group-host-1' }), makeGroup({ id: 'group-host-2' })] + } + }) + + const before = useAppStore.getState() + expect(activateBrowserPagePaletteResult({ ...target, executionHostId: 'ssh:host-2' })).toEqual({ + status: 'failed', + reason: 'missing-tab' + }) + expect(useAppStore.getState()).toBe(before) + expect(mocks.activateAndRevealWorktree).not.toHaveBeenCalled() + }) + + it('keeps browser workspaces with distinct unified tabs in multiple groups activatable', () => { + seedStore({ + unifiedTabsByWorktree: { + 'wt-1': [makeBrowserTab(), makeBrowserTab({ id: 'second-view', groupId: 'group-2' })] + }, + groupsByWorktree: { + 'wt-1': [ + makeGroup(), + makeGroup({ id: 'group-2', activeTabId: 'second-view', tabOrder: ['second-view'] }) + ] + } + }) + expect(activateBrowserPagePaletteResult(target).status).toBe('activated') + }) + it('activates pages in remote folder workspaces', () => { const worktreeId = folderWorkspaceKey('folder-1') seedStore({ diff --git a/src/renderer/src/lib/browser-page-palette-activation.ts b/src/renderer/src/lib/browser-page-palette-activation.ts index 5ba04e6cd3e..72f76722cd7 100644 --- a/src/renderer/src/lib/browser-page-palette-activation.ts +++ b/src/renderer/src/lib/browser-page-palette-activation.ts @@ -1,5 +1,8 @@ import { useAppStore } from '@/store' -import { activateBrowserWorkspaceTab } from '@/lib/browser-workspace-tab-activation' +import { + activateBrowserWorkspaceTab, + getActivatableBrowserWorkspaceTab +} from '@/lib/browser-workspace-tab-activation' import type { ExecutionHostId } from '../../../shared/execution-host' import { isBlankBrowserUrl } from './browser-palette-search' import { activateAndRevealWorktree } from './worktree-activation' @@ -29,18 +32,21 @@ export function activateBrowserPagePaletteResult({ worktreeId }: BrowserPagePaletteActivationTarget): BrowserPagePaletteActivationResult { const initialState = useAppStore.getState() - const page = (initialState.browserPagesByWorkspace[workspaceId] ?? []).find( - (candidate) => candidate.id === pageId - ) - const workspace = (initialState.browserTabsByWorktree[worktreeId] ?? []).find( - (candidate) => candidate.id === workspaceId - ) const worktree = initialState.getKnownWorktreeById(worktreeId, executionHostId) // Why worktree first: removing a worktree also purges its browser workspaces // and pages, so a page-first check would report a dead workspace as a stale page. if (!worktree) { return { status: 'failed', reason: 'missing-worktree' } } + const page = (initialState.browserPagesByWorkspace[workspaceId] ?? []).find( + (candidate) => + candidate.id === pageId && + candidate.workspaceId === workspaceId && + candidate.worktreeId === worktreeId + ) + const workspace = (initialState.browserTabsByWorktree[worktreeId] ?? []).find( + (candidate) => candidate.id === workspaceId && candidate.worktreeId === worktreeId + ) if (!page || !workspace) { return { status: 'failed', reason: 'missing-page' } } @@ -52,6 +58,11 @@ export function activateBrowserPagePaletteResult({ : 'webview' const targetHostId = executionHostId ?? worktree.hostId + if ( + !getActivatableBrowserWorkspaceTab({ worktreeId, workspaceId, executionHostId: targetHostId }) + ) { + return { status: 'failed', reason: 'missing-tab' } + } const activated = activateAndRevealWorktree( worktree.id, targetHostId ? { executionHostId: targetHostId } : {} @@ -66,7 +77,8 @@ export function activateBrowserPagePaletteResult({ !activateBrowserWorkspaceTab({ worktreeId: worktree.id, workspaceId: workspace.id, - pageId + pageId, + ...(targetHostId ? { executionHostId: targetHostId } : {}) }) ) { return { status: 'failed', reason: 'missing-tab' } diff --git a/src/renderer/src/lib/browser-palette-page-entries.test.ts b/src/renderer/src/lib/browser-palette-page-entries.test.ts index c05a29a40bd..a05ae42c6d5 100644 --- a/src/renderer/src/lib/browser-palette-page-entries.test.ts +++ b/src/renderer/src/lib/browser-palette-page-entries.test.ts @@ -210,13 +210,7 @@ describe('buildSearchableBrowserPages', () => { ]) }) - it('re-hosts a same-id page entry when the sibling row is missing from the catalog', () => { - // Why: host qualification is gated on both same-id rows being present. With one reaped, a - // local-stamped tab still renders but carries the surviving row's host — so a wrong-host - // Cmd-J activation means the catalog lost a row, not that host qualification regressed. - // This characterizes today's fallback, it does not bless it: overriding a tab's own 'local' - // stamp may be the wrong answer, and changing it is tracked as the unified-tab-host-ownership - // follow-up. Update this expectation with that change rather than treating it as a contract. + it('does not re-host a tab whose stamped owner is absent from the catalog', () => { const sharedId = 'repo-shared::/workspace' const remote = makeWorktree({ id: sharedId, hostId: 'runtime:host-b' }) const entries = buildSearchableBrowserPages({ @@ -239,9 +233,7 @@ describe('buildSearchableBrowserPages', () => { activeTabType: 'terminal' }) - expect(entries.map((entry) => [entry.page.id, entry.executionHostId])).toEqual([ - ['page-local', 'runtime:host-b'] - ]) + expect(entries).toEqual([]) }) it('does not route one ambiguous legacy browser bucket to both hosts', () => { @@ -265,6 +257,25 @@ describe('buildSearchableBrowserPages', () => { ).toEqual([]) }) + it('omits a browser row whose backing tab id is duplicated', () => { + const browserTab = browserUnifiedTab('shared-tab', 'ws-1', 'wt-1') + expect( + buildSearchableBrowserPages({ + worktrees: [worktreeA], + repoMap, + worktreeOrder, + browserTabsByWorktree: { 'wt-1': [makeWorkspace()] }, + browserPagesByWorkspace: { 'ws-1': [makePage()] }, + unifiedTabsByWorktree: { + 'wt-1': [browserTab, { ...browserTab, contentType: 'terminal' }] + }, + activeBrowserTabId: null, + activeWorktreeId: null, + activeTabType: 'terminal' + }) + ).toEqual([]) + }) + it('builds one entry per page across every workspace in a worktree', () => { const entries = buildFixture() @@ -373,14 +384,50 @@ describe('buildSearchableBrowserPages', () => { }) expect(entries.map((entry) => entry.lastActiveAt)).toEqual([4000, 9000]) + expect(entries.map((entry) => entry.lastFocusedAt)).toEqual([4000, undefined]) + }) + + it('moves the workspace-focus proxy when the active browser page changes', () => { + const browserTab: Tab = { + id: 'tab-ws-1', + entityId: 'ws-1', + groupId: 'group-1', + worktreeId: 'wt-1', + contentType: 'browser', + label: 'Example', + customLabel: null, + color: null, + sortOrder: 0, + createdAt: 0, + lastFocusedAt: 8_000 + } + const pages = [makePage({ createdAt: 1_000 }), makePage({ id: 'page-2', createdAt: 2_000 })] + const build = (activePageId: string) => + buildSearchableBrowserPages({ + worktrees: [worktreeA], + repoMap, + worktreeOrder, + browserTabsByWorktree: { + 'wt-1': [makeWorkspace({ activePageId, pageIds: ['page-1', 'page-2'] })] + }, + browserPagesByWorkspace: { 'ws-1': pages }, + unifiedTabsByWorktree: { 'wt-1': [browserTab] }, + activeBrowserTabId: null, + activeWorktreeId: null, + activeTabType: 'browser' + }) + + expect(build('page-1').map((entry) => entry.lastActiveAt)).toEqual([8_000, 2_000]) + expect(build('page-2').map((entry) => entry.lastActiveAt)).toEqual([1_000, 8_000]) + expect(build('page-1').map((entry) => entry.lastFocusedAt)).toEqual([8_000, undefined]) + expect(build('page-2').map((entry) => entry.lastFocusedAt)).toEqual([undefined, 8_000]) }) it('feeds Cmd+J browser search the same ranking as the inline builder did', () => { const results = searchBrowserPages(buildFixture(), 'docs') - // Current page first, then the two url-only matches in the active worktree, - // then the other worktree's title match. - expect(results.map((result) => result.pageId)).toEqual(['page-1', 'page-2', 'page-3', 'page-4']) + // Primary title proofs lead URL-only proofs even across worktrees. + expect(results.map((result) => result.pageId)).toEqual(['page-1', 'page-4', 'page-2', 'page-3']) expect(results[0].isCurrentPage).toBe(true) }) }) diff --git a/src/renderer/src/lib/browser-palette-page-entries.ts b/src/renderer/src/lib/browser-palette-page-entries.ts index 93b5d71f442..b1b7173d7dd 100644 --- a/src/renderer/src/lib/browser-palette-page-entries.ts +++ b/src/renderer/src/lib/browser-palette-page-entries.ts @@ -1,18 +1,23 @@ -import { getWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' import type { BrowserPage, BrowserWorkspace } from '../../../shared/browser-workspace-types' import type { Tab, WorkspaceVisibleTabType } from '../../../shared/tab-types' import type { Worktree } from '../../../shared/worktree/types' import type { ExecutionHostId } from '../../../shared/execution-host' -import { isPaletteCurrentWorktree, resolvePaletteRepoForWorktree } from './palette-repo-resolution' +import { + getPaletteWorktreeIdentity, + isPaletteCurrentWorktree, + resolvePaletteRepoForWorktree +} from './palette-repo-resolution' import { buildSearchableBrowserPageDocument, type SearchableBrowserPage } from './browser-palette-search' import { findAmbiguousWorktreeIds, + findDuplicateIds, getUnifiedTabPaletteExecutionHostId, isUnifiedTabOwnedByWorktree } from './unified-tab-host-ownership' +import { maxValidPaletteActivityTimestamp } from './palette-match/palette-ranking' type BrowserPaletteActiveTabType = WorkspaceVisibleTabType @@ -48,11 +53,21 @@ export function buildSearchableBrowserPages({ }: BuildSearchableBrowserPagesOptions): SearchableBrowserPage[] { const entries: SearchableBrowserPage[] = [] const ambiguousWorktreeIds = findAmbiguousWorktreeIds(ownershipWorktrees ?? worktrees) + const allUnifiedTabs = Object.values(unifiedTabsByWorktree ?? {}).flatMap((tabs) => tabs ?? []) + const duplicateTabIds = findDuplicateIds(allUnifiedTabs) + const duplicateWorkspaceIds = findDuplicateIds( + allUnifiedTabs + .filter((tab) => tab.contentType === 'browser') + .map((tab) => ({ id: tab.entityId })) + ) + const duplicateStoredWorkspaceIds = findDuplicateIds( + Object.values(browserTabsByWorktree).flatMap((workspaces) => workspaces ?? []) + ) for (const worktree of worktrees) { const repoName = resolvePaletteRepoForWorktree(worktree, repoMap, repoMapByHostIdentity)?.displayName ?? '' const worktreeSortIndex = - worktreeOrder.get(getWorktreeHostIdentity(worktree)) ?? + worktreeOrder.get(getPaletteWorktreeIdentity(worktree)) ?? worktreeOrder.get(worktree.id) ?? Number.MAX_SAFE_INTEGER const focusedAtByWorkspaceId = new Map() @@ -67,17 +82,35 @@ export function buildSearchableBrowserPages({ } } for (const workspace of browserTabsByWorktree[worktree.id] ?? []) { - const unifiedTab = unifiedTabs.find( - (tab) => - tab.contentType === 'browser' && - tab.entityId === workspace.id && - isUnifiedTabOwnedByWorktree(tab, worktree, ambiguousWorktreeIds) + if ( + duplicateWorkspaceIds.has(workspace.id) || + duplicateStoredWorkspaceIds.has(workspace.id) + ) { + continue + } + const workspaceTabs = unifiedTabs.filter( + (tab) => tab.contentType === 'browser' && tab.entityId === workspace.id ) - if (!unifiedTab && ambiguousWorktreeIds.has(worktree.id)) { + const unifiedTab = workspaceTabs.find((tab) => + isUnifiedTabOwnedByWorktree(tab, worktree, ambiguousWorktreeIds) + ) + if (!unifiedTab && (workspaceTabs.length > 0 || ambiguousWorktreeIds.has(worktree.id))) { + continue + } + if (unifiedTab && duplicateTabIds.has(unifiedTab.id)) { continue } const workspaceFocusedAt = focusedAtByWorkspaceId.get(workspace.id) - for (const page of browserPagesByWorkspace[workspace.id] ?? []) { + const pages = browserPagesByWorkspace[workspace.id] ?? [] + const duplicatePageIds = findDuplicateIds(pages) + for (const page of pages) { + if ( + duplicatePageIds.has(page.id) || + page.workspaceId !== workspace.id || + page.worktreeId !== worktree.id + ) { + continue + } entries.push({ page, workspace, @@ -95,8 +128,12 @@ export function buildSearchableBrowserPages({ activeWorktreeId, activeWorkspaceExecutionHostId ), - // Never older than the page itself: it was opened while the workspace was focused. - lastActiveAt: workspaceFocusedAt ? Math.max(workspaceFocusedAt, page.createdAt) : null, + // Workspace focus is a lossy proxy for only its currently active page. + lastFocusedAt: workspace.activePageId === page.id ? workspaceFocusedAt : undefined, + lastActiveAt: + workspace.activePageId === page.id && workspaceFocusedAt + ? maxValidPaletteActivityTimestamp([workspaceFocusedAt, page.createdAt]) + : maxValidPaletteActivityTimestamp([page.createdAt]), document: buildSearchableBrowserPageDocument({ page, workspace, worktree, repoName }) }) } diff --git a/src/renderer/src/lib/browser-palette-search.ts b/src/renderer/src/lib/browser-palette-search.ts index 0cd9f7e0618..dca4b7a639b 100644 --- a/src/renderer/src/lib/browser-palette-search.ts +++ b/src/renderer/src/lib/browser-palette-search.ts @@ -5,6 +5,7 @@ import { isClipboardTextByteLengthOverLimit } from '../../../shared/clipboard-te import { compareBaseSensitivityLocaleText } from './locale-text-collators' import { comparePaletteTabResults, + isOmniboxPaletteTabFieldAllowed, matchPaletteTabDocument, preparePaletteTabQuery } from './palette-match/tab-match' @@ -17,6 +18,13 @@ import type { ExecutionHostId } from '../../../shared/execution-host' import type { MatchRange } from './palette-match/normalized-text' import type { PaletteDocument, PaletteDocumentRank } from './palette-match/palette-document' import type { PaletteResultQualityClass } from './palette-match/match-quality' +import { + createPaletteSearchContext, + encodePaletteIdentity, + preparePaletteActivity, + type PaletteActivityRank, + type PaletteSearchContext +} from './palette-match/palette-ranking' const NO_RANGES: readonly MatchRange[] = [] @@ -31,6 +39,7 @@ export type SearchableBrowserPage = { isCurrentWorktree: boolean /** Last time the owning browser workspace was focused; null when never focused. */ lastActiveAt?: number | null + lastFocusedAt?: number /** Normalized field index, built once per entry rather than per keystroke. */ document: PaletteDocument } @@ -38,6 +47,7 @@ export type SearchableBrowserPage = { export type BrowserPaletteSearchResult = { /** Worktree ids collide across hosts; activation must not resolve by id alone. */ executionHostId?: ExecutionHostId + paletteIdentity: string pageId: string workspaceId: string worktreeId: string @@ -46,6 +56,8 @@ export type BrowserPaletteSearchResult = { /** Raw page URL, so callers can dedupe a row against another list of destinations. */ url: string secondaryText: string + /** Matched formatted/raw URLs with highlight offsets into each `text`; exposes hits beyond the displayed URL. */ + secondaryMatches: readonly { text: string; ranges: readonly MatchRange[] }[] workspaceLabel: string | null repoName: string worktreeName: string @@ -62,6 +74,7 @@ export type BrowserPaletteSearchResult = { qualityClass: PaletteResultQualityClass | null rank: PaletteDocumentRank | null lastActiveAt?: number | null + activity: PaletteActivityRank } export const BROWSER_PALETTE_QUERY_MAX_BYTES = 2 * 1024 @@ -145,11 +158,22 @@ function positionScore(entry: SearchableBrowserPage): number { return entry.worktreeSortIndex * 100 - (entry.isCurrentWorktree ? 1000 : 0) } -function baseResult(entry: SearchableBrowserPage): BrowserPaletteSearchResult { +function baseResult( + entry: SearchableBrowserPage, + context: PaletteSearchContext +): BrowserPaletteSearchResult { const formattedUrl = formatBrowserPaletteUrl(entry.page.url) const executionHostId = entry.executionHostId ?? entry.worktree.hostId + const activity = preparePaletteActivity(entry.lastActiveAt, context) return { ...(executionHostId ? { executionHostId } : {}), + paletteIdentity: encodePaletteIdentity([ + 'browser-page', + executionHostId ?? '', + entry.worktree.id, + entry.workspace.id, + entry.page.id + ]), pageId: entry.page.id, workspaceId: entry.workspace.id, worktreeId: entry.worktree.id, @@ -157,6 +181,7 @@ function baseResult(entry: SearchableBrowserPage): BrowserPaletteSearchResult { faviconUrl: entry.page.faviconUrl, url: entry.page.url, secondaryText: formattedUrl, + secondaryMatches: [], workspaceLabel: entry.workspace.label ?? null, repoName: entry.repoName, // Why resolve: a cleared display name leaves the raw field undefined at runtime. @@ -173,14 +198,17 @@ function baseResult(entry: SearchableBrowserPage): BrowserPaletteSearchResult { score: positionScore(entry), qualityClass: null, rank: null, - lastActiveAt: entry.lastActiveAt ?? null + lastActiveAt: activity.timestamp || null, + activity } } export function searchBrowserPages( entries: readonly SearchableBrowserPage[], - query: string + query: string, + options: { context?: PaletteSearchContext; fieldMode?: 'all' | 'omnibox' } = {} ): BrowserPaletteSearchResult[] { + const context = options.context ?? createPaletteSearchContext(Date.now()) if (isBrowserPaletteQueryTooLarge(query)) { return [] } @@ -190,14 +218,16 @@ export function searchBrowserPages( // listing, so the invalid case is filtered out by the token guard below. return query.trim() ? [] - : entries.map((entry) => baseResult(entry)).sort(compareEmptyQueryResults) + : entries.map((entry) => baseResult(entry, context)).sort(compareEmptyQueryResults) } const results: BrowserPaletteSearchResult[] = [] for (const entry of entries) { - const base = baseResult(entry) + const base = baseResult(entry, context) const secondaryTexts = browserPaletteSecondaryTexts(entry.page) - const match = matchPaletteTabDocument(entry.document, prepared) + const match = matchPaletteTabDocument(entry.document, prepared, { + isFieldAllowed: options.fieldMode === 'omnibox' ? isOmniboxPaletteTabFieldAllowed : undefined + }) if (!match) { continue } @@ -205,6 +235,10 @@ export function searchBrowserPages( ...base, secondaryText: match.secondary !== null ? secondaryTexts[match.secondary.index] : base.secondaryText, + secondaryMatches: match.secondaryMatches.map((secondary) => ({ + text: secondaryTexts[secondary.index] ?? '', + ranges: secondary.ranges + })), workspaceRanges: match.workspaceRanges, titleRanges: match.titleRanges, secondaryRanges: match.secondary?.ranges ?? NO_RANGES, @@ -222,14 +256,14 @@ export function searchBrowserPages( { rank: a.rank, positionScore: a.score, - id: a.pageId, - lastActiveAt: a.lastActiveAt ?? undefined + identity: a.paletteIdentity, + activity: a.activity }, { rank: b.rank, positionScore: b.score, - id: b.pageId, - lastActiveAt: b.lastActiveAt ?? undefined + identity: b.paletteIdentity, + activity: b.activity } ) : compareEmptyQueryResults(a, b) diff --git a/src/renderer/src/lib/browser-workspace-tab-activation.test.ts b/src/renderer/src/lib/browser-workspace-tab-activation.test.ts new file mode 100644 index 00000000000..363a99ec893 --- /dev/null +++ b/src/renderer/src/lib/browser-workspace-tab-activation.test.ts @@ -0,0 +1,134 @@ +// @vitest-environment happy-dom + +import { afterEach, expect, it } from 'vitest' +import { useAppStore } from '@/store' +import type { Tab } from '../../../shared/tab-types' +import type { Worktree } from '../../../shared/worktree/types' +import type { FolderWorkspace } from '../../../shared/folder-workspace-types' +import { folderWorkspaceKey } from '../../../shared/workspace-scope' +import { getActivatableBrowserWorkspaceTab } from './browser-workspace-tab-activation' + +const initialState = useAppStore.getInitialState() +afterEach(() => useAppStore.setState(initialState, true)) + +function makeWorktree(overrides: Partial & Pick): Worktree { + return { + repoId: 'repo', + path: '/workspace', + head: '', + branch: 'main', + isBare: false, + isMainWorktree: true, + displayName: 'Workspace', + comment: '', + linkedIssue: null, + linkedPR: null, + linkedLinearIssue: null, + isArchived: false, + isUnread: false, + isPinned: false, + sortOrder: 0, + lastActivityAt: 0, + ...overrides + } +} + +const browserTab: Tab = { + id: 'unified-browser', + entityId: 'workspace', + groupId: 'group', + worktreeId: 'wt', + contentType: 'browser', + label: 'Browser', + customLabel: null, + color: null, + sortOrder: 0, + createdAt: 0 +} + +function seedState(worktreesByRepo: Record, tab: Tab): void { + useAppStore.setState( + { ...initialState, worktreesByRepo, unifiedTabsByWorktree: { wt: [tab] } }, + true + ) +} + +function makeFolderWorkspace(executionHostId: 'local' | 'ssh:remote'): FolderWorkspace { + return { + id: 'shared-folder', + projectGroupId: 'group', + name: 'Shared folder', + folderPath: '/workspace', + executionHostId, + linkedTask: null, + comment: '', + isArchived: false, + isUnread: false, + isPinned: false, + sortOrder: 0, + lastActivityAt: 0, + createdAt: 0, + updatedAt: 0 + } +} + +it('refuses a hostless browser tab for a remote worktree whose ID also exists locally', () => { + seedState( + { + local: [makeWorktree({ id: 'wt' })], + remote: [makeWorktree({ id: 'wt', repoId: 'repo-remote', hostId: 'ssh:remote' })] + }, + browserTab + ) + + expect( + getActivatableBrowserWorkspaceTab({ + worktreeId: 'wt', + workspaceId: 'workspace', + executionHostId: 'ssh:remote' + }) + ).toBeNull() +}) + +it('refuses hostless activation when the caller omits a host for an ambiguous worktree id', () => { + seedState( + { + local: [makeWorktree({ id: 'wt' })], + remote: [makeWorktree({ id: 'wt', repoId: 'repo-remote', hostId: 'ssh:remote' })] + }, + { ...browserTab, executionHostId: 'ssh:remote' } + ) + + expect( + getActivatableBrowserWorkspaceTab({ worktreeId: 'wt', workspaceId: 'workspace' }) + ).toBeNull() +}) + +it('accepts a hostless browser tab when the worktree ID is unambiguous', () => { + seedState({ remote: [makeWorktree({ id: 'wt', hostId: 'ssh:remote' })] }, browserTab) + + expect( + getActivatableBrowserWorkspaceTab({ + worktreeId: 'wt', + workspaceId: 'workspace', + executionHostId: 'ssh:remote' + }) + ).toEqual(browserTab) +}) + +it('includes folder workspaces when rejecting ambiguous hostless activation', () => { + const worktreeId = folderWorkspaceKey('shared-folder') + useAppStore.setState( + { + ...initialState, + folderWorkspaces: [makeFolderWorkspace('local'), makeFolderWorkspace('ssh:remote')], + worktreesByRepo: {}, + unifiedTabsByWorktree: { + [worktreeId]: [{ ...browserTab, worktreeId, executionHostId: 'ssh:remote' }] + } + }, + true + ) + + expect(getActivatableBrowserWorkspaceTab({ worktreeId, workspaceId: 'workspace' })).toBeNull() +}) diff --git a/src/renderer/src/lib/browser-workspace-tab-activation.ts b/src/renderer/src/lib/browser-workspace-tab-activation.ts index f37780ed741..2008cca90e4 100644 --- a/src/renderer/src/lib/browser-workspace-tab-activation.ts +++ b/src/renderer/src/lib/browser-workspace-tab-activation.ts @@ -1,27 +1,58 @@ import { useAppStore } from '@/store' +import type { Tab } from '../../../shared/tab-types' +import type { ExecutionHostId } from '../../../shared/execution-host' +import { + findAmbiguousWorktreeIds, + getPaletteOwnershipWorktreeIds, + isUnifiedTabOwnedByWorktree +} from './unified-tab-host-ownership' -/** - * Bring a browser workspace forward as the surface the reader is in. - * - * Why the unified tab and not just the browser state: the pane renders whatever its group's active - * tab is, so selecting the workspace alone leaves the page live behind a tab that never shows it. - * Returns false when the workspace has no unified tab yet, which is the caller's cue that there is - * nothing to bring forward. - */ -export function activateBrowserWorkspaceTab(params: { +type BrowserWorkspaceTabTarget = { worktreeId: string workspaceId: string pageId?: string -}): boolean { + executionHostId?: ExecutionHostId +} + +export function getActivatableBrowserWorkspaceTab(params: BrowserWorkspaceTabTarget): Tab | null { const state = useAppStore.getState() - const unifiedTab = (state.unifiedTabsByWorktree[params.worktreeId] ?? []).find( + // A hostless tab cannot be attributed when the same worktree ID exists on several hosts. + const ambiguousWorktreeIds = findAmbiguousWorktreeIds(getPaletteOwnershipWorktreeIds(state)) + if (!params.executionHostId && ambiguousWorktreeIds.has(params.worktreeId)) { + return null + } + const worktree = state.getKnownWorktreeById(params.worktreeId, params.executionHostId) + if (!worktree) { + return null + } + // setActiveBrowserTab resolves its backing tab globally by workspace ID. + const tabs = Object.values(state.unifiedTabsByWorktree).flat() + const browserTabs = tabs.filter( (candidate) => candidate.contentType === 'browser' && candidate.entityId === params.workspaceId ) + const unifiedTab = browserTabs[0] + if ( + browserTabs.some( + (tab) => + tab.worktreeId !== params.worktreeId || + (worktree && !isUnifiedTabOwnedByWorktree(tab, worktree, ambiguousWorktreeIds)) + ) || + !unifiedTab || + tabs.filter((candidate) => candidate.id === unifiedTab.id).length !== 1 + ) { + return null + } + return unifiedTab +} + +export function activateBrowserWorkspaceTab(params: BrowserWorkspaceTabTarget): boolean { + const unifiedTab = getActivatableBrowserWorkspaceTab(params) if (!unifiedTab) { return false } + const state = useAppStore.getState() state.focusGroup(params.worktreeId, unifiedTab.groupId) - state.activateTab(unifiedTab.id) + state.activateTab(unifiedTab.id, { worktreeId: params.worktreeId }) state.setActiveBrowserTab(params.workspaceId) if (params.pageId) { state.setActiveBrowserPage(params.workspaceId, params.pageId) diff --git a/src/renderer/src/lib/cmd-j-host-qualified-candidate-ownership.test.ts b/src/renderer/src/lib/cmd-j-host-qualified-candidate-ownership.test.ts index 2612fd85a6f..de9a652f4dd 100644 --- a/src/renderer/src/lib/cmd-j-host-qualified-candidate-ownership.test.ts +++ b/src/renderer/src/lib/cmd-j-host-qualified-candidate-ownership.test.ts @@ -274,13 +274,77 @@ describe('Cmd-J host-qualified candidate ownership', () => { }) expect( - searchWorkspaceTabs(entries, 'shell').map((result) => [result.tabId, result.executionHostId]) + searchWorkspaceTabs(entries, 'shell') + .map((result) => [result.tabId, result.executionHostId]) + .sort(([left], [right]) => String(left).localeCompare(String(right))) ).toEqual([ ['local-terminal', 'local'], ['remote-terminal', RUNTIME_HOST_ID] ]) }) + it('omits editor rows whose bare file id cannot be activated safely', () => { + const entries = buildSearchableWorkspaceTabs({ + worktrees: pairedWorktrees(), + repoMap: new Map(), + worktreeOrder: new Map(), + unifiedTabsByWorktree: { + [SHARED_WORKTREE_ID]: [ + makeTab({ + id: 'shared-editor', + entityId: 'shared-file', + contentType: 'editor', + executionHostId: 'local' + }), + makeTab({ + id: 'shared-editor', + entityId: 'shared-file', + contentType: 'editor', + executionHostId: RUNTIME_HOST_ID + }) + ] + }, + tabsByWorktree: {}, + openFiles: [ + { + id: 'shared-file', + filePath: '/local/local-atlas.ts', + relativePath: 'local/local-atlas.ts', + worktreeId: SHARED_WORKTREE_ID, + language: 'typescript', + isDirty: false, + mode: 'edit' + }, + { + id: 'shared-file', + filePath: '/remote/remote-atlas.ts', + relativePath: 'remote/remote-atlas.ts', + worktreeId: SHARED_WORKTREE_ID, + language: 'typescript', + isDirty: false, + runtimeEnvironmentId: 'paired-host', + mode: 'edit' + } + ], + agentStatusByPaneKey: {}, + retainedAgentsByPaneKey: {}, + sleepingAgentSessionsByPaneKey: {}, + activeGroupIdByWorktree: {}, + groupsByWorktree: {}, + activeWorktreeId: null, + activeTabType: 'terminal', + activeTabId: null, + activeTabIdByWorktree: {}, + activeFileId: null, + activeFileIdByWorktree: {}, + activeTabTypeByWorktree: {}, + generatedTitlesEnabled: true + }) + + expect(searchWorkspaceTabs(entries, 'local-atlas')).toEqual([]) + expect(searchWorkspaceTabs(entries, 'remote-atlas')).toEqual([]) + }) + it('retains one unambiguous legacy tab without guessing between sibling hosts', () => { const legacyWorktree = makeWorktree({ hostId: undefined }) const entries = buildSearchableSimulatorTabs({ @@ -335,7 +399,7 @@ describe('Cmd-J host-qualified candidate ownership', () => { generatedTitlesEnabled: true, groupsByWorktree: {}, openFiles: [], - ownershipWorktrees, + folderWorkspaces: [], repo: null, tabsByWorktree: { [SHARED_WORKTREE_ID]: [ @@ -352,7 +416,8 @@ describe('Cmd-J host-qualified candidate ownership', () => { ] }, unifiedTabsByWorktree, - worktree: ownershipWorktrees[0] + worktree: ownershipWorktrees[0], + worktreesByRepo: { repo: ownershipWorktrees } }, { agentStatusByPaneKey: {}, diff --git a/src/renderer/src/lib/cmd-j-section-leadership.test.ts b/src/renderer/src/lib/cmd-j-section-leadership.test.ts index 410bf676e0d..da41a7172e3 100644 --- a/src/renderer/src/lib/cmd-j-section-leadership.test.ts +++ b/src/renderer/src/lib/cmd-j-section-leadership.test.ts @@ -11,13 +11,14 @@ import type { PaletteDocumentRank } from './palette-match/palette-document' function rank(overrides: Partial = {}): PaletteDocumentRank { return { - exactIntent: 1, + destination: 2, + recovery: 0, + wordMatch: 0, + coverage: 0, containerOnlyTokenCount: 0, - wholeQuery: 3, - worstQuality: 5, - usesSupportingEvidence: 0, - fuzzyTokenCount: 0, - fieldHopCount: 1, + recoveryTokenCount: 0, + strength: 0, + placement: 2, ...overrides } } @@ -110,38 +111,48 @@ describe('intent section leadership', () => { describe('ranked item comparison', () => { it('compares match rank lexicographically before list order', () => { - const strong = { rank: rank({ wholeQuery: 0 }), order: 99, id: 'b' } - const weak = { rank: rank({ wholeQuery: 2 }), order: 0, id: 'a' } + const strong = { rank: rank({ strength: 0 }), order: 99, identity: 'b' } + const weak = { rank: rank({ strength: 2 }), order: 0, identity: 'a' } expect(comparePaletteRankedItems(strong, weak)).toBeLessThan(0) }) it('prefers recently active item when match rank ties', () => { - const recent = { rank: rank(), order: 10, id: 'z', lastActiveAt: 2000 } - const older = { rank: rank(), order: 0, id: 'a', lastActiveAt: 1000 } + const recent = { + rank: rank(), + order: 10, + identity: 'z', + activity: { ageBucket: 0, timestamp: 2000 } + } + const older = { + rank: rank(), + order: 0, + identity: 'a', + activity: { ageBucket: 0, timestamp: 1000 } + } expect(comparePaletteRankedItems(recent, older)).toBeLessThan(0) }) it('falls back to the section order when match rank and recency tie', () => { - const first = { rank: rank(), order: 1, id: 'z', lastActiveAt: 1000 } - const second = { rank: rank(), order: 2, id: 'a', lastActiveAt: 1000 } + const first = { rank: rank(), order: 1, identity: 'z' } + const second = { rank: rank(), order: 2, identity: 'a' } expect(comparePaletteRankedItems(first, second)).toBeLessThan(0) }) it('breaks a full tie on the stable id', () => { - const a = { rank: rank(), order: 1, id: 'a' } - const b = { rank: rank(), order: 1, id: 'b' } + const a = { rank: rank(), order: 1, identity: 'a' } + const b = { rank: rank(), order: 1, identity: 'b' } expect(comparePaletteRankedItems(a, b)).toBeLessThan(0) }) it('keeps unmatched rows behind matched ones', () => { - const matched = { rank: rank(), order: 9, id: 'z' } - const unmatched = { rank: null, order: 0, id: 'a' } + const matched = { rank: rank(), order: 9, identity: 'z' } + const unmatched = { rank: null, order: 0, identity: 'a' } expect(comparePaletteRankedItems(matched, unmatched)).toBeLessThan(0) }) it('orders empty-query rows by their section order alone', () => { - const a = { rank: null, order: 0, id: 'z' } - const b = { rank: null, order: 1, id: 'a' } + const a = { rank: null, order: 0, identity: 'z' } + const b = { rank: null, order: 1, identity: 'a' } expect(comparePaletteRankedItems(a, b)).toBeLessThan(0) }) }) diff --git a/src/renderer/src/lib/cmd-j-section-leadership.ts b/src/renderer/src/lib/cmd-j-section-leadership.ts index 093cacce567..d484daec840 100644 --- a/src/renderer/src/lib/cmd-j-section-leadership.ts +++ b/src/renderer/src/lib/cmd-j-section-leadership.ts @@ -2,8 +2,11 @@ import { paletteResultQualityClassRank, type PaletteResultQualityClass } from './palette-match/match-quality' -import { comparePaletteDocumentRank } from './palette-match/palette-document' import type { PaletteDocumentRank } from './palette-match/palette-document' +import { + comparePaletteEntityRanks, + type PaletteActivityRank +} from './palette-match/palette-ranking' // Why a shared class and not raw scores: each section's score encodes its own list // position, so only a small common vocabulary can say which section holds the @@ -30,32 +33,34 @@ export type PaletteRankedItem = { rank: PaletteDocumentRank | null /** Existing smart-recency / list position, used only after match rank ties. */ order: number - id: string - /** Timestamp of most recent activity (focus or agent interaction). */ - lastActiveAt?: number + identity: string + activity?: PaletteActivityRank } /** Match rank first, then recent activity, then positional order, then stable id. */ export function comparePaletteRankedItems(a: PaletteRankedItem, b: PaletteRankedItem): number { if (a.rank && b.rank) { - const byRank = comparePaletteDocumentRank(a.rank, b.rank) - if (byRank !== 0) { - return byRank - } + return comparePaletteEntityRanks( + { + rank: a.rank, + activity: a.activity ?? { ageBucket: null, timestamp: 0 }, + position: a.order, + identity: a.identity + }, + { + rank: b.rank, + activity: b.activity ?? { ageBucket: null, timestamp: 0 }, + position: b.order, + identity: b.identity + } + ) } else if (a.rank !== b.rank) { return a.rank ? -1 : 1 } - if (a.lastActiveAt !== b.lastActiveAt) { - const aTime = a.lastActiveAt ?? 0 - const bTime = b.lastActiveAt ?? 0 - if (aTime !== bTime) { - return bTime - aTime - } - } if (a.order !== b.order) { return a.order - b.order } - return a.id.localeCompare(b.id) + return a.identity < b.identity ? -1 : a.identity > b.identity ? 1 : 0 } /** Ties prefer Open Tabs, matching the documented section-leadership rule. */ diff --git a/src/renderer/src/lib/file-preview.test.ts b/src/renderer/src/lib/file-preview.test.ts index 561d92bbc8a..4c770c85d9e 100644 --- a/src/renderer/src/lib/file-preview.test.ts +++ b/src/renderer/src/lib/file-preview.test.ts @@ -179,15 +179,27 @@ describe('openFileInBrowserTab', () => { } mocks.unifiedTabsByWorktree = { 'wt-1': [ - { id: 'tab-terminal', contentType: 'terminal', entityId: 'term-1', groupId: 'group-1' }, - { id: 'tab-doc', contentType: 'browser', entityId: 'browser-9', groupId: 'group-1' } + { + id: 'tab-terminal', + worktreeId: 'wt-1', + contentType: 'terminal', + entityId: 'term-1', + groupId: 'group-1' + }, + { + id: 'tab-doc', + worktreeId: 'wt-1', + contentType: 'browser', + entityId: 'browser-9', + groupId: 'group-1' + } ] } openFileInBrowserTab({ filePath: '/home/alice/report.html', worktreeId: 'wt-1' }) expect(mocks.focusGroup).toHaveBeenCalledWith('wt-1', 'group-1') - expect(mocks.activateTab).toHaveBeenCalledWith('tab-doc') + expect(mocks.activateTab).toHaveBeenCalledWith('tab-doc', { worktreeId: 'wt-1' }) expect(mocks.setActiveBrowserTab).toHaveBeenCalledWith('browser-9') expect(mocks.createBrowserTab).not.toHaveBeenCalled() }) diff --git a/src/renderer/src/lib/http-link-destinations.test.ts b/src/renderer/src/lib/http-link-destinations.test.ts new file mode 100644 index 00000000000..11c35df86fd --- /dev/null +++ b/src/renderer/src/lib/http-link-destinations.test.ts @@ -0,0 +1,64 @@ +import { describe, expect, it } from 'vitest' +import { buildHttpLinkActions, httpLinkActionDestinationsFor } from './http-link-destinations' + +describe('httpLinkActionDestinationsFor', () => { + it.each([ + ['local', { kind: 'local' } as const, false], + ['capable runtime', { kind: 'runtime', runtimeEnvironmentId: 'env-1' } as const, true], + ['eligible SSH', { kind: 'ssh', connectionId: 'ssh-1' } as const, true] + ])( + 'offers both destinations for a %s owner and follows the preference', + (_label, owner, canOpen) => { + expect(httpLinkActionDestinationsFor({ openLinksInApp: true }, owner, canOpen)).toEqual({ + primary: 'orca', + alternate: 'system' + }) + expect(httpLinkActionDestinationsFor({ openLinksInApp: false }, owner, canOpen)).toEqual({ + primary: 'system', + alternate: 'orca' + }) + } + ) + + it.each([ + ['incapable runtime', { kind: 'runtime', runtimeEnvironmentId: 'env-1' } as const], + ['ineligible SSH', { kind: 'ssh', connectionId: 'ssh-1' } as const], + ['unknown owner', { kind: 'unknown' } as const] + ])('offers only the system browser for an %s', (_label, owner) => { + expect(httpLinkActionDestinationsFor({ openLinksInApp: true }, owner, false)).toEqual({ + primary: 'system' + }) + }) +}) + +describe('buildHttpLinkActions', () => { + it('labels each offered destination and routes the run to it', () => { + const opened: (string | undefined)[] = [] + const actions = buildHttpLinkActions( + { primary: 'orca', alternate: 'system' }, + (destination) => { + opened.push(destination) + } + ) + + expect(actions.primary.label).toBe('Orca Browser') + expect(actions.primary.external).toBe(false) + expect(actions.alternate?.label).toBe('System Browser') + expect(actions.alternate?.external).toBe(true) + + void actions.primary.run() + void actions.alternate?.run() + expect(opened).toEqual(['orca', 'system']) + }) + + it('omits the alternate row when only one destination is offered', () => { + const actions = buildHttpLinkActions({ primary: 'system' }, () => {}) + expect(actions.alternate).toBeUndefined() + }) + + it('falls back to a generic label when no destination is known', () => { + const actions = buildHttpLinkActions(undefined, () => {}) + expect(actions.primary.label).toBe('Open link') + expect(actions.alternate).toBeUndefined() + }) +}) diff --git a/src/renderer/src/lib/http-link-destinations.ts b/src/renderer/src/lib/http-link-destinations.ts new file mode 100644 index 00000000000..8fecfe99cc8 --- /dev/null +++ b/src/renderer/src/lib/http-link-destinations.ts @@ -0,0 +1,149 @@ +import { translate } from '@/i18n/i18n' +import { openHttpLink, type HttpLinkSourceOwner } from '@/lib/http-link-routing' + +// Catalog keys keep their original terminal namespace: they are opaque ids with +// shipped translations, and the popover is now shared with native chat. + +export type HttpLinkDestination = 'orca' | 'system' + +export type HttpLinkActionDestinations = { + primary: HttpLinkDestination + alternate?: HttpLinkDestination +} + +export type HttpLinkAction = { + external?: boolean + label: string + run: () => void | Promise +} + +export function canSourceOwnerOpenInOrca( + sourceOwner: HttpLinkSourceOwner, + canOpenOwnedBrowser: boolean +): boolean { + return ( + sourceOwner.kind === 'local' || + ((sourceOwner.kind === 'runtime' || sourceOwner.kind === 'ssh') && canOpenOwnedBrowser) + ) +} + +/** Which destinations a clicked link offers, primary first; a remote source that + * cannot reach Orca's managed browser offers only the system browser. */ +export function httpLinkActionDestinationsFor( + settings: { openLinksInApp?: boolean } | null | undefined, + sourceOwner: HttpLinkSourceOwner, + canOpenOwnedBrowser: boolean +): HttpLinkActionDestinations { + if (!canSourceOwnerOpenInOrca(sourceOwner, canOpenOwnedBrowser)) { + return { primary: 'system' } + } + return settings?.openLinksInApp === true + ? { primary: 'orca', alternate: 'system' } + : { primary: 'system', alternate: 'orca' } +} + +export function httpLinkDestinationLabel(destination: HttpLinkDestination): string { + return destination === 'orca' + ? translate( + 'auto.components.terminal.pane.TerminalLinkActionPopover.orcaBrowser', + 'Orca Browser' + ) + : translate( + 'auto.components.terminal.pane.TerminalLinkActionPopover.systemBrowser', + 'System Browser' + ) +} + +/** One action per offered destination; surfaces share the labels and the open call. */ +export function buildHttpLinkActions( + destinations: HttpLinkActionDestinations | undefined, + open: (destination: HttpLinkDestination | undefined) => void | Promise +): { primary: HttpLinkAction; alternate?: HttpLinkAction } { + const primaryDestination = destinations?.primary + const primary: HttpLinkAction = { + external: primaryDestination === 'system', + label: primaryDestination + ? httpLinkDestinationLabel(primaryDestination) + : translate('auto.components.terminal.pane.TerminalLinkActionPopover.openLink', 'Open link'), + run: () => open(primaryDestination) + } + const alternateDestination = destinations?.alternate + if (!alternateDestination) { + return { primary } + } + return { + primary, + alternate: { + external: alternateDestination === 'system', + label: httpLinkDestinationLabel(alternateDestination), + run: () => open(alternateDestination) + } + } +} + +export type HttpLinkRoutingPreferenceRequester = ( + url: string +) => boolean | Promise | null | undefined + +export type RoutedHttpLinkOptions = { + worktreeId: string + sourceOwner?: HttpLinkSourceOwner + modifierHeld?: boolean + forceDestination?: HttpLinkDestination + requestOpenLinksInAppPreference?: HttpLinkRoutingPreferenceRequester +} + +export function openRoutedHttpLink(url: string, deps: RoutedHttpLinkOptions): void { + // Why: the clicked link's owner beats the global active runtime for both local and remote routes. + const sourceOwner = deps.sourceOwner ?? { kind: 'local' } + if (deps.forceDestination) { + openHttpLink(url, { + allowRemoteInApp: true, + worktreeId: deps.worktreeId, + forceInApp: deps.forceDestination === 'orca', + forceSystemBrowser: deps.forceDestination === 'system', + sourceOwner + }) + return + } + if (deps.modifierHeld) { + // Why: the modifier states a destination outright, so it also skips the + // one-time routing prompt; openHttpLink resolves which destination it means. + openHttpLink(url, { + allowRemoteInApp: true, + worktreeId: deps.worktreeId, + modifierHeld: true, + sourceOwner + }) + return + } + + // Why: remote sources use the persisted routing preference and never prompt the viewing client. + const preferenceDecision = + sourceOwner.kind === 'local' ? deps.requestOpenLinksInAppPreference?.(url) : null + if (preferenceDecision === null || preferenceDecision === undefined) { + openHttpLink(url, { allowRemoteInApp: true, worktreeId: deps.worktreeId, sourceOwner }) + return + } + + // Why: the first link click may need an async preference dialog. + // Suppress the browser's default link handling first, then route after the + // persisted choice is available. + void Promise.resolve(preferenceDecision) + .then((openInOrca) => { + openHttpLink(url, { + allowRemoteInApp: true, + worktreeId: deps.worktreeId, + forceSystemBrowser: !openInOrca, + sourceOwner + }) + }) + .catch(() => { + openHttpLink(url, { + allowRemoteInApp: true, + worktreeId: deps.worktreeId, + forceSystemBrowser: true, + sourceOwner + }) + }) +} diff --git a/src/renderer/src/lib/native-chat-initial-view-mode.ts b/src/renderer/src/lib/native-chat-initial-view-mode.ts index 258fd8f190f..8f89ecee038 100644 --- a/src/renderer/src/lib/native-chat-initial-view-mode.ts +++ b/src/renderer/src/lib/native-chat-initial-view-mode.ts @@ -1,4 +1,5 @@ import type { GlobalSettings } from '../../../shared/global-settings-types' +import { agentTabsDefaultToNativeChat } from '../../../shared/structured-native-chat-launch-route' import type { Tab } from '../../../shared/tab-types' import type { TuiAgent } from '../../../shared/tui-agent' import { canMirrorLaunchDraftToNativeChat } from '@/lib/native-chat-launch-draft-mirrorability' @@ -27,7 +28,7 @@ export function decideInitialAgentTabViewMode(args: { launchDraftText?: string nativeChatTranscriptIsLocalReadable?: boolean }): Tab['viewMode'] { - if (args.experimentalNativeChat !== true || args.openAgentTabsInChatByDefault !== true) { + if (!agentTabsDefaultToNativeChat(args)) { return undefined } if (!isNativeChatSupportedAgent(args.agent)) { diff --git a/src/renderer/src/lib/palette-match/cmd-j-ranking-contract.test.ts b/src/renderer/src/lib/palette-match/cmd-j-ranking-contract.test.ts new file mode 100644 index 00000000000..fb7f030cf68 --- /dev/null +++ b/src/renderer/src/lib/palette-match/cmd-j-ranking-contract.test.ts @@ -0,0 +1,211 @@ +import { describe, expect, it } from 'vitest' +import { buildPaletteDocument, comparePaletteDocumentRank } from './palette-document' +import { matchPaletteDocument } from './match-document' +import { preparePaletteQuery } from './palette-query' +import { buildPaletteTabDocument } from './tab-document' +import { matchPaletteTabDocument } from './tab-match' + +function ready(query: string) { + const prepared = preparePaletteQuery(query) + if (prepared.state !== 'ready') { + throw new Error(`Expected ready query: ${query}`) + } + return prepared +} + +function matchTitleAndPath(title: string, path: string, query = 'atlas') { + return matchPaletteTabDocument( + buildPaletteTabDocument({ + id: title, + title, + secondaryTexts: [path], + worktreeName: 'workspace', + branch: 'main', + repoName: 'repo' + }), + ready(query) + ) +} + +describe('Cmd+J semantic proof contract', () => { + it('puts a path word boundary above a mid-word title, but a title word above that path', () => { + const path = matchTitleAndPath('megatlascope', '/notes/atlas/') + const title = matchTitleAndPath('Atlas planning', '/notes/atlas/') + expect(path?.secondaryMatches).toHaveLength(1) + expect(title?.titleRanges).toHaveLength(1) + expect(path && title && comparePaletteDocumentRank(title.rank, path.rank)).toBeLessThan(0) + }) + + it('chooses a literal secondary proof over a primary typo', () => { + const match = matchTitleAndPath('atlaz', '/notes/atlas/') + expect(match?.secondaryMatches).toHaveLength(1) + expect(match?.rank).toMatchObject({ recovery: 0, wordMatch: 0, coverage: 1 }) + }) + + it('uses the stronger secondary proof when another token already requires container coverage', () => { + const match = matchPaletteTabDocument( + buildPaletteTabDocument({ + id: 'tab', + title: 'alphabet', + secondaryTexts: ['/alpha'], + worktreeName: 'beta', + branch: 'main', + repoName: 'repo' + }), + ready('alpha beta') + ) + expect(match?.rank).toMatchObject({ coverage: 2, strength: 0 }) + expect(match?.titleRanges).toEqual([]) + expect(match?.secondaryMatches).toEqual([{ index: 0, ranges: [{ start: 1, end: 6 }] }]) + expect(match?.worktreeRanges).toEqual([{ start: 0, end: 4 }]) + }) + + it('chooses the same semantic proof regardless of field source order', () => { + const field = (id: string, role: 'secondary' | 'container') => ({ + id, + profile: 'structured-label' as const, + text: 'alpha', + role, + destinationEligible: false + }) + const match = (visibleFields: ReturnType[]) => { + const query = ready('alpha beta') + return matchPaletteDocument({ + document: buildPaletteDocument({ + id: 'order-invariant', + visibleFields: [ + ...visibleFields, + { + id: 'beta', + profile: 'structured-label', + text: 'beta', + role: 'container', + destinationEligible: false + } + ], + evidence: [] + }), + tokens: query.tokens, + normalizedQuery: query.normalized + }) + } + + const containerFirst = match([field('container', 'container'), field('secondary', 'secondary')]) + const secondaryFirst = match([field('secondary', 'secondary'), field('container', 'container')]) + expect(containerFirst?.rank).toEqual(secondaryFirst?.rank) + expect(containerFirst?.assignments.map((assignment) => assignment.fieldId)).toEqual([ + 'secondary', + 'beta' + ]) + expect(secondaryFirst?.assignments.map((assignment) => assignment.fieldId)).toEqual([ + 'secondary', + 'beta' + ]) + }) + + it('restores contained secondary fields and preserves every selected representation', () => { + const restored = matchTitleAndPath('foobar', 'bar', 'b') + expect(restored?.secondaryMatches[0]?.ranges).toEqual([{ start: 0, end: 1 }]) + + const multi = matchPaletteTabDocument( + buildPaletteTabDocument({ + id: 'editor', + title: 'main.ts', + secondaryTexts: ['src/main.ts', '/home/me/project/src/main.ts'], + worktreeName: 'workspace', + branch: 'main', + repoName: 'repo' + }), + ready('src/main.ts /home/me') + ) + expect(multi?.secondaryMatches.map((proof) => proof.index)).toEqual([0, 1]) + }) + + it('promotes eligible equality but not repository equality', () => { + const eligible = matchTitleAndPath('notes', '/tmp/atlas', '/tmp/atlas') + const ineligible = matchPaletteTabDocument( + buildPaletteTabDocument({ + id: 'repo-hit', + title: 'notes', + secondaryTexts: [], + worktreeName: 'workspace', + branch: 'main', + repoName: '/tmp/atlas' + }), + ready('/tmp/atlas') + ) + expect(eligible?.rank.destination).toBe(1) + expect(ineligible?.rank.destination).toBe(2) + }) + + it('recognizes only a single complete compatible sigilled number', () => { + const document = buildPaletteDocument({ + id: 'review', + visibleFields: [ + { + id: 'name', + profile: 'structured-label', + text: 'migration', + role: 'primary', + destinationEligible: true + } + ], + evidence: [ + { + unit: { id: 'pr', kind: 'pr', text: '#123', accessibilityLabel: 'Pull request' }, + fields: [ + { + id: 'pr-number', + profile: 'identifier', + text: '#123', + evidenceId: 'pr', + renderOffset: 0, + identifier: { kind: 'number', sigil: '#' } + } + ] + } + ] + }) + const run = (query: string) => { + const prepared = ready(query) + return matchPaletteDocument({ + document, + tokens: prepared.tokens, + normalizedQuery: prepared.normalized, + tokenCountBeforeDeduplication: prepared.tokenCountBeforeDeduplication + }) + } + expect(run('#123')?.rank.destination).toBe(0) + expect(run('#123 #123')?.rank.destination).toBe(2) + expect(run('#123 migration')?.rank.destination).toBe(2) + expect(run('123')?.rank.destination).toBe(2) + expect(run('!123')).toBeNull() + }) + + it('uses the proof with fewer container-only tokens', () => { + const match = matchPaletteTabDocument( + buildPaletteTabDocument({ + id: 'tab', + title: 'atlas', + secondaryTexts: [], + worktreeName: 'atlas sprint', + branch: 'main', + repoName: 'repo' + }), + ready('atlas sprint') + ) + expect(match?.rank).toMatchObject({ + coverage: 2, + containerOnlyTokenCount: 1, + placement: 2 + }) + expect(match?.qualityClass).toBe('exact-visible') + expect(match?.titleRanges).toHaveLength(1) + expect(match?.worktreeRanges).toHaveLength(1) + }) + + it('finds a later word-boundary phrase after an incidental first occurrence', () => { + const match = matchTitleAndPath('xatlas sprint Atlas sprint notes', '', 'atlas sprint') + expect(match?.rank.placement).toBe(1) + }) +}) diff --git a/src/renderer/src/lib/palette-match/indexed-field.ts b/src/renderer/src/lib/palette-match/indexed-field.ts index 0e87ae37895..fbfee097fd7 100644 --- a/src/renderer/src/lib/palette-match/indexed-field.ts +++ b/src/renderer/src/lib/palette-match/indexed-field.ts @@ -23,26 +23,41 @@ export type PaletteIdentifierOptions = { sigil?: PaletteIdentifierSigil } -export type PaletteFieldSource = { +export type PaletteFieldRole = 'primary' | 'secondary' | 'alias' | 'container' + +type PaletteFieldSourceBase = { id: string profile: PaletteFieldProfile text: string - /** null marks a visible identity field; identity fields combine freely. */ - evidenceId?: string | null identifier?: PaletteIdentifierOptions - /** Container-level fields (e.g. worktree/branch for tabs) demote when matched alone. */ - isContainer?: boolean } +export type PaletteVisibleFieldSource = PaletteFieldSourceBase & { + evidenceId?: null + role: PaletteFieldRole + destinationEligible: boolean +} + +export type PaletteEvidenceFieldSource = PaletteFieldSourceBase & { + evidenceId: string + role?: never + destinationEligible?: never +} + +export type PaletteFieldSource = PaletteVisibleFieldSource | PaletteEvidenceFieldSource + export type PaletteIndexedField = { id: string + /** Stable source order used to break otherwise-equivalent match proofs. */ + sourceOrder: number profile: PaletteFieldProfile text: NormalizedText atoms: readonly PaletteAtom[] words: readonly PaletteWord[] evidenceId: string | null identifier: PaletteIdentifierOptions | null - isContainer: boolean + role: PaletteFieldRole | null + destinationEligible: boolean } const IDENTIFIER_PREFIX_KINDS: ReadonlySet = new Set([ @@ -114,7 +129,10 @@ export function paletteProfileAllowedQualities( return QUALITIES_BY_PROFILE[profile] } -export function indexPaletteField(source: PaletteFieldSource): PaletteIndexedField | null { +export function indexPaletteField( + source: PaletteFieldSource, + sourceOrder = 0 +): PaletteIndexedField | null { const trimmed = source.text.trim() if (!trimmed) { return null @@ -123,13 +141,15 @@ export function indexPaletteField(source: PaletteFieldSource): PaletteIndexedFie const segments = segmentPaletteText(text) return { id: source.id, + sourceOrder, profile: source.profile, text, atoms: segments.atoms, words: segments.words, evidenceId: source.evidenceId ?? null, identifier: source.identifier ?? null, - isContainer: Boolean(source.isContainer) + role: source.role ?? null, + destinationEligible: source.destinationEligible === true } } @@ -142,7 +162,7 @@ export function indexPaletteFields( if (!source) { continue } - const field = indexPaletteField(source) + const field = indexPaletteField(source, fields.length) if (field && !seenIds.has(field.id)) { seenIds.add(field.id) fields.push(field) diff --git a/src/renderer/src/lib/palette-match/match-document.ts b/src/renderer/src/lib/palette-match/match-document.ts index 7b2363c1e16..8f4e5cb32b5 100644 --- a/src/renderer/src/lib/palette-match/match-document.ts +++ b/src/renderer/src/lib/palette-match/match-document.ts @@ -1,328 +1,297 @@ import { matchPaletteField, type PaletteFieldMatch } from './match-field' -import { - isFuzzyPaletteMatchQuality, - paletteMatchQualityRank, - resolvePaletteResultQualityClass, - type PaletteMatchQuality -} from './match-quality' -import { mergeMatchRanges, type MatchRange } from './normalized-text' +import { resolvePaletteResultQualityClass, type PaletteMatchQuality } from './match-quality' import { createPaletteQueryToken, type PaletteQueryToken } from './palette-query' import { comparePaletteDocumentRank, type PaletteDocument, type PaletteDocumentMatch, - type PaletteDocumentRank, - type PaletteSupportingEvidence, type PaletteTokenAssignment } from './palette-document' +import type { PaletteIndexedField } from './indexed-field' +import { + addRankedAssignment, + collectCompleteVisibleAssignments, + collectRecognizedIdentifierAssignments, + collectScopeAssignments, + selectThresholdAssignment, + summarizeCandidates, + type RankedAssignment +} from './palette-assignment-ranking' +import { buildRangesByField, buildSupportingEvidence } from './palette-match-rendering' +import { assignmentsAreContainerOnly } from './palette-assignment-inspection' +import { compareSelectedSourceOrder } from './palette-selection-source-order' -type FieldHit = { fieldId: string; match: PaletteFieldMatch } +type FieldHit = { field: PaletteIndexedField; match: PaletteFieldMatch } -/** One token's chosen coverage; a `repo/branch` composite carries two hits. */ -type TokenCandidate = { hits: readonly FieldHit[]; quality: PaletteMatchQuality } - -type TokenCandidates = { - visible: TokenCandidate | null - byEvidenceId: Map +/** One token's proof; a repo/branch composite deliberately retains both hits. */ +export type TokenCandidate = { + hits: readonly FieldHit[] + quality: PaletteMatchQuality + recovery: number + wordMatch: number + coverage: number + strength: number + containerOnly: number } -function better(a: TokenCandidate | null, b: TokenCandidate): TokenCandidate { - if (!a) { - return b +export type TokenCandidates = { + visible: TokenCandidate[] + byEvidenceId: Map +} + +export type PaletteMatchDiagnostics = { + selectionCandidateVisits: number +} + +const STRENGTH: Record = { + 'field-exact': 0, + 'word-exact': 0, + 'field-prefix': 1, + 'word-prefix': 1, + 'boundary-substring': 2, + 'literal-substring': 3, + compact: 4, + typo: 5 +} + +function fieldCoverage(field: PaletteIndexedField): number { + if (field.evidenceId) { + return 3 } - return paletteMatchQualityRank(a.quality) <= paletteMatchQualityRank(b.quality) ? a : b + if (field.role === 'primary') { + return 0 + } + if (field.role === 'secondary' || field.role === 'alias') { + return 1 + } + return 2 } function toCandidate(hits: readonly FieldHit[]): TokenCandidate { let quality = hits[0].match.quality - for (const hit of hits) { - if (paletteMatchQualityRank(hit.match.quality) > paletteMatchQualityRank(quality)) { + let strength = STRENGTH[quality] + let recovery = strength >= STRENGTH.compact ? 1 : 0 + let wordMatch = strength >= STRENGTH['literal-substring'] ? 1 : 0 + let coverage = fieldCoverage(hits[0].field) + for (let index = 1; index < hits.length; index += 1) { + const hit = hits[index] + const value = STRENGTH[hit.match.quality] + if (value > strength) { + strength = value quality = hit.match.quality } + if (value >= STRENGTH.compact) { + recovery = 1 + } + if (value >= STRENGTH['literal-substring']) { + wordMatch = 1 + } + coverage = Math.max(coverage, fieldCoverage(hit.field)) + } + return { + hits, + quality, + recovery, + wordMatch, + coverage, + containerOnly: hits.every((hit) => hit.field.role === 'container') ? 1 : 0, + strength } - return { hits, quality } } function matchCompositePairs( document: PaletteDocument, - token: PaletteQueryToken -): TokenCandidate | null { + token: PaletteQueryToken, + isFieldAllowed?: (field: PaletteIndexedField) => boolean +): TokenCandidate[] { if (!token.repoBranch || !document.compositePairs.length) { - return null + return [] } const left = createPaletteQueryToken(token.repoBranch.repo, token.index) const right = createPaletteQueryToken(token.repoBranch.branch, token.index) - let best: TokenCandidate | null = null + const candidates: TokenCandidate[] = [] for (const pair of document.compositePairs) { - const leftField = document.fields.find((field) => field.id === pair.leftFieldId) - const rightField = document.fields.find((field) => field.id === pair.rightFieldId) - if (!leftField || !rightField) { + const leftField = document.fieldById.get(pair.leftFieldId) + const rightField = document.fieldById.get(pair.rightFieldId) + if ( + !leftField || + !rightField || + (isFieldAllowed && (!isFieldAllowed(leftField) || !isFieldAllowed(rightField))) + ) { continue } const leftMatch = matchPaletteField(leftField, left) const rightMatch = matchPaletteField(rightField, right) - if (!leftMatch || !rightMatch) { - continue + if (leftMatch && rightMatch) { + candidates.push( + toCandidate([ + { field: leftField, match: leftMatch }, + { field: rightField, match: rightMatch } + ]) + ) } - best = better( - best, - toCandidate([ - { fieldId: leftField.id, match: leftMatch }, - { fieldId: rightField.id, match: rightMatch } - ]) - ) } - return best + return candidates } function collectTokenCandidates( document: PaletteDocument, - token: PaletteQueryToken + token: PaletteQueryToken, + isFieldAllowed?: (field: PaletteIndexedField) => boolean ): TokenCandidates | null { const candidates: TokenCandidates = { - visible: matchCompositePairs(document, token), + visible: matchCompositePairs(document, token, isFieldAllowed), byEvidenceId: new Map() } - let found = candidates.visible !== null + let found = candidates.visible.length > 0 for (const field of document.fields) { + if (isFieldAllowed && !isFieldAllowed(field)) { + continue + } const match = matchPaletteField(field, token) if (!match) { continue } found = true - const candidate = toCandidate([{ fieldId: field.id, match }]) + const candidate = toCandidate([{ field, match }]) if (!field.evidenceId) { - candidates.visible = better(candidates.visible, candidate) + candidates.visible.push(candidate) } else { - candidates.byEvidenceId.set( - field.evidenceId, - better(candidates.byEvidenceId.get(field.evidenceId) ?? null, candidate) - ) + const bucket = candidates.byEvidenceId.get(field.evidenceId) + if (bucket) { + bucket.push(candidate) + } else { + candidates.byEvidenceId.set(field.evidenceId, [candidate]) + } } } return found ? candidates : null } -function scoreWholeQuery(document: PaletteDocument, normalizedQuery: string): number { - let best = 3 - for (const field of document.visibleFields) { - const text = field.text.normalized - if (text === normalizedQuery) { - return 0 - } - if (text.startsWith(normalizedQuery)) { - best = Math.min(best, 1) - continue - } - const index = text.indexOf(normalizedQuery) - if (index > 0 && field.words.some((word) => word.start === index)) { - best = Math.min(best, 2) - } - } - return best -} - -function buildAssignments( - candidates: readonly TokenCandidates[], +function toTokenAssignments( tokens: readonly PaletteQueryToken[], - evidenceId: string | null -): { assignments: PaletteTokenAssignment[]; usesEvidence: boolean } | null { + selected: readonly TokenCandidate[] +): PaletteTokenAssignment[] { const assignments: PaletteTokenAssignment[] = [] - let usesEvidence = false - - for (let index = 0; index < candidates.length; index += 1) { - const candidate = candidates[index] - const evidence = evidenceId ? (candidate.byEvidenceId.get(evidenceId) ?? null) : null - const chosen = evidence ? better(candidate.visible, evidence) : candidate.visible - if (!chosen) { - return null - } - if (evidence && chosen === evidence) { - usesEvidence = true - } - for (const hit of chosen.hits) { + selected.forEach((candidate, index) => { + for (const hit of candidate.hits) { assignments.push({ tokenIndex: tokens[index].index, - fieldId: hit.fieldId, + fieldId: hit.field.id, quality: hit.match.quality, ranges: hit.match.ranges }) } - } - - return { assignments, usesEvidence } + }) + return assignments } -function rankAssignments(args: { - document: PaletteDocument - assignments: readonly PaletteTokenAssignment[] - usesEvidence: boolean - wholeQuery: number - exactIntent: boolean -}): { rank: PaletteDocumentRank; worstQuality: PaletteMatchQuality; isContainerOnly: boolean } { - let worstQuality: PaletteMatchQuality = 'field-exact' - let fuzzyTokenCount = 0 - const fields = new Set() - let containerOnlyTokenCount = 0 - let tokenIndex = -1 - let tokenHasDirectField = false - let matchedTokenCount = 0 - - for (const assignment of args.assignments) { - if (paletteMatchQualityRank(assignment.quality) > paletteMatchQualityRank(worstQuality)) { - worstQuality = assignment.quality - } - if (isFuzzyPaletteMatchQuality(assignment.quality)) { - fuzzyTokenCount += 1 - } - fields.add(assignment.fieldId) - if (assignment.tokenIndex !== tokenIndex) { - if (tokenIndex !== -1 && !tokenHasDirectField) { - containerOnlyTokenCount += 1 - } - tokenIndex = assignment.tokenIndex - tokenHasDirectField = false - matchedTokenCount += 1 - } - const field = args.document.fieldById.get(assignment.fieldId) - if (field && !field.isContainer) { - tokenHasDirectField = true - } - } - - if (tokenIndex !== -1 && !tokenHasDirectField) { - containerOnlyTokenCount += 1 - } - const isContainerOnly = - containerOnlyTokenCount > 0 && containerOnlyTokenCount === matchedTokenCount - - return { - worstQuality, - isContainerOnly, - rank: { - exactIntent: args.exactIntent ? 0 : 1, - containerOnlyTokenCount, - wholeQuery: args.wholeQuery, - worstQuality: paletteMatchQualityRank(worstQuality), - usesSupportingEvidence: args.usesEvidence ? 1 : 0, - fuzzyTokenCount, - fieldHopCount: fields.size - } - } -} - -function buildSupportingEvidence( - document: PaletteDocument, - assignments: readonly PaletteTokenAssignment[], - evidenceId: string | null -): PaletteSupportingEvidence[] { - const unit = evidenceId ? document.evidenceUnits.get(evidenceId) : undefined - if (!unit) { - return [] - } - const ranges: MatchRange[] = [] - for (const assignment of assignments) { - const offset = document.renderOffsetByFieldId.get(assignment.fieldId) - if (offset === undefined) { - continue - } - for (const range of assignment.ranges) { - // Why clamp: a range is only meaningful against the unit text the row renders, and - // an out-of-range end would highlight past the end of that string. - const start = Math.min(range.start + offset, unit.text.length) - const end = Math.min(range.end + offset, unit.text.length) - if (start < end) { - ranges.push({ start, end }) - } - } - } - if (!ranges.length) { - return [] - } - return [ - { - id: unit.id, - kind: unit.kind, - text: unit.text, - ranges: mergeMatchRanges(ranges), - accessibilityLabel: unit.accessibilityLabel - } - ] -} - -function buildRangesByField( - assignments: readonly PaletteTokenAssignment[] -): Map { - const byField = new Map() - for (const assignment of assignments) { - const bucket = byField.get(assignment.fieldId) - if (bucket) { - bucket.push(...assignment.ranges) - } else { - byField.set(assignment.fieldId, [...assignment.ranges]) - } - } - const merged = new Map() - for (const [fieldId, ranges] of byField) { - merged.set(fieldId, mergeMatchRanges(ranges)) - } - return merged -} - -/** - * Accepts a document only when every token has an allowed field match reachable - * from visible identity text plus at most one supporting-evidence unit. - */ export function matchPaletteDocument(args: { document: PaletteDocument tokens: readonly PaletteQueryToken[] normalizedQuery: string + tokenCountBeforeDeduplication?: number exactIntent?: boolean + isFieldAllowed?: (field: PaletteIndexedField) => boolean + diagnostics?: PaletteMatchDiagnostics }): PaletteDocumentMatch | null { - const { document, tokens } = args const candidates: TokenCandidates[] = [] - for (const token of tokens) { - const candidate = collectTokenCandidates(document, token) - if (!candidate) { + for (const token of args.tokens) { + const collected = collectTokenCandidates(args.document, token, args.isFieldAllowed) + if (!collected) { return null } - candidates.push(candidate) + candidates.push(collected) } - const wholeQuery = scoreWholeQuery(document, args.normalizedQuery) - const evidenceIds: (string | null)[] = [null, ...document.evidenceUnits.keys()] - let best: PaletteDocumentMatch | null = null - - for (const evidenceId of evidenceIds) { - const built = buildAssignments(candidates, tokens, evidenceId) - if (!built) { - continue - } - const usedEvidenceId = built.usesEvidence ? evidenceId : null - const { rank, worstQuality, isContainerOnly } = rankAssignments({ - document, - assignments: built.assignments, - usesEvidence: built.usesEvidence, - wholeQuery, - exactIntent: args.exactIntent === true + const visibleSummaries = candidates.map((candidate) => + summarizeCandidates(candidate.visible, args.diagnostics) + ) + const evidenceSummaries = candidates.map( + (candidate) => + new Map( + [...candidate.byEvidenceId].map(([evidenceId, entries]) => [ + evidenceId, + summarizeCandidates(entries, args.diagnostics) + ]) + ) + ) + const ranked: RankedAssignment[] = [ + ...collectCompleteVisibleAssignments({ + document: args.document, + candidates, + normalizedQuery: args.normalizedQuery, + diagnostics: args.diagnostics }) - if (best && comparePaletteDocumentRank(best.rank, rank) <= 0) { - continue + ] + addRankedAssignment( + ranked, + args.document, + selectThresholdAssignment(visibleSummaries, args.diagnostics), + args.normalizedQuery, + null + ) + const matchedEvidenceIds = new Set() + for (const candidate of candidates) { + for (const evidenceId of candidate.byEvidenceId.keys()) { + matchedEvidenceIds.add(evidenceId) } - best = { - qualityClass: args.exactIntent + } + for (const evidenceId of matchedEvidenceIds) { + ranked.push( + ...collectScopeAssignments({ + document: args.document, + visibleSummaries, + evidenceSummaries, + normalizedQuery: args.normalizedQuery, + evidenceId, + diagnostics: args.diagnostics + }) + ) + } + if ((args.tokenCountBeforeDeduplication ?? args.tokens.length) === 1) { + ranked.push( + ...collectRecognizedIdentifierAssignments({ + document: args.document, + candidates, + normalizedQuery: args.normalizedQuery, + diagnostics: args.diagnostics + }) + ) + } + if (!ranked.length) { + return null + } + ranked.sort((a, b) => { + const rank = comparePaletteDocumentRank(a.rank, b.rank) + if (rank !== 0) { + return rank + } + return compareSelectedSourceOrder(a.selected, b.selected) + }) + const winner = ranked[0] + const winnerRank = args.exactIntent ? { ...winner.rank, destination: 0 } : winner.rank + const assignments = toTokenAssignments(args.tokens, winner.selected) + const worstQuality = winner.selected.reduce( + (worst, candidate) => + STRENGTH[candidate.quality] > STRENGTH[worst] ? candidate.quality : worst, + 'field-exact' + ) + const usesSupportingEvidence = assignments.some( + (assignment) => args.document.fieldById.get(assignment.fieldId)?.evidenceId + ) + return { + qualityClass: + winnerRank.destination === 0 ? 'exact-intent' : resolvePaletteResultQualityClass({ worstQuality, - usesSupportingEvidence: built.usesEvidence, - isContainerOnly + usesSupportingEvidence, + isContainerOnly: assignmentsAreContainerOnly(args.document, assignments) }), - rank, - assignments: built.assignments, - rangesByField: buildRangesByField(built.assignments), - supportingEvidence: buildSupportingEvidence(document, built.assignments, usedEvidenceId) - } + rank: winnerRank, + assignments, + rangesByField: buildRangesByField(assignments), + supportingEvidence: buildSupportingEvidence(args.document, assignments, winner.evidenceId) } - - return best } diff --git a/src/renderer/src/lib/palette-match/match-field-allocation.test.ts b/src/renderer/src/lib/palette-match/match-field-allocation.test.ts index 5b0021d2e71..85634d27282 100644 --- a/src/renderer/src/lib/palette-match/match-field-allocation.test.ts +++ b/src/renderer/src/lib/palette-match/match-field-allocation.test.ts @@ -23,6 +23,8 @@ describe('palette field quality allocation', () => { id: String(i), profile: profiles[i % profiles.length], text: 'scan daily 1234 workspace', + role: 'primary', + destinationEligible: true, ...(i % 2 === 0 ? { identifier: { kind: 'number' as const } } : {}) })! ) @@ -56,6 +58,8 @@ describe('palette quality restrictions remain local to each match', () => { id: 'id', profile: 'identifier', text: '12345', + role: 'primary', + destinationEligible: true, identifier: { kind } })! const prefix = createPaletteQueryToken('123', 0) @@ -75,7 +79,13 @@ describe('palette quality restrictions remain local to each match', () => { it.each(['structured-label', 'identifier', 'path', 'prose', 'exact-alias'])( 'preserves typo restrictions for %s without mutating the profile', (profile) => { - const field = indexPaletteField({ id: 'id', profile, text: 'scan' })! + const field = indexPaletteField({ + id: 'id', + profile, + text: 'scan', + role: 'primary', + destinationEligible: true + })! expect(matchPaletteField(field, createPaletteQueryToken('s', 0))).toEqual({ quality: 'field-prefix', ranges: [{ start: 0, end: 1 }] diff --git a/src/renderer/src/lib/palette-match/palette-assignment-inspection.ts b/src/renderer/src/lib/palette-match/palette-assignment-inspection.ts new file mode 100644 index 00000000000..760f827f6d7 --- /dev/null +++ b/src/renderer/src/lib/palette-match/palette-assignment-inspection.ts @@ -0,0 +1,16 @@ +import type { PaletteDocument, PaletteTokenAssignment } from './palette-document' + +export function assignmentsAreContainerOnly( + document: PaletteDocument, + assignments: readonly PaletteTokenAssignment[] +): boolean { + const tokenRoles = new Map() + for (const assignment of assignments) { + const isContainer = document.fieldById.get(assignment.fieldId)?.role === 'container' + tokenRoles.set( + assignment.tokenIndex, + (tokenRoles.get(assignment.tokenIndex) ?? true) && isContainer + ) + } + return tokenRoles.size > 0 && [...tokenRoles.values()].every(Boolean) +} diff --git a/src/renderer/src/lib/palette-match/palette-assignment-ranking.ts b/src/renderer/src/lib/palette-match/palette-assignment-ranking.ts new file mode 100644 index 00000000000..c14525ec6d7 --- /dev/null +++ b/src/renderer/src/lib/palette-match/palette-assignment-ranking.ts @@ -0,0 +1,302 @@ +import type { PaletteDocument, PaletteDocumentRank } from './palette-document' +import type { PaletteIndexedField } from './indexed-field' +import type { PaletteMatchDiagnostics, TokenCandidate, TokenCandidates } from './match-document' + +type CandidateMetric = 'recovery' | 'wordMatch' | 'coverage' | 'containerOnly' | 'strength' + +const CANDIDATE_METRICS: readonly CandidateMetric[] = [ + 'recovery', + 'wordMatch', + 'coverage', + 'containerOnly', + 'strength' +] + +const SELECTION_STEPS: readonly { + key: CandidateMetric + aggregate: 'maximum' | 'total' +}[] = [ + { key: 'recovery', aggregate: 'maximum' }, + { key: 'wordMatch', aggregate: 'maximum' }, + { key: 'coverage', aggregate: 'maximum' }, + { key: 'containerOnly', aggregate: 'total' }, + { key: 'recovery', aggregate: 'total' }, + { key: 'strength', aggregate: 'maximum' } +] + +function isDominatedBy(candidate: TokenCandidate, alternative: TokenCandidate): boolean { + // Visible fields precede evidence in source order, so equality is dominated too. + return CANDIDATE_METRICS.every((key) => alternative[key] <= candidate[key]) +} + +function phrasePlacement(field: PaletteIndexedField, normalizedQuery: string): number { + const text = field.text.normalized + if (text.startsWith(normalizedQuery)) { + return 0 + } + let index = text.indexOf(normalizedQuery, 1) + while (index !== -1) { + if (field.words.some((word) => word.start === index)) { + return 1 + } + index = text.indexOf(normalizedQuery, index + 1) + } + return 2 +} + +export function selectThresholdAssignment( + candidates: readonly TokenCandidate[][], + diagnostics?: PaletteMatchDiagnostics +): TokenCandidate[] | null { + if (candidates.some((entries) => entries.length === 0)) { + return null + } + let remaining = candidates.map((entries) => [...entries]) + for (const { key, aggregate } of SELECTION_STEPS) { + if (aggregate === 'total') { + remaining = remaining.map((entries) => { + let optimum = Number.POSITIVE_INFINITY + for (const candidate of entries) { + if (diagnostics) { + diagnostics.selectionCandidateVisits += 1 + } + optimum = Math.min(optimum, candidate[key]) + } + return entries.filter((candidate) => { + if (diagnostics) { + diagnostics.selectionCandidateVisits += 1 + } + return candidate[key] === optimum + }) + }) + continue + } + let optimum = 0 + for (const entries of remaining) { + let minimum = Number.POSITIVE_INFINITY + for (const candidate of entries) { + if (diagnostics) { + diagnostics.selectionCandidateVisits += 1 + } + minimum = Math.min(minimum, candidate[key]) + } + optimum = Math.max(optimum, minimum) + } + remaining = remaining.map((entries) => + entries.filter((candidate) => { + if (diagnostics) { + diagnostics.selectionCandidateVisits += 1 + } + return candidate[key] <= optimum + }) + ) + } + return remaining.map((entries) => entries[0]) +} + +function candidateMetricKey(candidate: TokenCandidate): number { + return ( + ((((candidate.recovery * 2 + candidate.wordMatch) * 4 + candidate.coverage) * 2 + + candidate.containerOnly) * + 6 + + candidate.strength) | + 0 + ) +} + +export function summarizeCandidates( + candidates: readonly TokenCandidate[], + diagnostics?: PaletteMatchDiagnostics +): TokenCandidate[] { + if (candidates.length < 2) { + return [...candidates] + } + const byMetric = new Map() + for (const candidate of candidates) { + if (diagnostics) { + diagnostics.selectionCandidateVisits += 1 + } + const key = candidateMetricKey(candidate) + if (!byMetric.has(key)) { + byMetric.set(key, candidate) + } + } + return [...byMetric.values()] +} + +function assignmentPlacement( + document: PaletteDocument, + selected: readonly TokenCandidate[], + normalizedQuery: string +): number { + const fieldId = selected[0]?.hits.length === 1 ? selected[0].hits[0].field.id : null + if (!fieldId) { + return 2 + } + if ( + selected.some( + (candidate) => candidate.hits.length !== 1 || candidate.hits[0].field.id !== fieldId + ) + ) { + return 2 + } + const field = document.fieldById.get(fieldId) + return field && !field.evidenceId ? phrasePlacement(field, normalizedQuery) : 2 +} + +function rankSelected( + selected: readonly TokenCandidate[], + destination: number, + placement: number +): PaletteDocumentRank { + return { + destination, + recovery: Math.max(...selected.map((candidate) => candidate.recovery)), + wordMatch: Math.max(...selected.map((candidate) => candidate.wordMatch)), + coverage: Math.max(...selected.map((candidate) => candidate.coverage)), + containerOnlyTokenCount: selected.filter((candidate) => candidate.containerOnly === 1).length, + recoveryTokenCount: selected.filter((candidate) => candidate.recovery > 0).length, + strength: Math.max(...selected.map((candidate) => candidate.strength)), + placement + } +} + +export type RankedAssignment = { + selected: readonly TokenCandidate[] + rank: PaletteDocumentRank + evidenceId: string | null +} + +export function addRankedAssignment( + target: RankedAssignment[], + document: PaletteDocument, + selected: readonly TokenCandidate[] | null, + normalizedQuery: string, + evidenceId: string | null, + destination = 2 +): void { + if (!selected) { + return + } + target.push({ + selected, + rank: rankSelected( + selected, + destination, + assignmentPlacement(document, selected, normalizedQuery) + ), + evidenceId + }) +} + +export function collectScopeAssignments(args: { + document: PaletteDocument + visibleSummaries: readonly TokenCandidate[][] + evidenceSummaries: ReadonlyMap[] + normalizedQuery: string + evidenceId: string + diagnostics?: PaletteMatchDiagnostics +}): RankedAssignment[] { + let addsUsefulCandidate = false + const scopeCandidates = args.visibleSummaries.map((visible, index) => { + const evidence = (args.evidenceSummaries[index].get(args.evidenceId) ?? []).filter( + (candidate) => !visible.some((alternative) => isDominatedBy(candidate, alternative)) + ) + if (!evidence.length) { + return visible + } + addsUsefulCandidate = true + return summarizeCandidates([...visible, ...evidence], args.diagnostics) + }) + if (!addsUsefulCandidate) { + return [] + } + const assignments: RankedAssignment[] = [] + const selected = selectThresholdAssignment(scopeCandidates, args.diagnostics) + if ( + !selected?.some((candidate) => + candidate.hits.some((hit) => hit.field.evidenceId === args.evidenceId) + ) + ) { + return assignments + } + addRankedAssignment(assignments, args.document, selected, args.normalizedQuery, args.evidenceId) + return assignments +} + +export function collectCompleteVisibleAssignments(args: { + document: PaletteDocument + candidates: readonly TokenCandidates[] + normalizedQuery: string + diagnostics?: PaletteMatchDiagnostics +}): RankedAssignment[] { + const candidateByField = args.candidates.map((tokenCandidates) => { + const byField = new Map() + for (const candidate of tokenCandidates.visible) { + if (args.diagnostics) { + args.diagnostics.selectionCandidateVisits += 1 + } + if (candidate.hits.length === 1) { + byField.set(candidate.hits[0].field.id, candidate) + } + } + return byField + }) + const assignments: RankedAssignment[] = [] + for (const field of args.document.visibleFields) { + const selected = candidateByField.map((byField) => byField.get(field.id)) + if (selected.some((candidate) => !candidate)) { + continue + } + addRankedAssignment( + assignments, + args.document, + selected as TokenCandidate[], + args.normalizedQuery, + null, + field.destinationEligible && field.text.normalized === args.normalizedQuery ? 1 : 2 + ) + } + return assignments +} + +export function collectRecognizedIdentifierAssignments(args: { + document: PaletteDocument + candidates: readonly TokenCandidates[] + normalizedQuery: string + diagnostics?: PaletteMatchDiagnostics +}): RankedAssignment[] { + const assignments: RankedAssignment[] = [] + if (args.normalizedQuery[0] !== '#' && args.normalizedQuery[0] !== '!') { + return assignments + } + for (const tokenCandidates of args.candidates) { + for (const entries of [tokenCandidates.visible, ...tokenCandidates.byEvidenceId.values()]) { + for (const candidate of entries) { + if (args.diagnostics) { + args.diagnostics.selectionCandidateVisits += 1 + } + if (candidate.hits.length !== 1) { + continue + } + const field = candidate.hits[0].field + if ( + field?.identifier?.kind === 'number' && + field.text.normalized === args.normalizedQuery && + candidate.hits[0].match.quality === 'field-exact' && + field.identifier.sigil === args.normalizedQuery[0] + ) { + addRankedAssignment( + assignments, + args.document, + [candidate], + args.normalizedQuery, + field.evidenceId, + 0 + ) + } + } + } + } + return assignments +} diff --git a/src/renderer/src/lib/palette-match/palette-document.ts b/src/renderer/src/lib/palette-match/palette-document.ts index 0934ff2cf41..8acad29ae96 100644 --- a/src/renderer/src/lib/palette-match/palette-document.ts +++ b/src/renderer/src/lib/palette-match/palette-document.ts @@ -1,6 +1,7 @@ import { indexPaletteFields, type PaletteFieldSource, + type PaletteEvidenceFieldSource as IndexedPaletteEvidenceFieldSource, type PaletteIndexedField } from './indexed-field' import type { MatchRange } from './normalized-text' @@ -18,8 +19,7 @@ export type PaletteEvidenceUnit = { accessibilityLabel: string } -export type PaletteEvidenceFieldSource = PaletteFieldSource & { - evidenceId: string +export type PaletteEvidenceFieldSource = IndexedPaletteEvidenceFieldSource & { /** Offset of this field's text inside its unit's rendered text. */ renderOffset: number } @@ -39,7 +39,6 @@ export type PaletteDocument = { renderOffsetByFieldId: ReadonlyMap /** Visible identity fields, cached because they carry the whole-query check. */ visibleFields: readonly PaletteIndexedField[] - fieldsByEvidenceId: ReadonlyMap fieldById: ReadonlyMap } @@ -74,25 +73,19 @@ export function buildPaletteDocument(input: PaletteDocumentInput): PaletteDocume evidenceUnits.set(entry.unit.id, entry.unit) for (const field of fields) { evidenceSources.push(field) - renderOffsetByFieldId.set(field.id, field.renderOffset) + if (!renderOffsetByFieldId.has(field.id)) { + renderOffsetByFieldId.set(field.id, field.renderOffset) + } } } const fields = indexPaletteFields([...input.visibleFields, ...evidenceSources]) - const fieldsByEvidenceId = new Map() const visibleFields: PaletteIndexedField[] = [] const fieldById = new Map() for (const field of fields) { fieldById.set(field.id, field) if (!field.evidenceId) { visibleFields.push(field) - continue - } - const bucket = fieldsByEvidenceId.get(field.evidenceId) - if (bucket) { - bucket.push(field) - } else { - fieldsByEvidenceId.set(field.evidenceId, [field]) } } @@ -106,7 +99,6 @@ export function buildPaletteDocument(input: PaletteDocumentInput): PaletteDocume evidenceUnits, renderOffsetByFieldId, visibleFields, - fieldsByEvidenceId, fieldById } } @@ -127,17 +119,18 @@ export type PaletteSupportingEvidence = { } export type PaletteDocumentRank = { - /** 0 when a recognized exact intent (such as a task URL) produced this row. */ - exactIntent: number - /** Tokens whose chosen assignment only matched container fields. */ + /** 0 recognized destination, 1 eligible equality, 2 structured, 3 fallback. */ + destination: number + recovery: number + wordMatch: number + coverage: number + /** Tokens proved only by container fields; fewer preserves direct-match relevance. */ containerOnlyTokenCount: number - /** 0 equality, 1 prefix, 2 word boundary, 3 none — whole query in visible text. */ - wholeQuery: number - worstQuality: number - /** 0 when every token landed on visible identity text. */ - usesSupportingEvidence: number - fuzzyTokenCount: number - fieldHopCount: number + /** Tokens that required compact or typo recovery; fewer breaks equal-severity ties. */ + recoveryTokenCount: number + strength: number + /** 0 prefix, 1 later word boundary, 2 distributed/other. */ + placement: number } export type PaletteDocumentMatch = { @@ -149,20 +142,70 @@ export type PaletteDocumentMatch = { } const RANK_KEYS: readonly (keyof PaletteDocumentRank)[] = [ - 'exactIntent', + 'destination', + 'recovery', + 'wordMatch', + 'coverage', 'containerOnlyTokenCount', - 'wholeQuery', - 'worstQuality', - 'usesSupportingEvidence', - 'fuzzyTokenCount', - 'fieldHopCount' + 'recoveryTokenCount', + 'strength', + 'placement' ] -export function comparePaletteDocumentRank(a: PaletteDocumentRank, b: PaletteDocumentRank): number { - for (const key of RANK_KEYS) { - if (a[key] !== b[key]) { - return a[key] - b[key] +const SEMANTIC_RANK_KEYS: readonly (keyof PaletteDocumentRank)[] = [ + 'destination', + 'recovery', + 'wordMatch', + 'coverage', + 'containerOnlyTokenCount', + 'recoveryTokenCount', + 'strength' +] + +function compareRankKeys( + a: PaletteDocumentRank, + b: PaletteDocumentRank, + keys: readonly (keyof PaletteDocumentRank)[] +): number { + for (const key of keys) { + const difference = a[key] - b[key] + if (difference !== 0) { + return difference } } return 0 } + +export function comparePaletteSemanticRank(a: PaletteDocumentRank, b: PaletteDocumentRank): number { + return compareRankKeys(a, b, SEMANTIC_RANK_KEYS) +} + +export function comparePaletteDocumentRank(a: PaletteDocumentRank, b: PaletteDocumentRank): number { + return compareRankKeys(a, b, RANK_KEYS) +} + +export function createRecognizedPaletteRank(): PaletteDocumentRank { + return { + destination: 0, + recovery: 0, + wordMatch: 0, + coverage: 0, + containerOnlyTokenCount: 0, + recoveryTokenCount: 0, + strength: 0, + placement: 0 + } +} + +export function createPaletteFallbackRank(): PaletteDocumentRank { + return { + destination: 3, + recovery: 0, + wordMatch: 0, + coverage: 0, + containerOnlyTokenCount: 0, + recoveryTokenCount: 0, + strength: 0, + placement: 0 + } +} diff --git a/src/renderer/src/lib/palette-match/palette-match-budget.ts b/src/renderer/src/lib/palette-match/palette-match-budget.ts index 864473a65e6..fad3669d1a8 100644 --- a/src/renderer/src/lib/palette-match/palette-match-budget.ts +++ b/src/renderer/src/lib/palette-match/palette-match-budget.ts @@ -21,18 +21,24 @@ export const PALETTE_MATCH_BUDGET = { * Ceiling on `matchPaletteField` calls per candidate for the worst query. * Deterministic — it counts work, not time — so it catches a fan-out * regression (re-matching every field per evidence unit, say) on any machine. - * Measured 45: 15 fields across the 3 tokens scanned before the first miss. + * Measured 240: 15 fields across every token in the accepted fixture. */ - fieldMatchesPerCandidate: 60, + fieldMatchesPerCandidate: 280, + /** Fixed-domain selection visits per accepted candidate. Measured 3,008. */ + selectionCandidateVisitsPerCandidate: 3_600, /** Milliseconds to normalize every document once (cold open), fastest sample. */ coldBuildMs: 900, /** Milliseconds to match the whole corpus against one prepared query, fastest sample. */ warmMatchMs: 220, + /** Milliseconds for warm worktree search plus entity-rank sorting. Measured 106.5 ms. */ + fullSearchSortMs: 180, /** * Megabytes of indexed text and offset tables the normalized documents retain. * Measured deterministically rather than from `heapUsed`, which is polluted by * whatever else shares the vitest worker. Process heap for the same corpus * measured ~40 MB in isolation. */ - documentPayloadMb: 24 + documentPayloadMb: 24, + /** Megabytes retained by the accepted query's match/range results. Measured 0.69 MB. */ + matchPayloadMb: 1 } as const diff --git a/src/renderer/src/lib/palette-match/palette-match-core.test.ts b/src/renderer/src/lib/palette-match/palette-match-core.test.ts index de7bd7696c7..ced386fe544 100644 --- a/src/renderer/src/lib/palette-match/palette-match-core.test.ts +++ b/src/renderer/src/lib/palette-match/palette-match-core.test.ts @@ -23,13 +23,22 @@ function run(input: PaletteDocumentInput, query: string) { return matchPaletteDocument({ document: buildPaletteDocument(input), tokens: prepared.tokens, - normalizedQuery: prepared.normalized + normalizedQuery: prepared.normalized, + tokenCountBeforeDeduplication: prepared.tokenCountBeforeDeduplication }) } const labelOnly = (text: string): PaletteDocumentInput => ({ id: 'doc', - visibleFields: [{ id: 'name', profile: 'structured-label', text }], + visibleFields: [ + { + id: 'name', + profile: 'structured-label', + text, + role: 'primary', + destinationEligible: true + } + ], evidence: [] }) @@ -50,8 +59,8 @@ describe('palette query preparation', () => { // Why: field text is always single-spaced, so an uncollapsed run could never satisfy // the whole-query equality tier and the exactly-named row silently lost its rank. expect(ready('scan daily').normalized).toBe('scan daily') - expect(run(labelOnly('scan daily'), 'scan daily')?.rank.wholeQuery).toBe( - run(labelOnly('scan daily'), 'scan daily')?.rank.wholeQuery + expect(run(labelOnly('scan daily'), 'scan daily')?.rank.placement).toBe( + run(labelOnly('scan daily'), 'scan daily')?.rank.placement ) }) @@ -185,7 +194,7 @@ describe('structured label matching', () => { it('applies light typo matching to long letter-only words', () => { expect(run(document, 'dayly')).not.toBeNull() - expect(run(document, 'scam')?.rank.fuzzyTokenCount).toBe(1) + expect(run(document, 'scam')?.rank.recovery).toBe(1) }) it('limits single Latin characters to word equality or prefix', () => { @@ -205,7 +214,15 @@ describe('structured label matching', () => { describe('identifier fields', () => { const review = (sigil: '#' | '!'): PaletteDocumentInput => ({ id: 'doc', - visibleFields: [{ id: 'name', profile: 'structured-label', text: 'reconnect flow' }], + visibleFields: [ + { + id: 'name', + profile: 'structured-label', + text: 'reconnect flow', + role: 'primary', + destinationEligible: true + } + ], evidence: [ { unit: { id: 'review', kind: 'pr', text: '#4123 · Fix reconnect', accessibilityLabel: 'PR' }, @@ -252,7 +269,7 @@ describe('identifier fields', () => { it('combines an identity token with one evidence token', () => { const match = run(review('#'), 'reconnect 4123') - expect(match?.rank.usesSupportingEvidence).toBe(1) + expect(match?.rank.coverage).toBe(3) expect(match?.supportingEvidence).toHaveLength(1) }) }) @@ -262,7 +279,15 @@ describe('duplicate evidence unit ids', () => { // host:port:pid, so a parent and a forked child both survive. const duplicateUnits: PaletteDocumentInput = { id: 'doc', - visibleFields: [{ id: 'name', profile: 'structured-label', text: 'checkout' }], + visibleFields: [ + { + id: 'name', + profile: 'structured-label', + text: 'checkout', + role: 'primary', + destinationEligible: true + } + ], evidence: [ { unit: { @@ -289,7 +314,7 @@ describe('duplicate evidence unit ids', () => { profile: 'structured-label', text: 'node', evidenceId: 'port:3000', - renderOffset: 7 + renderOffset: 0 } ] } @@ -322,7 +347,15 @@ describe('duplicate evidence unit ids', () => { describe('evidence limits', () => { const twoUnits: PaletteDocumentInput = { id: 'doc', - visibleFields: [{ id: 'name', profile: 'structured-label', text: 'checkout' }], + visibleFields: [ + { + id: 'name', + profile: 'structured-label', + text: 'checkout', + role: 'primary', + destinationEligible: true + } + ], evidence: [ { unit: { id: 'port:3000', kind: 'port', text: '3000 · node', accessibilityLabel: 'Port' }, @@ -367,7 +400,7 @@ describe('evidence limits', () => { it('prefers visible evidence over supporting evidence', () => { const match = run(twoUnits, 'checkout') - expect(match?.rank.usesSupportingEvidence).toBe(0) + expect(match?.rank.coverage).toBe(0) expect(match?.supportingEvidence).toHaveLength(0) }) }) @@ -391,14 +424,26 @@ describe('container field matching', () => { const tabDoc: PaletteDocumentInput = { id: 'tab-1', visibleFields: [ - { id: 'title', profile: 'structured-label', text: 'README.md' }, - { id: 'worktree', profile: 'structured-label', text: 'STA-4360-feature', isContainer: true } + { + id: 'title', + profile: 'structured-label', + text: 'README.md', + role: 'primary', + destinationEligible: true + }, + { + id: 'worktree', + profile: 'structured-label', + text: 'STA-4360-feature', + role: 'container', + destinationEligible: false + } ], evidence: [] } const match = run(tabDoc, '4360') expect(match).not.toBeNull() - expect(match?.rank.containerOnlyTokenCount).toBe(1) + expect(match?.rank.coverage).toBe(2) expect(match?.qualityClass).toBe('exact-evidence') }) @@ -406,14 +451,26 @@ describe('container field matching', () => { const tabDoc: PaletteDocumentInput = { id: 'tab-1', visibleFields: [ - { id: 'title', profile: 'structured-label', text: 'wsl-transcript-4360.ts' }, - { id: 'worktree', profile: 'structured-label', text: 'STA-4360-feature', isContainer: true } + { + id: 'title', + profile: 'structured-label', + text: 'wsl-transcript-4360.ts', + role: 'primary', + destinationEligible: true + }, + { + id: 'worktree', + profile: 'structured-label', + text: 'STA-4360-feature', + role: 'container', + destinationEligible: false + } ], evidence: [] } const match = run(tabDoc, '4360') expect(match).not.toBeNull() - expect(match?.rank.containerOnlyTokenCount).toBe(0) + expect(match?.rank.coverage).toBe(0) expect(match?.qualityClass).toBe('exact-visible') }) @@ -422,8 +479,20 @@ describe('container field matching', () => { { id: 'direct', visibleFields: [ - { id: 'title', profile: 'structured-label', text: 'alpha' }, - { id: 'path', profile: 'structured-label', text: 'beta' } + { + id: 'title', + profile: 'structured-label', + text: 'alpha', + role: 'primary', + destinationEligible: true + }, + { + id: 'path', + profile: 'structured-label', + text: 'beta', + role: 'secondary', + destinationEligible: true + } ], evidence: [] }, @@ -433,8 +502,20 @@ describe('container field matching', () => { { id: 'mixed', visibleFields: [ - { id: 'title', profile: 'structured-label', text: 'alpha' }, - { id: 'worktree', profile: 'structured-label', text: 'beta', isContainer: true } + { + id: 'title', + profile: 'structured-label', + text: 'alpha', + role: 'primary', + destinationEligible: true + }, + { + id: 'worktree', + profile: 'structured-label', + text: 'beta', + role: 'container', + destinationEligible: false + } ], evidence: [] }, @@ -443,8 +524,8 @@ describe('container field matching', () => { expect(direct).not.toBeNull() expect(mixed).not.toBeNull() - expect(direct?.rank.containerOnlyTokenCount).toBe(0) - expect(mixed?.rank.containerOnlyTokenCount).toBe(1) + expect(direct?.rank.coverage).toBe(1) + expect(mixed?.rank.coverage).toBe(2) if (direct && mixed) { expect(comparePaletteDocumentRank(direct.rank, mixed.rank)).toBeLessThan(0) } diff --git a/src/renderer/src/lib/palette-match/palette-match-performance.test.ts b/src/renderer/src/lib/palette-match/palette-match-performance.test.ts index 13fbced916f..0a7a9d93a61 100644 --- a/src/renderer/src/lib/palette-match/palette-match-performance.test.ts +++ b/src/renderer/src/lib/palette-match/palette-match-performance.test.ts @@ -1,15 +1,19 @@ import { describe, expect, it, vi } from 'vitest' import { PALETTE_MATCH_BUDGET } from './palette-match-budget' -import { matchPaletteDocument } from './match-document' +import { matchPaletteDocument, type PaletteMatchDiagnostics } from './match-document' import * as matchFieldModule from './match-field' import { preparePaletteQuery } from './palette-query' import { buildWorktreePaletteDocuments } from '../worktree-palette-document' -import type { PaletteDocument } from './palette-document' +import { searchWorktreeDocuments } from '../worktree-palette-search' +import { comparePaletteEntityRanks, createPaletteSearchContext } from './palette-ranking' +import { buildPaletteDocument, type PaletteDocument } from './palette-document' import type { PaletteQueryToken } from './palette-query' import type { Repo } from '../../../../shared/repo-types' import type { Worktree } from '../../../../shared/worktree/types' const { candidateCount, tokenCount } = PALETTE_MATCH_BUDGET +const QUERY_TOKENS = Array.from({ length: tokenCount }, (_, index) => `token${index}`) +const QUERY_TEXT = QUERY_TOKENS.join(' ') const LONG_COMMENT = `Blocked on the staging relay while the host reconnects; see the runbook for the escalation path and the rollback steps before retrying the deploy. `.repeat( @@ -35,11 +39,11 @@ function makeWorktree(index: number): Worktree { repoId: 'repo-1', path: `/work/wt-${index}`, head: `${index}`.padStart(7, 'a'), - branch: `refs/heads/feature/workspace-${index}-rebuild`, + branch: `refs/heads/${QUERY_TOKENS.join('-')}`, isBare: false, isMainWorktree: false, - displayName: `scan daily 1.4.${index} · 2026-08-13 · ${`${index}`.padStart(7, '9')}`, - comment: LONG_COMMENT, + displayName: `${QUERY_TEXT} workspace ${index}`, + comment: `${QUERY_TEXT}. ${LONG_COMMENT}`, linkedIssue: 1000 + index, linkedPR: 2000 + index, linkedLinearIssue: `ORC-${index}`, @@ -47,7 +51,7 @@ function makeWorktree(index: number): Worktree { provider: 'linear', type: 'issue', number: index, - title: `Rework the palette ranking pipeline for workspace ${index}`, + title: `${QUERY_TEXT} work item ${index}`, url: `https://linear.app/acme/issue/ORC-${index}`, linearIdentifier: `ORC-${index}` }, @@ -56,7 +60,7 @@ function makeWorktree(index: number): Worktree { automationId: 'auto-1', automationNameSnapshot: 'Nightly review', automationRunId: `run-${index}`, - automationRunTitleSnapshot: `Scan daily sweep ${index}`, + automationRunTitleSnapshot: `${QUERY_TEXT} sweep ${index}`, createdAt: Date.UTC(2026, 7, 13), executionTargetType: 'local', executionTargetId: 'repo-1', @@ -77,7 +81,7 @@ const ports = new Map( const issueCache = Object.fromEntries( worktrees.map((worktree, index) => [ `/repos/orca::${worktree.id}`, - { data: { number: 1000 + index, title: `Cached issue title ${index}` } } + { data: { number: 1000 + index, title: `${QUERY_TEXT} issue ${index}` } } ]) ) @@ -85,12 +89,10 @@ const sources = { repoMap, issueCache, workspacePortsByWorktreeId: ports, - hostLabelByWorktreeId: new Map(worktrees.map((worktree) => [worktree.id, 'bastion-eu'])) + hostLabelByWorktreeId: new Map(worktrees.map((worktree) => [worktree.id, QUERY_TEXT])) } -const WORST_QUERY = Array.from({ length: tokenCount }, (_, index) => - index === 0 ? 'scan' : index === 1 ? 'daily' : `token${index}` -).join(' ') +const WORST_QUERY = QUERY_TEXT function prepareWorstQuery(): { tokens: readonly PaletteQueryToken[]; normalized: string } { const prepared = preparePaletteQuery(WORST_QUERY) @@ -102,14 +104,43 @@ function prepareWorstQuery(): { tokens: readonly PaletteQueryToken[]; normalized const preparedQuery = prepareWorstQuery() -function matchEveryDocument(documents: ReadonlyMap): void { +function matchEveryDocument( + documents: ReadonlyMap, + diagnostics?: PaletteMatchDiagnostics +): ReturnType[] { + const matches: ReturnType[] = [] for (const document of documents.values()) { - matchPaletteDocument({ - document, - tokens: preparedQuery.tokens, - normalizedQuery: preparedQuery.normalized - }) + matches.push( + matchPaletteDocument({ + document, + tokens: preparedQuery.tokens, + normalizedQuery: preparedQuery.normalized, + diagnostics + }) + ) } + return matches +} + +function retainedMatchPayloadBytes(matches: ReturnType[]): number { + let bytes = 0 + for (const match of matches) { + if (!match) { + continue + } + for (const assignment of match.assignments) { + bytes += assignment.fieldId.length * 2 + 16 + bytes += assignment.ranges.length * 16 + } + for (const [fieldId, ranges] of match.rangesByField) { + bytes += fieldId.length * 2 + ranges.length * 16 + } + for (const evidence of match.supportingEvidence) { + bytes += (evidence.id.length + evidence.kind.length + evidence.text.length) * 2 + bytes += evidence.ranges.length * 16 + } + } + return bytes } /** @@ -143,7 +174,7 @@ describe('palette matcher performance budget', () => { const documents = buildWorktreePaletteDocuments(worktrees, sources) // Warm the matcher before timing so JIT compilation is not part of the samples. - matchEveryDocument(documents) + expect(matchEveryDocument(documents).filter(Boolean)).toHaveLength(candidateCount) const samples = timeRepeatedly(() => matchEveryDocument(documents), 10) expect(fastestSample(samples)).toBeLessThan(PALETTE_MATCH_BUDGET.warmMatchMs) @@ -163,6 +194,154 @@ describe('palette matcher performance budget', () => { } }) + it('bounds candidate selection work per accepted candidate', () => { + const documents = buildWorktreePaletteDocuments(worktrees, sources) + const diagnostics: PaletteMatchDiagnostics = { selectionCandidateVisits: 0 } + matchEveryDocument(documents, diagnostics) + expect(diagnostics.selectionCandidateVisits / documents.size).toBeLessThan( + PALETTE_MATCH_BUDGET.selectionCandidateVisitsPerCandidate + ) + }) + + it('does not revisit an all-visible assignment for unmatched evidence units', () => { + const buildDocument = (evidenceCount: number): PaletteDocument => + buildPaletteDocument({ + id: `visible-${evidenceCount}`, + visibleFields: [ + { + id: 'title', + profile: 'structured-label', + text: 'atlas', + role: 'primary', + destinationEligible: true + } + ], + evidence: Array.from({ length: evidenceCount }, (_, index) => ({ + unit: { + id: `evidence-${index}`, + kind: 'comment', + text: `unrelated ${index}`, + accessibilityLabel: 'Comment' + }, + fields: [ + { + id: `evidence-field-${index}`, + profile: 'prose' as const, + text: `unrelated ${index}`, + evidenceId: `evidence-${index}`, + renderOffset: 0 + } + ] + })) + }) + const query = preparePaletteQuery('atlas') + if (query.state !== 'ready') { + throw new Error('Expected ready query') + } + const selectionVisits = (document: PaletteDocument): number => { + const diagnostics: PaletteMatchDiagnostics = { selectionCandidateVisits: 0 } + matchPaletteDocument({ + document, + tokens: query.tokens, + normalizedQuery: query.normalized, + diagnostics + }) + return diagnostics.selectionCandidateVisits + } + + expect(selectionVisits(buildDocument(100))).toBe(selectionVisits(buildDocument(0))) + }) + + it('does not revisit an all-visible assignment for dominated evidence matches', () => { + const buildDocument = (evidenceCount: number): PaletteDocument => + buildPaletteDocument({ + id: `visible-matched-${evidenceCount}`, + visibleFields: [ + { + id: 'title', + profile: 'structured-label', + text: 'atlas', + role: 'primary', + destinationEligible: true + } + ], + evidence: Array.from({ length: evidenceCount }, (_, index) => ({ + unit: { + id: `evidence-${index}`, + kind: 'comment', + text: 'atlas', + accessibilityLabel: 'Comment' + }, + fields: [ + { + id: `evidence-field-${index}`, + profile: 'prose' as const, + text: 'atlas', + evidenceId: `evidence-${index}`, + renderOffset: 0 + } + ] + })) + }) + const query = preparePaletteQuery('atlas') + if (query.state !== 'ready') { + throw new Error('Expected ready query') + } + const selectionVisits = (document: PaletteDocument): number => { + const diagnostics: PaletteMatchDiagnostics = { selectionCandidateVisits: 0 } + const match = matchPaletteDocument({ + document, + tokens: query.tokens, + normalizedQuery: query.normalized, + diagnostics + }) + expect(match?.supportingEvidence).toEqual([]) + return diagnostics.selectionCandidateVisits + } + + expect(selectionVisits(buildDocument(100))).toBe(selectionVisits(buildDocument(0))) + }) + + it('keeps retained match and range payload within budget', () => { + const documents = buildWorktreePaletteDocuments(worktrees, sources) + const matches = matchEveryDocument(documents) + expect(retainedMatchPayloadBytes(matches) / (1024 * 1024)).toBeLessThan( + PALETTE_MATCH_BUDGET.matchPayloadMb + ) + }) + + it('searches and sorts the accepted corpus within budget', () => { + const documents = buildWorktreePaletteDocuments(worktrees, sources) + const context = createPaletteSearchContext(Date.UTC(2026, 8, 5)) + const searchAndSort = (): void => { + searchWorktreeDocuments({ + worktrees, + query: WORST_QUERY, + documents, + repoMap, + context + }).sort((a, b) => + comparePaletteEntityRanks( + { + rank: a.rank!, + activity: a.activity, + position: 0, + identity: `${a.worktreeHostId ?? ''}:${a.worktreeId}` + }, + { + rank: b.rank!, + activity: b.activity, + position: 0, + identity: `${b.worktreeHostId ?? ''}:${b.worktreeId}` + } + ) + ) + } + searchAndSort() + const samples = timeRepeatedly(searchAndSort, 10) + expect(fastestSample(samples)).toBeLessThan(PALETTE_MATCH_BUDGET.fullSearchSortMs) + }) + it('keeps the retained document payload within budget', () => { const documents = buildWorktreePaletteDocuments(worktrees, sources) expect(documents.size).toBe(candidateCount) diff --git a/src/renderer/src/lib/palette-match/palette-match-rendering.ts b/src/renderer/src/lib/palette-match/palette-match-rendering.ts new file mode 100644 index 00000000000..a104012e0a9 --- /dev/null +++ b/src/renderer/src/lib/palette-match/palette-match-rendering.ts @@ -0,0 +1,50 @@ +import { mergeMatchRanges, type MatchRange } from './normalized-text' +import type { + PaletteDocument, + PaletteSupportingEvidence, + PaletteTokenAssignment +} from './palette-document' + +export function buildSupportingEvidence( + document: PaletteDocument, + assignments: readonly PaletteTokenAssignment[], + evidenceId: string | null +): PaletteSupportingEvidence[] { + const unit = evidenceId ? document.evidenceUnits.get(evidenceId) : undefined + if (!unit) { + return [] + } + const ranges: MatchRange[] = [] + for (const assignment of assignments) { + const offset = document.renderOffsetByFieldId.get(assignment.fieldId) + if (offset === undefined) { + continue + } + for (const range of assignment.ranges) { + const start = Math.min(range.start + offset, unit.text.length) + const end = Math.min(range.end + offset, unit.text.length) + if (start < end) { + ranges.push({ start, end }) + } + } + } + if (!ranges.length) { + return [] + } + return [{ ...unit, ranges: mergeMatchRanges(ranges) }] +} + +export function buildRangesByField( + assignments: readonly PaletteTokenAssignment[] +): Map { + const byField = new Map() + for (const assignment of assignments) { + const bucket = byField.get(assignment.fieldId) + if (bucket) { + bucket.push(...assignment.ranges) + } else { + byField.set(assignment.fieldId, [...assignment.ranges]) + } + } + return new Map([...byField].map(([id, ranges]) => [id, mergeMatchRanges(ranges)])) +} diff --git a/src/renderer/src/lib/palette-match/palette-query.ts b/src/renderer/src/lib/palette-match/palette-query.ts index f0149bce59b..66ee5ccf737 100644 --- a/src/renderer/src/lib/palette-match/palette-query.ts +++ b/src/renderer/src/lib/palette-match/palette-query.ts @@ -31,7 +31,13 @@ export type PaletteQueryToken = { export type PreparedPaletteQuery = | { state: 'empty' } | { state: 'invalid'; reason: 'too-large' | 'too-many-tokens' } - | { state: 'ready'; normalized: string; tokens: readonly PaletteQueryToken[] } + | { + state: 'ready' + normalized: string + tokens: readonly PaletteQueryToken[] + /** Count before duplicate-token removal; destination recognition uses the complete query. */ + tokenCountBeforeDeduplication: number + } function splitComponents(text: string): string[] { const components: string[] = [] @@ -82,10 +88,7 @@ export function preparePaletteQuery(query: string): PreparedPaletteQuery { if (isWorktreePaletteQueryTooLarge(query)) { return { state: 'invalid', reason: 'too-large' } } - // Why collapse runs: field text is always single-spaced, so an uncollapsed double - // space can never satisfy the whole-query equality/prefix tier and the exact-name - // match silently loses its rank. Safe here — this string feeds only scoreWholeQuery - // and carries no offset mapping back into the source text. + // Field text is single-spaced, and this value has no source-offset mapping to preserve. const normalized = normalizePaletteText(query).normalized.replace(/ +/g, ' ').trim() if (!normalized) { return { state: 'empty' } @@ -93,7 +96,8 @@ export function preparePaletteQuery(query: string): PreparedPaletteQuery { const seen = new Set() const tokens: PaletteQueryToken[] = [] - for (const raw of normalized.split(' ')) { + const rawTokens = normalized.split(' ').filter(Boolean) + for (const raw of rawTokens) { if (!raw || seen.has(raw)) { continue } @@ -107,7 +111,12 @@ export function preparePaletteQuery(query: string): PreparedPaletteQuery { if (tokens.length > PALETTE_QUERY_MAX_TOKENS) { return { state: 'invalid', reason: 'too-many-tokens' } } - return { state: 'ready', normalized, tokens } + return { + state: 'ready', + normalized, + tokens, + tokenCountBeforeDeduplication: rawTokens.length + } } export function isLetterOnlyWord(word: string): boolean { diff --git a/src/renderer/src/lib/palette-match/palette-ranking.test.ts b/src/renderer/src/lib/palette-match/palette-ranking.test.ts new file mode 100644 index 00000000000..7dd6a026774 --- /dev/null +++ b/src/renderer/src/lib/palette-match/palette-ranking.test.ts @@ -0,0 +1,167 @@ +import { describe, expect, it } from 'vitest' +import type { PaletteDocumentRank } from './palette-document' +import { + comparePaletteEntityRanks, + createPaletteSearchContext, + encodePaletteIdentity, + maxValidPaletteActivityTimestamp, + preparePaletteActivity +} from './palette-ranking' + +const HOUR = 60 * 60 * 1000 +const DAY = 24 * HOUR +const WEEK = 7 * DAY +const NOW = 100 * DAY + +function rank(overrides: Partial = {}): PaletteDocumentRank { + return { + destination: 2, + recovery: 0, + wordMatch: 0, + coverage: 0, + containerOnlyTokenCount: 0, + recoveryTokenCount: 0, + strength: 0, + placement: 2, + ...overrides + } +} + +function item(args: { + rank?: PaletteDocumentRank + timestamp?: number | null + position?: number | readonly number[] + identity?: string +}) { + const context = createPaletteSearchContext(NOW) + return { + rank: args.rank ?? rank(), + activity: preparePaletteActivity(args.timestamp, context), + position: args.position ?? 0, + identity: args.identity ?? 'id' + } +} + +describe('palette activity preparation', () => { + it.each([ + [NOW, 0], + [NOW - HOUR + 1, 0], + [NOW - HOUR, 1], + [NOW - DAY, 2], + [NOW - WEEK, 3], + [NOW - 2 * WEEK, 4], + [NOW - 3 * WEEK, 5], + [NOW - 80 * DAY, 13] + ])('places timestamp %s in bucket %s', (timestamp, bucket) => { + expect(preparePaletteActivity(timestamp, createPaletteSearchContext(NOW)).ageBucket).toBe( + bucket + ) + }) + + it('keeps every known old timestamp ahead of invalid or unknown activity', () => { + const context = createPaletteSearchContext(NOW) + const old = preparePaletteActivity(1, context) + for (const invalid of [undefined, null, 0, -1, Number.NaN, Number.POSITIVE_INFINITY]) { + const unknown = preparePaletteActivity(invalid, context) + expect(old.ageBucket).not.toBeNull() + expect(unknown).toEqual({ ageBucket: null, timestamp: 0 }) + } + }) + + it('clamps future clocks to the evaluation clock', () => { + expect(preparePaletteActivity(NOW + DAY, createPaletteSearchContext(NOW))).toEqual({ + ageBucket: 0, + timestamp: NOW + }) + }) + + it('ignores invalid values while reducing activity signals', () => { + expect( + maxValidPaletteActivityTimestamp([100, Number.NaN, 300, Number.POSITIVE_INFINITY, -1]) + ).toBe(300) + }) +}) + +describe('palette entity comparator', () => { + it('keeps semantics ahead of recency', () => { + const oldExact = item({ rank: rank({ strength: 0 }), timestamp: NOW - 80 * DAY }) + const recentWeak = item({ rank: rank({ strength: 1 }), timestamp: NOW }) + expect(comparePaletteEntityRanks(oldExact, recentWeak)).toBeLessThan(0) + }) + + it('uses age bucket before placement and placement before timestamp within a bucket', () => { + const recentLater = item({ rank: rank({ placement: 2 }), timestamp: NOW - 30 * 60 * 1000 }) + const olderPrefix = item({ rank: rank({ placement: 0 }), timestamp: NOW - 2 * HOUR }) + expect(comparePaletteEntityRanks(recentLater, olderPrefix)).toBeLessThan(0) + + const sameBucketNewer = item({ rank: rank({ placement: 2 }), timestamp: NOW - 10 * 60 * 1000 }) + const sameBucketPrefix = item({ rank: rank({ placement: 0 }), timestamp: NOW - 50 * 60 * 1000 }) + expect(comparePaletteEntityRanks(sameBucketPrefix, sameBucketNewer)).toBeLessThan(0) + }) + + it('uses timestamp, position tuple, and fixed code-unit identity for successive ties', () => { + expect( + comparePaletteEntityRanks( + item({ timestamp: NOW - 1, position: 9, identity: 'z' }), + item({ timestamp: NOW - 2, position: 0, identity: 'a' }) + ) + ).toBeLessThan(0) + expect( + comparePaletteEntityRanks( + item({ timestamp: NOW, position: [0, 9], identity: 'z' }), + item({ timestamp: NOW, position: [1, 0], identity: 'a' }) + ) + ).toBeLessThan(0) + expect( + comparePaletteEntityRanks( + item({ timestamp: NOW, position: 0, identity: 'A' }), + item({ timestamp: NOW, position: 0, identity: 'a' }) + ) + ).toBeLessThan(0) + }) + + it('is permutation-invariant for unique qualified identities', () => { + const rows = [ + item({ + timestamp: NOW - 2 * HOUR, + identity: encodePaletteIdentity(['browser', 'host-b', '1']) + }), + item({ + timestamp: NOW - 20 * 60 * 1000, + identity: encodePaletteIdentity(['tab', 'host-a', '1']) + }), + item({ + timestamp: NOW - 20 * 60 * 1000, + identity: encodePaletteIdentity(['tab', 'host-b', '1']) + }) + ] + const expected = [...rows].sort(comparePaletteEntityRanks).map((row) => row.identity) + expect( + rows + .toReversed() + .sort(comparePaletteEntityRanks) + .map((row) => row.identity) + ).toEqual(expected) + }) + + it('separates future-clamped clocks once evaluation passes the earlier stamp', () => { + const earlierFuture = NOW + HOUR + const laterFuture = NOW + 2 * HOUR + const before = createPaletteSearchContext(NOW) + const afterEarlier = createPaletteSearchContext(NOW + HOUR + 1) + const build = (timestamp: number, context: ReturnType) => ({ + rank: rank(), + activity: preparePaletteActivity(timestamp, context), + position: 0, + identity: String(timestamp) + }) + + expect(build(earlierFuture, before).activity).toEqual(build(laterFuture, before).activity) + expect( + comparePaletteEntityRanks( + build(laterFuture, afterEarlier), + build(earlierFuture, afterEarlier) + ) + ).toBeLessThan(0) + }) +}) diff --git a/src/renderer/src/lib/palette-match/palette-ranking.ts b/src/renderer/src/lib/palette-match/palette-ranking.ts new file mode 100644 index 00000000000..89a99b07548 --- /dev/null +++ b/src/renderer/src/lib/palette-match/palette-ranking.ts @@ -0,0 +1,109 @@ +import { comparePaletteSemanticRank, type PaletteDocumentRank } from './palette-document' + +const HOUR_MS = 60 * 60 * 1000 +const DAY_MS = 24 * HOUR_MS +const WEEK_MS = 7 * DAY_MS + +export type PaletteSearchContext = { nowMs: number } + +export type PaletteActivityRank = { + ageBucket: number | null + timestamp: number +} + +export type PaletteEntityRankInput = { + rank: PaletteDocumentRank + activity: PaletteActivityRank + position: number | readonly number[] + identity: string +} + +export function createPaletteSearchContext(nowMs: number): PaletteSearchContext { + if (!Number.isFinite(nowMs) || nowMs <= 0) { + throw new Error('Palette search context requires a finite positive nowMs') + } + return { nowMs } +} + +export function preparePaletteActivity( + value: number | null | undefined, + context: PaletteSearchContext +): PaletteActivityRank { + if (!Number.isFinite(value) || (value ?? 0) <= 0) { + return { ageBucket: null, timestamp: 0 } + } + const timestamp = Math.min(value as number, context.nowMs) + const ageMs = context.nowMs - timestamp + const ageBucket = + ageMs < HOUR_MS + ? 0 + : ageMs < DAY_MS + ? 1 + : ageMs < WEEK_MS + ? 2 + : 3 + Math.floor((ageMs - WEEK_MS) / WEEK_MS) + return { ageBucket, timestamp } +} + +/** Latest usable activity signal before evaluation-time future clamping. */ +export function maxValidPaletteActivityTimestamp( + values: readonly (number | null | undefined)[] +): number | null { + let maximum: number | null = null + for (const value of values) { + if ( + typeof value === 'number' && + Number.isFinite(value) && + value > 0 && + (maximum === null || value > maximum) + ) { + maximum = value + } + } + return maximum +} + +function compareCodeUnits(a: string, b: string): number { + return a < b ? -1 : a > b ? 1 : 0 +} + +/** Length-prefixing keeps identities collision-safe even when parts contain separators. */ +export function encodePaletteIdentity(parts: readonly string[]): string { + return parts.map((part) => `${part.length}:${part}`).join('') +} + +export function comparePaletteEntityRanks( + a: PaletteEntityRankInput, + b: PaletteEntityRankInput +): number { + const semantic = comparePaletteSemanticRank(a.rank, b.rank) + if (semantic !== 0) { + return semantic + } + + if (a.activity.ageBucket !== b.activity.ageBucket) { + if (a.activity.ageBucket === null) { + return 1 + } + if (b.activity.ageBucket === null) { + return -1 + } + return a.activity.ageBucket - b.activity.ageBucket + } + if (a.rank.placement !== b.rank.placement) { + return a.rank.placement - b.rank.placement + } + if (a.activity.timestamp !== b.activity.timestamp) { + return b.activity.timestamp - a.activity.timestamp + } + const aPosition = typeof a.position === 'number' ? [a.position] : a.position + const bPosition = typeof b.position === 'number' ? [b.position] : b.position + const count = Math.max(aPosition.length, bPosition.length) + for (let index = 0; index < count; index += 1) { + const difference = (aPosition[index] ?? 0) - (bPosition[index] ?? 0) + if (difference !== 0) { + return difference + } + } + return compareCodeUnits(a.identity, b.identity) +} diff --git a/src/renderer/src/lib/palette-match/palette-selection-source-order.ts b/src/renderer/src/lib/palette-match/palette-selection-source-order.ts new file mode 100644 index 00000000000..3b5857829de --- /dev/null +++ b/src/renderer/src/lib/palette-match/palette-selection-source-order.ts @@ -0,0 +1,21 @@ +import type { TokenCandidate } from './match-document' + +export function compareSelectedSourceOrder( + a: readonly TokenCandidate[], + b: readonly TokenCandidate[] +): number { + for (let tokenIndex = 0; tokenIndex < a.length; tokenIndex += 1) { + const aHits = a[tokenIndex].hits + const bHits = b[tokenIndex].hits + for (let hitIndex = 0; hitIndex < Math.max(aHits.length, bHits.length); hitIndex += 1) { + if (hitIndex >= aHits.length || hitIndex >= bHits.length) { + return aHits.length - bHits.length + } + const difference = aHits[hitIndex].field.sourceOrder - bHits[hitIndex].field.sourceOrder + if (difference !== 0) { + return difference + } + } + } + return 0 +} diff --git a/src/renderer/src/lib/palette-match/tab-document.ts b/src/renderer/src/lib/palette-match/tab-document.ts index 8c289931afc..b8ed0d2291e 100644 --- a/src/renderer/src/lib/palette-match/tab-document.ts +++ b/src/renderer/src/lib/palette-match/tab-document.ts @@ -1,6 +1,5 @@ -import { normalizePaletteText } from './normalized-text' import { buildPaletteDocument, type PaletteDocument } from './palette-document' -import type { PaletteFieldSource } from './indexed-field' +import type { PaletteVisibleFieldSource } from './indexed-field' export const PALETTE_TAB_TITLE_FIELD_ID = 'title' export const PALETTE_TAB_WORKTREE_FIELD_ID = 'worktree' @@ -39,75 +38,67 @@ export function parsePaletteTabIndexedFieldId(fieldId: string, prefix: string): return Number.isInteger(index) ? index : null } -/** - * Tab rows repeat the same string across fields — a browser title that is its own - * URL, or a relative path contained in its absolute one. Indexing both would - * inflate field-hop counts without adding a way to explain the match. - */ -function dedupeSecondaryTexts( - title: string, - secondaryTexts: readonly string[] -): { index: number; text: string }[] { - const seen = new Set([normalizePaletteText(title.trim()).normalized]) - const kept: { index: number; text: string }[] = [] - for (const [index, text] of secondaryTexts.entries()) { - const trimmed = text.trim() - if (!trimmed) { - continue - } - const normalized = normalizePaletteText(trimmed).normalized - if (seen.has(normalized) || [...seen].some((existing) => existing.includes(normalized))) { - continue - } - seen.add(normalized) - kept.push({ index, text: trimmed }) - } - return kept -} - /** * Every tab field is visible identity text, so tokens combine freely — a tab has * no hidden supporting evidence in phase 2. */ export function buildPaletteTabDocument(input: PaletteTabDocumentInput): PaletteDocument { - const fields: PaletteFieldSource[] = [ - { id: PALETTE_TAB_TITLE_FIELD_ID, profile: 'structured-label', text: input.title }, + const fields: PaletteVisibleFieldSource[] = [ + { + id: PALETTE_TAB_TITLE_FIELD_ID, + profile: 'structured-label', + text: input.title, + role: 'primary', + destinationEligible: true + }, { id: PALETTE_TAB_WORKTREE_FIELD_ID, profile: 'structured-label', text: input.worktreeName, - isContainer: true + role: 'container', + destinationEligible: false }, { id: PALETTE_TAB_BRANCH_FIELD_ID, profile: 'structured-label', text: input.branch, - isContainer: true + role: 'container', + destinationEligible: false }, { id: PALETTE_TAB_REPO_FIELD_ID, profile: 'structured-label', text: input.repoName, - isContainer: true + role: 'container', + destinationEligible: false }, { id: PALETTE_TAB_WORKSPACE_FIELD_ID, profile: 'structured-label', text: input.workspaceLabel ?? '', - isContainer: true + role: 'container', + destinationEligible: false } ] - for (const secondary of dedupeSecondaryTexts(input.title, input.secondaryTexts)) { + for (const [index, text] of input.secondaryTexts.entries()) { fields.push({ - id: paletteTabSecondaryFieldId(secondary.index), + id: paletteTabSecondaryFieldId(index), profile: 'path', - text: secondary.text + text, + role: 'secondary', + destinationEligible: true }) } for (const [index, alias] of (input.typeAliases ?? []).entries()) { - fields.push({ id: paletteTabAliasFieldId(index), profile: 'exact-alias', text: alias }) + fields.push({ + id: paletteTabAliasFieldId(index), + profile: 'exact-alias', + text: alias, + role: 'alias', + destinationEligible: false + }) } return buildPaletteDocument({ diff --git a/src/renderer/src/lib/palette-match/tab-match.ts b/src/renderer/src/lib/palette-match/tab-match.ts index f40dd6de03d..32057054c12 100644 --- a/src/renderer/src/lib/palette-match/tab-match.ts +++ b/src/renderer/src/lib/palette-match/tab-match.ts @@ -12,14 +12,16 @@ import { } from './tab-document' import type { MatchRange } from './normalized-text' import type { PaletteResultQualityClass } from './match-quality' -import { - comparePaletteDocumentRank, - type PaletteDocument, - type PaletteDocumentRank -} from './palette-document' +import type { PaletteDocument, PaletteDocumentRank } from './palette-document' +import type { PaletteIndexedField } from './indexed-field' +import { comparePaletteEntityRanks, type PaletteActivityRank } from './palette-ranking' const NO_RANGES: readonly MatchRange[] = [] +export function isOmniboxPaletteTabFieldAllowed(field: Pick): boolean { + return field.id !== PALETTE_TAB_WORKTREE_FIELD_ID && field.id !== PALETTE_TAB_REPO_FIELD_ID +} + export type PaletteTabIndexedMatch = { index: number; ranges: readonly MatchRange[] } export type PaletteTabMatch = { @@ -30,40 +32,45 @@ export type PaletteTabMatch = { branchRanges: readonly MatchRange[] repoRanges: readonly MatchRange[] workspaceRanges: readonly MatchRange[] + secondaryMatches: readonly PaletteTabIndexedMatch[] + typeAliasMatches: readonly PaletteTabIndexedMatch[] + /** First display-preferred proof retained for older row adapters. */ secondary: PaletteTabIndexedMatch | null typeAlias: PaletteTabIndexedMatch | null } -function firstIndexed( +function indexedMatches( rangesByField: ReadonlyMap, prefix: string -): PaletteTabIndexedMatch | null { - let best: PaletteTabIndexedMatch | null = null +): PaletteTabIndexedMatch[] { + const matches: PaletteTabIndexedMatch[] = [] for (const [fieldId, ranges] of rangesByField) { const index = parsePaletteTabIndexedFieldId(fieldId, prefix) - if (index === null) { - continue - } - if (!best || index < best.index) { - best = { index, ranges } + if (index !== null) { + matches.push({ index, ranges }) } } - return best + return matches.sort((a, b) => a.index - b.index) } export function matchPaletteTabDocument( document: PaletteDocument, - query: Extract + query: Extract, + options: { isFieldAllowed?: (field: PaletteIndexedField) => boolean } = {} ): PaletteTabMatch | null { const match = matchPaletteDocument({ document, tokens: query.tokens, - normalizedQuery: query.normalized + normalizedQuery: query.normalized, + tokenCountBeforeDeduplication: query.tokenCountBeforeDeduplication, + isFieldAllowed: options.isFieldAllowed }) if (!match) { return null } const ranges = match.rangesByField + const secondaryMatches = indexedMatches(ranges, PALETTE_TAB_SECONDARY_FIELD_PREFIX) + const typeAliasMatches = indexedMatches(ranges, PALETTE_TAB_ALIAS_FIELD_PREFIX) return { qualityClass: match.qualityClass, rank: match.rank, @@ -72,8 +79,10 @@ export function matchPaletteTabDocument( branchRanges: ranges.get(PALETTE_TAB_BRANCH_FIELD_ID) ?? NO_RANGES, repoRanges: ranges.get(PALETTE_TAB_REPO_FIELD_ID) ?? NO_RANGES, workspaceRanges: ranges.get(PALETTE_TAB_WORKSPACE_FIELD_ID) ?? NO_RANGES, - secondary: firstIndexed(ranges, PALETTE_TAB_SECONDARY_FIELD_PREFIX), - typeAlias: firstIndexed(ranges, PALETTE_TAB_ALIAS_FIELD_PREFIX) + secondaryMatches, + typeAliasMatches, + secondary: secondaryMatches[0] ?? null, + typeAlias: typeAliasMatches[0] ?? null } } @@ -96,26 +105,14 @@ export type PaletteTabRankInputs = { rank: PaletteDocumentRank /** Existing positional score: current tab, current worktree, then list order. */ positionScore: number - id: string - /** Timestamp of most recent activity (focus or agent interaction). */ - lastActiveAt?: number + identity: string + activity: PaletteActivityRank } -/** Lexicographic match rank first, then recent activity, then positional order. */ +/** Shared semantic, bucketed-recency, placement, position, and identity order. */ export function comparePaletteTabResults(a: PaletteTabRankInputs, b: PaletteTabRankInputs): number { - const byRank = comparePaletteDocumentRank(a.rank, b.rank) - if (byRank !== 0) { - return byRank - } - if (a.lastActiveAt !== b.lastActiveAt) { - const aTime = a.lastActiveAt ?? 0 - const bTime = b.lastActiveAt ?? 0 - if (aTime !== bTime) { - return bTime - aTime - } - } - if (a.positionScore !== b.positionScore) { - return a.positionScore - b.positionScore - } - return a.id.localeCompare(b.id) + return comparePaletteEntityRanks( + { rank: a.rank, activity: a.activity, position: a.positionScore, identity: a.identity }, + { rank: b.rank, activity: b.activity, position: b.positionScore, identity: b.identity } + ) } diff --git a/src/renderer/src/lib/palette-repo-resolution.ts b/src/renderer/src/lib/palette-repo-resolution.ts index 0ceddfac948..298b6e77b4d 100644 --- a/src/renderer/src/lib/palette-repo-resolution.ts +++ b/src/renderer/src/lib/palette-repo-resolution.ts @@ -4,39 +4,59 @@ import { type ExecutionHostId } from '../../../shared/execution-host' import { getRepoHostIdentityForParts } from '../../../shared/repo-host-identity' -import { - composeWorktreeHostIdentity, - getWorktreeHostIdentity -} from '../../../shared/worktree/host-qualified-identity' +import { composeWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' import type { Worktree } from '../../../shared/worktree/types' import { isExecutionHostAliasForWorktree } from './worktree-execution-host-alias' type PaletteWorktreeIdentity = Pick +export function getPaletteWorktreeExecutionHostId( + worktree: PaletteWorktreeIdentity +): ExecutionHostId | undefined { + const runtimeOwner = worktree.runtimeOwnerEnvironmentId?.trim() + return runtimeOwner ? toRuntimeExecutionHostId(runtimeOwner) : worktree.hostId +} + +export function getPaletteWorktreeIdentity(worktree: PaletteWorktreeIdentity): string { + return composeWorktreeHostIdentity(getPaletteWorktreeExecutionHostId(worktree), worktree.id) +} + export type PaletteWorktreeIndex = { byHostIdentity: ReadonlyMap byBareId: ReadonlyMap } +export function dedupePaletteWorktrees( + worktrees: readonly T[] +): T[] { + const byIdentity = new Map() + for (const worktree of worktrees) { + byIdentity.set(getPaletteWorktreeIdentity(worktree), worktree) + } + return [...byIdentity.values()] +} + export function buildPaletteWorktreeIndex( worktrees: readonly T[] ): PaletteWorktreeIndex { const byHostIdentity = new Map() const byBareId = new Map() + const byPhysicalHostIdentity = new Map() for (const worktree of worktrees) { - byHostIdentity.set(getWorktreeHostIdentity(worktree), worktree) - if (worktree.runtimeOwnerEnvironmentId) { - byHostIdentity.set( - composeWorktreeHostIdentity( - toRuntimeExecutionHostId(worktree.runtimeOwnerEnvironmentId), - worktree.id - ), - worktree - ) - } + byHostIdentity.set(getPaletteWorktreeIdentity(worktree), worktree) if (!byBareId.has(worktree.id)) { byBareId.set(worktree.id, worktree) } + const physicalIdentity = composeWorktreeHostIdentity(worktree.hostId, worktree.id) + byPhysicalHostIdentity.set( + physicalIdentity, + byPhysicalHostIdentity.has(physicalIdentity) ? null : worktree + ) + } + for (const [physicalIdentity, worktree] of byPhysicalHostIdentity) { + if (worktree && !byHostIdentity.has(physicalIdentity)) { + byHostIdentity.set(physicalIdentity, worktree) + } } return { byHostIdentity, byBareId } } diff --git a/src/renderer/src/lib/pane-manager/pane-manager-pane-creation.ts b/src/renderer/src/lib/pane-manager/pane-manager-pane-creation.ts index c4c3a28c4ca..53ca7f58166 100644 --- a/src/renderer/src/lib/pane-manager/pane-manager-pane-creation.ts +++ b/src/renderer/src/lib/pane-manager/pane-manager-pane-creation.ts @@ -1,3 +1,4 @@ +import { focusPanePreservingOverlays } from './pane-overlay-focus' import type { ManagedPane, ManagedPaneInternal, PaneManagerOptions } from './pane-manager-types' import type { PaneManagerHost } from './pane-manager-host' import { applyPaneOpacity } from './pane-divider' @@ -23,7 +24,7 @@ export function createInitialManagedPane( applyPaneOpacity(host.panes.values(), host.getActivePaneId(), host.getStyleOptions()) if (opts?.focus !== false) { - pane.terminal.focus() + focusPanePreservingOverlays(pane) } host.publishPaneCreated(pane) diff --git a/src/renderer/src/lib/pane-manager/pane-manager-types.ts b/src/renderer/src/lib/pane-manager/pane-manager-types.ts index 00637ae7f97..a26be1e3a9f 100644 --- a/src/renderer/src/lib/pane-manager/pane-manager-types.ts +++ b/src/renderer/src/lib/pane-manager/pane-manager-types.ts @@ -77,6 +77,7 @@ export type PaneManagerOptions = { openLinkHint: string ) => string | null | undefined | Promise initialRenderingSuspended?: boolean + retainHiddenWebgl?: boolean terminalGpuAcceleration?: GlobalSettings['terminalGpuAcceleration'] // Why: diagnostic label for log correlation. safeFit and other internal // helpers log warnings that are hard to correlate without knowing which diff --git a/src/renderer/src/lib/pane-manager/pane-manager.ts b/src/renderer/src/lib/pane-manager/pane-manager.ts index 798d5feaf25..1ce8f850c5d 100644 --- a/src/renderer/src/lib/pane-manager/pane-manager.ts +++ b/src/renderer/src/lib/pane-manager/pane-manager.ts @@ -1,13 +1,11 @@ +import { focusPanePreservingOverlays } from './pane-overlay-focus' import type { PaneManagerOptions, PaneStyleOptions, ManagedPane, ManagedPaneInternal, PaneRenderingDiagnostics, - DropZone, - PaneExternalDropHandler, - PaneExternalDropResolver, - PaneExternalDropTarget + DropZone } from './pane-manager-types' import type { SplitPaneAroundLeafIdsOptions } from './pane-subtree-split' import type { PaneManagerHost } from './pane-manager-host' @@ -68,7 +66,7 @@ export type { PaneExternalDropTarget, PaneExternalDropResolver, PaneExternalDropHandler -} +} from './pane-manager-types' export class PaneManager { private root: HTMLElement @@ -235,7 +233,7 @@ export class PaneManager { applyPaneOpacity(this.panes.values(), this.activePaneId, this.styleOptions) if (opts?.focus !== false) { - pane.terminal.focus() + focusPanePreservingOverlays(pane) } if (changed) { @@ -313,10 +311,12 @@ export class PaneManager { suspendRendering(): void { this.renderingSuspended = true - suspendPaneRendering(this.panes.values(), { - owner: this, - livePanes: () => (this.destroyed ? [] : this.panes.values()) - }) + suspendPaneRendering( + this.panes.values(), + this.options.retainHiddenWebgl === false + ? undefined + : { owner: this, livePanes: () => (this.destroyed ? [] : this.panes.values()) } + ) } resumeRendering(): void { diff --git a/src/renderer/src/lib/pane-manager/pane-overlay-focus.test.ts b/src/renderer/src/lib/pane-manager/pane-overlay-focus.test.ts new file mode 100644 index 00000000000..6b51c1ba449 --- /dev/null +++ b/src/renderer/src/lib/pane-manager/pane-overlay-focus.test.ts @@ -0,0 +1,131 @@ +// @vitest-environment happy-dom +import { afterEach, describe, expect, it, vi } from 'vitest' +import { PaneManager } from './pane-manager' +import { createInitialManagedPane } from './pane-manager-pane-creation' +import type { PaneManagerHost } from './pane-manager-host' +import type { ManagedPaneInternal } from './pane-manager-types' + +vi.mock('./pane-lifecycle', () => ({ + openTerminal: vi.fn(), + createPaneDOM: vi.fn(), + disposePane: vi.fn(), + setLigaturesEnabled: vi.fn() +})) + +afterEach(() => { + document.body.innerHTML = '' +}) + +function fixture() { + const root = document.createElement('div') + document.body.append(root) + const container = document.createElement('div') + const textarea = document.createElement('textarea') + container.append(textarea) + const pane = { + id: 1, + container, + terminal: { focus: vi.fn(() => textarea.focus()) } + } as unknown as ManagedPaneInternal + const panes = new Map([[pane.id, pane]]) + const publishPaneCreated = vi.fn() + const onActivePaneChange = vi.fn() + const manager = Object.create(PaneManager.prototype) as PaneManager + Object.assign(manager, { + panes, + activePaneId: null, + styleOptions: {}, + options: { onActivePaneChange } + }) + const host = { + options: {}, + root, + panes, + createPaneInternal: () => pane, + setActivePaneId: vi.fn(), + getActivePaneId: () => pane.id, + getStyleOptions: () => ({}), + publishPaneCreated + } as unknown as PaneManagerHost + return { root, container, textarea, pane, host, manager, publishPaneCreated, onActivePaneChange } +} + +function overlay(role: string) { + const element = document.createElement('div') + element.setAttribute('role', role) + element.tabIndex = -1 + document.body.append(element) + element.focus() + return element +} + +describe.each(['initial', 'active'] as const)('%s pane focus', (operation) => { + function focus(f: ReturnType, requested = true) { + if (operation === 'initial') { + createInitialManagedPane(f.host, { focus: requested }) + expect(f.publishPaneCreated).toHaveBeenCalledWith(f.pane) + } else { + f.root.append(f.container) + f.manager.setActivePane(f.pane.id, { focus: requested }) + expect(f.manager.getActivePane()?.id).toBe(f.pane.id) + expect(f.onActivePaneChange).toHaveBeenCalledTimes(1) + } + } + + it.each(['menu', 'dialog', 'alertdialog', 'listbox'])('preserves a visible %s', (role) => { + const f = fixture() + const popup = overlay(role) + focus(f) + expect(document.activeElement).toBe(popup) + expect(f.pane.terminal.focus).not.toHaveBeenCalled() + }) + + it('focuses a terminal hosted inside a dialog', () => { + const f = fixture() + overlay('dialog').append(f.root) + focus(f) + expect(document.activeElement).toBe(f.textarea) + }) + + it('preserves a nested popup over a dialog-hosted terminal', () => { + const f = fixture() + const dialog = overlay('dialog') + dialog.append(f.root) + const popup = overlay('menu') + dialog.append(popup) + popup.focus() + focus(f) + expect(document.activeElement).toBe(popup) + }) + + it('allows focus with only persistent sidebar chrome', () => { + const f = fixture() + overlay('listbox').setAttribute('data-worktree-sidebar', '') + focus(f) + expect(document.activeElement).toBe(f.textarea) + }) + + it('allows focus after the overlay closes', () => { + const f = fixture() + overlay('menu').style.display = 'none' + focus(f) + expect(document.activeElement).toBe(f.textarea) + }) + + it('preserves a popup nested inside sidebar chrome', () => { + const f = fixture() + const sidebar = overlay('listbox') + sidebar.setAttribute('data-worktree-sidebar', '') + const popup = overlay('menu') + sidebar.append(popup) + popup.focus() + focus(f) + expect(document.activeElement).toBe(popup) + }) + + it('honors an explicit no-focus request', () => { + const f = fixture() + focus(f, false) + expect(f.pane.terminal.focus).not.toHaveBeenCalled() + }) +}) diff --git a/src/renderer/src/lib/pane-manager/pane-overlay-focus.ts b/src/renderer/src/lib/pane-manager/pane-overlay-focus.ts new file mode 100644 index 00000000000..0dcf5c07ce9 --- /dev/null +++ b/src/renderer/src/lib/pane-manager/pane-overlay-focus.ts @@ -0,0 +1,18 @@ +import { hasVisibleOverlay } from '../visible-overlay' +import type { ManagedPane } from './pane-manager-types' + +export function focusPanePreservingOverlays( + pane: Pick +): void { + if ( + typeof document !== 'undefined' && + hasVisibleOverlay({ + ignoreMatches: '[role="listbox"][data-worktree-sidebar]', + ignoreContaining: pane.container, + ignoreDismissed: true + }) + ) { + return + } + pane.terminal.focus() +} diff --git a/src/renderer/src/lib/pane-manager/terminal-webgl-hidden-retention.test.ts b/src/renderer/src/lib/pane-manager/terminal-webgl-hidden-retention.test.ts index 7d6bfdf26c2..708beed44f5 100644 --- a/src/renderer/src/lib/pane-manager/terminal-webgl-hidden-retention.test.ts +++ b/src/renderer/src/lib/pane-manager/terminal-webgl-hidden-retention.test.ts @@ -1,6 +1,7 @@ import { beforeEach, describe, expect, it, vi } from 'vitest' -import type { ManagedPaneInternal } from './pane-manager-types' +import type { ManagedPaneInternal, PaneManagerOptions } from './pane-manager-types' import { resumePaneRendering, suspendPaneRendering } from './pane-rendering-control' +import { PaneManager } from './pane-manager' import { releaseHiddenWebglRetention, resetHiddenWebglRetentionForTest, @@ -27,6 +28,13 @@ function retentionFor(owner: object, panes: ManagedPaneInternal[]) { return { owner, livePanes: () => panes } } +// panes is private, and the retention branch is only reachable through a mounted pane. +function managerWithPane(pane: ManagedPaneInternal, options: Partial) { + const manager = new PaneManager({} as HTMLElement, options as PaneManagerOptions) + Object.assign(manager, { panes: new Map([[1, pane]]) }) + return manager +} + describe('terminal-webgl-hidden-retention', () => { beforeEach(() => { resetHiddenWebglRetentionForTest() @@ -50,6 +58,25 @@ describe('terminal-webgl-hidden-retention', () => { expect(panes[0].webglAddon).toBeNull() }) + it('disposes a floating manager context on hide so reopen cannot reuse a corrupt atlas', () => { + const pane = createPane() + const addon = pane.webglAddon + managerWithPane(pane, { retainHiddenWebgl: false }).suspendRendering() + expect(addon?.dispose).toHaveBeenCalledTimes(1) + expect(pane.webglAddon).toBeNull() + expect(pane.webglAttachmentDeferred).toBe(true) + expect(retainedHiddenWebglOwnerCountForTest()).toBe(0) + }) + + // Why: pins the option's polarity — an inverted default would silently strand + // every ordinary worktree on the dispose branch. + it('retains an ordinary manager context on hide', () => { + const pane = createPane() + managerWithPane(pane, {}).suspendRendering() + expect(pane.webglAddon).not.toBeNull() + expect(retainedHiddenWebglOwnerCountForTest()).toBe(1) + }) + // Why: the retained branch's blur is already pinned above; only the dispose branch changed. it('blurs a suspended pane on the dispose branch', () => { const panes = [createPane()] diff --git a/src/renderer/src/lib/recent-workspace-tab-rows.test.ts b/src/renderer/src/lib/recent-workspace-tab-rows.test.ts index 54b56a72bb3..ba7ead3bd90 100644 --- a/src/renderer/src/lib/recent-workspace-tab-rows.test.ts +++ b/src/renderer/src/lib/recent-workspace-tab-rows.test.ts @@ -1,14 +1,11 @@ import { describe, expect, it } from 'vitest' import { - buildFocusedGroupTabRecency, - focusedGroupTabKey, orderRecentWorkspaceTabs, resolveRecentWorkspaceTabStatus, type RecentWorkspaceTabRow } from './recent-workspace-tab-rows' import type { TabPaneInputSources } from '@/components/sidebar/smart-attention' import type { AgentStatusEntry, AgentStatusState } from '../../../shared/agent-status-types' -import type { TabGroup } from '../../../shared/tab-types' const NOW = 1_700_000_000_000 const LEAF_ID = '11111111-2222-4333-8444-555555555555' @@ -59,255 +56,52 @@ function sources( } } -function order( - rows: RecentWorkspaceTabRow[], - paneSources: TabPaneInputSources, - overrides: { - lastVisitedAtByWorktreeId?: Record - focusedGroupTabRecency?: Map - } = {} -): string[] { - return orderRecentWorkspaceTabs({ - rows, - paneSources, - now: NOW, - lastVisitedAtByWorktreeId: overrides.lastVisitedAtByWorktreeId ?? {}, - focusedGroupTabRecency: overrides.focusedGroupTabRecency ?? new Map() - }) -} - describe('orderRecentWorkspaceTabs', () => { - it('puts blocked agents above freshly finished ones, whatever their timestamps', () => { - const rows = [row('done'), row('blocked')] - const paneSources = sources([ - entry('done', 'done', NOW - 1_000), - entry('blocked', 'blocked', NOW - 600_000) + it('orders individual tab visits across worktrees and hosts', () => { + const rows = [ + row('old', { lastFocusedAt: NOW - 3 * 86400_000 }), + row('recent', { lastFocusedAt: NOW - 60_000, worktreeHostId: 'ssh:builder' }), + row('newest', { lastFocusedAt: NOW, worktreeId: 'folder:/project' }) + ] + expect(orderRecentWorkspaceTabs({ rows })).toEqual(['newest', 'recent', 'old']) + }) + + it('keeps unknown and invalid visit times below visited tabs with stable ties', () => { + const rows = [ + row('unknown'), + row('nan', { lastFocusedAt: Number.NaN }), + row('first', { lastFocusedAt: NOW }), + row('infinite', { lastFocusedAt: Infinity }), + row('second', { lastFocusedAt: NOW }) + ] + expect(orderRecentWorkspaceTabs({ rows })).toEqual([ + 'first', + 'second', + 'unknown', + 'nan', + 'infinite' ]) - - expect(order(rows, paneSources)).toEqual(['blocked', 'done']) + expect(rows[0].id).toBe('unknown') }) - it('orders within a tier by attention timestamp, newest first', () => { - const rows = [row('older'), row('newer')] - const paneSources = sources([ - entry('older', 'waiting', NOW - 500_000), - entry('newer', 'waiting', NOW - 1_000) - ]) - - expect(order(rows, paneSources)).toEqual(['newer', 'older']) - }) - - it('demotes an interrupted done below a live blocked row', () => { - const rows = [row('interrupted'), row('blocked')] - const paneSources = sources([ - entry('interrupted', 'done', NOW - 1_000, { interrupted: true }), - entry('blocked', 'blocked', NOW - 900_000) - ]) - - expect(order(rows, paneSources)).toEqual(['blocked', 'interrupted']) - }) - - it('drops a stale done out of the attention tier after the freshness window', () => { - const rows = [row('stale'), row('visited')] - const paneSources = sources([entry('stale', 'done', NOW - 40 * 60_000)]) - - expect( - order(rows, paneSources, { - lastVisitedAtByWorktreeId: { 'wt-visited': NOW - 1_000 } - }) - ).toEqual(['visited', 'stale']) - }) - - it('ranks non-attention rows by worktree focus recency', () => { - const rows = [row('cold'), row('warm')] - - expect( - order(rows, sources([]), { - lastVisitedAtByWorktreeId: { - 'wt-cold': NOW - 900_000, - 'wt-warm': NOW - 1_000 - } - }) - ).toEqual(['warm', 'cold']) - }) - - it('uses host-qualified recency for same-id worktree rows', () => { + it('keeps duplicate ids on different hosts as separate occurrences', () => { const rows = [ - row('local', { worktreeId: 'repo::/app', worktreeHostId: 'local' }), - row('ssh', { worktreeId: 'repo::/app', worktreeHostId: 'ssh:builder' }) + row('same', { occurrenceId: 'local', lastFocusedAt: NOW - 1 }), + row('same', { occurrenceId: 'ssh', worktreeHostId: 'ssh:builder', lastFocusedAt: NOW }) ] - - expect( - order(rows, sources([]), { - lastVisitedAtByWorktreeId: { - 'local|repo::/app': NOW - 1_000, - 'ssh:builder|repo::/app': NOW - } - }) - ).toEqual(['ssh', 'local']) + expect(orderRecentWorkspaceTabs({ rows })).toEqual(['ssh', 'local']) }) - it('prefers any visited worktree over a never-visited one', () => { - const rows = [row('never'), row('ancient')] - - expect( - order(rows, sources([]), { - lastVisitedAtByWorktreeId: { 'wt-ancient': 1 } - }) - ).toEqual(['ancient', 'never']) - }) - - it('breaks a same-worktree tie with the focused group MRU tail', () => { - const rows = [ - row('first', { worktreeId: 'wt-1' }), - row('second', { worktreeId: 'wt-1' }), - row('third', { worktreeId: 'wt-1' }) - ] - - expect( - order(rows, sources([]), { - lastVisitedAtByWorktreeId: { 'wt-1': NOW }, - focusedGroupTabRecency: new Map([ - [focusedGroupTabKey('wt-1', 'unified-first'), 0], - [focusedGroupTabKey('wt-1', 'unified-third'), 1], - [focusedGroupTabKey('wt-1', 'unified-second'), 2] - ]) - }) - ).toEqual(['second', 'third', 'first']) - }) - - it('keeps input order across worktrees instead of comparing their unrelated MRU ordinals', () => { - // Callers pass worktree-grouped positional order; both worktrees are never-visited, so only - // the per-worktree focus ordinals differ — beta's larger ordinal must not hoist it over alpha. - const rows = [ - row('alpha-1', { worktreeId: 'wt-alpha', unifiedTabId: 'unified-alpha-1' }), - row('alpha-2', { worktreeId: 'wt-alpha', unifiedTabId: 'unified-alpha-2' }), - row('beta-1', { worktreeId: 'wt-beta', unifiedTabId: 'unified-beta-1' }) - ] - - expect( - order(rows, sources([]), { - focusedGroupTabRecency: new Map([ - [focusedGroupTabKey('wt-alpha', 'unified-alpha-1'), 0], - [focusedGroupTabKey('wt-alpha', 'unified-alpha-2'), 1], - [focusedGroupTabKey('wt-beta', 'unified-beta-1'), 5] - ]) - }) - ).toEqual(['alpha-2', 'alpha-1', 'beta-1']) - }) - - it('keeps an interleaved worktree block together before applying its focused-group MRU', () => { - const rows = [ - row('alpha-old', { worktreeId: 'wt-alpha', unifiedTabId: 'unified-alpha-old' }), - row('beta', { worktreeId: 'wt-beta', unifiedTabId: 'unified-beta' }), - row('alpha-new', { worktreeId: 'wt-alpha', unifiedTabId: 'unified-alpha-new' }) - ] - - expect( - order(rows, sources([]), { - focusedGroupTabRecency: new Map([ - [focusedGroupTabKey('wt-alpha', 'unified-alpha-old'), 0], - [focusedGroupTabKey('wt-alpha', 'unified-alpha-new'), 1], - [focusedGroupTabKey('wt-beta', 'unified-beta'), 0] - ]) - }) - ).toEqual(['alpha-new', 'alpha-old', 'beta']) - }) - - it('keeps duplicate tab ids in separate worktrees on their own MRU ordinals', () => { - const rows = [ - row('alpha-1', { worktreeId: 'wt-alpha', unifiedTabId: 'shared-tab' }), - row('alpha-2', { worktreeId: 'wt-alpha', unifiedTabId: 'alpha-only' }), - row('beta', { worktreeId: 'wt-beta', unifiedTabId: 'shared-tab' }) - ] - - expect( - order(rows, sources([]), { - lastVisitedAtByWorktreeId: { 'wt-alpha': NOW, 'wt-beta': NOW - 1 }, - focusedGroupTabRecency: new Map([ - [focusedGroupTabKey('wt-alpha', 'shared-tab'), 0], - [focusedGroupTabKey('wt-alpha', 'alpha-only'), 1], - // Beta's ordinal for the same tab id must not hoist alpha's occurrence. - [focusedGroupTabKey('wt-beta', 'shared-tab'), 9] - ]) - }) - ).toEqual(['alpha-2', 'alpha-1', 'beta']) - }) - - it('keeps same-id worktrees on two hosts in separate order blocks', () => { - const rows = [ - row('local-old', { worktreeId: 'wt-1', worktreeHostId: 'local', unifiedTabId: 'local-old' }), - row('ssh', { worktreeId: 'wt-1', worktreeHostId: 'ssh:box', unifiedTabId: 'ssh' }), - row('local-new', { worktreeId: 'wt-1', worktreeHostId: 'local', unifiedTabId: 'local-new' }) - ] - - expect( - order(rows, sources([]), { - focusedGroupTabRecency: new Map([ - [focusedGroupTabKey('wt-1', 'local-old'), 0], - [focusedGroupTabKey('wt-1', 'local-new'), 1], - [focusedGroupTabKey('wt-1', 'ssh'), 5] - ]) - }) - ).toEqual(['local-new', 'local-old', 'ssh']) - }) - - it('keeps input (positional) order when nothing else separates two rows', () => { - const rows = [row('a', { worktreeId: 'wt-1' }), row('b', { worktreeId: 'wt-1' })] - - expect(order(rows, sources([]), { lastVisitedAtByWorktreeId: { 'wt-1': NOW } })).toEqual([ - 'a', - 'b' - ]) - }) - - it('returns each occurrence identity when palette ids collide', () => { - const rows = [ - row('workspace-tab:duplicate', { - occurrenceId: 'recent-tab:alpha', - worktreeId: 'wt-alpha' - }), - row('workspace-tab:duplicate', { - occurrenceId: 'recent-tab:beta', - worktreeId: 'wt-beta' - }) - ] - - expect(order(rows, sources([]))).toEqual(['recent-tab:alpha', 'recent-tab:beta']) - }) - - it('treats rows without a terminal tab as idle', () => { - const rows = [row('browser', { terminalTab: null, unifiedTabId: null }), row('blocked')] - - expect(order(rows, sources([entry('blocked', 'blocked', NOW)]))).toEqual(['blocked', 'browser']) - }) - - it('promotes a hookless pane whose live title reads as a permission prompt', () => { - const rows = [ - row('titled', { - terminalTab: { id: 'titled', title: 'OMP - action required' } - }) - ] - const paneSources = sources([], { - ptyIdsByTabId: { titled: ['pty-1'] }, - runtimePaneTitlesByTabId: { titled: { 1: 'OMP - action required' } } + it('retains permission badges without promoting an old permission title', () => { + const old = row('old', { + lastFocusedAt: NOW - 3 * 86400_000, + terminalTab: { id: 'old', title: 'OMP - action required' } }) - - expect(order(rows, paneSources)).toEqual(['titled']) - expect(resolveRecentWorkspaceTabStatus(rows[0], paneSources, NOW)).toBe('permission') - }) - - it('does not let a slept tab leak its stale title into the ranking', () => { - const rows = [ - row('slept', { - terminalTab: { id: 'slept', title: 'OMP - action required' } - }) - ] - const paneSources = sources([], { - runtimePaneTitlesByTabId: { slept: { 1: 'OMP - action required' } } - }) - - expect(resolveRecentWorkspaceTabStatus(rows[0], paneSources, NOW)).toBe('inactive') + const paneSources = sources([], { ptyIdsByTabId: { old: ['pty-1'] } }) + expect(resolveRecentWorkspaceTabStatus(old, paneSources, NOW)).toBe('permission') + expect( + orderRecentWorkspaceTabs({ rows: [old, row('recent', { lastFocusedAt: NOW })] }) + ).toEqual(['recent', 'old']) }) }) @@ -385,31 +179,3 @@ describe('resolveRecentWorkspaceTabStatus', () => { expect(resolveRecentWorkspaceTabStatus(live, sources([]), NOW)).toBe('inactive') }) }) - -describe('buildFocusedGroupTabRecency', () => { - function group(id: string, recentTabIds: string[]): TabGroup { - return { - id, - worktreeId: 'wt-1', - activeTabId: recentTabIds.at(-1) ?? null, - tabOrder: recentTabIds, - recentTabIds - } - } - - it('indexes only the focused group of each worktree', () => { - const recency = buildFocusedGroupTabRecency( - { 'wt-1': 'group-a' }, - { 'wt-1': [group('group-a', ['t1', 't2']), group('group-b', ['t3'])] } - ) - - expect([...recency]).toEqual([ - [focusedGroupTabKey('wt-1', 't1'), 0], - [focusedGroupTabKey('wt-1', 't2'), 1] - ]) - }) - - it('skips worktrees with no focused group', () => { - expect(buildFocusedGroupTabRecency({}, { 'wt-1': [group('group-a', ['t1'])] }).size).toBe(0) - }) -}) diff --git a/src/renderer/src/lib/recent-workspace-tab-rows.ts b/src/renderer/src/lib/recent-workspace-tab-rows.ts index cde50e5fd13..2f4b172c475 100644 --- a/src/renderer/src/lib/recent-workspace-tab-rows.ts +++ b/src/renderer/src/lib/recent-workspace-tab-rows.ts @@ -9,19 +9,11 @@ import { import { tabHasLivePty } from './tab-has-live-pty' import { isExplicitAgentStatusFresh } from './pane-agent-evidence' import type { WorktreeStatus } from './worktree-status' -import type { TabGroup } from '../../../shared/tab-types' import type { TerminalTab } from '../../../shared/terminal-tab-types' import type { ExecutionHostId } from '../../../shared/execution-host' import { AGENT_STATUS_STALE_AFTER_MS } from '../../../shared/agent-status-types' -import { getWorktreeVisitTimestamp } from './worktree-visit-recency' -import { composeWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' -/** - * Row model for Cmd+J's empty-query "Recent chats & terminals" section. - * See docs/cmd-j-recent-chats.md — ranking is a two-tier collapse of the sidebar's - * attention model, deliberately blind to agent activity (`updatedAt`) so a chatty - * agent can't pin itself to the top. - */ +/** Row model for Cmd+J's empty-query recent tabs section. */ export type RecentWorkspaceTabRow = { /** Palette item id. */ id: string @@ -39,33 +31,13 @@ export type RecentWorkspaceTabRow = { /** Terminal tab whose panes carry agent state. Null for editor, browser and simulator rows. */ terminalTab: Pick | null worktreeLastActivityAt: number + lastFocusedAt?: number | null } export type RecentWorkspaceTabOrderInputs = { rows: readonly RecentWorkspaceTabRow[] - paneSources: TabPaneInputSources - now: number - lastVisitedAtByWorktreeId: Record - /** `focusedGroupTabKey` → ordinal in that worktree's focused group; higher is more recent. */ - focusedGroupTabRecency: ReadonlyMap } -type RankedRow = { - occurrenceId: string - needsAttention: boolean - attentionClass: SmartClass - attentionTimestamp: number - visitedAt: number | undefined - focusOrdinal: number - worktreeId: string - worktreeOrder: number -} - -/** Classes 1 (blocked/waiting) and 2 (freshly done) are the rows that want the user. */ -const NEEDS_ATTENTION_MAX_CLASS = 2 - -const NO_FOCUS_ORDINAL = -1 - const STATUS_BY_ATTENTION_CLASS: Record = { 1: 'permission', 2: 'done', @@ -128,100 +100,15 @@ export function resolveRecentWorkspaceTabStatus( return tabHasLivePty(paneSources.ptyIdsByTabId, row.terminalTab.id) ? 'active' : 'inactive' } -/** - * Ordinals are per-worktree, so the key must be too: two worktrees can publish the same tab id and - * a bare key would let one overwrite the other's MRU position. - */ -export function focusedGroupTabKey(worktreeId: string, unifiedTabId: string): string { - // NUL separator: a worktree id embeds a filesystem path, so a printable one would be ambiguous. - return `${worktreeId}\u0000${unifiedTabId}` -} - -/** `TabGroup.recentTabIds` keeps most-recent at the tail, so the index is the ordinal. */ -export function buildFocusedGroupTabRecency( - activeGroupIdByWorktree: Record, - groupsByWorktree: Record -): Map { - const recency = new Map() - for (const [worktreeId, groups] of Object.entries(groupsByWorktree)) { - const activeGroupId = activeGroupIdByWorktree[worktreeId] - if (!activeGroupId) { - continue - } - // Why: MRU only means something inside the focused group; other groups keep positional order. - const focusedGroup = groups?.find((group) => group.id === activeGroupId) - focusedGroup?.recentTabIds?.forEach((tabId, index) => - recency.set(focusedGroupTabKey(worktreeId, tabId), index) - ) - } - return recency -} - -function compareRankedRows(a: RankedRow, b: RankedRow): number { - if (a.needsAttention !== b.needsAttention) { - return a.needsAttention ? -1 : 1 - } - if (a.needsAttention) { - return a.attentionClass !== b.attentionClass - ? a.attentionClass - b.attentionClass - : b.attentionTimestamp - a.attentionTimestamp - } - if (a.visitedAt !== b.visitedAt) { - // Why: presence before value — a visited worktree outranks a never-visited one whatever - // its timestamp, matching orderEmptyQueryWorktrees. - if (a.visitedAt === undefined) { - return 1 - } - if (b.visitedAt === undefined) { - return -1 - } - return b.visitedAt - a.visitedAt - } - if (a.worktreeOrder !== b.worktreeOrder) { - return a.worktreeOrder - b.worktreeOrder - } - return b.focusOrdinal - a.focusOrdinal -} - -/** - * Rank rows into ids, most-wanted first: - * tier 1 — needs attention: class 1 (blocked/waiting) then 2 (fresh done), newest first - * tier 2 — everything else: worktree focus recency, then focused-group MRU - * Equal worktree tiers preserve first-seen worktree order, then use that worktree's MRU. - */ -export function orderRecentWorkspaceTabs(inputs: RecentWorkspaceTabOrderInputs): string[] { - const { rows, paneSources, now, lastVisitedAtByWorktreeId, focusedGroupTabRecency } = inputs - // Host-qualified: the same worktree id on two hosts is two workspaces and must not share a block. - const worktreeOrder = new Map() - for (const row of rows) { - const identity = composeWorktreeHostIdentity(row.worktreeHostId, row.worktreeId) - if (!worktreeOrder.has(identity)) { - worktreeOrder.set(identity, worktreeOrder.size) - } - } - return rows - .map((row): RankedRow => { - const attention = resolveRecentWorkspaceTabAttention(row, paneSources, now) - return { - occurrenceId: row.occurrenceId ?? row.id, - needsAttention: attention.cls <= NEEDS_ATTENTION_MAX_CLASS, - attentionClass: attention.cls, - attentionTimestamp: attention.attentionTimestamp, - visitedAt: getWorktreeVisitTimestamp(lastVisitedAtByWorktreeId, { - id: row.worktreeId, - hostId: row.worktreeHostId - }), - worktreeId: row.worktreeId, - worktreeOrder: - worktreeOrder.get(composeWorktreeHostIdentity(row.worktreeHostId, row.worktreeId)) ?? - Number.MAX_SAFE_INTEGER, - focusOrdinal: - row.unifiedTabId === null - ? NO_FOCUS_ORDINAL - : (focusedGroupTabRecency.get(focusedGroupTabKey(row.worktreeId, row.unifiedTabId)) ?? - NO_FOCUS_ORDINAL) - } - }) - .sort(compareRankedRows) - .map((row) => row.occurrenceId) +/** Unknown visit times stay at the bottom in their existing order. */ +export function orderRecentWorkspaceTabs({ rows }: RecentWorkspaceTabOrderInputs): string[] { + const visitedAt = (row: RecentWorkspaceTabRow): number => + typeof row.lastFocusedAt === 'number' && + Number.isFinite(row.lastFocusedAt) && + row.lastFocusedAt > 0 + ? row.lastFocusedAt + : 0 + return [...rows] + .sort((a, b) => visitedAt(b) - visitedAt(a)) + .map((row) => row.occurrenceId ?? row.id) } diff --git a/src/renderer/src/lib/simulator-palette-active-tab.ts b/src/renderer/src/lib/simulator-palette-active-tab.ts new file mode 100644 index 00000000000..569939acf73 --- /dev/null +++ b/src/renderer/src/lib/simulator-palette-active-tab.ts @@ -0,0 +1,42 @@ +import type { ExecutionHostId } from '../../../shared/execution-host' +import type { TabGroup, WorkspaceVisibleTabType } from '../../../shared/tab-types' +import type { Worktree } from '../../../shared/worktree/types' +import { isPaletteCurrentWorktree } from './palette-repo-resolution' + +export function getActiveSimulatorTabId({ + worktreeId, + worktreeHostId, + worktreeRuntimeOwnerEnvironmentId, + activeWorktreeId, + activeWorkspaceExecutionHostId, + activeTabType, + activeGroupId, + groups +}: { + worktreeId: string + worktreeHostId?: Worktree['hostId'] + worktreeRuntimeOwnerEnvironmentId?: Worktree['runtimeOwnerEnvironmentId'] + activeWorktreeId: string | null + activeWorkspaceExecutionHostId?: ExecutionHostId | null + activeTabType: WorkspaceVisibleTabType + activeGroupId?: string + groups?: readonly TabGroup[] +}): string | null { + if ( + !isPaletteCurrentWorktree( + { + id: worktreeId, + hostId: worktreeHostId, + runtimeOwnerEnvironmentId: worktreeRuntimeOwnerEnvironmentId + }, + activeWorktreeId, + activeWorkspaceExecutionHostId + ) || + activeTabType !== 'simulator' + ) { + return null + } + return activeGroupId + ? (groups?.find((group) => group.id === activeGroupId)?.activeTabId ?? null) + : null +} diff --git a/src/renderer/src/lib/simulator-palette-search.test.ts b/src/renderer/src/lib/simulator-palette-search.test.ts index 82922a3aa7c..6a084709c0e 100644 --- a/src/renderer/src/lib/simulator-palette-search.test.ts +++ b/src/renderer/src/lib/simulator-palette-search.test.ts @@ -309,7 +309,10 @@ describe('simulator-palette-search', () => { ] const hit = searchSimulatorTabs(entries, 'emulator checkout')[0] - expect(hit?.typeAliasMatch).toEqual({ text: 'emulator', ranges: [{ start: 0, end: 8 }] }) + expect(hit?.typeAliasMatch).toEqual({ + text: 'mobile emulator tab', + ranges: [{ start: 7, end: 15 }] + }) expect(hit?.worktreeRanges).toEqual([{ start: 0, end: 8 }]) }) diff --git a/src/renderer/src/lib/simulator-palette-search.ts b/src/renderer/src/lib/simulator-palette-search.ts index ea0e0aa2c2f..092d3b335dd 100644 --- a/src/renderer/src/lib/simulator-palette-search.ts +++ b/src/renderer/src/lib/simulator-palette-search.ts @@ -1,12 +1,17 @@ -import { getWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' import type { ExecutionHostId } from '../../../shared/execution-host' import type { Tab, TabGroup, WorkspaceVisibleTabType } from '../../../shared/tab-types' import type { Worktree } from '../../../shared/worktree/types' -import { isPaletteCurrentWorktree, resolvePaletteRepoForWorktree } from './palette-repo-resolution' +import { + getPaletteWorktreeIdentity, + isPaletteCurrentWorktree, + resolvePaletteRepoForWorktree +} from './palette-repo-resolution' +import { getActiveSimulatorTabId } from './simulator-palette-active-tab' import { isClipboardTextByteLengthOverLimit } from '../../../shared/clipboard-text' import { compareBaseSensitivityLocaleText } from './locale-text-collators' import { comparePaletteTabResults, + isOmniboxPaletteTabFieldAllowed, matchPaletteTabDocument, preparePaletteTabQuery } from './palette-match/tab-match' @@ -18,8 +23,17 @@ import { import type { MatchRange } from './palette-match/normalized-text' import type { PaletteDocument, PaletteDocumentRank } from './palette-match/palette-document' import type { PaletteResultQualityClass } from './palette-match/match-quality' +import { + createPaletteSearchContext, + encodePaletteIdentity, + maxValidPaletteActivityTimestamp, + preparePaletteActivity, + type PaletteActivityRank, + type PaletteSearchContext +} from './palette-match/palette-ranking' import { findAmbiguousWorktreeIds, + findDuplicateIds, getUnifiedTabPaletteExecutionHostId, isUnifiedTabOwnedByWorktree } from './unified-tab-host-ownership' @@ -40,11 +54,13 @@ export type SearchableSimulatorTab = { export type SimulatorPaletteSearchResult = { /** Worktree ids collide across hosts; activation must not resolve by id alone. */ executionHostId?: ExecutionHostId + paletteIdentity: string tabId: string worktreeId: string groupId: string title: string secondaryText: string + secondaryMatches: readonly { text: string; ranges: readonly MatchRange[] }[] repoName: string worktreeName: string branchName: string @@ -54,16 +70,16 @@ export type SimulatorPaletteSearchResult = { worktreeRanges: readonly MatchRange[] branchRanges: readonly MatchRange[] typeAliasMatch?: { text: string; ranges: readonly MatchRange[] } | null + typeAliasMatches: readonly { text: string; ranges: readonly MatchRange[] }[] isCurrentTab: boolean isCurrentWorktree: boolean score: number qualityClass: PaletteResultQualityClass | null rank: PaletteDocumentRank | null lastActiveAt?: number | null + activity: PaletteActivityRank } -type SimulatorPaletteActiveTabType = WorkspaceVisibleTabType - export const SIMULATOR_PALETTE_QUERY_MAX_BYTES = 2 * 1024 // Why search-only: the row icon already says "emulator"; a fixed secondary label @@ -94,7 +110,7 @@ export type BuildSearchableSimulatorTabsOptions = { groupsByWorktree: Record activeWorktreeId: string | null activeWorkspaceExecutionHostId?: ExecutionHostId | null - activeTabType: SimulatorPaletteActiveTabType + activeTabType: WorkspaceVisibleTabType } function compareText(a: string, b: string): number { @@ -134,15 +150,30 @@ export function simulatorPaletteTabTitle(tab: Tab): string { return tab.label || 'Mobile Emulator' } -function baseResult(entry: SearchableSimulatorTab): SimulatorPaletteSearchResult { +function baseResult( + entry: SearchableSimulatorTab, + context: PaletteSearchContext +): SimulatorPaletteSearchResult { + const executionHostId = getUnifiedTabPaletteExecutionHostId(entry.tab, entry.worktree) + const activity = preparePaletteActivity( + maxValidPaletteActivityTimestamp([entry.tab.lastFocusedAt, entry.tab.createdAt]), + context + ) return { - executionHostId: getUnifiedTabPaletteExecutionHostId(entry.tab, entry.worktree), + ...(executionHostId ? { executionHostId } : {}), + paletteIdentity: encodePaletteIdentity([ + 'simulator-tab', + executionHostId ?? '', + entry.worktree.id, + entry.tab.id + ]), tabId: entry.tab.id, worktreeId: entry.worktree.id, groupId: entry.tab.groupId, title: simulatorPaletteTabTitle(entry.tab), // Why empty: the smartphone icon already says the type; a fixed label crowds the row. secondaryText: '', + secondaryMatches: [], repoName: entry.repoName, // Why resolve: a cleared display name leaves the raw field undefined at runtime. worktreeName: resolveWorktreeDisplayName(entry.worktree), @@ -152,57 +183,17 @@ function baseResult(entry: SearchableSimulatorTab): SimulatorPaletteSearchResult repoRanges: NO_RANGES, worktreeRanges: NO_RANGES, branchRanges: NO_RANGES, + typeAliasMatches: [], isCurrentTab: entry.isCurrentTab, isCurrentWorktree: entry.isCurrentWorktree, score: positionScore(entry), qualityClass: null, rank: null, - // Never older than the tab itself: creation is a focus event too. - lastActiveAt: entry.tab.lastFocusedAt - ? Math.max(entry.tab.lastFocusedAt, entry.tab.createdAt) - : null + lastActiveAt: activity.timestamp || null, + activity } } -function getActiveUnifiedTabId({ - worktreeId, - worktreeHostId, - worktreeRuntimeOwnerEnvironmentId, - activeWorktreeId, - activeWorkspaceExecutionHostId, - activeTabType, - activeGroupId, - groups -}: Pick< - BuildSearchableSimulatorTabsOptions, - 'activeTabType' | 'activeWorktreeId' | 'activeWorkspaceExecutionHostId' -> & { - worktreeId: string - worktreeHostId?: Worktree['hostId'] - worktreeRuntimeOwnerEnvironmentId?: Worktree['runtimeOwnerEnvironmentId'] - activeGroupId?: string - groups?: readonly TabGroup[] -}): string | null { - if ( - !isPaletteCurrentWorktree( - { - id: worktreeId, - hostId: worktreeHostId, - runtimeOwnerEnvironmentId: worktreeRuntimeOwnerEnvironmentId - }, - activeWorktreeId, - activeWorkspaceExecutionHostId - ) || - activeTabType !== 'simulator' - ) { - return null - } - const activeGroup = activeGroupId - ? groups?.find((group) => group.id === activeGroupId) - : undefined - return activeGroup?.activeTabId ?? null -} - export function buildSearchableSimulatorTabs({ worktrees, ownershipWorktrees, @@ -222,10 +213,10 @@ export function buildSearchableSimulatorTabs({ const repoName = resolvePaletteRepoForWorktree(worktree, repoMap, repoMapByHostIdentity)?.displayName ?? '' const worktreeSortIndex = - worktreeOrder.get(getWorktreeHostIdentity(worktree)) ?? + worktreeOrder.get(getPaletteWorktreeIdentity(worktree)) ?? worktreeOrder.get(worktree.id) ?? Number.MAX_SAFE_INTEGER - const activeUnifiedTabId = getActiveUnifiedTabId({ + const activeUnifiedTabId = getActiveSimulatorTabId({ worktreeId: worktree.id, worktreeHostId: worktree.hostId, worktreeRuntimeOwnerEnvironmentId: worktree.runtimeOwnerEnvironmentId, @@ -236,8 +227,10 @@ export function buildSearchableSimulatorTabs({ groups: groupsByWorktree[worktree.id] }) const tabs = unifiedTabsByWorktree[worktree.id] ?? [] + const duplicateTabIds = findDuplicateIds(tabs) for (const tab of tabs) { if ( + duplicateTabIds.has(tab.id) || tab.contentType !== 'simulator' || !isUnifiedTabOwnedByWorktree(tab, worktree, ambiguousWorktreeIds) ) { @@ -273,8 +266,10 @@ export function buildSearchableSimulatorTabs({ export function searchSimulatorTabs( entries: readonly SearchableSimulatorTab[], - query: string + query: string, + options: { context?: PaletteSearchContext; fieldMode?: 'all' | 'omnibox' } = {} ): SimulatorPaletteSearchResult[] { + const context = options.context ?? createPaletteSearchContext(Date.now()) if (isSimulatorPaletteQueryTooLarge(query)) { return [] } @@ -282,19 +277,21 @@ export function searchSimulatorTabs( if (!prepared) { return query.trim() ? [] - : entries.map((entry) => baseResult(entry)).sort(compareEmptyQueryResults) + : entries.map((entry) => baseResult(entry, context)).sort(compareEmptyQueryResults) } const results: SimulatorPaletteSearchResult[] = [] for (const entry of entries) { - const match = matchPaletteTabDocument(entry.document, prepared) + const match = matchPaletteTabDocument(entry.document, prepared, { + isFieldAllowed: options.fieldMode === 'omnibox' ? isOmniboxPaletteTabFieldAllowed : undefined + }) if (!match) { continue } const alias = match.typeAlias !== null ? SIMULATOR_TYPE_SEARCH_ALIASES[match.typeAlias.index] : undefined results.push({ - ...baseResult(entry), + ...baseResult(entry, context), titleRanges: match.titleRanges, repoRanges: match.repoRanges, worktreeRanges: match.worktreeRanges, @@ -302,6 +299,10 @@ export function searchSimulatorTabs( // Ranges are into the alias string, not the row: the icon explains the hit, // so nothing on the row is highlighted from them. typeAliasMatch: alias ? { text: alias, ranges: match.typeAlias?.ranges ?? NO_RANGES } : null, + typeAliasMatches: match.typeAliasMatches.map((typeAlias) => ({ + text: SIMULATOR_TYPE_SEARCH_ALIASES[typeAlias.index] ?? '', + ranges: typeAlias.ranges + })), qualityClass: match.qualityClass, rank: match.rank }) @@ -313,14 +314,14 @@ export function searchSimulatorTabs( { rank: a.rank, positionScore: a.score, - id: a.tabId, - lastActiveAt: a.lastActiveAt ?? undefined + identity: a.paletteIdentity, + activity: a.activity }, { rank: b.rank, positionScore: b.score, - id: b.tabId, - lastActiveAt: b.lastActiveAt ?? undefined + identity: b.paletteIdentity, + activity: b.activity } ) : compareEmptyQueryResults(a, b) diff --git a/src/renderer/src/lib/simulator-tab-palette-activation.test.ts b/src/renderer/src/lib/simulator-tab-palette-activation.test.ts index b7d78d067fe..dc398524c0e 100644 --- a/src/renderer/src/lib/simulator-tab-palette-activation.test.ts +++ b/src/renderer/src/lib/simulator-tab-palette-activation.test.ts @@ -132,20 +132,43 @@ describe('activateSimulatorTabPaletteResult', () => { }) }) - it('picks the host that owns the row when the worktree id exists on two hosts', () => { + it('rejects colliding child ids before mutating either host', () => { seedStore({ worktreesByRepo: { 'repo-1': [makeWorktree({ hostId: 'ssh:host-1' })], 'repo-2': [makeWorktree({ repoId: 'repo-2', hostId: 'ssh:host-2', path: '/tmp/wt-1-b' })] + }, + unifiedTabsByWorktree: { + 'wt-1': [ + makeTab({ executionHostId: 'ssh:host-1', groupId: 'group-host-1' }), + makeTab({ executionHostId: 'ssh:host-2', groupId: 'group-host-2' }) + ] + }, + groupsByWorktree: { + 'wt-1': [makeGroup({ id: 'group-host-1' }), makeGroup({ id: 'group-host-2' })] } }) - expect( - activateSimulatorTabPaletteResult({ ...target, executionHostId: 'ssh:host-2' }).status - ).toBe('activated') - expect(mocks.activateAndRevealWorktree).toHaveBeenCalledWith('wt-1', { - executionHostId: 'ssh:host-2' + const before = useAppStore.getState() + expect(activateSimulatorTabPaletteResult({ ...target, executionHostId: 'ssh:host-2' })).toEqual( + { status: 'failed', reason: 'missing-tab' } + ) + expect(useAppStore.getState()).toBe(before) + expect(mocks.activateAndRevealWorktree).not.toHaveBeenCalled() + }) + + it('rejects a hostless tab when its worktree id exists on multiple hosts', () => { + seedStore({ + worktreesByRepo: { + local: [makeWorktree()], + remote: [makeWorktree({ repoId: 'repo-2', hostId: 'ssh:remote', path: '/tmp/remote' })] + } }) + + expect(activateSimulatorTabPaletteResult({ ...target, executionHostId: 'ssh:remote' })).toEqual( + { status: 'failed', reason: 'missing-tab' } + ) + expect(mocks.activateAndRevealWorktree).not.toHaveBeenCalled() }) it('reports an unknown worktree without activating', () => { diff --git a/src/renderer/src/lib/simulator-tab-palette-activation.ts b/src/renderer/src/lib/simulator-tab-palette-activation.ts index abae54d775d..fbc2e0e7e38 100644 --- a/src/renderer/src/lib/simulator-tab-palette-activation.ts +++ b/src/renderer/src/lib/simulator-tab-palette-activation.ts @@ -1,6 +1,11 @@ import { useAppStore } from '@/store' import type { ExecutionHostId } from '../../../shared/execution-host' import { activateAndRevealWorktree } from './worktree-activation' +import { + findAmbiguousWorktreeIds, + getPaletteOwnershipWorktreeIds, + isUnifiedTabOwnedByWorktree +} from './unified-tab-host-ownership' export type SimulatorTabPaletteActivationFailure = 'missing-tab' | 'missing-worktree' @@ -20,17 +25,27 @@ export function activateSimulatorTabPaletteResult({ worktreeId }: SimulatorTabPaletteActivationTarget): SimulatorTabPaletteActivationResult { const initialState = useAppStore.getState() - const tab = (initialState.unifiedTabsByWorktree[worktreeId] ?? []).find( - (candidate) => candidate.id === tabId && candidate.contentType === 'simulator' + const ambiguousWorktreeIds = findAmbiguousWorktreeIds( + getPaletteOwnershipWorktreeIds(initialState) ) - if (!tab) { - return { status: 'failed', reason: 'missing-tab' } + if (!executionHostId && ambiguousWorktreeIds.has(worktreeId)) { + return { status: 'failed', reason: 'missing-worktree' } } - const worktree = initialState.getKnownWorktreeById(worktreeId, executionHostId) if (!worktree) { return { status: 'failed', reason: 'missing-worktree' } } + const tabs = (initialState.unifiedTabsByWorktree[worktreeId] ?? []).filter( + (candidate) => candidate.id === tabId + ) + const tab = tabs[0] + if ( + tabs.length !== 1 || + tab.contentType !== 'simulator' || + !isUnifiedTabOwnedByWorktree(tab, worktree, ambiguousWorktreeIds) + ) { + return { status: 'failed', reason: 'missing-tab' } + } const targetHostId = executionHostId ?? worktree.hostId const activated = activateAndRevealWorktree( @@ -43,7 +58,7 @@ export function activateSimulatorTabPaletteResult({ const state = useAppStore.getState() state.focusGroup(worktreeId, tab.groupId) - state.activateTab(tab.id) + state.activateTab(tab.id, { worktreeId }) state.setActiveTab(tab.id) state.setActiveTabType('simulator') return { status: 'activated', tabId: tab.id } diff --git a/src/renderer/src/lib/unified-tab-anchor-insertion.ts b/src/renderer/src/lib/unified-tab-anchor-insertion.ts new file mode 100644 index 00000000000..2b9055b919f --- /dev/null +++ b/src/renderer/src/lib/unified-tab-anchor-insertion.ts @@ -0,0 +1,22 @@ +import { useAppStore } from '../store' + +/** Move `tabId` to sit immediately after `anchorTabId`; no-op unless both share a group. */ +export function insertUnifiedTabAfterAnchor( + worktreeId: string, + tabId: string, + anchorTabId: string +): void { + if (tabId === anchorTabId) { + return + } + const state = useAppStore.getState() + const group = (state.groupsByWorktree[worktreeId] ?? []).find( + (candidate) => candidate.tabOrder.includes(tabId) && candidate.tabOrder.includes(anchorTabId) + ) + if (!group) { + return + } + const order = group.tabOrder.filter((id) => id !== tabId) + order.splice(order.indexOf(anchorTabId) + 1, 0, tabId) + state.reorderUnifiedTabs(group.id, order, { recordInteraction: false }) +} diff --git a/src/renderer/src/lib/unified-tab-host-ownership.ts b/src/renderer/src/lib/unified-tab-host-ownership.ts index 41fcb366de1..9eeac23c123 100644 --- a/src/renderer/src/lib/unified-tab-host-ownership.ts +++ b/src/renderer/src/lib/unified-tab-host-ownership.ts @@ -1,20 +1,42 @@ import type { Tab } from '../../../shared/tab-types' import type { Worktree } from '../../../shared/worktree/types' -import { LOCAL_EXECUTION_HOST_ID, type ExecutionHostId } from '../../../shared/execution-host' +import type { OpenFile } from '@/store/slices/editor' +import { + LOCAL_EXECUTION_HOST_ID, + toRuntimeExecutionHostId, + toSshExecutionHostId, + type ExecutionHostId +} from '../../../shared/execution-host' import { isExecutionHostAliasForWorktree } from './worktree-execution-host-alias' +import { folderWorkspaceKey } from '../../../shared/workspace-scope' +import type { AppState } from '@/store/types' +import { dedupePaletteWorktrees } from './palette-repo-resolution' + +export function getPaletteOwnershipWorktreeIds( + state: Pick +): Pick[] { + return [ + ...dedupePaletteWorktrees(Object.values(state.worktreesByRepo).flat()), + ...(state.folderWorkspaces ?? []).map((workspace) => ({ id: folderWorkspaceKey(workspace.id) })) + ] +} + +export function findDuplicateIds(items: readonly { id: string }[]): ReadonlySet { + const seen = new Set() + const duplicates = new Set() + for (const item of items) { + if (seen.has(item.id)) { + duplicates.add(item.id) + } + seen.add(item.id) + } + return duplicates +} export function findAmbiguousWorktreeIds( worktrees: readonly Pick[] ): ReadonlySet { - const seen = new Set() - const ambiguous = new Set() - for (const worktree of worktrees) { - if (seen.has(worktree.id)) { - ambiguous.add(worktree.id) - } - seen.add(worktree.id) - } - return ambiguous + return findDuplicateIds(worktrees) } export function getActiveExecutionHostIdForWorktree( @@ -43,6 +65,38 @@ export function isUnifiedTabOwnedByWorktree( return !ambiguousWorktreeIds.has(worktree.id) } +export function isOpenFileOwnedByWorktree( + file: Pick< + OpenFile, + 'externalSshTargetId' | 'operationProvenance' | 'runtimeEnvironmentId' | 'worktreeId' + >, + worktree: Pick +): boolean { + if (file.worktreeId !== worktree.id) { + return false + } + const operationHost = file.operationProvenance?.generation.route.executionHostId + if (operationHost) { + return isExecutionHostAliasForWorktree(operationHost, worktree) + } + if (file.externalSshTargetId) { + return isExecutionHostAliasForWorktree(toSshExecutionHostId(file.externalSshTargetId), worktree) + } + if (file.runtimeEnvironmentId) { + return isExecutionHostAliasForWorktree( + toRuntimeExecutionHostId(file.runtimeEnvironmentId), + worktree + ) + } + return isExecutionHostAliasForWorktree(LOCAL_EXECUTION_HOST_ID, worktree) +} + +export function hasOpenFileExecutionHostEvidence( + file: Pick +): boolean { + return Boolean(file.operationProvenance || file.externalSshTargetId || file.runtimeEnvironmentId) +} + export function getUnifiedTabPaletteExecutionHostId( tab: Pick | undefined, worktree: Pick diff --git a/src/renderer/src/lib/visible-overlay.test.ts b/src/renderer/src/lib/visible-overlay.test.ts index b6cf072d2d9..4fa99e0be23 100644 --- a/src/renderer/src/lib/visible-overlay.test.ts +++ b/src/renderer/src/lib/visible-overlay.test.ts @@ -30,6 +30,16 @@ describe('hasVisibleOverlay', () => { expect(hasVisibleOverlay()).toBe(false) }) + it('ignores the persistent workspace list while preserving its nested popups', () => { + mount('
') + + expect(hasVisibleOverlay()).toBe(false) + + mount('
') + + expect(hasVisibleOverlay()).toBe(true) + }) + it('ignores a display:none overlay', () => { mount('
') diff --git a/src/renderer/src/lib/visible-overlay.ts b/src/renderer/src/lib/visible-overlay.ts index 19a8315cccf..c7e4e721069 100644 --- a/src/renderer/src/lib/visible-overlay.ts +++ b/src/renderer/src/lib/visible-overlay.ts @@ -1,14 +1,23 @@ -const OVERLAY_SELECTOR = '[role="dialog"], [role="alertdialog"], [role="listbox"], [role="menu"]' +// The always-mounted worktree sidebar is page chrome, not an Escape-owning popup. +const OVERLAY_SELECTOR = + '[role="dialog"], [role="alertdialog"], [role="listbox"]:not([data-worktree-sidebar]), [role="menu"]' type VisibleOverlayOptions = { /** Overlays inside a match are treated as page content, not as a layer above it. */ ignoreSelector?: string + /** Ignore matching chrome itself while retaining overlays nested within it. */ + ignoreMatches?: string + /** A terminal hosted inside an overlay may still take focus within that overlay. */ + ignoreContaining?: Element + /** An overlay animating out no longer outranks focus that was queued before it closed. */ + ignoreDismissed?: boolean } /** * Whether a dialog, alert dialog, listbox, or menu is on screen. Page-level Escape * handlers ask this before acting: the overlay owns the first Escape, and a page - * that preventDefaults instead vetoes the overlay's own dismissal. + * that preventDefaults instead vetoes the overlay's own dismissal. Terminal focus + * asks the same question: a live overlay outranks a queued pane focus. */ export function hasVisibleOverlay(options?: VisibleOverlayOptions): boolean { return Array.from(document.querySelectorAll(OVERLAY_SELECTOR)).some((element) => { @@ -21,6 +30,19 @@ export function hasVisibleOverlay(options?: VisibleOverlayOptions): boolean { if (options?.ignoreSelector && element.closest(options.ignoreSelector)) { return false } + if (options?.ignoreMatches && element.matches(options.ignoreMatches)) { + return false + } + if (options?.ignoreContaining && element.contains(options.ignoreContaining)) { + return false + } + // Why: overlays stay mounted and painted through their exit animation, so a + // dismissed one would otherwise keep owning a queued focus for ~300ms. Escape + // callers opt out: they run before the attribute flips, so it only ever hides + // a still-open overlay from them. + if (options?.ignoreDismissed && element.getAttribute('data-state') === 'closed') { + return false + } const style = window.getComputedStyle(element) return ( style.display !== 'none' && diff --git a/src/renderer/src/lib/worker-terminal-takeover-report.ts b/src/renderer/src/lib/worker-terminal-takeover-report.ts index 4a761046194..56c8fbdc0c9 100644 --- a/src/renderer/src/lib/worker-terminal-takeover-report.ts +++ b/src/renderer/src/lib/worker-terminal-takeover-report.ts @@ -15,9 +15,30 @@ const lastReportByPaneKey = new Map() export function reportWorkerTerminalUserInput( paneKey: string, runtimeEnvironmentId: string | null +): void { + reportTakeover({ paneKey }, runtimeEnvironmentId) +} + +/** + * The same takeover, for a worker that IS a structured agent session. + * + * Addressed by SESSION, never by pane key: a structured worker's pane key is a random identity + * credential held only in main, and handing it to a renderer to echo back would make it learnable + * by anyone who can see a chat pane. The owning runtime resolves the session to its own pane key. + */ +export function reportStructuredSessionUserInput( + sessionId: string, + runtimeEnvironmentId: string | null +): void { + reportTakeover({ sessionId }, runtimeEnvironmentId) +} + +function reportTakeover( + subject: { paneKey: string } | { sessionId: string }, + runtimeEnvironmentId: string | null ): void { const now = Date.now() - const gateKey = JSON.stringify([runtimeEnvironmentId, paneKey]) + const gateKey = JSON.stringify([runtimeEnvironmentId, subject]) const last = lastReportByPaneKey.get(gateKey) if (last !== undefined && now - last < REPORT_INTERVAL_MS) { return @@ -30,7 +51,7 @@ export function reportWorkerTerminalUserInput( } } lastReportByPaneKey.set(gateKey, now) - void sendTakeoverReport(paneKey, runtimeEnvironmentId).catch(() => { + void sendTakeoverReport(subject, runtimeEnvironmentId).catch(() => { if (lastReportByPaneKey.get(gateKey) === now) { lastReportByPaneKey.delete(gateKey) } @@ -38,7 +59,7 @@ export function reportWorkerTerminalUserInput( } async function sendTakeoverReport( - paneKey: string, + subject: { paneKey: string } | { sessionId: string }, runtimeEnvironmentId: string | null ): Promise { const target = @@ -46,12 +67,10 @@ async function sendTakeoverReport( ? ({ kind: 'environment', environmentId: runtimeEnvironmentId } as const) : ({ kind: 'local' } as const) const report = () => - callRuntimeRpc( - target, - 'orchestration.workerTerminalUserInput', - { paneKey }, - { suppressFeatureInteraction: true, reuseRecentCompatibilityFailure: true } - ) + callRuntimeRpc(target, 'orchestration.workerTerminalUserInput', subject, { + suppressFeatureInteraction: true, + reuseRecentCompatibilityFailure: true + }) try { await report() } catch { diff --git a/src/renderer/src/lib/workspace-tab-agent-metadata.test.ts b/src/renderer/src/lib/workspace-tab-agent-metadata.test.ts index 7ad4526bd54..9b5b1da6982 100644 --- a/src/renderer/src/lib/workspace-tab-agent-metadata.test.ts +++ b/src/renderer/src/lib/workspace-tab-agent-metadata.test.ts @@ -1,6 +1,8 @@ import { describe, expect, it } from 'vitest' import type { AgentStatusEntry } from '../../../shared/agent-status-types' import { + buildAgentMetadataTabIndex, + collectAgentMetadataFromIndex, collectAgentMetadataForTerminal, maxAgentActivityAt, type AgentMetadata @@ -55,6 +57,93 @@ describe('collectAgentMetadataForTerminal', () => { expect(metadata?.lastActivityAt).toBe(5000) }) + + it('uses the reader clock for mirrored agent evidence and does not refresh replays', () => { + const collect = (updatedAt: number) => + collectAgentMetadataForTerminal({ + terminalTabId: 'tab-1', + worktreeId: 'wt-1', + agentStatusByPaneKey: { + 'tab-1:leaf-1': makeEntry({ + updatedAt, + evidenceObservedAt: 100, + mirroredEvidenceReceivedAt: 5_000 + }) + }, + retainedAgentsByPaneKey: {}, + sleepingAgentSessionsByPaneKey: {} + })[0]?.lastActivityAt + + expect(collect(200)).toBe(5_000) + expect(collect(50_000)).toBe(5_000) + }) + + it('uses authority observation time for locally observed evidence', () => { + const [metadata] = collectAgentMetadataForTerminal({ + terminalTabId: 'tab-1', + worktreeId: 'wt-1', + agentStatusByPaneKey: { + 'tab-1:leaf-1': makeEntry({ updatedAt: 50_000, evidenceObservedAt: 4_000 }) + }, + retainedAgentsByPaneKey: {}, + sleepingAgentSessionsByPaneKey: {} + }) + + expect(metadata?.lastActivityAt).toBe(4_000) + }) +}) + +describe('host-qualified agent metadata joins', () => { + it('does not borrow snippets or activity between same-id worktrees', () => { + const index = buildAgentMetadataTabIndex({ + agentStatusByPaneKey: { + 'shared-tab:local-pane': makeEntry({ + paneKey: 'shared-tab:local-pane', + tabId: 'shared-tab', + prompt: 'local atlas prompt', + updatedAt: 1_000, + connectionId: null + }), + 'shared-tab:remote-pane': makeEntry({ + paneKey: 'shared-tab:remote-pane', + tabId: 'shared-tab', + prompt: 'remote atlas prompt', + updatedAt: 2_000, + connectionId: 'private-target' + }), + 'shared-tab:unstamped-pane': makeEntry({ + paneKey: 'shared-tab:unstamped-pane', + tabId: 'shared-tab', + prompt: 'unknown owner prompt', + updatedAt: 3_000 + }) + }, + retainedAgentsByPaneKey: {}, + sleepingAgentSessionsByPaneKey: {} + }) + const ambiguous = new Set(['wt-1']) + const local = collectAgentMetadataFromIndex( + index, + 'shared-tab', + { id: 'wt-1', hostId: 'local' }, + ambiguous + ) + const remote = collectAgentMetadataFromIndex( + index, + 'shared-tab', + { + id: 'wt-1', + hostId: 'ssh:private-target', + runtimeOwnerEnvironmentId: 'paired-host' + }, + ambiguous + ) + + expect(local.map((entry) => entry.snippetCandidates[0])).toEqual(['local atlas prompt']) + expect(remote.map((entry) => entry.snippetCandidates[0])).toEqual(['remote atlas prompt']) + expect(maxAgentActivityAt(local)).toBe(1_000) + expect(maxAgentActivityAt(remote)).toBe(2_000) + }) }) function makeMetadata(overrides: Partial = {}): AgentMetadata { diff --git a/src/renderer/src/lib/workspace-tab-agent-metadata.ts b/src/renderer/src/lib/workspace-tab-agent-metadata.ts index a2a91838191..2583d76a312 100644 --- a/src/renderer/src/lib/workspace-tab-agent-metadata.ts +++ b/src/renderer/src/lib/workspace-tab-agent-metadata.ts @@ -1,6 +1,10 @@ import type { RetainedAgentEntry } from '@/store/slices/agent-status' import type { AgentStatusEntry } from '../../../shared/agent-status-types' import type { SleepingAgentSessionRecord } from '../../../shared/agent-session-resume' +import { agentStatusEvidenceObservedAt } from '../../../shared/agent-status-freshness' +import type { Worktree } from '../../../shared/worktree/types' +import { LOCAL_EXECUTION_HOST_ID, toSshExecutionHostId } from '../../../shared/execution-host' +import { isExecutionHostAliasForWorktree } from './worktree-execution-host-alias' export type AgentMetadata = { paneKey: string @@ -13,7 +17,11 @@ export type AgentMetadata = { export function maxAgentActivityAt(metadata: readonly AgentMetadata[]): number | null { let max: number | null = null for (const entry of metadata) { - if (entry.lastActivityAt > 0 && (max === null || entry.lastActivityAt > max)) { + if ( + Number.isFinite(entry.lastActivityAt) && + entry.lastActivityAt > 0 && + (max === null || entry.lastActivityAt > max) + ) { max = entry.lastActivityAt } } @@ -98,7 +106,7 @@ function collectLiveMetadata( addText(textParts, historyEntry.prompt) addText(snippetCandidates, historyEntry.prompt) } - return { textParts, snippetCandidates, lastActivityAt: entry.updatedAt } + return { textParts, snippetCandidates, lastActivityAt: agentStatusEvidenceObservedAt(entry) } } function collectSleepingMetadata( @@ -185,6 +193,7 @@ export function collectAgentMetadataForTerminal({ type IndexedAgentEntry = { paneKey: string worktreeId: string | null | undefined + connectionId: string | null | undefined metadata: AgentMetadata } @@ -218,6 +227,7 @@ export function buildAgentMetadataTabIndex( pushToIndex(index, tabId, { paneKey, worktreeId: entry.worktreeId, + connectionId: entry.connectionId, metadata: { paneKey, ...collectLiveMetadata(entry) } }) } @@ -237,6 +247,7 @@ export function buildAgentMetadataTabIndex( pushToIndex(index, tabId, { paneKey, worktreeId: retained.worktreeId, + connectionId: retained.entry.connectionId, metadata: { paneKey, ...meta } }) } @@ -252,6 +263,7 @@ export function buildAgentMetadataTabIndex( pushToIndex(index, tabId, { paneKey, worktreeId: record.worktreeId, + connectionId: record.connectionId, metadata: { paneKey, ...collectSleepingMetadata(record) } }) } @@ -262,11 +274,28 @@ export function buildAgentMetadataTabIndex( export function collectAgentMetadataFromIndex( index: AgentMetadataTabIndex, terminalTabId: string, - worktreeId: string + worktree: Pick, + ambiguousWorktreeIds: ReadonlySet ): AgentMetadata[] { const entries = index.get(terminalTabId) if (!entries) { return [] } - return entries.filter((e) => !e.worktreeId || e.worktreeId === worktreeId).map((e) => e.metadata) + return entries + .filter((entry) => { + if (entry.worktreeId && entry.worktreeId !== worktree.id) { + return false + } + if (entry.connectionId) { + return isExecutionHostAliasForWorktree(toSshExecutionHostId(entry.connectionId), worktree) + } + if (entry.connectionId === undefined && ambiguousWorktreeIds.has(worktree.id)) { + return false + } + return ( + !ambiguousWorktreeIds.has(worktree.id) || + isExecutionHostAliasForWorktree(LOCAL_EXECUTION_HOST_ID, worktree) + ) + }) + .map((entry) => entry.metadata) } diff --git a/src/renderer/src/lib/workspace-tab-agent-snippet-match.ts b/src/renderer/src/lib/workspace-tab-agent-snippet-match.ts index ae8b0d497c2..2971c1e1bc7 100644 --- a/src/renderer/src/lib/workspace-tab-agent-snippet-match.ts +++ b/src/renderer/src/lib/workspace-tab-agent-snippet-match.ts @@ -5,11 +5,9 @@ import { type MatchRange } from './palette-match/normalized-text' import { - PALETTE_MATCH_QUALITIES, - paletteMatchQualityRank, - type PaletteMatchQuality -} from './palette-match/match-quality' -import type { PaletteDocumentRank } from './palette-match/palette-document' + createPaletteFallbackRank, + type PaletteDocumentRank +} from './palette-match/palette-document' import type { PaletteQueryToken } from './palette-match/palette-query' import type { AgentMetadata } from './workspace-tab-agent-metadata' @@ -19,15 +17,7 @@ import type { AgentMetadata } from './workspace-tab-agent-metadata' * fallback preserves the pre-existing ability to find a terminal by what its agent * said, as a strictly last-place tier that never contributes to token coverage. */ -const AGENT_SNIPPET_RANK: PaletteDocumentRank = { - exactIntent: 1, - containerOnlyTokenCount: Number.MAX_SAFE_INTEGER, - wholeQuery: 3, - worstQuality: paletteMatchQualityRank(PALETTE_MATCH_QUALITIES.at(-1) as PaletteMatchQuality) + 1, - usesSupportingEvidence: 1, - fuzzyTokenCount: 0, - fieldHopCount: Number.MAX_SAFE_INTEGER -} +const AGENT_SNIPPET_RANK: PaletteDocumentRank = createPaletteFallbackRank() export type WorkspaceTabAgentSnippetMatch = { text: string diff --git a/src/renderer/src/lib/workspace-tab-palette-activation.store.test.ts b/src/renderer/src/lib/workspace-tab-palette-activation.store.test.ts new file mode 100644 index 00000000000..60b10097f86 --- /dev/null +++ b/src/renderer/src/lib/workspace-tab-palette-activation.store.test.ts @@ -0,0 +1,212 @@ +// @vitest-environment happy-dom + +import { afterEach, expect, it, vi } from 'vitest' +import { useAppStore } from '@/store' +import type { Tab } from '../../../shared/tab-types' +import type { Worktree } from '../../../shared/worktree/types' + +vi.mock('./worktree-activation', () => ({ activateAndRevealWorktree: () => true })) + +import { activateWorkspaceTabPaletteResult } from './workspace-tab-palette-activation' + +const initialState = useAppStore.getInitialState() +afterEach(() => useAppStore.setState(initialState, true)) + +it('keeps the selected diff active when an editor for the same file shares its group', () => { + const worktree: Worktree = { + id: 'wt', + repoId: 'repo', + path: '/workspace', + head: '', + branch: 'main', + isBare: false, + isMainWorktree: true, + displayName: 'Workspace', + comment: '', + linkedIssue: null, + linkedPR: null, + linkedLinearIssue: null, + isArchived: false, + isUnread: false, + isPinned: false, + sortOrder: 0, + lastActivityAt: 0 + } + const editor: Tab = { + id: 'editor', + entityId: 'file', + groupId: 'group', + worktreeId: 'wt', + contentType: 'editor', + label: 'app.ts', + customLabel: null, + color: null, + sortOrder: 0, + createdAt: 0 + } + useAppStore.setState( + { + ...initialState, + worktreesByRepo: { repo: [worktree] }, + activeWorktreeId: 'wt', + groupsByWorktree: { + wt: [{ id: 'group', worktreeId: 'wt', activeTabId: 'editor', tabOrder: ['editor', 'diff'] }] + }, + activeGroupIdByWorktree: { wt: 'group' }, + unifiedTabsByWorktree: { wt: [editor, { ...editor, id: 'diff', contentType: 'diff' }] }, + openFiles: [ + { + id: 'file', + worktreeId: 'wt', + filePath: '/workspace/app.ts', + relativePath: 'app.ts', + language: 'typescript', + isDirty: false, + mode: 'edit' + } + ] + }, + true + ) + + expect( + activateWorkspaceTabPaletteResult({ + worktreeId: 'wt', + groupId: 'group', + tabId: 'diff', + entityId: 'file', + contentType: 'diff' + }) + ).toEqual({ status: 'activated' }) + expect(useAppStore.getState().groupsByWorktree.wt[0].activeTabId).toBe('diff') + expect(useAppStore.getState().activeFileId).toBe('file') +}) + +function makeWorktree(overrides: Partial & Pick): Worktree { + return { + repoId: 'repo', + path: '/workspace', + head: '', + branch: 'main', + isBare: false, + isMainWorktree: true, + displayName: 'Workspace', + comment: '', + linkedIssue: null, + linkedPR: null, + linkedLinearIssue: null, + isArchived: false, + isUnread: false, + isPinned: false, + sortOrder: 0, + lastActivityAt: 0, + ...overrides + } +} + +const COLLIDING_WORKTREES = { + local: [makeWorktree({ id: 'wt' })], + remote: [makeWorktree({ id: 'wt', repoId: 'repo-remote', hostId: 'ssh:remote' })] +} + +it('refuses a hostless tab for a remote target whose worktree ID also exists locally', () => { + const terminal: Tab = { + id: 'unified-terminal', + entityId: 'terminal', + groupId: 'group', + worktreeId: 'wt', + contentType: 'terminal', + label: 'Terminal', + customLabel: null, + color: null, + sortOrder: 0, + createdAt: 0 + } + useAppStore.setState( + { + ...initialState, + worktreesByRepo: COLLIDING_WORKTREES, + groupsByWorktree: { + wt: [ + { + id: 'group', + worktreeId: 'wt', + activeTabId: 'unified-terminal', + tabOrder: ['unified-terminal'] + } + ] + }, + activeGroupIdByWorktree: { wt: 'group' }, + unifiedTabsByWorktree: { wt: [terminal] } + }, + true + ) + + expect( + activateWorkspaceTabPaletteResult({ + worktreeId: 'wt', + groupId: 'group', + tabId: 'unified-terminal', + entityId: 'terminal', + contentType: 'terminal', + executionHostId: 'ssh:remote' + }) + ).toEqual({ status: 'failed', reason: 'missing-tab' }) +}) + +it('refuses a hostless backing file for a remote target whose worktree ID also exists locally', () => { + const editor: Tab = { + id: 'unified-editor', + entityId: 'file', + groupId: 'group', + worktreeId: 'wt', + contentType: 'editor', + executionHostId: 'ssh:remote', + label: 'app.ts', + customLabel: null, + color: null, + sortOrder: 0, + createdAt: 0 + } + useAppStore.setState( + { + ...initialState, + worktreesByRepo: COLLIDING_WORKTREES, + groupsByWorktree: { + wt: [ + { + id: 'group', + worktreeId: 'wt', + activeTabId: 'unified-editor', + tabOrder: ['unified-editor'] + } + ] + }, + activeGroupIdByWorktree: { wt: 'group' }, + unifiedTabsByWorktree: { wt: [editor] }, + openFiles: [ + { + id: 'file', + worktreeId: 'wt', + filePath: '/workspace/app.ts', + relativePath: 'app.ts', + language: 'typescript', + isDirty: false, + mode: 'edit' + } + ] + }, + true + ) + + expect( + activateWorkspaceTabPaletteResult({ + worktreeId: 'wt', + groupId: 'group', + tabId: 'unified-editor', + entityId: 'file', + contentType: 'editor', + executionHostId: 'ssh:remote' + }) + ).toEqual({ status: 'failed', reason: 'missing-file' }) +}) diff --git a/src/renderer/src/lib/workspace-tab-palette-activation.test.ts b/src/renderer/src/lib/workspace-tab-palette-activation.test.ts index 3ba4cf97afd..1660a13930b 100644 --- a/src/renderer/src/lib/workspace-tab-palette-activation.test.ts +++ b/src/renderer/src/lib/workspace-tab-palette-activation.test.ts @@ -6,7 +6,7 @@ const mocks = vi.hoisted(() => { worktreesByRepo: Record groupsByWorktree: Record[]> unifiedTabsByWorktree: Record[]> - openFiles: { id: string; worktreeId: string }[] + openFiles: { id: string; worktreeId: string; externalSshTargetId?: string }[] repos: unknown[] settings: Record activeGroupIdByWorktree: Record @@ -105,6 +105,7 @@ function makeResult( occupantAgent: null, title: 'Terminal', secondaryText: '', + secondaryMatches: [], repoName: 'repo/orca', worktreeName: 'Palette Worktree', branchName: 'main', @@ -113,12 +114,15 @@ function makeResult( repoRanges: [], worktreeRanges: [], branchRanges: [], + typeAliasMatches: [], isCurrentTab: false, isCurrentWorktree: false, score: 0, qualityClass: null, rank: null, + paletteIdentity: 'terminal\u0000wt-1\u0000group-1\u0000unified-terminal-1', lastActiveAt: null, + activity: { ageBucket: null, timestamp: 0 }, ...overrides } } @@ -175,7 +179,9 @@ describe('activateWorkspaceTabPaletteResult', () => { expect(mocks.activateAndRevealWorktree).toHaveBeenCalledWith('wt-1') expect(mocks.store.focusGroup).toHaveBeenCalledWith('wt-1', 'group-1') - expect(mocks.store.activateTab).toHaveBeenCalledWith('unified-terminal-1') + expect(mocks.store.activateTab).toHaveBeenCalledWith('unified-terminal-1', { + worktreeId: 'wt-1' + }) expect(mocks.store.setActiveTab).toHaveBeenCalledWith('terminal-1') expect(mocks.store.setActiveTabType).toHaveBeenCalledWith('terminal') expect(mocks.focusTerminalTabSurface).toHaveBeenCalledWith('terminal-1') @@ -183,6 +189,13 @@ describe('activateWorkspaceTabPaletteResult', () => { it('scopes activation to the host carried by the search result', () => { const executionHostId = 'runtime:host-1' as const + mocks.store.getKnownWorktreeById.mockReturnValue({ + id: 'wt-1', + repoId: 'repo-1', + path: '/tmp/wt-1', + runtimeOwnerEnvironmentId: 'host-1' + }) + mocks.store.unifiedTabsByWorktree['wt-1'][0].executionHostId = executionHostId expect(activateWorkspaceTabPaletteResult({ ...makeResult(), executionHostId })).toEqual({ status: 'activated' @@ -192,6 +205,31 @@ describe('activateWorkspaceTabPaletteResult', () => { expect(mocks.activateAndRevealWorktree).toHaveBeenCalledWith('wt-1', { executionHostId }) }) + it('rejects colliding child ids before mutating either host', () => { + const executionHostId = 'runtime:host-1' as const + mocks.store.getKnownWorktreeById.mockImplementation((_worktreeId, hostId) => + hostId === executionHostId + ? { + id: 'wt-1', + repoId: 'repo-1', + path: '/remote/wt-1', + runtimeOwnerEnvironmentId: 'host-1' + } + : { id: 'wt-1', repoId: 'repo-1', path: '/local/wt-1', hostId: 'local' } + ) + mocks.store.unifiedTabsByWorktree['wt-1'] = [ + { ...mocks.store.unifiedTabsByWorktree['wt-1'][0], executionHostId: 'local' }, + { ...mocks.store.unifiedTabsByWorktree['wt-1'][0], executionHostId } + ] + + expect(activateWorkspaceTabPaletteResult({ ...makeResult(), executionHostId })).toEqual({ + status: 'failed', + reason: 'missing-tab' + }) + expect(mocks.activateAndRevealWorktree).not.toHaveBeenCalled() + expect(mocks.store.activateTab).not.toHaveBeenCalled() + }) + it('activates tabs in known folder or detected workspaces', () => { mocks.store.worktreesByRepo = {} mocks.store.getKnownWorktreeById.mockReturnValue({ id: 'wt-1', repoId: 'repo-1' }) @@ -254,7 +292,7 @@ describe('activateWorkspaceTabPaletteResult', () => { expect(mocks.store.focusGroup).toHaveBeenCalledWith('wt-1', 'group-2') expect(mocks.store.setActiveFile).toHaveBeenCalledWith('/tmp/wt-1/src/app.ts') - expect(mocks.store.activateTab).toHaveBeenLastCalledWith('diff-tab-1') + expect(mocks.store.activateTab).toHaveBeenLastCalledWith('diff-tab-1', { worktreeId: 'wt-1' }) expect(mocks.store.setActiveTabType).toHaveBeenCalledWith('editor') expect(mocks.focusTerminalTabSurface).not.toHaveBeenCalled() }) @@ -305,7 +343,7 @@ describe('activateWorkspaceTabPaletteResult', () => { expect(mocks.store.focusGroup).toHaveBeenCalledWith('wt-1', 'group-2') expect(mocks.store.setActiveFile).toHaveBeenCalledWith(entityId) - expect(mocks.store.activateTab).toHaveBeenLastCalledWith(tabId) + expect(mocks.store.activateTab).toHaveBeenLastCalledWith(tabId, { worktreeId: 'wt-1' }) expect(mocks.store.setActiveTabType).toHaveBeenCalledWith('editor') }) @@ -328,6 +366,18 @@ describe('activateWorkspaceTabPaletteResult', () => { expect(mocks.store.focusGroup).not.toHaveBeenCalled() }) + it('rejects a sole backing file whose explicit owner differs from the target', () => { + mocks.store.unifiedTabsByWorktree['wt-1'][0].contentType = 'editor' + mocks.store.openFiles = [ + { id: 'terminal-1', worktreeId: 'wt-1', externalSshTargetId: 'other-host' } + ] + expect(activateWorkspaceTabPaletteResult(makeResult({ contentType: 'editor' }))).toEqual({ + status: 'failed', + reason: 'missing-file' + }) + expect(mocks.activateAndRevealWorktree).not.toHaveBeenCalled() + }) + it('treats missing editor backing files and worktrees as stale', () => { mocks.store.unifiedTabsByWorktree = { 'wt-1': [ diff --git a/src/renderer/src/lib/workspace-tab-palette-activation.ts b/src/renderer/src/lib/workspace-tab-palette-activation.ts index 688dae2fe7d..249d788c47c 100644 --- a/src/renderer/src/lib/workspace-tab-palette-activation.ts +++ b/src/renderer/src/lib/workspace-tab-palette-activation.ts @@ -8,6 +8,13 @@ import { useAppStore } from '@/store' import type { AppState } from '@/store/types' import type { ExecutionHostId } from '../../../shared/execution-host' import { activateAndRevealWorktree } from './worktree-activation' +import { + findAmbiguousWorktreeIds, + getPaletteOwnershipWorktreeIds, + hasOpenFileExecutionHostEvidence, + isOpenFileOwnedByWorktree, + isUnifiedTabOwnedByWorktree +} from './unified-tab-host-ownership' import type { WorkspaceTabPaletteSearchResult } from './workspace-tab-palette-search' export type WorkspaceTabPaletteActivationFailure = @@ -30,6 +37,7 @@ type WorkspaceTabPaletteActivationState = Pick< AppState, | 'activateTab' | 'focusGroup' + | 'folderWorkspaces' | 'getKnownWorktreeById' | 'groupsByWorktree' | 'openFiles' @@ -37,13 +45,19 @@ type WorkspaceTabPaletteActivationState = Pick< | 'setActiveTab' | 'setActiveTabType' | 'unifiedTabsByWorktree' + | 'worktreesByRepo' > function validateTarget( state: WorkspaceTabPaletteActivationState, result: WorkspaceTabPaletteActivationTarget ): WorkspaceTabPaletteActivationFailure | null { - if (!state.getKnownWorktreeById(result.worktreeId, result.executionHostId)) { + const ambiguousWorktreeIds = findAmbiguousWorktreeIds(getPaletteOwnershipWorktreeIds(state)) + if (!result.executionHostId && ambiguousWorktreeIds.has(result.worktreeId)) { + return 'missing-worktree' + } + const worktree = state.getKnownWorktreeById(result.worktreeId, result.executionHostId) + if (!worktree) { return 'missing-worktree' } const group = (state.groupsByWorktree[result.worktreeId] ?? []).find( @@ -52,24 +66,32 @@ function validateTarget( if (!group) { return 'missing-group' } - const tab = (state.unifiedTabsByWorktree[result.worktreeId] ?? []).find( + const tabs = (state.unifiedTabsByWorktree[result.worktreeId] ?? []).filter( + (candidate) => candidate.id === result.tabId + ) + const tab = tabs.find( (candidate) => - candidate.id === result.tabId && candidate.entityId === result.entityId && candidate.groupId === result.groupId && candidate.worktreeId === result.worktreeId && - candidate.contentType === result.contentType + candidate.contentType === result.contentType && + isUnifiedTabOwnedByWorktree(candidate, worktree, ambiguousWorktreeIds) ) - if (!tab) { + if (tabs.length !== 1 || !tab) { return 'missing-tab' } - if ( - result.contentType !== 'terminal' && - !state.openFiles.some( - (file) => file.id === result.entityId && file.worktreeId === result.worktreeId - ) - ) { - return 'missing-file' + if (result.contentType !== 'terminal') { + const files = state.openFiles.filter((file) => file.id === result.entityId) + if (files.length !== 1 || files[0].worktreeId !== result.worktreeId) { + return 'missing-file' + } + const file = files[0] + // A hostless file falls back to local ownership, which only decides the match when IDs collide. + const requiresOwnershipCheck = + hasOpenFileExecutionHostEvidence(file) || ambiguousWorktreeIds.has(worktree.id) + if (requiresOwnershipCheck && !isOpenFileOwnedByWorktree(file, worktree)) { + return 'missing-file' + } } return null } @@ -100,7 +122,7 @@ export function activateWorkspaceTabPaletteResult( const runtimeEnvironmentId = getRuntimeEnvironmentIdForWorktree(state, result.worktreeId) state.focusGroup(result.worktreeId, result.groupId) - state.activateTab(result.tabId) + state.activateTab(result.tabId, { worktreeId: result.worktreeId }) if (result.contentType === 'terminal') { if (isWebRuntimeSessionActive(runtimeEnvironmentId)) { @@ -117,6 +139,8 @@ export function activateWorkspaceTabPaletteResult( } state.setActiveFile(result.entityId) + // setActiveFile may pick an editor tab for the same entity instead of this diff. + state.activateTab(result.tabId, { worktreeId: result.worktreeId }) state.setActiveTabType('editor') return { status: 'activated' } } diff --git a/src/renderer/src/lib/workspace-tab-palette-content-type.ts b/src/renderer/src/lib/workspace-tab-palette-content-type.ts new file mode 100644 index 00000000000..0e65e2f034a --- /dev/null +++ b/src/renderer/src/lib/workspace-tab-palette-content-type.ts @@ -0,0 +1,8 @@ +import type { TabContentType } from '../../../shared/tab-types' +import type { WorkspaceTabContentType } from './workspace-tab-palette-search' + +export function isWorkspaceTabContentType( + contentType: TabContentType +): contentType is WorkspaceTabContentType { + return ['terminal', 'editor', 'diff', 'conflict-review', 'check-details'].includes(contentType) +} diff --git a/src/renderer/src/lib/workspace-tab-palette-entry-builder.ts b/src/renderer/src/lib/workspace-tab-palette-entry-builder.ts index fdb9bac0830..375caec36cd 100644 --- a/src/renderer/src/lib/workspace-tab-palette-entry-builder.ts +++ b/src/renderer/src/lib/workspace-tab-palette-entry-builder.ts @@ -1,12 +1,15 @@ -import { getWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' import { resolveTerminalTabTitle, resolveUnifiedTabLabel } from '../../../shared/tab-title-resolution' -import type { Tab, TabContentType } from '../../../shared/tab-types' +import type { Tab } from '../../../shared/tab-types' import { getEditorDisplayLabel } from '@/components/editor/editor-labels' import { buildPaletteTabDocument } from './palette-match/tab-document' -import { isPaletteCurrentWorktree, resolvePaletteRepoForWorktree } from './palette-repo-resolution' +import { + getPaletteWorktreeIdentity, + isPaletteCurrentWorktree, + resolvePaletteRepoForWorktree +} from './palette-repo-resolution' import { resolveOpenTabOccupantAgent } from './open-tab-occupant-agent' import { resolveWorktreeBranchLabel, @@ -23,9 +26,15 @@ import type { } from './workspace-tab-palette-search' import { findAmbiguousWorktreeIds, + findDuplicateIds, getUnifiedTabPaletteExecutionHostId, + hasOpenFileExecutionHostEvidence, + isOpenFileOwnedByWorktree, isUnifiedTabOwnedByWorktree } from './unified-tab-host-ownership' +import type { OpenFile } from '@/store/slices/editor' +import type { TerminalTab } from '../../../shared/terminal-tab-types' +import { isWorkspaceTabContentType } from './workspace-tab-palette-content-type' function getActiveUnifiedTabId({ worktreeId, @@ -87,12 +96,6 @@ function isCurrentWorkspaceTab({ : (activeFileIdByWorktree[tab.worktreeId] ?? activeFileId) === tab.entityId } -function isWorkspaceTabContentType( - contentType: TabContentType -): contentType is WorkspaceTabContentType { - return ['terminal', 'editor', 'diff', 'conflict-review', 'check-details'].includes(contentType) -} - export function buildSearchableWorkspaceTabEntries({ worktrees, ownershipWorktrees, @@ -121,7 +124,15 @@ export function buildSearchableWorkspaceTabEntries({ }: BuildSearchableWorkspaceTabsOptions): SearchableWorkspaceTab[] { const entries: SearchableWorkspaceTab[] = [] const seenTabIdentities = new Set() - const openFilesById = new Map(openFiles.map((file) => [file.id, file])) + const openFilesById = new Map() + for (const file of openFiles) { + const bucket = openFilesById.get(file.id) + if (bucket) { + bucket.push(file) + } else { + openFilesById.set(file.id, [file]) + } + } const agentIndex = buildAgentMetadataTabIndex({ agentStatusByPaneKey, retainedAgentsByPaneKey, @@ -135,7 +146,7 @@ export function buildSearchableWorkspaceTabEntries({ const worktreeName = resolveWorktreeDisplayName(worktree) const branch = resolveWorktreeBranchLabel(worktree) const worktreeSortIndex = - worktreeOrder.get(getWorktreeHostIdentity(worktree)) ?? + worktreeOrder.get(getPaletteWorktreeIdentity(worktree)) ?? worktreeOrder.get(worktree.id) ?? Number.MAX_SAFE_INTEGER const isCurrentWorktree = isPaletteCurrentWorktree( @@ -156,10 +167,16 @@ export function buildSearchableWorkspaceTabEntries({ for (const group of groups) { group.tabOrder.forEach((tabId, index) => tabOrder.set(tabId, index)) } - const terminalTabs = new Map((tabsByWorktree[worktree.id] ?? []).map((tab) => [tab.id, tab])) + const terminalTabs = new Map() + for (const terminalTab of tabsByWorktree[worktree.id] ?? []) { + terminalTabs.set(terminalTab.id, terminalTabs.has(terminalTab.id) ? null : terminalTab) + } - for (const rawTab of unifiedTabsByWorktree[worktree.id] ?? []) { + const unifiedTabs = unifiedTabsByWorktree[worktree.id] ?? [] + const duplicateTabIds = findDuplicateIds(unifiedTabs) + for (const rawTab of unifiedTabs) { if ( + duplicateTabIds.has(rawTab.id) || !isWorkspaceTabContentType(rawTab.contentType) || !isUnifiedTabOwnedByWorktree(rawTab, worktree, ambiguousWorktreeIds) ) { @@ -195,6 +212,9 @@ export function buildSearchableWorkspaceTabEntries({ } if (tab.contentType === 'terminal') { const terminalTab = terminalTabs.get(tab.entityId) + if (terminalTab === null) { + continue + } const terminalTitle = terminalTab ? resolveTerminalTabTitle(terminalTab, generatedTitlesEnabled, 'Terminal') : 'Terminal' @@ -225,7 +245,12 @@ export function buildSearchableWorkspaceTabEntries({ repoName, typeAliases: ['terminal tab', 'terminal'] }), - agentMetadata: collectAgentMetadataFromIndex(agentIndex, tab.entityId, worktree.id), + agentMetadata: collectAgentMetadataFromIndex( + agentIndex, + tab.entityId, + worktree, + ambiguousWorktreeIds + ), occupantAgent: resolveOpenTabOccupantAgent({ tabId: tab.entityId, title, @@ -240,8 +265,19 @@ export function buildSearchableWorkspaceTabEntries({ }) continue } - const file = openFilesById.get(tab.entityId) - if (!file || file.worktreeId !== worktree.id) { + const files = openFilesById.get(tab.entityId) + if (files?.length !== 1) { + continue + } + const file = files.find( + (candidate) => + candidate.worktreeId === worktree.id && + (!( + hasOpenFileExecutionHostEvidence(candidate) || ambiguousWorktreeIds.has(worktree.id) + ) || + isOpenFileOwnedByWorktree(candidate, worktree)) + ) + if (!file) { continue } const title = getEditorDisplayLabel(file) diff --git a/src/renderer/src/lib/workspace-tab-palette-results.test.ts b/src/renderer/src/lib/workspace-tab-palette-results.test.ts index 8c34265d3a1..12c93319029 100644 --- a/src/renderer/src/lib/workspace-tab-palette-results.test.ts +++ b/src/renderer/src/lib/workspace-tab-palette-results.test.ts @@ -4,6 +4,7 @@ import type { Worktree } from '../../../shared/worktree/types' import { buildPaletteTabDocument } from './palette-match/tab-document' import { searchWorkspaceTabs } from './workspace-tab-palette-results' import type { SearchableWorkspaceTab } from './workspace-tab-palette-search' +import { createPaletteSearchContext } from './palette-match/palette-ranking' const REPO_NAME = 'octo/rocket' const WORKTREE_NAME = 'Aurora Workspace' @@ -52,15 +53,21 @@ function makeEntry({ contentType = 'terminal', createdAt = 0, worktree = makeWorktree(), - agentLastActivityAt + agentLastActivityAt, + agentSnippet, + title = id, + secondaryText = '' }: { id?: string contentType?: 'terminal' | 'editor' createdAt?: number worktree?: Worktree agentLastActivityAt?: number + agentSnippet?: string + title?: string + secondaryText?: string } = {}): SearchableWorkspaceTab { - const title = id + const secondarySearchTexts = secondaryText ? [secondaryText] : [] return { tab: makeTab(id, contentType, createdAt) as SearchableWorkspaceTab['tab'], worktree, @@ -70,26 +77,26 @@ function makeEntry({ tabSortIndex: 0, occupantAgent: null, title, - secondaryText: '', + secondaryText, titleSearchText: title, - secondarySearchTexts: [], + secondarySearchTexts, document: buildPaletteTabDocument({ id, title, - secondaryTexts: [], + secondaryTexts: secondarySearchTexts, worktreeName: WORKTREE_NAME, branch: BRANCH_NAME, repoName: REPO_NAME }), agentMetadata: - agentLastActivityAt === undefined + agentLastActivityAt === undefined && !agentSnippet ? [] : [ { paneKey: `${id}-pane`, textParts: [], - snippetCandidates: [], - lastActivityAt: agentLastActivityAt + snippetCandidates: agentSnippet ? [agentSnippet] : [], + lastActivityAt: agentLastActivityAt ?? 0 } ], isCurrentTab: false, @@ -98,19 +105,19 @@ function makeEntry({ } describe('searchWorkspaceTabs lastActiveAt', () => { - it('is null when neither agent activity nor worktree activity is known', () => { + it('uses tab creation when no later activity is known', () => { const [result] = searchWorkspaceTabs([makeEntry({ createdAt: 4000 })], '') - expect(result.lastActiveAt).toBeNull() + expect(result.lastActiveAt).toBe(4000) }) - it('falls back to worktree PTY activity for editor tabs with no agent metadata', () => { + it('does not borrow worktree PTY activity for editor tabs', () => { const entry = makeEntry({ contentType: 'editor', createdAt: 1000, worktree: makeWorktree({ lastActivityAt: 5000 }) }) const [result] = searchWorkspaceTabs([entry], '') - expect(result.lastActiveAt).toBe(5000) + expect(result.lastActiveAt).toBe(1000) }) it('prefers agent activity over worktree activity when agent activity is newer', () => { @@ -175,9 +182,88 @@ describe('searchWorkspaceTabs lastActiveAt', () => { expect(result.lastActiveAt).toBe(8000) }) + + it('keeps valid creation when other activity signals are invalid', () => { + const entry = makeEntry({ createdAt: 4_000, agentLastActivityAt: Number.POSITIVE_INFINITY }) + entry.tab.lastFocusedAt = Number.NaN + + const [result] = searchWorkspaceTabs([entry], '', { + context: createPaletteSearchContext(10_000) + }) + + expect(result.lastActiveAt).toBe(4_000) + }) + + it('uses the same future-clamped timestamp for rank activity and row display', () => { + const [result] = searchWorkspaceTabs([makeEntry({ createdAt: 20_000 })], 'tab', { + context: createPaletteSearchContext(10_000) + }) + + expect(result.activity).toEqual({ ageBucket: 0, timestamp: 10_000 }) + expect(result.lastActiveAt).toBe(10_000) + }) }) describe('searchWorkspaceTabs ranking', () => { + it.each(['atl', 'atlas'])('keeps the Atlas reference fixture order for %s', (query) => { + const now = 100 * 24 * 60 * 60 * 1000 + const age = (milliseconds: number): number => now - milliseconds + const entries = [ + makeEntry({ + id: 'old-prefix-2d', + title: 'atlas-follow-up-draft-2026-09-01.md', + createdAt: age(2 * 24 * 60 * 60 * 1000) + }), + makeEntry({ + id: 'old-prefix-3d', + title: 'atlas-meeting-todo.md', + createdAt: age(3 * 24 * 60 * 60 * 1000) + }), + makeEntry({ + id: 'recent-title', + title: 'Clarify Atlas action items', + createdAt: age(30_000) + }), + makeEntry({ + id: 'recent-path', + title: 'questions-and-answers.md', + secondaryText: 'notes/atlas/questions.md', + createdAt: age(30 * 60 * 1000) + }), + makeEntry({ + id: 'older-path', + title: 'worklog.md', + secondaryText: 'notes/atlas/worklog.md', + createdAt: age(9 * 60 * 60 * 1000) + }), + makeEntry({ + id: 'older-title', + title: 'Advance Atlas security review', + createdAt: age(19 * 60 * 60 * 1000) + }), + makeEntry({ + id: 'snippet', + title: 'Agent conversation', + agentSnippet: 'Discuss atlas rollout', + createdAt: age(47 * 60 * 60 * 1000) + }) + ] + + const results = searchWorkspaceTabs(entries, query, { + context: createPaletteSearchContext(now) + }) + + expect(results.map((result) => result.tabId)).toEqual([ + 'recent-title', + 'older-title', + 'old-prefix-2d', + 'old-prefix-3d', + 'recent-path', + 'older-path', + 'snippet' + ]) + }) + it('ranks a multi-token direct-plus-container hit above a container-only whole-query hit', () => { const directEntry = makeEntry({ id: 'direct-tab' }) const containerEntry = makeEntry({ id: 'container-tab' }) @@ -201,9 +287,36 @@ describe('searchWorkspaceTabs ranking', () => { const results = searchWorkspaceTabs([containerEntry, directEntry], 'auth aurora') expect(results.map((result) => result.tabId)).toEqual(['direct-tab', 'container-tab']) + expect(results.map((result) => result.rank?.coverage)).toEqual([2, 2]) expect(results.map((result) => result.rank?.containerOnlyTokenCount)).toEqual([1, 2]) }) + it('ranks one recovered token above an otherwise-equal all-recovered match', () => { + const oneRecovery = makeEntry({ id: 'one-recovery' }) + const twoRecoveries = makeEntry({ id: 'two-recoveries' }) + oneRecovery.document = buildPaletteTabDocument({ + id: 'one-recovery', + title: 'alphx bravo', + secondaryTexts: [], + worktreeName: 'workspace', + branch: 'main', + repoName: 'repo' + }) + twoRecoveries.document = buildPaletteTabDocument({ + id: 'two-recoveries', + title: 'alphx bravx', + secondaryTexts: [], + worktreeName: 'workspace', + branch: 'main', + repoName: 'repo' + }) + + const results = searchWorkspaceTabs([twoRecoveries, oneRecovery], 'alpha bravo') + + expect(results.map((result) => result.tabId)).toEqual(['one-recovery', 'two-recoveries']) + expect(results.map((result) => result.rank?.recoveryTokenCount)).toEqual([1, 2]) + }) + it('ranks direct tab title matches ahead of container-only worktree matches', () => { const directEntry = makeEntry({ id: 'README-4360' }) const containerEntry = makeEntry({ id: 'unrelated-file' }) @@ -228,9 +341,9 @@ describe('searchWorkspaceTabs ranking', () => { const results = searchWorkspaceTabs([containerEntry, directEntry], '4360') expect(results).toHaveLength(2) expect(results[0].tabId).toBe('README-4360') - expect(results[0].rank?.containerOnlyTokenCount).toBe(0) + expect(results[0].rank?.coverage).toBe(0) expect(results[1].tabId).toBe('unrelated-file') - expect(results[1].rank?.containerOnlyTokenCount).toBe(1) + expect(results[1].rank?.coverage).toBe(2) }) it('breaks tie between two container-matching tabs using lastActiveAt recency', () => { diff --git a/src/renderer/src/lib/workspace-tab-palette-results.ts b/src/renderer/src/lib/workspace-tab-palette-results.ts index 9eeeff525f6..f60d33872d5 100644 --- a/src/renderer/src/lib/workspace-tab-palette-results.ts +++ b/src/renderer/src/lib/workspace-tab-palette-results.ts @@ -1,6 +1,7 @@ import { compareBaseSensitivityLocaleText } from './locale-text-collators' import { comparePaletteTabResults, + isOmniboxPaletteTabFieldAllowed, matchPaletteTabDocument, preparePaletteTabQuery, isPaletteTabQueryRejected @@ -15,6 +16,14 @@ import type { ExecutionHostId } from '../../../shared/execution-host' import type { MatchRange } from './palette-match/normalized-text' import type { PaletteDocumentRank } from './palette-match/palette-document' import type { PaletteResultQualityClass } from './palette-match/match-quality' +import { + createPaletteSearchContext, + encodePaletteIdentity, + maxValidPaletteActivityTimestamp, + preparePaletteActivity, + type PaletteActivityRank, + type PaletteSearchContext +} from './palette-match/palette-ranking' import type { TuiAgent } from '../../../shared/tui-agent' import { getUnifiedTabPaletteExecutionHostId } from './unified-tab-host-ownership' import type { @@ -27,6 +36,7 @@ const NO_RANGES: readonly MatchRange[] = [] export type WorkspaceTabPaletteSearchResult = { /** Worktree ids collide across hosts; activation must not resolve by id alone. */ executionHostId?: ExecutionHostId + paletteIdentity: string tabId: string entityId: string worktreeId: string @@ -35,6 +45,7 @@ export type WorkspaceTabPaletteSearchResult = { occupantAgent: TuiAgent | null title: string secondaryText: string + secondaryMatches: readonly { text: string; ranges: readonly MatchRange[] }[] repoName: string worktreeName: string branchName: string @@ -44,6 +55,7 @@ export type WorkspaceTabPaletteSearchResult = { worktreeRanges: readonly MatchRange[] branchRanges: readonly MatchRange[] typeAliasMatch?: { text: string; ranges: readonly MatchRange[] } | null + typeAliasMatches: readonly { text: string; ranges: readonly MatchRange[] }[] isCurrentTab: boolean isCurrentWorktree: boolean score: number @@ -51,6 +63,7 @@ export type WorkspaceTabPaletteSearchResult = { rank: PaletteDocumentRank | null /** Most recent activity for this tab, or null when nothing is known. */ lastActiveAt: number | null + activity: PaletteActivityRank } function compareText(a: string, b: string): number { @@ -87,20 +100,27 @@ function positionScore(entry: SearchableWorkspaceTab): number { } function resolveWorkspaceTabLastActiveAt(entry: SearchableWorkspaceTab): number | null { - // Why: explicit tab activity outranks the worktree fallback; creation only clamps stale signals. - const tabLocalActivity = - Math.max(maxAgentActivityAt(entry.agentMetadata) ?? 0, entry.tab.lastFocusedAt ?? 0) || null - const candidate = tabLocalActivity || entry.worktree.lastActivityAt || null - if (candidate == null) { - return null - } - return Math.max(candidate, entry.tab.createdAt) + return maxValidPaletteActivityTimestamp([ + maxAgentActivityAt(entry.agentMetadata), + entry.tab.lastFocusedAt, + entry.tab.createdAt + ]) } -function baseResult(entry: SearchableWorkspaceTab): WorkspaceTabPaletteSearchResult { +function baseResult( + entry: SearchableWorkspaceTab, + context: PaletteSearchContext +): WorkspaceTabPaletteSearchResult { const executionHostId = getUnifiedTabPaletteExecutionHostId(entry.tab, entry.worktree) + const activity = preparePaletteActivity(resolveWorkspaceTabLastActiveAt(entry), context) return { ...(executionHostId ? { executionHostId } : {}), + paletteIdentity: encodePaletteIdentity([ + 'workspace-tab', + executionHostId ?? '', + entry.worktree.id, + entry.tab.id + ]), tabId: entry.tab.id, entityId: entry.tab.entityId, worktreeId: entry.worktree.id, @@ -109,6 +129,7 @@ function baseResult(entry: SearchableWorkspaceTab): WorkspaceTabPaletteSearchRes occupantAgent: entry.occupantAgent, title: entry.title, secondaryText: entry.secondaryText, + secondaryMatches: [], repoName: entry.repoName, // Why resolve: a cleared display name leaves the raw field undefined at runtime. worktreeName: resolveWorktreeDisplayName(entry.worktree), @@ -118,21 +139,25 @@ function baseResult(entry: SearchableWorkspaceTab): WorkspaceTabPaletteSearchRes repoRanges: NO_RANGES, worktreeRanges: NO_RANGES, branchRanges: NO_RANGES, + typeAliasMatches: [], isCurrentTab: entry.isCurrentTab, isCurrentWorktree: entry.isCurrentWorktree, score: positionScore(entry), qualityClass: null, rank: null, - lastActiveAt: resolveWorkspaceTabLastActiveAt(entry) + lastActiveAt: activity.timestamp || null, + activity } } function matchEntry( entry: SearchableWorkspaceTab, - query: NonNullable> + query: NonNullable>, + context: PaletteSearchContext, + fieldMode: 'all' | 'omnibox' ): WorkspaceTabPaletteSearchResult | null { - const match = matchPaletteTabDocument(entry.document, query) - if (!match) { + const unrestrictedMatch = matchPaletteTabDocument(entry.document, query) + if (!unrestrictedMatch) { // Why kept separate: agent text is not part of the structured field set, so it // never contributes to token coverage — it only recovers a row nothing else found. const snippet = matchWorkspaceTabAgentSnippet(entry.agentMetadata, query) @@ -140,7 +165,7 @@ function matchEntry( return null } return { - ...baseResult(entry), + ...baseResult(entry, context), secondaryText: snippet.text, secondaryRanges: snippet.ranges, qualityClass: 'fuzzy-evidence', @@ -148,6 +173,17 @@ function matchEntry( } } + const match = + fieldMode !== 'omnibox' || + (unrestrictedMatch.worktreeRanges.length === 0 && unrestrictedMatch.repoRanges.length === 0) + ? unrestrictedMatch + : matchPaletteTabDocument(entry.document, query, { + isFieldAllowed: isOmniboxPaletteTabFieldAllowed + }) + if (!match) { + return null + } + const secondaryText = match.secondary !== null ? (entry.secondarySearchTexts[match.secondary.index] ?? entry.secondaryText) @@ -156,8 +192,12 @@ function matchEntry( match.typeAlias !== null ? (entry.typeSearchAliases ?? [])[match.typeAlias.index] : undefined return { - ...baseResult(entry), + ...baseResult(entry, context), secondaryText, + secondaryMatches: match.secondaryMatches.map((secondary) => ({ + text: entry.secondarySearchTexts[secondary.index] ?? '', + ranges: secondary.ranges + })), titleRanges: match.titleRanges, secondaryRanges: match.secondary?.ranges ?? NO_RANGES, repoRanges: match.repoRanges, @@ -166,6 +206,10 @@ function matchEntry( // Ranges are into the alias string, not the row: the content icon explains the // hit, so nothing on the row is highlighted from them. typeAliasMatch: alias ? { text: alias, ranges: match.typeAlias?.ranges ?? NO_RANGES } : null, + typeAliasMatches: match.typeAliasMatches.map((typeAlias) => ({ + text: (entry.typeSearchAliases ?? [])[typeAlias.index] ?? '', + ranges: typeAlias.ranges + })), qualityClass: match.qualityClass, rank: match.rank } @@ -173,19 +217,24 @@ function matchEntry( export function searchWorkspaceTabs( entries: readonly SearchableWorkspaceTab[], - query: string + query: string, + options: { + context?: PaletteSearchContext + fieldMode?: 'all' | 'omnibox' + } = {} ): WorkspaceTabPaletteSearchResult[] { + const context = options.context ?? createPaletteSearchContext(Date.now()) if (isPaletteTabQueryRejected(query)) { return [] } const prepared = preparePaletteTabQuery(query) if (!prepared) { - return entries.map((entry) => baseResult(entry)).sort(compareEmptyQueryResults) + return entries.map((entry) => baseResult(entry, context)).sort(compareEmptyQueryResults) } const results: WorkspaceTabPaletteSearchResult[] = [] for (const entry of entries) { - const result = matchEntry(entry, prepared) + const result = matchEntry(entry, prepared, context, options.fieldMode ?? 'all') if (result) { results.push(result) } @@ -197,14 +246,14 @@ export function searchWorkspaceTabs( { rank: a.rank, positionScore: a.score, - id: a.tabId, - lastActiveAt: a.lastActiveAt ?? undefined + identity: a.paletteIdentity, + activity: a.activity }, { rank: b.rank, positionScore: b.score, - id: b.tabId, - lastActiveAt: b.lastActiveAt ?? undefined + identity: b.paletteIdentity, + activity: b.activity } ) : compareEmptyQueryResults(a, b) diff --git a/src/renderer/src/lib/workspace-tab-palette-search.test.ts b/src/renderer/src/lib/workspace-tab-palette-search.test.ts index d31c1128fe0..e12ea7308f6 100644 --- a/src/renderer/src/lib/workspace-tab-palette-search.test.ts +++ b/src/renderer/src/lib/workspace-tab-palette-search.test.ts @@ -143,9 +143,7 @@ describe('workspace-tab-palette-search', () => { expect(result.executionHostId).toBe('ssh:box') }) - it('keeps the resolvable twin when the first record under an id has no open file', () => { - // Why: dropping the id on sight would lose the row entirely — the leading - // record dies at the open-file lookup and the survivor never gets its turn. + it('omits colliding tab ids even when only one record has an open file', () => { const orphaned = makeUnifiedTab({ id: 'unified-editor-dup', contentType: 'editor', @@ -161,12 +159,26 @@ describe('workspace-tab-palette-search', () => { openFiles: [makeOpenFile()] }) - expect(entries.map((entry) => entry.tab.id)).toEqual(['unified-editor-dup']) - expect(entries[0]?.secondaryText).toBe(SRC_APP_RELATIVE_PATH) + expect(entries).toEqual([]) }) - it('emits one entry per tab id when a session persisted the same id twice', () => { - // Why: the palette keys rows by tab id, and duplicated persisted records used - // to render the row twice under one React key, stranding a ghost row. + + it('omits an editor row whose explicit file host disagrees with its unique worktree', () => { + const remote = makeWorktree({ hostId: 'ssh:remote' }) + const editor = makeUnifiedTab({ + id: 'remote-editor', + entityId: SRC_APP_PATH, + contentType: 'editor', + executionHostId: 'ssh:remote' + }) + const entries = buildEntries({ + worktrees: [remote], + unifiedTabsByWorktree: { 'wt-1': [editor] }, + openFiles: [makeOpenFile({ externalSshTargetId: 'other-host' })] + }) + + expect(entries).toEqual([]) + }) + it('omits a tab id when a session persisted it twice', () => { const duplicate = makeUnifiedTab({ id: 'unified-terminal-dup' }) const entries = buildEntries({ unifiedTabsByWorktree: { @@ -175,7 +187,7 @@ describe('workspace-tab-palette-search', () => { }) const tabIds = entries.map((entry) => entry.tab.id) - expect(tabIds).toEqual(['unified-terminal-1', 'unified-terminal-dup']) + expect(tabIds).toEqual(['unified-terminal-1']) const results = searchWorkspaceTabs(entries, 'unified') expect(results.map((result) => result.tabId)).toEqual(tabIds) @@ -560,9 +572,8 @@ describe('workspace-tab-palette-search', () => { title: 'Fix login race', secondaryText: '', secondaryRanges: [], - // The bare alias matches exactly, so it outranks "terminal tab"; its range - // indexes the alias string, not the row. - typeAliasMatch: { text: 'terminal', ranges: [{ start: 0, end: 8 }] } + // Equal-strength aliases use the builder's stable display order. + typeAliasMatch: { text: 'terminal tab', ranges: [{ start: 0, end: 8 }] } }) }) diff --git a/src/renderer/src/lib/worktree-activation-store-contract.ts b/src/renderer/src/lib/worktree-activation-store-contract.ts index 6f2eea5e215..63ce180ffed 100644 --- a/src/renderer/src/lib/worktree-activation-store-contract.ts +++ b/src/renderer/src/lib/worktree-activation-store-contract.ts @@ -71,4 +71,8 @@ export type InitialTerminalOptions = { * workspace", wake) has to hand back a usable surface. Activation sets this unless the * caller says it provides its own surface; background worktree creation leaves it unset. */ reseedEmptiedWorkspace?: boolean + /** Set by callers that open their own primary surface (a structured native chat session). + * Setup/issue work still runs, but work that needs no host terminal must not seed a shell + * beside the chat the caller is about to create. */ + callerProvidesSurface?: boolean } diff --git a/src/renderer/src/lib/worktree-activation-structured-chat-surface.test.ts b/src/renderer/src/lib/worktree-activation-structured-chat-surface.test.ts new file mode 100644 index 00000000000..33be3888a24 --- /dev/null +++ b/src/renderer/src/lib/worktree-activation-structured-chat-surface.test.ts @@ -0,0 +1,121 @@ +import { afterEach, describe, expect, it, vi } from 'vitest' +import { useAppStore } from '@/store' +import { activateAndRevealWorktree } from './worktree-activation' +import { ensureWorktreeHasInitialTerminal } from './worktree-initial-terminal-seeding' +import { + makeCreatedAgentWorktree as makeWorktree, + seedEmptyActivatableWorktree +} from '@/lib/worktree-activation-created-agent-test-state' +import { + createMockStore, + registerWorktreeActivationReset, + setSetupScriptLaunchMode +} from './worktree-activation-test-harness' + +const initialAppStoreState = useAppStore.getState() + +registerWorktreeActivationReset() + +afterEach(() => { + vi.restoreAllMocks() + useAppStore.setState(initialAppStoreState, true) +}) + +const setup = { + runnerScriptPath: '/tmp/repo/.git/orca/setup-runner.sh', + envVars: { ORCA_WORKTREE_PATH: '/tmp/worktrees/wt-1' } +} + +// Why: a native-chat create used to land the user on a bare "Terminal 1" beside the chat, +// because the returned setup script counted as work needing a shell to attach to. +describe('seeding beside a caller-provided chat surface', () => { + it('runs a new-tab setup script without seeding a shell', () => { + let createdIndex = 0 + const createTab = vi.fn(() => ({ id: `tab-${++createdIndex}` })) + const store = createMockStore({ createTab }) + + const primaryTabId = ensureWorktreeHasInitialTerminal( + store, + 'wt-1', + undefined, + setup, + undefined, + undefined, + { callerProvidesSurface: true } + ) + + expect(primaryTabId).toBeNull() + expect(createTab).toHaveBeenCalledTimes(1) + expect(store.setTabCustomTitle).toHaveBeenCalledWith('tab-1', 'Setup', { + recordInteraction: false + }) + expect(store.queueTabStartupCommand).toHaveBeenCalledWith('tab-1', { + command: 'bash /tmp/repo/.git/orca/setup-runner.sh', + env: setup.envVars + }) + }) + + it('still seeds a shell when setup runs as a split', () => { + let createdIndex = 0 + const createTab = vi.fn(() => ({ id: `tab-${++createdIndex}` })) + const store = createMockStore({ createTab }) + setSetupScriptLaunchMode('split-vertical') + + ensureWorktreeHasInitialTerminal(store, 'wt-1', undefined, setup, undefined, undefined, { + callerProvidesSurface: true + }) + + expect(createTab).toHaveBeenCalledTimes(1) + expect(store.queueTabSetupSplit).toHaveBeenCalledWith('tab-1', expect.anything()) + }) + + it('still seeds a shell for issue automation, which splits from it', () => { + let createdIndex = 0 + const createTab = vi.fn(() => ({ id: `tab-${++createdIndex}` })) + const store = createMockStore({ createTab }) + + ensureWorktreeHasInitialTerminal( + store, + 'wt-1', + undefined, + undefined, + { command: 'orca issue run' }, + undefined, + { callerProvidesSurface: true } + ) + + expect(createTab).toHaveBeenCalledTimes(1) + expect(store.queueTabIssueCommandSplit).toHaveBeenCalledWith('tab-1', { + command: 'orca issue run', + env: undefined + }) + }) + + it('still seeds a shell when the caller owns no surface', () => { + let createdIndex = 0 + const createTab = vi.fn(() => ({ id: `tab-${++createdIndex}` })) + const store = createMockStore({ createTab }) + + ensureWorktreeHasInitialTerminal(store, 'wt-1', undefined, setup) + + expect(createTab).toHaveBeenCalledTimes(2) + expect(store.setTabCustomTitle).toHaveBeenCalledWith('tab-2', 'Setup', { + recordInteraction: false + }) + }) + + it('activation forwards providesInitialSurface so setup alone adds one tab', () => { + const worktree = makeWorktree() + seedEmptyActivatableWorktree(worktree) + + const result = activateAndRevealWorktree(worktree.id, { + providesInitialSurface: true, + notifyHostRuntime: false, + setup + }) + + expect(result).not.toBe(false) + expect(result === false ? 'unused' : result.primaryTabId).toBeNull() + expect(useAppStore.getState().tabsByWorktree[worktree.id]).toHaveLength(1) + }) +}) diff --git a/src/renderer/src/lib/worktree-activation.ts b/src/renderer/src/lib/worktree-activation.ts index ac68b0f649a..d2304c28b9d 100644 --- a/src/renderer/src/lib/worktree-activation.ts +++ b/src/renderer/src/lib/worktree-activation.ts @@ -286,6 +286,7 @@ export function activateAndRevealWorktree( { ...(opts?.backendStartupTerminalSpawned ? { backendStartupTerminalSpawned: true } : {}), ...(opts?.createNewTerminalForStartup ? { createNewTerminalForStartup: true } : {}), + ...(opts?.providesInitialSurface === true ? { callerProvidesSurface: true } : {}), reseedEmptiedWorkspace: opts?.providesInitialSurface !== true } ) diff --git a/src/renderer/src/lib/worktree-creation-chat-setup.test.ts b/src/renderer/src/lib/worktree-creation-chat-setup.test.ts new file mode 100644 index 00000000000..3c6be8b26af --- /dev/null +++ b/src/renderer/src/lib/worktree-creation-chat-setup.test.ts @@ -0,0 +1,84 @@ +import { afterEach, describe, expect, it, vi } from 'vitest' +import { useAppStore } from '@/store' +import { executeWorktreeCreation } from './worktree-creation-flow-execute' +import { launchStructuredWorktreeSession } from './worktree-creation-structured-session' +import { + makeCreatedAgentWorktree, + seedEmptyActivatableWorktree +} from './worktree-activation-created-agent-test-state' +import { registerWorktreeActivationReset } from './worktree-activation-test-harness' +import type { WorktreeCreationRequest } from './pending-worktree-creation' + +vi.mock('./worktree-creation-structured-session', () => ({ + launchStructuredWorktreeSession: vi.fn(async (args) => ({ + accepted: true, + cancelled: false, + visibilityUnknown: false, + activation: args.activation, + primaryTabId: args.primaryTabId + })) +})) +vi.mock('./worktree-creation-completion', () => ({ completeWorktreeCreation: vi.fn() })) + +const initialState = useAppStore.getState() +registerWorktreeActivationReset() +afterEach(() => { + vi.restoreAllMocks() + useAppStore.setState(initialState, true) +}) + +describe('native chat creation completed in the background', () => { + it.each(['claude', 'codex'] as const)( + 'runs setup once without an idle shell or focus change for %s', + async (agent) => { + const worktree = makeCreatedAgentWorktree() + seedEmptyActivatableWorktree(worktree) + const request: WorktreeCreationRequest = { + repoId: worktree.repoId, + name: 'feature', + setupDecision: 'run', + agent, + agentLaunchRoute: 'structured-native-chat', + pendingFirstAgentMessageRename: false, + note: '', + startupPlan: null, + quickPrompt: '', + quickTelemetry: null + } + const setup = { runnerScriptPath: '/tmp/setup-runner.sh', envVars: {} } + useAppStore.setState({ + activeView: 'tasks', + activeWorktreeId: 'previous-worktree', + activeTabId: 'previous-tab', + createWorktree: vi.fn().mockResolvedValue({ worktree, setup }), + pendingWorktreeCreations: { + 'creation-1': { + creationId: 'creation-1', + phase: 'fetching', + status: 'creating', + startedAt: 1, + indeterminate: false, + loaderVisible: true, + request + } + } + }) + + await executeWorktreeCreation('creation-1', request) + + const state = useAppStore.getState() + const tabs = state.tabsByWorktree[worktree.id] + expect(tabs).toHaveLength(1) + expect(tabs[0].customTitle).toBe('Setup') + expect(state.pendingStartupByTabId[tabs[0].id]).toMatchObject({ + command: 'bash /tmp/setup-runner.sh' + }) + expect(state.activeView).toBe('tasks') + expect(state.activeWorktreeId).toBe('previous-worktree') + expect(state.activeTabId).toBe('previous-tab') + expect(launchStructuredWorktreeSession).toHaveBeenCalledWith( + expect.objectContaining({ primaryTabId: null, shouldActivateOnCompletion: false }) + ) + } + ) +}) diff --git a/src/renderer/src/lib/worktree-creation-flow-execute.ts b/src/renderer/src/lib/worktree-creation-flow-execute.ts index 9b6ba569b1e..27ee8da4827 100644 --- a/src/renderer/src/lib/worktree-creation-flow-execute.ts +++ b/src/renderer/src/lib/worktree-creation-flow-execute.ts @@ -77,7 +77,7 @@ export async function executeWorktreeCreation( preparedRequest.linkedGitLabMR, preparedRequest.linkedGitLabIssue, backendStartup, - structuredLaunch ? false : preparedRequest.pendingFirstAgentMessageRename, + preparedRequest.pendingFirstAgentMessageRename, creationId, preparedRequest.linkedLinearIssueWorkspaceId, preparedRequest.linkedLinearIssueOrganizationUrlKey, @@ -203,6 +203,7 @@ export async function executeWorktreeCreation( result.defaultTabs, { activateCreatedTabs: false, + ...(structuredLaunch ? { callerProvidesSurface: true } : {}), ...(backendSpawned ? { backendStartupTerminalSpawned: true } : {}) } ) diff --git a/src/renderer/src/lib/worktree-initial-terminal-seeding.ts b/src/renderer/src/lib/worktree-initial-terminal-seeding.ts index 6b4214a3fd3..e36a79cb364 100644 --- a/src/renderer/src/lib/worktree-initial-terminal-seeding.ts +++ b/src/renderer/src/lib/worktree-initial-terminal-seeding.ts @@ -132,6 +132,32 @@ export function ensureWorktreeHasInitialTerminal( } const hasExplicitLaunchWork = Boolean(sequencedStartup || setup || issueCommand) + // Why: a caller opening its own primary surface (a structured native chat) asked for that surface + // alone. Setup launched in its own tab needs no shell to attach to, so seeding one leaves a stray + // "Terminal 1" beside the chat. Splits and issue automation still need a pane to split from. + const setupNeedsHostTerminal = + setup !== undefined && + (useAppStore.getState().settings?.setupScriptLaunchMode ?? 'new-tab') !== 'new-tab' + if ( + opts?.callerProvidesSurface === true && + renderableTabCount === 0 && + !sequencedStartup && + !issueCommand && + !setupNeedsHostTerminal && + !defaultTabs?.tabs.length && + opts?.createNewTerminalForStartup !== true + ) { + queueSetupAndIssueCommands( + store, + worktreeId, + null, + setup, + undefined, + wrappedSetupCommandStr, + opts + ) + return null + } // Why: only startup hydration honours the closed-last-tab tombstone. Every explicit // activation (sidebar, palette, automation resume, wake) re-seeds a surface instead, // because closing the last terminal normally deactivates the workspace too diff --git a/src/renderer/src/lib/worktree-palette-document.ts b/src/renderer/src/lib/worktree-palette-document.ts index cf419b8e370..8efa20fba22 100644 --- a/src/renderer/src/lib/worktree-palette-document.ts +++ b/src/renderer/src/lib/worktree-palette-document.ts @@ -1,4 +1,3 @@ -import { getWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' import { getRepoExecutionHostId, LOCAL_EXECUTION_HOST_ID } from '../../../shared/execution-host' import { issueCacheKey as getIssueCacheKey } from '@/store/github/cache-identity' import { buildPaletteDocument, type PaletteDocument } from './palette-match/palette-document' @@ -20,7 +19,10 @@ import type { HostedReviewInfo } from '../../../shared/hosted-review' import type { Repo } from '../../../shared/repo-types' import type { Worktree } from '../../../shared/worktree/types' import { isGitHubPRSuppressed } from '../../../shared/worktree/github-pr-suppression' -import { resolvePaletteRepoForWorktree } from './palette-repo-resolution' +import { + getPaletteWorktreeIdentity, + resolvePaletteRepoForWorktree +} from './palette-repo-resolution' export const WORKTREE_PALETTE_NAME_FIELD_ID = 'name' export const WORKTREE_PALETTE_BRANCH_FIELD_ID = 'branch' @@ -147,17 +149,23 @@ export function buildWorktreePaletteDocument( { id: WORKTREE_PALETTE_NAME_FIELD_ID, profile: 'structured-label', - text: resolveWorktreeDisplayName(worktree) + text: resolveWorktreeDisplayName(worktree), + role: 'primary', + destinationEligible: true }, { id: WORKTREE_PALETTE_BRANCH_FIELD_ID, profile: 'structured-label', - text: resolveWorktreeBranchLabel(worktree) + text: resolveWorktreeBranchLabel(worktree), + role: 'secondary', + destinationEligible: true }, { id: WORKTREE_PALETTE_REPO_FIELD_ID, profile: 'structured-label', - text: repo?.displayName ?? '' + text: repo?.displayName ?? '', + role: 'secondary', + destinationEligible: false }, { id: WORKTREE_PALETTE_HOST_FIELD_ID, @@ -167,9 +175,11 @@ export function buildWorktreePaletteDocument( // Why both keys: the palette keys this map by host identity so two same-id // workspaces keep distinct chips, but a bare-id map is still a valid input. text: - sources.hostLabelByWorktreeId?.get(getWorktreeHostIdentity(worktree)) ?? + sources.hostLabelByWorktreeId?.get(getPaletteWorktreeIdentity(worktree)) ?? sources.hostLabelByWorktreeId?.get(worktree.id) ?? - '' + '', + role: 'secondary', + destinationEligible: false } ], compositePairs: [ @@ -193,7 +203,7 @@ export function buildWorktreePaletteDocuments( // the bare id lets the second host overwrite the first and one workspace becomes // unsearchable by its own name. documents.set( - getWorktreeHostIdentity(worktree), + getPaletteWorktreeIdentity(worktree), buildWorktreePaletteDocument(worktree, sources) ) } diff --git a/src/renderer/src/lib/worktree-palette-multi-keyword.test.ts b/src/renderer/src/lib/worktree-palette-multi-keyword.test.ts index 8e72fef0a0d..5e869b6085f 100644 --- a/src/renderer/src/lib/worktree-palette-multi-keyword.test.ts +++ b/src/renderer/src/lib/worktree-palette-multi-keyword.test.ts @@ -108,14 +108,14 @@ describe('evidence and ranking', () => { it('prefers visible identity over supporting evidence', () => { const [result] = search('docs') - expect(result.rank?.usesSupportingEvidence).toBe(0) + expect(result.rank?.coverage).toBeLessThan(3) expect(result.qualityClass).toBe('exact-visible') }) it('ranks an exact identifier above an incidental numeric substring', () => { const [exact] = search('#4123') expect(exact.supportingText?.labelKind).toBe('pr') - expect(exact.qualityClass).toBe('exact-evidence') + expect(exact.qualityClass).toBe('exact-intent') // A port prefix is the only reading of `412`, and it ranks below the exact hit. const [partial] = search('412') expect(partial.supportingText?.labelKind).toBe('port') @@ -126,7 +126,7 @@ describe('evidence and ranking', () => { const [result] = search('reconect') expect(result.worktreeId).toBe('wt-reconnect') expect(result.qualityClass).toBe('fuzzy-evidence') - expect(result.rank?.fuzzyTokenCount).toBe(1) + expect(result.rank?.recovery).toBe(1) }) }) diff --git a/src/renderer/src/lib/worktree-palette-runtime-owner-identity.test.ts b/src/renderer/src/lib/worktree-palette-runtime-owner-identity.test.ts new file mode 100644 index 00000000000..54c93135e78 --- /dev/null +++ b/src/renderer/src/lib/worktree-palette-runtime-owner-identity.test.ts @@ -0,0 +1,99 @@ +import { expect, it } from 'vitest' +import type { Repo } from '../../../shared/repo-types' +import type { Worktree } from '../../../shared/worktree/types' +import { + buildPaletteWorktreeIndex, + dedupePaletteWorktrees, + resolvePaletteWorktree +} from './palette-repo-resolution' +import { buildWorktreePaletteDocuments } from './worktree-palette-document' +import { searchWorktreeDocuments } from './worktree-palette-search' +import { + findAmbiguousWorktreeIds, + getPaletteOwnershipWorktreeIds +} from './unified-tab-host-ownership' + +function makeWorktree(runtimeOwnerEnvironmentId: string, displayName: string): Worktree { + return { + id: 'repo::/srv/same', + repoId: 'repo', + path: '/srv/same', + head: 'abc123', + branch: 'main', + isBare: false, + isMainWorktree: false, + displayName, + comment: '', + linkedIssue: null, + linkedPR: null, + linkedLinearIssue: null, + isArchived: false, + isUnread: false, + isPinned: false, + sortOrder: 0, + lastActivityAt: 0, + hostId: 'ssh:same-private-target', + runtimeOwnerEnvironmentId + } +} + +it('keeps same-target SSH worktrees from separate paired runtimes distinct', () => { + const worktrees = [ + makeWorktree('hub-a', 'AlphaOwner workspace'), + makeWorktree('hub-b', 'BetaOwner workspace') + ] + const repoMap = new Map() + const documents = buildWorktreePaletteDocuments(worktrees, { repoMap }) + const results = searchWorktreeDocuments({ worktrees, query: 'workspace', documents, repoMap }) + const alphaResults = searchWorktreeDocuments({ + worktrees, + query: 'alphaowner', + documents, + repoMap + }) + const betaResults = searchWorktreeDocuments({ + worktrees, + query: 'betaowner', + documents, + repoMap + }) + const index = buildPaletteWorktreeIndex(worktrees) + + expect(dedupePaletteWorktrees(worktrees)).toHaveLength(2) + expect(documents.size).toBe(2) + expect(results.map((result) => result.worktreeHostId)).toEqual(['runtime:hub-a', 'runtime:hub-b']) + expect(alphaResults.map((result) => result.worktreeHostId)).toEqual(['runtime:hub-a']) + expect(betaResults.map((result) => result.worktreeHostId)).toEqual(['runtime:hub-b']) + expect( + results.map( + (result) => + resolvePaletteWorktree(index, result.worktreeId, result.worktreeHostId) + ?.runtimeOwnerEnvironmentId + ) + ).toEqual(['hub-a', 'hub-b']) +}) + +it('keeps a physical-host alias only when one runtime owns it', () => { + const hubA = makeWorktree('hub-a', 'AlphaOwner workspace') + const hubB = makeWorktree('hub-b', 'BetaOwner workspace') + const uniqueIndex = buildPaletteWorktreeIndex([hubA]) + const ambiguousIndex = buildPaletteWorktreeIndex([hubA, hubB]) + + expect( + resolvePaletteWorktree(uniqueIndex, hubA.id, 'ssh:same-private-target') + ?.runtimeOwnerEnvironmentId + ).toBe('hub-a') + expect(resolvePaletteWorktree(ambiguousIndex, hubA.id, 'ssh:same-private-target')).toBeUndefined() +}) + +it('keeps both runtime owners in the tab ownership ambiguity inventory', () => { + const hubA = makeWorktree('hub-a', 'AlphaOwner workspace') + const hubB = makeWorktree('hub-b', 'BetaOwner workspace') + const ownershipWorktrees = getPaletteOwnershipWorktreeIds({ + worktreesByRepo: { repo: [hubA, hubB] }, + folderWorkspaces: [] + }) + + expect(ownershipWorktrees).toHaveLength(2) + expect(findAmbiguousWorktreeIds(ownershipWorktrees).has(hubA.id)).toBe(true) +}) diff --git a/src/renderer/src/lib/worktree-palette-search.test.ts b/src/renderer/src/lib/worktree-palette-search.test.ts index 633df3a395d..fe2fdd0e5bd 100644 --- a/src/renderer/src/lib/worktree-palette-search.test.ts +++ b/src/renderer/src/lib/worktree-palette-search.test.ts @@ -103,7 +103,9 @@ describe('worktree-palette-search', () => { hostRanges: [], supportingText: null, qualityClass: null, - rank: null + rank: null, + lastActiveAt: null, + activity: { ageBucket: null, timestamp: 0 } }) }) @@ -519,7 +521,7 @@ describe('worktree-palette-search', () => { expect(results.map((result) => result.worktreeId)).toEqual(['wt-linear']) }) - it('matches workspace ports by port number before issue and PR numbers', () => { + it('promotes an exact sigilled issue number above an ordinary port number', () => { const results = searchWorktrees( [makeWorktree({ id: 'wt-port', linkedIssue: 3000 })], '3000', @@ -528,12 +530,12 @@ describe('worktree-palette-search', () => { ) expect(results).toHaveLength(1) - expect(results[0].matchedFields).toEqual(['port']) + expect(results[0].matchedFields).toEqual(['issue']) expect(results[0].supportingText).toEqual({ - labelKind: 'port', - text: '3000 · vite', - matchRanges: [{ start: 0, end: 4 }], - accessibilityLabel: 'Listening port' + labelKind: 'issue', + text: '#3000', + matchRanges: [{ start: 1, end: 5 }], + accessibilityLabel: 'Linked issue' }) }) diff --git a/src/renderer/src/lib/worktree-palette-search.ts b/src/renderer/src/lib/worktree-palette-search.ts index 8cb637bd77a..89cd0cf06a3 100644 --- a/src/renderer/src/lib/worktree-palette-search.ts +++ b/src/renderer/src/lib/worktree-palette-search.ts @@ -1,4 +1,3 @@ -import { getWorktreeHostIdentity } from '../../../shared/worktree/host-qualified-identity' import { matchPaletteDocument } from './palette-match/match-document' import { preparePaletteQuery } from './palette-match/palette-query' import type { MatchRange } from './palette-match/normalized-text' @@ -25,11 +24,21 @@ import { import type { HostedReviewInfo } from '../../../shared/hosted-review' import type { Repo } from '../../../shared/repo-types' import type { Worktree } from '../../../shared/worktree/types' -import { resolvePaletteRepoForWorktree } from './palette-repo-resolution' +import { + getPaletteWorktreeExecutionHostId, + getPaletteWorktreeIdentity, + resolvePaletteRepoForWorktree +} from './palette-repo-resolution' import { matchWorktreePaletteTaskUrl, parseCmdJTaskSourceUrl } from './worktree-palette-task-url-match' +import { + createPaletteSearchContext, + preparePaletteActivity, + type PaletteActivityRank, + type PaletteSearchContext +} from './palette-match/palette-ranking' export type { MatchRange } @@ -56,6 +65,9 @@ export type PaletteSearchResult = { /** null for the empty query, where every worktree is listed without a match. */ qualityClass: PaletteResultQualityClass | null rank: PaletteDocumentRank | null + /** Normalized against the evaluation context for ranking and the age badge. */ + lastActiveAt: number | null + activity: PaletteActivityRank } const NO_RANGES: readonly MatchRange[] = [] @@ -76,8 +88,11 @@ export function getWorktreePaletteSearchScope(args: { export function makeEmptyPaletteSearchResult( worktreeId: string, - worktreeHostId?: Worktree['hostId'] + worktreeHostId?: Worktree['hostId'], + context = createPaletteSearchContext(Date.now()), + lastActivityAt?: number | null ): PaletteSearchResult { + const activity = preparePaletteActivity(lastActivityAt, context) return { worktreeId, ...(worktreeHostId ? { worktreeHostId } : {}), @@ -88,7 +103,9 @@ export function makeEmptyPaletteSearchResult( hostRanges: NO_RANGES, supportingText: null, qualityClass: null, - rank: null + rank: null, + lastActiveAt: activity.timestamp || null, + activity } } @@ -125,8 +142,11 @@ function toSupportingText(match: PaletteDocumentMatch): PaletteSupportingText | export function toWorktreePaletteSearchResult( worktreeId: string, match: PaletteDocumentMatch, - worktreeHostId?: Worktree['hostId'] + worktreeHostId?: Worktree['hostId'], + context = createPaletteSearchContext(Date.now()), + lastActivityAt?: number | null ): PaletteSearchResult { + const activity = preparePaletteActivity(lastActivityAt, context) const supportingText = toSupportingText(match) const matchedFields: PaletteMatchedField[] = [] for (const fieldId of match.rangesByField.keys()) { @@ -149,7 +169,9 @@ export function toWorktreePaletteSearchResult( hostRanges: match.rangesByField.get(WORKTREE_PALETTE_HOST_FIELD_ID) ?? NO_RANGES, supportingText, qualityClass: match.qualityClass, - rank: match.rank + rank: match.rank, + lastActiveAt: activity.timestamp || null, + activity } } @@ -160,17 +182,24 @@ export type WorktreePaletteSearchArgs = { repoMap: ReadonlyMap repoMapByHostIdentity?: ReadonlyMap checksReviewByWorktree?: ReadonlyMap + context?: PaletteSearchContext } /** Matches prepared documents; callers memoize `documents` across keystrokes. */ export function searchWorktreeDocuments(args: WorktreePaletteSearchArgs): PaletteSearchResult[] { + const context = args.context ?? createPaletteSearchContext(Date.now()) const prepared = preparePaletteQuery(args.query) if (prepared.state === 'invalid') { return [] } if (prepared.state === 'empty') { return args.worktrees.map((worktree) => - makeEmptyPaletteSearchResult(worktree.id, worktree.hostId) + makeEmptyPaletteSearchResult( + worktree.id, + getPaletteWorktreeExecutionHostId(worktree), + context, + worktree.lastActivityAt + ) ) } @@ -185,22 +214,36 @@ export function searchWorktreeDocuments(args: WorktreePaletteSearchArgs): Palett review: args.checksReviewByWorktree?.get(worktree) }) if (match) { - results.push(match) + const activity = preparePaletteActivity(worktree.lastActivityAt, context) + results.push({ + ...match, + lastActiveAt: activity.timestamp || null, + activity + }) } continue } - const document = args.documents.get(getWorktreeHostIdentity(worktree)) + const document = args.documents.get(getPaletteWorktreeIdentity(worktree)) if (!document) { continue } const match = matchPaletteDocument({ document, tokens: prepared.tokens, - normalizedQuery: prepared.normalized + normalizedQuery: prepared.normalized, + tokenCountBeforeDeduplication: prepared.tokenCountBeforeDeduplication }) if (match) { - results.push(toWorktreePaletteSearchResult(worktree.id, match, worktree.hostId)) + results.push( + toWorktreePaletteSearchResult( + worktree.id, + match, + getPaletteWorktreeExecutionHostId(worktree), + context, + worktree.lastActivityAt + ) + ) } } return results diff --git a/src/renderer/src/lib/worktree-palette-task-url-match.ts b/src/renderer/src/lib/worktree-palette-task-url-match.ts index 543c6a48ee0..2e433bf5f59 100644 --- a/src/renderer/src/lib/worktree-palette-task-url-match.ts +++ b/src/renderer/src/lib/worktree-palette-task-url-match.ts @@ -24,6 +24,7 @@ import { import { isWorktreePaletteQueryTooLarge } from './worktree-palette-query-bounds' import { buildWorktreePaletteTaskUrlResult } from './worktree-palette-task-url-result' import type { PaletteSearchResult } from './worktree-palette-search' +import { getPaletteWorktreeExecutionHostId } from './palette-repo-resolution' export type CmdJTaskSourceUrl = | { provider: 'github'; link: GitHubIssueOrPRLink } @@ -278,13 +279,14 @@ export function matchWorktreePaletteTaskUrl(args: { review?: HostedReviewInfo | null }): PaletteSearchResult | null { const { worktree, intent, repo, review } = args + const worktreeHostId = getPaletteWorktreeExecutionHostId(worktree) if (intent.provider === 'github') { if (!worktreeMatchesGitHubUrl(worktree, intent.link, repo, review)) { return null } return buildWorktreePaletteTaskUrlResult({ worktreeId: worktree.id, - ...(worktree.hostId ? { worktreeHostId: worktree.hostId } : {}), + ...(worktreeHostId ? { worktreeHostId } : {}), labelKind: intent.link.type === 'pr' ? 'pr' : 'issue', text: `${intent.link.type === 'pr' ? 'PR' : 'Issue'} #${intent.link.number}` }) @@ -295,7 +297,7 @@ export function matchWorktreePaletteTaskUrl(args: { } return buildWorktreePaletteTaskUrlResult({ worktreeId: worktree.id, - ...(worktree.hostId ? { worktreeHostId: worktree.hostId } : {}), + ...(worktreeHostId ? { worktreeHostId } : {}), labelKind: 'issue', text: intent.intent.identifier }) @@ -306,7 +308,7 @@ export function matchWorktreePaletteTaskUrl(args: { } return buildWorktreePaletteTaskUrlResult({ worktreeId: worktree.id, - ...(worktree.hostId ? { worktreeHostId: worktree.hostId } : {}), + ...(worktreeHostId ? { worktreeHostId } : {}), labelKind: intent.link.type === 'mr' ? 'mr' : 'issue', text: `${intent.link.type === 'mr' ? 'MR' : 'Issue'} #${intent.link.number}` }) @@ -316,7 +318,7 @@ export function matchWorktreePaletteTaskUrl(args: { } return buildWorktreePaletteTaskUrlResult({ worktreeId: worktree.id, - ...(worktree.hostId ? { worktreeHostId: worktree.hostId } : {}), + ...(worktreeHostId ? { worktreeHostId } : {}), labelKind: 'issue', text: intent.parsed.issueKey }) diff --git a/src/renderer/src/lib/worktree-palette-task-url-result.ts b/src/renderer/src/lib/worktree-palette-task-url-result.ts index 27f06757387..989fa9c4d8c 100644 --- a/src/renderer/src/lib/worktree-palette-task-url-result.ts +++ b/src/renderer/src/lib/worktree-palette-task-url-result.ts @@ -1,5 +1,6 @@ import type { PaletteSearchResult, PaletteSupportingText } from './worktree-palette-search' import type { Worktree } from '../../../shared/worktree/types' +import { createRecognizedPaletteRank } from './palette-match/palette-document' const ACCESSIBILITY_LABELS: Record = { comment: 'Workspace comment', @@ -36,14 +37,8 @@ export function buildWorktreePaletteTaskUrlResult(args: { accessibilityLabel: ACCESSIBILITY_LABELS[args.labelKind] }, qualityClass: 'exact-intent', - rank: { - exactIntent: 0, - containerOnlyTokenCount: 0, - wholeQuery: 0, - worstQuality: 0, - usesSupportingEvidence: 1, - fuzzyTokenCount: 0, - fieldHopCount: 1 - } + rank: createRecognizedPaletteRank(), + lastActiveAt: null, + activity: { ageBucket: null, timestamp: 0 } } } diff --git a/src/renderer/src/lib/worktree-setup-issue-command-queue.ts b/src/renderer/src/lib/worktree-setup-issue-command-queue.ts index 3c474d98c6a..3a66305fb67 100644 --- a/src/renderer/src/lib/worktree-setup-issue-command-queue.ts +++ b/src/renderer/src/lib/worktree-setup-issue-command-queue.ts @@ -14,7 +14,8 @@ export type IssueCommandLaunch = export function queueSetupAndIssueCommands( store: WorktreeActivationStore, worktreeId: string, - terminalTabId: string, + /** Null when the caller opens its own primary surface: setup still gets its own tab, but there is no shell to split from or return focus to. */ + terminalTabId: string | null, setup: WorktreeSetupLaunch | undefined, issueCommand: IssueCommandLaunch | undefined, wrappedSetupCommandStr: string | undefined, @@ -36,13 +37,13 @@ export function queueSetupAndIssueCommands( ...(opts?.activateCreatedTabs === false ? { activate: false } : {}) }) // Why: createTab auto-activates the new tab; revert so focus stays on the primary terminal while Setup runs in the background. - if (opts?.activateCreatedTabs !== false) { + if (opts?.activateCreatedTabs !== false && terminalTabId) { store.setActiveTab(terminalTabId) } // Why: customTitle overrides the auto "Terminal N" label everywhere the tab renders, so it's the authoritative label source. store.setTabCustomTitle(setupTab.id, 'Setup', { recordInteraction: false }) store.queueTabStartupCommand(setupTab.id, setupCommand) - } else { + } else if (terminalTabId) { store.queueTabSetupSplit(terminalTabId, { ...setupCommand, direction: mode === 'split-horizontal' ? 'horizontal' : 'vertical' @@ -51,7 +52,7 @@ export function queueSetupAndIssueCommands( } // Why: issue automation runs in its own split, queued independently from setup so both can start in parallel (separate concerns). - if (issueCommand) { + if (issueCommand && terminalTabId) { // Why: WorktreeSetupLaunch carries a runner-script file to shell out to; the TaskPage variant is already an expanded command string. const queuedIssueCommand = 'runnerScriptPath' in issueCommand diff --git a/src/renderer/src/runtime/remote-runtime-terminal-binary-controller.ts b/src/renderer/src/runtime/remote-runtime-terminal-binary-controller.ts index b8abe9da61a..c54c14378de 100644 --- a/src/renderer/src/runtime/remote-runtime-terminal-binary-controller.ts +++ b/src/renderer/src/runtime/remote-runtime-terminal-binary-controller.ts @@ -25,34 +25,28 @@ export abstract class RemoteRuntimeTerminalBinaryController extends RemoteRuntim this.failConnection(new Error('Remote terminal stream received a malformed frame.')) return } + const isOutput = + frame.opcode === TerminalStreamOpcode.Output || + frame.opcode === TerminalStreamOpcode.OutputSpan const stream = this.streams.get(frame.streamId) if (!stream) { - if ( - frame.opcode === TerminalStreamOpcode.Output || - frame.opcode === TerminalStreamOpcode.OutputSpan - ) { + if (isOutput) { // Why: the renderer already disposed this stream; unsubscribe releases server credit that cannot reach a parser. this.sendFrame(frame.streamId, TerminalStreamOpcode.Unsubscribe) } return } - if ( - (frame.opcode === TerminalStreamOpcode.Output || - frame.opcode === TerminalStreamOpcode.OutputSpan) && - shouldDropE2eRemoteTerminalOutput(stream, frame.payload.byteLength) - ) { + if (isOutput && shouldDropE2eRemoteTerminalOutput(stream, frame.payload.byteLength)) { this.queueOutputAcknowledgement(stream, frame.payload.byteLength) return } - stream.watchdog.recordInbound() + // Control frames prove transport activity, not delivery of command output. + stream.watchdog.recordInbound(isOutput) if (frame.opcode === TerminalStreamOpcode.WriteUnavailable) { stream.callbacks.onWriteUnavailable?.() return } - if ( - frame.opcode === TerminalStreamOpcode.Output || - frame.opcode === TerminalStreamOpcode.OutputSpan - ) { + if (isOutput) { this.handleOutputFrame(frame, stream) return } diff --git a/src/renderer/src/runtime/remote-runtime-terminal-response-controller.ts b/src/renderer/src/runtime/remote-runtime-terminal-response-controller.ts index a8f76c78e70..5682a1a3bdd 100644 --- a/src/renderer/src/runtime/remote-runtime-terminal-response-controller.ts +++ b/src/renderer/src/runtime/remote-runtime-terminal-response-controller.ts @@ -40,7 +40,7 @@ export abstract class RemoteRuntimeTerminalResponseController extends RemoteRunt if (!stream) { return } - stream.watchdog.recordInbound() + stream.watchdog.recordInbound(false) if (event.type === 'end' && shouldHoldE2eRemoteTerminalEnd(stream.terminal)) { return } diff --git a/src/renderer/src/runtime/remote-runtime-terminal-stall-recovery.test.ts b/src/renderer/src/runtime/remote-runtime-terminal-stall-recovery.test.ts index c6e64e4e410..b28d6507de1 100644 --- a/src/renderer/src/runtime/remote-runtime-terminal-stall-recovery.test.ts +++ b/src/renderer/src/runtime/remote-runtime-terminal-stall-recovery.test.ts @@ -236,7 +236,28 @@ describe('remote terminal stalled stream recovery', () => { stream.close() }) - it('restarts a stream when the authoritative snapshot advanced without live output', async () => { + // Why: a control frame landing just after Enter is transport activity, not the command's answer. + it.each<[string, (streamId: number) => void]>([ + ['no intervening frame', () => {}], + ['a resize acknowledgement', (id) => emitControlFrame(id, TerminalStreamOpcode.Resized)], + ['a metadata frame', (id) => emitControlFrame(id, TerminalStreamOpcode.Metadata)], + [ + 'a fit-override change', + (id) => + emitStreamEvent({ + type: 'fit-override-changed', + streamId: id, + mode: 'mobile-fit', + cols: 80, + rows: 24 + }) + ], + [ + 'a driver change', + (id) => emitStreamEvent({ type: 'driver-changed', streamId: id, driver: { kind: 'idle' } }) + ], + ['an unsolicited snapshot', (id) => emitSnapshot(id, undefined, 'baseline', 8)] + ])('recovers missing live output despite %s', async (_label, emitIntervening) => { const { getRemoteRuntimeTerminalMultiplexer } = await import('./remote-runtime-terminal-multiplexer') const onTransportClose = vi.fn() @@ -250,8 +271,10 @@ describe('remote terminal stalled stream recovery', () => { sendBinary.mockClear() expect(stream.sendInput('echo missing\r')).toBe(true) + emitIntervening(stream.streamId) await vi.advanceTimersByTimeAsync(REMOTE_TERMINAL_COMMAND_RESPONSE_TIMEOUT_MS) const request = sentFrames(TerminalStreamOpcode.SnapshotRequest)[0] + expect(request).toBeDefined() const payload = request ? decodeTerminalStreamJson<{ requestId: number }>(request.payload) : null @@ -445,6 +468,21 @@ describe('remote terminal stalled stream recovery', () => { ) } + function emitControlFrame(streamId: number, opcode: TerminalStreamOpcode): void { + callbacks?.onBinary( + encodeTerminalStreamFrame({ + opcode, + streamId, + seq: 0, + payload: encodeTerminalStreamJson({ cols: 80, rows: 24 }) + }) + ) + } + + function emitStreamEvent(result: Record): void { + callbacks?.onResponse({ ok: true, result }) + } + function emitSnapshot( streamId: number, requestId: number | undefined, diff --git a/src/renderer/src/runtime/remote-terminal-stream-watchdog.ts b/src/renderer/src/runtime/remote-terminal-stream-watchdog.ts index be9e1b9d404..290a366ecc2 100644 --- a/src/renderer/src/runtime/remote-terminal-stream-watchdog.ts +++ b/src/renderer/src/runtime/remote-terminal-stream-watchdog.ts @@ -13,7 +13,8 @@ export type RemoteTerminalStreamWatchdog = { recordOutputAcknowledged: (bytes: number) => void completeCommandResponseProbe: () => void recordCommandInput: (text: string) => void - recordInbound: () => void + /** Only live output answers a pending command; control frames prove transport activity alone. */ + recordInbound: (carriesOutput: boolean) => void dispose: () => void } @@ -111,9 +112,11 @@ export function createRemoteTerminalStreamWatchdog( REMOTE_TERMINAL_COMMAND_RESPONSE_TIMEOUT_MS ) }, - recordInbound() { + recordInbound(carriesOutput) { lastInboundAtMs = Date.now() - clearResponseTimer() + if (carriesOutput) { + clearResponseTimer() + } }, dispose() { disposed = true diff --git a/src/renderer/src/runtime/sync-runtime-graph-agent-status-projection-join.test.ts b/src/renderer/src/runtime/sync-runtime-graph-agent-status-projection-join.test.ts new file mode 100644 index 00000000000..bd4f278c04e --- /dev/null +++ b/src/renderer/src/runtime/sync-runtime-graph-agent-status-projection-join.test.ts @@ -0,0 +1,146 @@ +import { afterEach, describe, expect, it, vi } from 'vitest' +import type { AppState } from '@/store/types' +import { + buildRuntimeMobileAgentStatusProjectionForTests, + resetRuntimeMobileAgentStatusProjectionCacheForTests +} from './sync-runtime-graph' + +function makeEntry(index: number, overrides: Record = {}): never { + return { + paneKey: `tab-${index}:leaf-0`, + state: 'working', + prompt: `prompt ${index}`, + updatedAt: 1740000000000 + index * 17, + stateStartedAt: 1740000000000, + agentType: 'claude', + terminalTitle: `agent ${index}`, + stateHistory: [{ state: 'working', prompt: 'step', startedAt: 1740000000000 }], + toolName: 'shell_command', + toolInput: 'ls -la', + lastAssistantMessage: 'answer', + ...overrides + } as never +} + +function mapOf(indices: readonly number[]): AppState['agentStatusByPaneKey'] { + const map: AppState['agentStatusByPaneKey'] = {} + for (const index of indices) { + map[`tab-${index}:leaf-0`] = makeEntry(index) + } + return map +} + +/** Re-spread with the same entry objects, as a status ping does. */ +function respread(map: AppState['agentStatusByPaneKey']): AppState['agentStatusByPaneKey'] { + return { ...map } +} + +function countJoins(run: () => string): { + result: string + joins: number + sorts: number +} { + const originalJoin = Array.prototype.join + const originalSort = Array.prototype.sort + let joins = 0 + let sorts = 0 + const joinSpy = vi.spyOn(Array.prototype, 'join').mockImplementation(function ( + this: unknown[], + separator?: string + ) { + joins += 1 + return originalJoin.call(this, separator) + }) + const sortSpy = vi.spyOn(Array.prototype, 'sort').mockImplementation(function ( + this: unknown[], + compare?: (a: unknown, b: unknown) => number + ) { + sorts += 1 + return originalSort.call(this, compare) + }) + try { + return { result: run(), joins, sorts } + } finally { + joinSpy.mockRestore() + sortSpy.mockRestore() + } +} + +describe('agent-status projection join short circuit', () => { + afterEach(() => { + resetRuntimeMobileAgentStatusProjectionCacheForTests() + }) + + it('skips the join when a re-spread reuses every entry', () => { + resetRuntimeMobileAgentStatusProjectionCacheForTests() + const map = mapOf([0, 1, 2]) + const first = buildRuntimeMobileAgentStatusProjectionForTests(map) + + // A new map identity with identical entry references — the common ping shape. + const { result, joins, sorts } = countJoins(() => + buildRuntimeMobileAgentStatusProjectionForTests(respread(map)) + ) + + expect(result).toBe(first) + expect(joins).toBe(0) + expect(sorts).toBe(0) + }) + + it('caches the new map identity on the short-circuit path', () => { + resetRuntimeMobileAgentStatusProjectionCacheForTests() + const map = mapOf([0, 1]) + buildRuntimeMobileAgentStatusProjectionForTests(map) + const again = respread(map) + buildRuntimeMobileAgentStatusProjectionForTests(again) + + // A repeat call with the same identity must hit the identity early-out, not re-walk the keys. + const { joins, sorts } = countJoins(() => + buildRuntimeMobileAgentStatusProjectionForTests(again) + ) + expect(joins).toBe(0) + expect(sorts).toBe(0) + }) + + it('still rebuilds when an entry changes', () => { + resetRuntimeMobileAgentStatusProjectionCacheForTests() + const map = mapOf([0, 1]) + const first = buildRuntimeMobileAgentStatusProjectionForTests(map) + + const changed = { ...map, 'tab-1:leaf-0': makeEntry(1, { state: 'idle' }) } + const { result, joins } = countJoins(() => + buildRuntimeMobileAgentStatusProjectionForTests(changed) + ) + + expect(result).not.toBe(first) + expect(joins).toBeGreaterThan(0) + }) + + it('still rebuilds when a pane is removed, even though every survivor is reused', () => { + // The reuse check alone cannot see a removal; only the entry-count check does. + resetRuntimeMobileAgentStatusProjectionCacheForTests() + const map = mapOf([0, 1]) + const first = buildRuntimeMobileAgentStatusProjectionForTests(map) + + const removed = { 'tab-0:leaf-0': map['tab-0:leaf-0'] } + const result = buildRuntimeMobileAgentStatusProjectionForTests(removed) + + expect(result).not.toBe(first) + expect(result).toBe( + buildRuntimeMobileAgentStatusProjectionForTests({ + 'tab-0:leaf-0': map['tab-0:leaf-0'] + }) + ) + }) + + it('still rebuilds when a pane is added', () => { + resetRuntimeMobileAgentStatusProjectionCacheForTests() + const map = mapOf([0]) + const first = buildRuntimeMobileAgentStatusProjectionForTests(map) + + const added = { ...map, 'tab-9:leaf-0': makeEntry(9) } + const result = buildRuntimeMobileAgentStatusProjectionForTests(added) + + expect(result).not.toBe(first) + expect(result).toContain('tab-9:leaf-0') + }) +}) diff --git a/src/renderer/src/runtime/sync-runtime-graph/agent-status-projection.ts b/src/renderer/src/runtime/sync-runtime-graph/agent-status-projection.ts index 5e14bd4a8ae..979df97762b 100644 --- a/src/renderer/src/runtime/sync-runtime-graph/agent-status-projection.ts +++ b/src/renderer/src/runtime/sync-runtime-graph/agent-status-projection.ts @@ -40,15 +40,30 @@ export function buildRuntimeMobileAgentStatusProjection( return cached.projection } + const nextEntries = Object.entries(agentStatusByPaneKey) + // Same key set, same entry objects: the sorted join would be character-identical to the cached + // string, so skip the O(N log N) sort and the O(bytes) join. Equal sizes plus every next key + // present in the cache proves the key sets match; a removal fails the size check and an addition + // fails the lookup. The cached Map is exactly what a rebuild would produce, so reuse it too. + if ( + cached != null && + nextEntries.length === cached.entries.size && + nextEntries.every(([paneKey, entry]) => cached.entries.get(paneKey)?.entry === entry) + ) { + graphState.cachedAgentStatusProjection = { + ...cached, + source: agentStatusByPaneKey + } + return cached.projection + } + // A status ping replaces one entry and re-spreads the map; reuse every other entry. const entries = new Map() const parts: string[] = [] // Code-unit order, not `localeCompare`: this projection is only ever compared with `===`, so it // must be deterministic, not locale-correct — and an ICU collator per comparison is ~4.5k calls // per ping at the 500-entry cap. - for (const [paneKey, entry] of Object.entries(agentStatusByPaneKey).sort(([a], [b]) => - a < b ? -1 : a > b ? 1 : 0 - )) { + for (const [paneKey, entry] of nextEntries.sort(([a], [b]) => (a < b ? -1 : a > b ? 1 : 0))) { const previous = cached?.entries.get(paneKey) const entryCache = previous?.entry === entry diff --git a/src/renderer/src/runtime/web-runtime-session-terminal-legacy-create.test.ts b/src/renderer/src/runtime/web-runtime-session-terminal-legacy-create.test.ts index 6b1ba5e6d7c..76ec9f5e150 100644 --- a/src/renderer/src/runtime/web-runtime-session-terminal-legacy-create.test.ts +++ b/src/renderer/src/runtime/web-runtime-session-terminal-legacy-create.test.ts @@ -131,6 +131,71 @@ describe('createWebRuntimeSessionTerminal', () => { ]) }) + it.each([ + { + agent: 'codex' as const, + predecessor: 'web-terminal-host-tab-1', + afterTabId: 'web-terminal-host-tab-1%3A%3Aleaf-1' + }, + { + agent: undefined, + predecessor: 'web-terminal-host-tab-1', + afterTabId: 'web-terminal-host-tab-1%3A%3Aleaf-1' + }, + { agent: undefined, predecessor: 'local-browser-tab', afterTabId: 'local-browser-tab' }, + { + agent: undefined, + predecessor: 'web-terminal-host-tab-1%3A%3Aleaf-1', + afterTabId: 'web-terminal-host-tab-1%3A%3Aleaf-1' + } + ])( + 'settles after $afterTabId for $agent creation without activating it', + async ({ agent, predecessor, afterTabId }) => { + const successor = 'web-terminal-host-tab-3' + const created = 'web-terminal-host-tab-2' + const reorderUnifiedTabs = vi.fn() + mocks.getState.mockReturnValue({ + ...mocks.getState(), + unifiedTabsByWorktree: { + [WORKTREE_ID]: [predecessor, successor, created].map((id) => ({ + id, + groupId: 'client-group' + })) + }, + groupsByWorktree: { + [WORKTREE_ID]: [{ id: 'client-group', tabOrder: [predecessor, successor, created] }] + }, + reorderUnifiedTabs, + moveUnifiedTabToGroup: mocks.moveUnifiedTabToGroup + }) + const runtimeCall = vi.fn(async (request: { method: string }) => ({ + id: request.method, + ok: true, + result: + request.method === 'session.tabs.createTerminal' + ? { tab: { id: 'host-tab-2::leaf-2' }, publicationEpoch: 'epoch-1', snapshotVersion: 2 } + : makeSnapshot() + })) + vi.stubGlobal('window', { api: { runtimeEnvironments: { call: runtimeCall } } }) + + await expect( + createWebRuntimeSessionTerminal({ + worktreeId: WORKTREE_ID, + afterTabId, + agent, + activate: false + }) + ).resolves.toEqual({ status: 'created' }) + + expect(reorderUnifiedTabs).toHaveBeenCalledExactlyOnceWith( + 'client-group', + [predecessor, created, successor], + { recordInteraction: false } + ) + expect(mocks.moveUnifiedTabToGroup).not.toHaveBeenCalled() + } + ) + it('can create a terminal without selecting the target worktree', async () => { const setStateResults: unknown[] = [] mocks.setState.mockImplementation((updater: (state: unknown) => unknown) => { diff --git a/src/renderer/src/runtime/web-runtime-terminal-create-operation.ts b/src/renderer/src/runtime/web-runtime-terminal-create-operation.ts index 5ca20bdb7f3..601888e5f9b 100644 --- a/src/renderer/src/runtime/web-runtime-terminal-create-operation.ts +++ b/src/renderer/src/runtime/web-runtime-terminal-create-operation.ts @@ -248,20 +248,26 @@ export async function createWebRuntimeSessionTerminalResult( // tab to THIS new terminal, instead of sticky-keeping the prior tab. recordWebSessionFocusIntent(intentOwner, args.worktreeId, createdTabId, createdLeafId) } + const placementTabId = + createdTabId && (args.targetGroupId || args.afterTabId) ? createdTabId : undefined await refreshWebRuntimeSessionTabsSnapshot(environmentId, args.worktreeId, { expectedEnvironmentPairingRevision: intentOwner.pairingRevision, // Why: the publication can beat the RPC response; replay it once after caller intent exists. acceptCurrentSnapshot: - Boolean(createdTabId) && (args.activate !== false || Boolean(args.targetGroupId)), + Boolean(createdTabId) && (args.activate !== false || Boolean(placementTabId)), // Why: a placement record needs a post-create list; a deduped in-flight one can predate it. - ...(args.targetGroupId && createdTabId ? { afterCurrentInFlight: true } : {}) + ...(placementTabId ? { afterCurrentInFlight: true } : {}) }) - if (args.targetGroupId && createdTabId) { + if (placementTabId) { await settleWebRuntimeTerminalPlacement( environmentId, args.worktreeId, - webTerminalPlacementParentTabId(createdTabId), - { groupId: args.targetGroupId, activate: args.activate !== false } + webTerminalPlacementParentTabId(placementTabId), + { + groupId: args.targetGroupId, + afterTabId: args.afterTabId, + activate: args.activate !== false + } ) } return { diff --git a/src/renderer/src/runtime/web-runtime-terminal-placement-settlement.ts b/src/renderer/src/runtime/web-runtime-terminal-placement-settlement.ts index 50cbb551f97..e1b10d8b5b2 100644 --- a/src/renderer/src/runtime/web-runtime-terminal-placement-settlement.ts +++ b/src/renderer/src/runtime/web-runtime-terminal-placement-settlement.ts @@ -1,13 +1,31 @@ +import { insertUnifiedTabAfterAnchor } from '../lib/unified-tab-anchor-insertion' import { useAppStore } from '../store' -import { forgetWebSessionTerminalPlacement } from './web-session-terminal-placement' -import { toWebTerminalSurfaceTabId } from './web-terminal-surface-id' +import { + forgetWebSessionTerminalPlacement, + webTerminalPlacementParentTabId +} from './web-session-terminal-placement' +import { + isWebTerminalSurfaceTabId, + toHostSessionTabId, + toWebTerminalSurfaceTabId +} from './web-terminal-surface-id' + +/** Snapshots key mirrored terminals by the parent tab, so an unknown `parent::leaf` anchor resolves to its parent. */ +function anchorUnifiedTabId(worktreeId: string, afterTabId: string): string { + const known = (useAppStore.getState().unifiedTabsByWorktree[worktreeId] ?? []).some( + (tab) => tab.id === afterTabId + ) + return known || !isWebTerminalSurfaceTabId(afterTabId) + ? afterTabId + : toWebTerminalSurfaceTabId(webTerminalPlacementParentTabId(toHostSessionTabId(afterTabId))) +} /** Settle the placement once the mirrored tab exists (bounded poll), then consume the record. */ export async function settleWebRuntimeTerminalPlacement( environmentId: string, worktreeId: string, hostTabId: string, - placement: { groupId: string; activate: boolean } + placement: { groupId?: string; afterTabId?: string; activate: boolean } ): Promise { const unifiedTabId = toWebTerminalSurfaceTabId(hostTabId) const findTab = () => @@ -20,18 +38,36 @@ export async function settleWebRuntimeTerminalPlacement( await new Promise((resolve) => setTimeout(resolve, 250)) } const tab = findTab() + if (!tab) { + return + } + const anchorId = placement.afterTabId + ? anchorUnifiedTabId(worktreeId, placement.afterTabId) + : undefined const state = useAppStore.getState() - const targetGroupExists = (state.groupsByWorktree[worktreeId] ?? []).some( - (group) => group.id === placement.groupId - ) - if (tab && targetGroupExists && tab.groupId !== placement.groupId) { + const groups = state.groupsByWorktree[worktreeId] ?? [] + // Why: the requested group can be closed while the mirrored tab is still in flight; the + // anchor's own group still expresses where the caller asked for this terminal. + const targetGroup = + groups.find((group) => group.id === placement.groupId) ?? + (anchorId === undefined + ? undefined + : groups.find((group) => group.tabOrder.includes(anchorId))) + if (!targetGroup) { + return + } + if (tab.groupId !== targetGroup.id) { // Why: a snapshot can adopt the tab before the record exists (the publication races the // RPC response); repair through the same client-owned move a user drag takes. - state.moveUnifiedTabToGroup(unifiedTabId, placement.groupId, { + state.moveUnifiedTabToGroup(unifiedTabId, targetGroup.id, { activate: placement.activate, recordInteraction: false }) } + if (anchorId) { + // The create caller owns this insertion; subsequent host snapshots preserve client order. + insertUnifiedTabAfterAnchor(worktreeId, unifiedTabId, anchorId) + } } finally { forgetWebSessionTerminalPlacement({ environmentId, worktreeId, hostTabId }) } diff --git a/src/renderer/src/store/copy-on-write-record.test.ts b/src/renderer/src/store/copy-on-write-record.test.ts new file mode 100644 index 00000000000..79c351e63fe --- /dev/null +++ b/src/renderer/src/store/copy-on-write-record.test.ts @@ -0,0 +1,30 @@ +import { describe, expect, it } from 'vitest' +import { copyOnWriteRecord } from './copy-on-write-record' + +describe('copyOnWriteRecord', () => { + it('returns the source untouched when nothing is written', () => { + const source = { a: 1 } + const record = copyOnWriteRecord(source) + record.delete('missing') + expect(record.read()).toBe(source) + expect(source).toEqual({ a: 1 }) + }) + + it('clones once and never mutates the source', () => { + const source = { a: 1, b: 2 } + const record = copyOnWriteRecord(source) + record.set('c', 3) + const afterFirstWrite = record.read() + record.delete('a') + expect(record.read()).toBe(afterFirstWrite) + expect(record.read()).toEqual({ b: 2, c: 3 }) + expect(source).toEqual({ a: 1, b: 2 }) + }) + + it('deletes a key added after the clone', () => { + const record = copyOnWriteRecord({}) + record.set('a', 1) + record.delete('a') + expect(record.read()).toEqual({}) + }) +}) diff --git a/src/renderer/src/store/copy-on-write-record.ts b/src/renderer/src/store/copy-on-write-record.ts new file mode 100644 index 00000000000..485f9a94c31 --- /dev/null +++ b/src/renderer/src/store/copy-on-write-record.ts @@ -0,0 +1,32 @@ +export type CopyOnWriteRecord = { + /** The source until the first write, then the one clone every later write reuses. */ + read: () => Record + /** No-op for an absent key, so deleting nothing never clones. */ + delete: (key: string) => void + set: (key: string, value: T) => void +} + +/** + * Lets a store patch touch a record only when it has something to change: an untouched + * source keeps its identity, so identity-keyed selectors and persist gates stay quiet. + */ +export function copyOnWriteRecord(source: Record): CopyOnWriteRecord { + let next = source + const mutable = (): Record => { + if (next === source) { + next = { ...source } + } + return next + } + return { + read: () => next, + delete: (key) => { + if (key in next) { + delete mutable()[key] + } + }, + set: (key, value) => { + mutable()[key] = value + } + } +} diff --git a/src/renderer/src/store/index.ts b/src/renderer/src/store/index.ts index 48976018f81..5f783852f82 100644 --- a/src/renderer/src/store/index.ts +++ b/src/renderer/src/store/index.ts @@ -1,4 +1,4 @@ -import { create } from 'zustand' +import { create, type StateCreator } from 'zustand' import type { AppState } from './types' import { createRepoSlice } from './slices/repos' import { createSparsePresetsSlice } from './slices/sparse-presets' @@ -53,62 +53,73 @@ import { } from '@/lib/http-link-routing' import { installStoreListenerCensus } from './store-listener-census' import { withReactCommitCascadeWriteProbe } from './react-commit-cascade-write-probe' +import { withStoreIdentityChurnProbe } from './store-identity-churn-probe' import { registerRendererMemoryProfileContributor, summarizeStateCollectionSizes } from '@/lib/renderer-memory-profile' import { estimateStateCollectionKB } from '@/lib/state-collection-byte-estimate' +// Why dev-only: nothing in the app arms the churn probe, so a shipped build would +// pay its wrapper frame on every write for a diagnostic it can never read. The +// cascade probe stays unconditional because crash telemetry arms it in the field. +const withDevelopmentStoreProbes = (createState: StateCreator) => + import.meta.env.DEV || e2eConfig.exposeStore + ? withStoreIdentityChurnProbe(createState) + : createState + export const useAppStore = create()( - withReactCommitCascadeWriteProbe((...a) => { - // Why: the inner api is only reachable here, before create() copies subscribe onto the hook. - installStoreListenerCensus(a[2]) - return { - ...createRepoSlice(...a), - ...createSparsePresetsSlice(...a), - ...createWorktreeSlice(...a), - ...createTerminalSlice(...a), - ...createTabsSlice(...a), - ...createUISlice(...a), - ...createSettingsSlice(...a), - ...createKeybindingsSlice(...a), - ...createGitHubSlice(...a), - ...createHostedReviewSlice(...a), - ...createLinearSlice(...a), - ...createPreflightSlice(...a), - ...createJiraSlice(...a), - ...createEditorSlice(...a), - ...createStatsSlice(...a), - ...createMemorySlice(...a), - ...createWorkspaceSpaceSlice(...a), - ...createClaudeUsageSlice(...a), - ...createCodexUsageSlice(...a), - ...createOpenCodeUsageSlice(...a), - ...createBrowserSlice(...a), - ...createRateLimitSlice(...a), - ...createSshSlice(...a), - ...createRuntimeEnvironmentSshSlice(...a), - ...createAgentStatusSlice(...a), - ...createPaneForegroundAgentSlice(...a), - ...createDiffCommentsSlice(...a), - ...createDetectedAgentsSlice(...a), - ...createRuntimeDetectedAgentsSlice(...a), - ...createWorktreeNavHistorySlice(...a), - ...createDictationSlice(...a), - ...createWorkspaceCleanupSlice(...a), - ...createWorkspaceCleanupBrowseSlice(...a), - ...createRuntimeStatusSlice(...a), - ...createPullRequestGenerationSlice(...a), - ...createCommitMessageGenerationSlice(...a), - ...createPinnedTabCloseConfirmSlice(...a), - ...createRecentlyClosedTabsSlice(...a), - ...createOrcaProfilesSlice(...a), - ...createNewIssueDraftSlice(...a), - ...createTaskCreationDraftsSlice(...a), - ...createRemoteServerUpdatesSlice(...a), - ...createTerminalQuickCommandHostsSlice(...a) - } - }) + withDevelopmentStoreProbes( + withReactCommitCascadeWriteProbe((...a) => { + // Why: the inner api is only reachable here, before create() copies subscribe onto the hook. + installStoreListenerCensus(a[2]) + return { + ...createRepoSlice(...a), + ...createSparsePresetsSlice(...a), + ...createWorktreeSlice(...a), + ...createTerminalSlice(...a), + ...createTabsSlice(...a), + ...createUISlice(...a), + ...createSettingsSlice(...a), + ...createKeybindingsSlice(...a), + ...createGitHubSlice(...a), + ...createHostedReviewSlice(...a), + ...createLinearSlice(...a), + ...createPreflightSlice(...a), + ...createJiraSlice(...a), + ...createEditorSlice(...a), + ...createStatsSlice(...a), + ...createMemorySlice(...a), + ...createWorkspaceSpaceSlice(...a), + ...createClaudeUsageSlice(...a), + ...createCodexUsageSlice(...a), + ...createOpenCodeUsageSlice(...a), + ...createBrowserSlice(...a), + ...createRateLimitSlice(...a), + ...createSshSlice(...a), + ...createRuntimeEnvironmentSshSlice(...a), + ...createAgentStatusSlice(...a), + ...createPaneForegroundAgentSlice(...a), + ...createDiffCommentsSlice(...a), + ...createDetectedAgentsSlice(...a), + ...createRuntimeDetectedAgentsSlice(...a), + ...createWorktreeNavHistorySlice(...a), + ...createDictationSlice(...a), + ...createWorkspaceCleanupSlice(...a), + ...createWorkspaceCleanupBrowseSlice(...a), + ...createRuntimeStatusSlice(...a), + ...createPullRequestGenerationSlice(...a), + ...createCommitMessageGenerationSlice(...a), + ...createPinnedTabCloseConfirmSlice(...a), + ...createRecentlyClosedTabsSlice(...a), + ...createOrcaProfilesSlice(...a), + ...createNewIssueDraftSlice(...a), + ...createTaskCreationDraftsSlice(...a), + ...createRemoteServerUpdatesSlice(...a), + ...createTerminalQuickCommandHostsSlice(...a) + } + }) + ) ) registerHttpLinkStoreAccessor(() => useAppStore.getState()) diff --git a/src/renderer/src/store/slices/agent-pane-authority.test.ts b/src/renderer/src/store/slices/agent-pane-authority.test.ts index 5e97e3c9064..01e99f27d6f 100644 --- a/src/renderer/src/store/slices/agent-pane-authority.test.ts +++ b/src/renderer/src/store/slices/agent-pane-authority.test.ts @@ -74,6 +74,20 @@ describe('agent pane authority', () => { expect(retirePaneAuthority).toHaveBeenCalledWith(TARGET) }) + it('re-retiring an already-retired pane keeps the retired-key map identity and epochs', () => { + const store = createTestStore() + store.getState().setAgentStatus(TARGET, { state: 'working', prompt: 'target' }) + store.getState().retireAgentPaneAuthority(TARGET) + const before = store.getState() + + store.getState().retireAgentPaneAuthority(TARGET) + + const after = store.getState() + expect(after.recentlyRetiredAgentStatusPaneKeys).toBe(before.recentlyRetiredAgentStatusPaneKeys) + expect(after.agentStatusEpoch).toBe(before.agentStatusEpoch) + expect(after.sortEpoch).toBe(before.sortEpoch) + }) + it('retires the pane activity cutoff with the rest of its pane-owned state', () => { const store = createTestStore() store.setState({ diff --git a/src/renderer/src/store/slices/agent-status-contract.ts b/src/renderer/src/store/slices/agent-status-contract.ts index 64dcb7f6919..39bbde535d1 100644 --- a/src/renderer/src/store/slices/agent-status-contract.ts +++ b/src/renderer/src/store/slices/agent-status-contract.ts @@ -92,6 +92,8 @@ export type AgentStatusPayload = ParsedAgentStatusPayload & { } export type AgentStatusTiming = { + /** Ordered authoritative sources may correct a prior publication clock. */ + allowOlderTimestamp?: boolean updatedAt?: number /** Observation clock for staleness; see `AgentStatusEntry.evidenceObservedAt`. */ evidenceObservedAt?: number diff --git a/src/renderer/src/store/slices/agent-status-live-entry-builder.ts b/src/renderer/src/store/slices/agent-status-live-entry-builder.ts index 12812b380c7..5c3ef89bdaf 100644 --- a/src/renderer/src/store/slices/agent-status-live-entry-builder.ts +++ b/src/renderer/src/store/slices/agent-status-live-entry-builder.ts @@ -74,7 +74,7 @@ export function buildAgentStatusLiveEntry( ): AgentStatusLiveEntryBuild | AgentStatusLiveEntryRejection { const { state, paneKey, payload, terminalTitle, timing, routing, metadata, updatedAt } = args const existing = state.agentStatusByPaneKey[paneKey] - if (existing && updatedAt < existing.updatedAt) { + if (existing && updatedAt < existing.updatedAt && !timing?.allowOlderTimestamp) { return { entry: null, reason: 'stale' } } const effectiveTitle = terminalTitle ?? existing?.terminalTitle diff --git a/src/renderer/src/store/slices/agent-status-pane-keyed-records.test.ts b/src/renderer/src/store/slices/agent-status-pane-keyed-records.test.ts new file mode 100644 index 00000000000..5b327561637 --- /dev/null +++ b/src/renderer/src/store/slices/agent-status-pane-keyed-records.test.ts @@ -0,0 +1,110 @@ +import { describe, expect, it } from 'vitest' +import { + RECENTLY_CLOSED_AGENT_STATUS_TAB_IDS_MAX, + RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX, + boundRecentlyClosedAgentStatusTabIds, + boundRecentlyRetiredAgentStatusPaneKeys +} from './agent-status-pane-keyed-records' + +function keyRecord(keys: readonly string[]): Record { + const record: Record = {} + for (const key of keys) { + record[key] = true + } + return record +} + +function fullRecord(max: number, prefix: string): Record { + return keyRecord(Array.from({ length: max }, (_, i) => `${prefix}${i}`)) +} + +describe('boundRecentlyRetiredAgentStatusPaneKeys', () => { + it('returns the existing record when there is nothing to add', () => { + const existing = keyRecord(['a', 'b']) + expect(boundRecentlyRetiredAgentStatusPaneKeys(existing, [])).toBe(existing) + const empty = keyRecord([]) + expect(boundRecentlyRetiredAgentStatusPaneKeys(empty, [])).toBe(empty) + }) + + it('returns the existing record when the additions already form its tail in order', () => { + const existing = keyRecord(['a', 'b', 'c']) + expect(boundRecentlyRetiredAgentStatusPaneKeys(existing, ['c'])).toBe(existing) + expect(boundRecentlyRetiredAgentStatusPaneKeys(existing, ['b', 'c'])).toBe(existing) + expect(boundRecentlyRetiredAgentStatusPaneKeys(existing, ['a', 'b', 'c'])).toBe(existing) + }) + + // Why: LRU order decides which key the cap evicts next. A key-set match is not a + // no-op when the re-added key is not already at the tail — it must move there. + it('re-retiring an existing non-tail key changes identity and moves it to the tail', () => { + const existing = keyRecord(['a', 'b', 'c']) + const next = boundRecentlyRetiredAgentStatusPaneKeys(existing, ['a']) + expect(next).not.toBe(existing) + expect(Object.keys(next)).toEqual(['b', 'c', 'a']) + expect(Object.keys(existing)).toEqual(['a', 'b', 'c']) + }) + + it('tail keys re-added in a different relative order are rebuilt in the new order', () => { + const existing = keyRecord(['a', 'b', 'c']) + const next = boundRecentlyRetiredAgentStatusPaneKeys(existing, ['c', 'b']) + expect(next).not.toBe(existing) + expect(Object.keys(next)).toEqual(['a', 'c', 'b']) + }) + + it('appends new keys after the existing ones', () => { + const existing = keyRecord(['a']) + const next = boundRecentlyRetiredAgentStatusPaneKeys(existing, ['b', 'c']) + expect(next).not.toBe(existing) + expect(Object.keys(next)).toEqual(['a', 'b', 'c']) + }) + + it('evicts the oldest keys once the cap is exceeded', () => { + const full = fullRecord(RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX, 'k') + const next = boundRecentlyRetiredAgentStatusPaneKeys(full, ['fresh']) + const keys = Object.keys(next) + expect(keys).toHaveLength(RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX) + expect(keys[0]).toBe('k1') + expect(keys.at(-1)).toBe('fresh') + expect(next.k0).toBeUndefined() + }) + + it('re-retiring the oldest key at the cap keeps it fenced and evicts the next oldest', () => { + const full = fullRecord(RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX, 'k') + const bumped = boundRecentlyRetiredAgentStatusPaneKeys(full, ['k0']) + expect(bumped).not.toBe(full) + expect(Object.keys(bumped).at(-1)).toBe('k0') + const afterFresh = boundRecentlyRetiredAgentStatusPaneKeys(bumped, ['fresh']) + expect(afterFresh.k0).toBe(true) + expect(afterFresh.k1).toBeUndefined() + }) + + it('never returns an over-cap record unchanged', () => { + const over = fullRecord(RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX + 1, 'k') + const last = `k${RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX}` + const next = boundRecentlyRetiredAgentStatusPaneKeys(over, [last]) + expect(next).not.toBe(over) + expect(Object.keys(next)).toHaveLength(RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX) + expect(next.k0).toBeUndefined() + }) +}) + +describe('boundRecentlyClosedAgentStatusTabIds', () => { + it('returns the existing record when the tab is already the most recent', () => { + const existing = keyRecord(['t1', 't2']) + expect(boundRecentlyClosedAgentStatusTabIds(existing, 't2')).toBe(existing) + }) + + it('moves a re-closed tab to the tail', () => { + const existing = keyRecord(['t1', 't2']) + const next = boundRecentlyClosedAgentStatusTabIds(existing, 't1') + expect(next).not.toBe(existing) + expect(Object.keys(next)).toEqual(['t2', 't1']) + }) + + it('evicts the oldest tab once the cap is exceeded', () => { + const full = fullRecord(RECENTLY_CLOSED_AGENT_STATUS_TAB_IDS_MAX, 't') + const next = boundRecentlyClosedAgentStatusTabIds(full, 'fresh') + expect(Object.keys(next)).toHaveLength(RECENTLY_CLOSED_AGENT_STATUS_TAB_IDS_MAX) + expect(next.t0).toBeUndefined() + expect(next.fresh).toBe(true) + }) +}) diff --git a/src/renderer/src/store/slices/agent-status-pane-keyed-records.ts b/src/renderer/src/store/slices/agent-status-pane-keyed-records.ts index af5cc4e83c7..f9199140c07 100644 --- a/src/renderer/src/store/slices/agent-status-pane-keyed-records.ts +++ b/src/renderer/src/store/slices/agent-status-pane-keyed-records.ts @@ -3,47 +3,66 @@ export const RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX = 1024 // delete-then-set for LRU recency, then evict oldest keys past the cap (Record iterates // insertion order); safe because a status for a tab closed >MAX tabs ago cannot still arrive. -export function boundRecentlyClosedAgentStatusTabIds( +function boundLruKeyRecord( existing: Record, - tabId: string + additions: ReadonlySet, + max: number ): Record { - const next: Record = {} - for (const key of Object.keys(existing)) { - if (key !== tabId) { - next[key] = true - } + if (isLruKeyRecordUnchanged(existing, additions, max)) { + return existing } - next[tabId] = true - const keys = Object.keys(next) - if (keys.length > RECENTLY_CLOSED_AGENT_STATUS_TAB_IDS_MAX) { - for (const stale of keys.slice(0, keys.length - RECENTLY_CLOSED_AGENT_STATUS_TAB_IDS_MAX)) { - delete next[stale] - } - } - return next -} - -export function boundRecentlyRetiredAgentStatusPaneKeys( - existing: Record, - paneKeys: readonly string[] -): Record { - const additions = new Set(paneKeys) const next: Record = {} for (const key of Object.keys(existing)) { if (!additions.has(key)) { next[key] = true } } - for (const paneKey of additions) { - next[paneKey] = true + for (const key of additions) { + next[key] = true } const keys = Object.keys(next) - for (const stale of keys.slice(0, -RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX)) { + for (const stale of keys.slice(0, -max)) { delete next[stale] } return next } +// The rebuild is a no-op only when nothing would be evicted and the additions are +// already the tail of `existing` in that same relative order. A matching key SET is +// not enough: re-adding a key moves it to the tail, and that order decides which key +// the cap evicts next, so a stale-order hit would un-fence a recently retired pane. +function isLruKeyRecordUnchanged( + existing: Record, + additions: ReadonlySet, + max: number +): boolean { + const keys = Object.keys(existing) + if (keys.length > max || additions.size > keys.length) { + return false + } + let index = keys.length - additions.size + for (const key of additions) { + if (keys[index++] !== key) { + return false + } + } + return true +} + +export function boundRecentlyClosedAgentStatusTabIds( + existing: Record, + tabId: string +): Record { + return boundLruKeyRecord(existing, new Set([tabId]), RECENTLY_CLOSED_AGENT_STATUS_TAB_IDS_MAX) +} + +export function boundRecentlyRetiredAgentStatusPaneKeys( + existing: Record, + paneKeys: readonly string[] +): Record { + return boundLruKeyRecord(existing, new Set(paneKeys), RECENTLY_RETIRED_AGENT_STATUS_PANE_KEYS_MAX) +} + export function movePaneKeyedRecord( record: Record, fromPaneKey: string, diff --git a/src/renderer/src/store/slices/browser/browser-close-actions.ts b/src/renderer/src/store/slices/browser/browser-close-actions.ts index c70b93f6eb2..3aa77f9e46c 100644 --- a/src/renderer/src/store/slices/browser/browser-close-actions.ts +++ b/src/renderer/src/store/slices/browser/browser-close-actions.ts @@ -12,6 +12,7 @@ import { getFallbackTabTypeForWorktree, isLocalBrowserPageOwner } from './browse import { closeRemoteBrowserPageInOwningEnvironment } from './browser-remote-close' import { releaseDocPreviewGrant } from '@/lib/doc-preview-grants' import { destroyWorkspaceWebviews } from '../browser-webview-cleanup' +import { omitRecordKeys } from '../worktrees/teardown/record-key-omission' export function createBrowserCloseActions( set: BrowserSliceSet, @@ -231,10 +232,15 @@ export function createBrowserCloseActions( destroyWorkspaceWebviews(browserPagesByWorkspace, workspace.id) } set((s) => { - const nextBrowserTabsByWorktree = { ...s.browserTabsByWorktree } - delete nextBrowserTabsByWorktree[worktreeId] - const nextActiveBrowserTabIdByWorktree = { ...s.activeBrowserTabIdByWorktree } - delete nextActiveBrowserTabIdByWorktree[worktreeId] + const removedWorktreeIds = [worktreeId] + const nextBrowserTabsByWorktree = omitRecordKeys( + s.browserTabsByWorktree, + removedWorktreeIds + ) + const nextActiveBrowserTabIdByWorktree = omitRecordKeys( + s.activeBrowserTabIdByWorktree, + removedWorktreeIds + ) // Why: reset the global browser surface only when the shut-down worktree is the active one AND had tabs. const shouldResetGlobalBrowser = s.activeWorktreeId === worktreeId && hadBrowserTabs return { diff --git a/src/renderer/src/store/slices/rate-limits.ts b/src/renderer/src/store/slices/rate-limits.ts index 3df495cf32d..7b045c74e3b 100644 --- a/src/renderer/src/store/slices/rate-limits.ts +++ b/src/renderer/src/store/slices/rate-limits.ts @@ -26,6 +26,7 @@ export const createRateLimitSlice: StateCreator { + it('keeps every tab-model map when only an orphan runtime terminal changed', () => { + const groupId = 'g-1' + const store = createTestStore() + store.setState({ + unifiedTabsByWorktree: { + [WORKTREE]: [ + { + id: 'sim-1', + entityId: 'sim-1', + groupId, + worktreeId: WORKTREE, + contentType: 'simulator', + label: 'Simulator', + customLabel: null, + color: null, + sortOrder: 0, + createdAt: 1 + } + ] + }, + groupsByWorktree: { + [WORKTREE]: [ + { + id: groupId, + worktreeId: WORKTREE, + activeTabId: 'sim-1', + tabOrder: ['sim-1'] + } + ] + }, + activeGroupIdByWorktree: { [WORKTREE]: groupId }, + layoutByWorktree: { [WORKTREE]: { type: 'leaf', groupId } }, + // Orphan: a runtime terminal with no unified row and no live PTY. + tabsByWorktree: { + [WORKTREE]: [ + { + id: 'orphan', + ptyId: null, + worktreeId: WORKTREE, + title: 'Terminal', + customTitle: null, + color: null, + sortOrder: 0, + createdAt: 1 + } + ] + }, + ptyIdsByTabId: { orphan: [] } + }) + const before = store.getState() + + const { patch } = projectWorktreeTabModelReconciliation(before, WORKTREE) + + expect(patch.tabsByWorktree?.[WORKTREE]).toEqual([]) + expect(patch.unifiedTabsByWorktree).toBe(before.unifiedTabsByWorktree) + expect(patch.groupsByWorktree).toBe(before.groupsByWorktree) + expect(patch.activeGroupIdByWorktree).toBe(before.activeGroupIdByWorktree) + expect(patch.layoutByWorktree).toBeUndefined() + }) +}) + +describe('writeBatchedWorkspaceRecordEntry identity', () => { + it('returns the same record when the entry already holds that value', () => { + const groups = [{ id: 'group-1' }] + const current = { [WORKTREE]: groups } + + const next = writeBatchedWorkspaceRecordEntry( + current, + 'groupsByWorktree', + WORKTREE, + groups, + undefined + ) + + // A new reference here rerenders every component selecting the map. + expect(next).toBe(current) + }) + + it('does not claim ownership of a map it never cloned', () => { + const groups = [{ id: 'group-1' }] + const current = { [WORKTREE]: groups } + const batch = createWorktreeTabModelReconciliationBatch({ openFiles: [] }) + + const unchanged = writeBatchedWorkspaceRecordEntry( + current, + 'groupsByWorktree', + WORKTREE, + groups, + batch + ) + expect(unchanged).toBe(current) + expect(batch.ownedStateKeys.has('groupsByWorktree')).toBe(false) + + // A later real change must therefore still copy rather than mutate the caller's map. + const changed = writeBatchedWorkspaceRecordEntry( + current, + 'groupsByWorktree', + WORKTREE, + [{ id: 'group-2' }], + batch + ) + expect(changed).not.toBe(current) + expect(current[WORKTREE]).toBe(groups) + expect(batch.ownedStateKeys.has('groupsByWorktree')).toBe(true) + }) + + it('copies when the value differs', () => { + const current = { [WORKTREE]: 'group-1' } + + const next = writeBatchedWorkspaceRecordEntry( + current, + 'activeGroupIdByWorktree', + WORKTREE, + 'group-2', + undefined + ) + + expect(next).not.toBe(current) + expect(next[WORKTREE]).toBe('group-2') + }) + + it('still stores an absent key, including an undefined value', () => { + const current: Record = { other: 'group-1' } + + const next = writeBatchedWorkspaceRecordEntry( + current, + 'activeGroupIdByWorktree', + WORKTREE, + undefined, + undefined + ) + + // The spread this replaces added the key; dropping it would change Object.keys. + expect(next).not.toBe(current) + expect(WORKTREE in next).toBe(true) + expect(next[WORKTREE]).toBeUndefined() + }) + + it('keeps mutating in place once the batch owns the map', () => { + const batch = createWorktreeTabModelReconciliationBatch({ openFiles: [] }) + batch.ownedStateKeys.add('groupsByWorktree') + const draft: Record = { [WORKTREE]: 'old' } + + const next = writeBatchedWorkspaceRecordEntry(draft, 'groupsByWorktree', WORKTREE, 'new', batch) + + expect(next).toBe(draft) + expect(draft[WORKTREE]).toBe('new') + }) +}) diff --git a/src/renderer/src/store/slices/tabs/tabs-reconciliation-batch.ts b/src/renderer/src/store/slices/tabs/tabs-reconciliation-batch.ts index 1eb11064203..0f6ed39dac3 100644 --- a/src/renderer/src/store/slices/tabs/tabs-reconciliation-batch.ts +++ b/src/renderer/src/store/slices/tabs/tabs-reconciliation-batch.ts @@ -49,6 +49,13 @@ export function writeBatchedWorkspaceRecordEntry( ;(current as Record)[worktreeId] = value return current } + // Why: the reconciliation gate writes every map when any one changed; spreading an + // already-equal entry would rerender its selectors for no data change. Nothing was + // cloned, so ownership is deliberately not claimed. Absent keys still get stored, + // matching the spread (`in` check). + if (worktreeId in current && Object.is(current[worktreeId], value)) { + return current + } const next = { ...current, [worktreeId]: value } as Record batch?.ownedStateKeys.add(stateKey) return next diff --git a/src/renderer/src/store/slices/tabs/tabs-reconciliation.ts b/src/renderer/src/store/slices/tabs/tabs-reconciliation.ts index 72d7be7195e..d5a005f537e 100644 --- a/src/renderer/src/store/slices/tabs/tabs-reconciliation.ts +++ b/src/renderer/src/store/slices/tabs/tabs-reconciliation.ts @@ -128,7 +128,11 @@ export function projectWorktreeTabModelReconciliation( return liveEditorIds.has(tab.entityId) } - const validTabs = reconciledUnifiedTabs.filter(isRenderableTab) + const renderableTabs = reconciledUnifiedTabs.filter(isRenderableTab) + // Why: hand the stored array back when nothing was filtered, so an unrelated + // change (orphans, layout) does not give `unifiedTabsByWorktree` a new identity. + const validTabs = + renderableTabs.length === reconciledUnifiedTabs.length ? reconciledUnifiedTabs : renderableTabs const validTabIds = new Set(validTabs.map((tab) => tab.id)) const nextGroupsWithEmpty = reconciledGroups.map((group) => { const tabOrder = group.tabOrder.filter((tabId) => validTabIds.has(tabId)) @@ -147,10 +151,14 @@ export function projectWorktreeTabModelReconciliation( ? group : { ...group, tabOrder, activeTabId, recentTabIds } }) - const nextGroups = + const prunedGroups = validTabs.length > 0 ? nextGroupsWithEmpty.filter((group) => group.tabOrder.length > 0) : nextGroupsWithEmpty + const groupsChanged = + prunedGroups.length !== groups.length || + prunedGroups.some((group, index) => group !== groups[index]) + const nextGroups = groupsChanged ? prunedGroups : groups const currentActiveGroupId = state.activeGroupIdByWorktree[worktreeId] ?? ensuredGroupState?.activeGroupIdByWorktree[worktreeId] @@ -160,10 +168,7 @@ export function projectWorktreeTabModelReconciliation( : (nextGroups.find((group) => group.activeTabId !== null)?.id ?? nextGroups[0]?.id ?? currentActiveGroupId) - const groupsChanged = - nextGroups.length !== groups.length || - nextGroups.some((group, index) => group !== groups[index]) - const tabsChanged = validTabs.length !== unifiedTabs.length || restoredLegacyTabs.length > 0 + const tabsChanged = validTabs !== unifiedTabs const activeGroupChanged = nextActiveGroupId !== currentActiveGroupId const baseNextLayout = restoredLegacyTabs.length > 0 && reconciliationGroup diff --git a/src/renderer/src/store/slices/worktrees/teardown/record-key-omission.test.ts b/src/renderer/src/store/slices/worktrees/teardown/record-key-omission.test.ts new file mode 100644 index 00000000000..7c7dcee026a --- /dev/null +++ b/src/renderer/src/store/slices/worktrees/teardown/record-key-omission.test.ts @@ -0,0 +1,26 @@ +import { describe, expect, it } from 'vitest' +import { omitRecordKeys } from './record-key-omission' + +describe('omitRecordKeys', () => { + it('returns the same record when none of the keys are present', () => { + const record = { a: 1 } + expect(omitRecordKeys(record, ['b', 'c'])).toBe(record) + expect(omitRecordKeys(record, new Set())).toBe(record) + }) + + it('copies once and drops every present key', () => { + const record = { a: 1, b: 2, c: 3 } + const next = omitRecordKeys(record, new Set(['a', 'c', 'missing'])) + expect(next).not.toBe(record) + expect(next).toEqual({ b: 2 }) + expect(record).toEqual({ a: 1, b: 2, c: 3 }) + }) + + it('drops a key whose value is undefined', () => { + expect(omitRecordKeys({ a: undefined }, ['a'])).toEqual({}) + }) + + it('normalizes a missing record to an empty one, as spread-then-delete did', () => { + expect(omitRecordKeys(undefined, ['a'])).toEqual({}) + }) +}) diff --git a/src/renderer/src/store/slices/worktrees/teardown/record-key-omission.ts b/src/renderer/src/store/slices/worktrees/teardown/record-key-omission.ts new file mode 100644 index 00000000000..5407df6bca2 --- /dev/null +++ b/src/renderer/src/store/slices/worktrees/teardown/record-key-omission.ts @@ -0,0 +1,30 @@ +/** + * Key removal that keeps a record's identity when it had none of the keys. + * + * Why identity matters here: teardown rewrites dozens of store maps at once, and + * a removed worktree has an entry in only a few of them. Copying the rest anyway + * gives every one a new reference, which rerenders every component selecting it + * for no data change. + * + * Why nullish input yields `{}`: some worktree-isolation callers hand over states + * with a slice omitted, and the spread-then-delete this replaces normalized those + * to an empty record. Production always initialises them, so the fresh object here + * costs nothing at runtime. + */ +export function omitRecordKeys( + record: Record | undefined, + keys: Iterable +): Record { + if (!record) { + return {} + } + let next: Record | null = null + for (const key of keys) { + if (!(key in record)) { + continue + } + next ??= { ...record } + delete next[key] + } + return next ?? record +} diff --git a/src/renderer/src/store/slices/worktrees/teardown/remove-worktree-map-identity.test.ts b/src/renderer/src/store/slices/worktrees/teardown/remove-worktree-map-identity.test.ts new file mode 100644 index 00000000000..74752b8a719 --- /dev/null +++ b/src/renderer/src/store/slices/worktrees/teardown/remove-worktree-map-identity.test.ts @@ -0,0 +1,85 @@ +import { describe, expect, it } from 'vitest' +import type { AppState } from '../../../types' +import { applyRemoveWorktreeSuccessState } from './remove-worktree-store-cleanup' + +const REMOVED_ID = 'repo-1::/repos/one/removed' +const SURVIVING_ID = 'repo-1::/repos/one/kept' + +/** Only the maps this test asserts on; the cleanup reads them defensively. */ +function buildState(): AppState { + return { + worktreesByRepo: { 'repo-1': [] }, + tabsByWorktree: { [REMOVED_ID]: [], [SURVIVING_ID]: [] }, + openFiles: [], + everActivatedWorktreeIds: new Set(), + lastVisitedAtByWorktreeId: {}, + deleteStateByWorktreeId: {}, + sortEpoch: 0, + // Worktree-keyed maps that hold nothing for the removed worktree. + gitStatusByWorktree: { [SURVIVING_ID]: 'clean' }, + gitStatusHugeByWorktree: {}, + showDotfilesByWorktree: { [SURVIVING_ID]: true }, + expandedDirs: {}, + fileSearchStateByWorktree: {}, + layoutByWorktree: { [SURVIVING_ID]: 'grid' }, + groupsByWorktree: {}, + unifiedTabsByWorktree: {}, + // Tab-keyed maps with no entry for the removed worktree's tabs. + terminalLayoutsByTabId: { 'other-tab': 'single' }, + ptyIdsByTabId: {}, + expandedPaneByTabId: {} + } as unknown as AppState +} + +function removeWorktree(state: AppState): AppState { + let current = state + applyRemoveWorktreeSuccessState( + (update) => { + const patch = typeof update === 'function' ? update(current) : update + current = { ...current, ...patch } + }, + REMOVED_ID, + new Set(['removed-tab']) + ) + return current +} + +describe('removeWorktree map identity', () => { + it('keeps the reference of every map that held nothing for the removed worktree', () => { + const before = buildState() + + const after = removeWorktree(before) + + // A new reference here rerenders every component selecting the map, for no data change. + for (const field of [ + 'gitStatusByWorktree', + 'gitStatusHugeByWorktree', + 'showDotfilesByWorktree', + 'expandedDirs', + 'fileSearchStateByWorktree', + 'layoutByWorktree', + 'groupsByWorktree', + 'unifiedTabsByWorktree', + 'terminalLayoutsByTabId', + 'ptyIdsByTabId', + 'expandedPaneByTabId' + ] as const) { + expect(after[field], field).toBe(before[field]) + } + }) + + it('still drops the removed worktree from the maps that did hold it', () => { + const before = buildState() + Object.assign(before, { + gitStatusByWorktree: { [REMOVED_ID]: 'dirty', [SURVIVING_ID]: 'clean' }, + terminalLayoutsByTabId: { 'removed-tab': 'single', 'other-tab': 'single' } + }) + + const after = removeWorktree(before) + + expect(after.gitStatusByWorktree).not.toBe(before.gitStatusByWorktree) + expect(after.gitStatusByWorktree).toEqual({ [SURVIVING_ID]: 'clean' }) + expect(after.terminalLayoutsByTabId).toEqual({ 'other-tab': 'single' }) + expect(after.tabsByWorktree).toEqual({ [SURVIVING_ID]: [] }) + }) +}) diff --git a/src/renderer/src/store/slices/worktrees/teardown/remove-worktree-store-cleanup.ts b/src/renderer/src/store/slices/worktrees/teardown/remove-worktree-store-cleanup.ts index bbd54e1c89c..415697b883e 100644 --- a/src/renderer/src/store/slices/worktrees/teardown/remove-worktree-store-cleanup.ts +++ b/src/renderer/src/store/slices/worktrees/teardown/remove-worktree-store-cleanup.ts @@ -4,6 +4,7 @@ import type { WorktreeSliceSet } from '../listing/worktree-slice-types' import { removeDeleteStatesForWorktreeIds } from './worktree-delete-state' import { removeWorktreeVisitEntries } from '@/lib/worktree-visit-recency' import { forgetAmbiguousOwnerWarnings } from '../listing/worktree-owner-settings' +import { omitRecordKeys } from './record-key-omission' export function applyRemoveWorktreeSuccessState( set: WorktreeSliceSet, @@ -15,99 +16,13 @@ export function applyRemoveWorktreeSuccessState( // re-arms the once-per-workspace warning if this id is ever added back. forgetAmbiguousOwnerWarnings([worktreeId]) set((s) => { - const next = { ...s.worktreesByRepo } - for (const repoId of Object.keys(next)) { - next[repoId] = next[repoId].filter((w) => w.id !== worktreeId) + const worktreeIds = [worktreeId] + const omitByWorktree = (m: Record | undefined) => omitRecordKeys(m, worktreeIds) + const omitByTabId = (m: Record | undefined) => omitRecordKeys(m, tabIds) + const nextWorktreesByRepo = { ...s.worktreesByRepo } + for (const repoId of Object.keys(nextWorktreesByRepo)) { + nextWorktreesByRepo[repoId] = nextWorktreesByRepo[repoId].filter((w) => w.id !== worktreeId) } - const nextTabs = { ...s.tabsByWorktree } - delete nextTabs[worktreeId] - const nextLayouts = { ...s.terminalLayoutsByTabId } - const nextPtyIdsByTabId = { ...s.ptyIdsByTabId } - const nextRuntimePaneTitlesByTabId = { ...s.runtimePaneTitlesByTabId } - const nextAutomaticAgentResumeClaimsByTabId = { - ...s.automaticAgentResumeClaimsByTabId - } - const nextNativeChatLaunchPromptByTabId = { ...s.nativeChatLaunchPromptByTabId } - const nextNativeChatLaunchDraftByTabId = { ...s.nativeChatLaunchDraftByTabId } - const nextUnverifiedPtyLossTabIds = { ...s.unverifiedPtyLossTabIds } - // Why: closeTab deletes these per-tab maps but removeWorktree missed them, leaking a split pane's expand flags. - const nextExpandedPaneByTabId = { ...s.expandedPaneByTabId } - const nextCanExpandPaneByTabId = { ...s.canExpandPaneByTabId } - for (const tabId of tabIds) { - delete nextLayouts[tabId] - delete nextPtyIdsByTabId[tabId] - delete nextRuntimePaneTitlesByTabId[tabId] - delete nextAutomaticAgentResumeClaimsByTabId[tabId] - delete nextNativeChatLaunchPromptByTabId[tabId] - delete nextNativeChatLaunchDraftByTabId[tabId] - delete nextUnverifiedPtyLossTabIds[tabId] - delete nextExpandedPaneByTabId[tabId] - delete nextCanExpandPaneByTabId[tabId] - } - const nextDeleteState = removeDeleteStatesForWorktreeIds( - s.deleteStateByWorktreeId, - new Set([worktreeId]) - ) - const nextLineage = { ...s.worktreeLineageById } - delete nextLineage[worktreeId] - const nextWorkspaceLineage = { ...s.workspaceLineageByChildKey } - delete nextWorkspaceLineage[worktreeWorkspaceKey(worktreeId)] - // Clean up editor files belonging to this worktree - const newOpenFiles = s.openFiles.filter((f) => f.worktreeId !== worktreeId) - const nextBrowserTabsByWorktree = { ...s.browserTabsByWorktree } - delete nextBrowserTabsByWorktree[worktreeId] - const nextActiveFileIdByWorktree = { ...s.activeFileIdByWorktree } - delete nextActiveFileIdByWorktree[worktreeId] - const nextActiveBrowserTabIdByWorktree = { ...s.activeBrowserTabIdByWorktree } - delete nextActiveBrowserTabIdByWorktree[worktreeId] - // Why: closeBrowserTab records a Cmd+Shift+T undo snapshot, but a deleted worktree's tabs can't be restored; purge it. - const nextRecentlyClosedBrowserTabsByWorktree = { - ...s.recentlyClosedBrowserTabsByWorktree - } - delete nextRecentlyClosedBrowserTabsByWorktree[worktreeId] - const nextActiveTabTypeByWorktree = { ...s.activeTabTypeByWorktree } - delete nextActiveTabTypeByWorktree[worktreeId] - const nextActiveTabIdByWorktree = { ...s.activeTabIdByWorktree } - delete nextActiveTabIdByWorktree[worktreeId] - const nextTabBarOrderByWorktree = { ...s.tabBarOrderByWorktree } - // Why: the tab strip persists visual order per worktree; drop the entry so stale tab IDs aren't retained. - delete nextTabBarOrderByWorktree[worktreeId] - const nextPendingReconnectTabByWorktree = { ...s.pendingReconnectTabByWorktree } - delete nextPendingReconnectTabByWorktree[worktreeId] - // Why: split-tab layout/group state is worktree-owned; leaving it makes a deleted worktree look restorable. - const nextUnifiedTabsByWorktree = { ...s.unifiedTabsByWorktree } - delete nextUnifiedTabsByWorktree[worktreeId] - const nextGroupsByWorktree = { ...s.groupsByWorktree } - delete nextGroupsByWorktree[worktreeId] - const nextLayoutByWorktree = { ...s.layoutByWorktree } - delete nextLayoutByWorktree[worktreeId] - const nextActiveGroupIdByWorktree = { ...s.activeGroupIdByWorktree } - delete nextActiveGroupIdByWorktree[worktreeId] - // Why: git status/compare caches stop refreshing once the worktree is deleted; remove them so no stale badges/diffs linger. - const nextGitStatusByWorktree = { ...s.gitStatusByWorktree } - delete nextGitStatusByWorktree[worktreeId] - const nextGitStatusHeadByWorktree = { ...s.gitStatusHeadByWorktree } - delete nextGitStatusHeadByWorktree[worktreeId] - const nextGitBranchLineTotalByWorktree = { ...s.gitBranchLineTotalByWorktree } - delete nextGitBranchLineTotalByWorktree[worktreeId] - const nextGitIgnoredPathsByWorktree = { ...s.gitIgnoredPathsByWorktree } - delete nextGitIgnoredPathsByWorktree[worktreeId] - const nextGitConflictOperationByWorktree = { ...s.gitConflictOperationByWorktree } - delete nextGitConflictOperationByWorktree[worktreeId] - const nextTrackedConflictPathsByWorktree = { ...s.trackedConflictPathsByWorktree } - delete nextTrackedConflictPathsByWorktree[worktreeId] - const nextGitBranchChangesByWorktree = { ...s.gitBranchChangesByWorktree } - delete nextGitBranchChangesByWorktree[worktreeId] - const nextGitBranchCompareSummaryByWorktree = { ...s.gitBranchCompareSummaryByWorktree } - delete nextGitBranchCompareSummaryByWorktree[worktreeId] - const nextGitBranchCompareRequestKeyByWorktree = { - ...s.gitBranchCompareRequestKeyByWorktree - } - delete nextGitBranchCompareRequestKeyByWorktree[worktreeId] - const nextGitBranchCompareRequestStatusHeadByWorktree = { - ...s.gitBranchCompareRequestStatusHeadByWorktree - } - delete nextGitBranchCompareRequestStatusHeadByWorktree[worktreeId] // Why: clean up per-file editor state for the removed worktree so stale drafts/view modes don't accumulate. const removedFileIds = new Set() for (const file of s.openFiles) { @@ -119,154 +34,109 @@ export function applyRemoveWorktreeSuccessState( removedFileIds.add(file.markdownPreviewSourceFileId) } } - const nextEditorDrafts = removedFileIds.size > 0 ? { ...s.editorDrafts } : s.editorDrafts - const nextMarkdownViewMode = - removedFileIds.size > 0 ? { ...s.markdownViewMode } : s.markdownViewMode - const nextMarkdownRichModeSizeOverride = - removedFileIds.size > 0 - ? { ...s.markdownRichModeSizeOverride } - : s.markdownRichModeSizeOverride - const nextEditorViewMode = removedFileIds.size > 0 ? { ...s.editorViewMode } : s.editorViewMode - const nextMarkdownFrontmatterVisible = - removedFileIds.size > 0 ? { ...s.markdownFrontmatterVisible } : s.markdownFrontmatterVisible - // Why: editorCursorLine is keyed by fileId; clear it with the other per-file state so it doesn't leak. - const nextEditorCursorLine = - removedFileIds.size > 0 ? { ...s.editorCursorLine } : s.editorCursorLine - if (removedFileIds.size > 0) { - for (const fileId of removedFileIds) { - delete nextEditorDrafts[fileId] - delete nextMarkdownViewMode[fileId] - delete nextMarkdownRichModeSizeOverride[fileId] - delete nextEditorViewMode[fileId] - delete nextMarkdownFrontmatterVisible[fileId] - delete nextEditorCursorLine[fileId] - } - } - const nextExpandedDirs = { ...s.expandedDirs } - delete nextExpandedDirs[worktreeId] - const nextShowDotfilesByWorktree = { ...s.showDotfilesByWorktree } - delete nextShowDotfilesByWorktree[worktreeId] - // Why: clear the huge-status marker so it doesn't linger after the worktree is gone. - const nextGitStatusHugeByWorktree = { ...s.gitStatusHugeByWorktree } - delete nextGitStatusHugeByWorktree[worktreeId] - const nextRightSidebarExplorerViewByWorktree = { - ...s.rightSidebarExplorerViewByWorktree - } - delete nextRightSidebarExplorerViewByWorktree[worktreeId] + const omitByFileId = (m: Record | undefined) => omitRecordKeys(m, removedFileIds) + // Why guarded: a removed worktree usually has no open file, and an unconditional + // filter would hand openFiles a new identity anyway — the sibling purge path + // already does this. + const nextOpenFiles = s.openFiles.some((f) => f.worktreeId === worktreeId) + ? s.openFiles.filter((f) => f.worktreeId !== worktreeId) + : s.openFiles // If the active file belonged to the removed worktree, clear it const activeFileCleared = s.activeFileId ? s.openFiles.some((f) => f.id === s.activeFileId && f.worktreeId === worktreeId) : false const removedActiveWorktree = s.activeWorktreeId === worktreeId - const nextEverActivatedWorktreeIds = s.everActivatedWorktreeIds.has(worktreeId) - ? new Set([...s.everActivatedWorktreeIds].filter((id) => id !== worktreeId)) - : s.everActivatedWorktreeIds - const nextLastVisitedAtByWorktreeId = removeWorktreeVisitEntries( - s.lastVisitedAtByWorktreeId, - new Set([worktreeId]), - executionHostId - ) return { - worktreesByRepo: next, - worktreeLineageById: nextLineage, - workspaceLineageByChildKey: nextWorkspaceLineage, - tabsByWorktree: nextTabs, - ptyIdsByTabId: nextPtyIdsByTabId, - runtimePaneTitlesByTabId: nextRuntimePaneTitlesByTabId, - automaticAgentResumeClaimsByTabId: nextAutomaticAgentResumeClaimsByTabId, - nativeChatLaunchPromptByTabId: nextNativeChatLaunchPromptByTabId, - nativeChatLaunchDraftByTabId: nextNativeChatLaunchDraftByTabId, - unverifiedPtyLossTabIds: nextUnverifiedPtyLossTabIds, - terminalLayoutsByTabId: nextLayouts, - expandedPaneByTabId: nextExpandedPaneByTabId, - canExpandPaneByTabId: nextCanExpandPaneByTabId, - deleteStateByWorktreeId: nextDeleteState, - baseStatusByWorktreeId: (() => { - const nextStatus = { ...s.baseStatusByWorktreeId } - delete nextStatus[worktreeId] - return nextStatus - })(), - remoteBranchConflictByWorktreeId: (() => { - const nextConflict = { ...s.remoteBranchConflictByWorktreeId } - delete nextConflict[worktreeId] - return nextConflict - })(), - fileSearchStateByWorktree: (() => { - const nextSearch = { ...s.fileSearchStateByWorktree } - // Why: file search state is worktree-scoped; clear it so another worktree can't inherit stale matches. - delete nextSearch[worktreeId] - return nextSearch - })(), + worktreesByRepo: nextWorktreesByRepo, + worktreeLineageById: omitByWorktree(s.worktreeLineageById), + workspaceLineageByChildKey: omitRecordKeys(s.workspaceLineageByChildKey, [ + worktreeWorkspaceKey(worktreeId) + ]), + tabsByWorktree: omitByWorktree(s.tabsByWorktree), + ptyIdsByTabId: omitByTabId(s.ptyIdsByTabId), + runtimePaneTitlesByTabId: omitByTabId(s.runtimePaneTitlesByTabId), + automaticAgentResumeClaimsByTabId: omitByTabId(s.automaticAgentResumeClaimsByTabId), + nativeChatLaunchPromptByTabId: omitByTabId(s.nativeChatLaunchPromptByTabId), + nativeChatLaunchDraftByTabId: omitByTabId(s.nativeChatLaunchDraftByTabId), + unverifiedPtyLossTabIds: omitByTabId(s.unverifiedPtyLossTabIds), + terminalLayoutsByTabId: omitByTabId(s.terminalLayoutsByTabId), + // Why: closeTab deletes these per-tab maps but removeWorktree missed them, leaking a split pane's expand flags. + expandedPaneByTabId: omitByTabId(s.expandedPaneByTabId), + canExpandPaneByTabId: omitByTabId(s.canExpandPaneByTabId), + deleteStateByWorktreeId: removeDeleteStatesForWorktreeIds( + s.deleteStateByWorktreeId, + new Set(worktreeIds) + ), + baseStatusByWorktreeId: omitByWorktree(s.baseStatusByWorktreeId), + remoteBranchConflictByWorktreeId: omitByWorktree(s.remoteBranchConflictByWorktreeId), + // Why: file search state is worktree-scoped; clear it so another worktree can't inherit stale matches. + fileSearchStateByWorktree: omitByWorktree(s.fileSearchStateByWorktree), // Why: these worktree-keyed maps are re-keyed on rename but were missed by removal, leaking one entry each. - remoteStatusesByWorktree: (() => { - const next = { ...s.remoteStatusesByWorktree } - delete next[worktreeId] - return next - })(), - recentlyClosedEditorTabsByWorktree: (() => { - const next = { ...s.recentlyClosedEditorTabsByWorktree } - delete next[worktreeId] - return next - })(), - recentlyClosedTerminalTabsByWorktree: (() => { - const next = { ...s.recentlyClosedTerminalTabsByWorktree } - delete next[worktreeId] - return next - })(), + remoteStatusesByWorktree: omitByWorktree(s.remoteStatusesByWorktree), + recentlyClosedEditorTabsByWorktree: omitByWorktree(s.recentlyClosedEditorTabsByWorktree), + recentlyClosedTerminalTabsByWorktree: omitByWorktree(s.recentlyClosedTerminalTabsByWorktree), // Why: a deleted worktree's tabs can never be reopened; purge the kind list with the snapshot stacks above. - recentlyClosedTabKindsByWorktree: (() => { - const next = { ...s.recentlyClosedTabKindsByWorktree } - delete next[worktreeId] - return next - })(), - defaultTerminalTabsAppliedByWorktreeId: (() => { - const next = { ...s.defaultTerminalTabsAppliedByWorktreeId } - delete next[worktreeId] - return next - })(), + recentlyClosedTabKindsByWorktree: omitByWorktree(s.recentlyClosedTabKindsByWorktree), + defaultTerminalTabsAppliedByWorktreeId: omitByWorktree( + s.defaultTerminalTabsAppliedByWorktreeId + ), activeWorktreeId: removedActiveWorktree ? null : s.activeWorktreeId, activeWorkspaceExecutionHostId: removedActiveWorktree ? null : s.activeWorkspaceExecutionHostId, activeTabId: s.activeTabId && tabIds.has(s.activeTabId) ? null : s.activeTabId, - openFiles: newOpenFiles, - browserTabsByWorktree: nextBrowserTabsByWorktree, - recentlyClosedBrowserTabsByWorktree: nextRecentlyClosedBrowserTabsByWorktree, - activeFileIdByWorktree: nextActiveFileIdByWorktree, - activeBrowserTabIdByWorktree: nextActiveBrowserTabIdByWorktree, - activeTabTypeByWorktree: nextActiveTabTypeByWorktree, - rightSidebarExplorerViewByWorktree: nextRightSidebarExplorerViewByWorktree, - activeTabIdByWorktree: nextActiveTabIdByWorktree, - tabBarOrderByWorktree: nextTabBarOrderByWorktree, - pendingReconnectTabByWorktree: nextPendingReconnectTabByWorktree, - unifiedTabsByWorktree: nextUnifiedTabsByWorktree, - groupsByWorktree: nextGroupsByWorktree, - layoutByWorktree: nextLayoutByWorktree, - activeGroupIdByWorktree: nextActiveGroupIdByWorktree, - editorDrafts: nextEditorDrafts, - markdownViewMode: nextMarkdownViewMode, - markdownRichModeSizeOverride: nextMarkdownRichModeSizeOverride, - editorViewMode: nextEditorViewMode, - markdownFrontmatterVisible: nextMarkdownFrontmatterVisible, - editorCursorLine: nextEditorCursorLine, - showDotfilesByWorktree: nextShowDotfilesByWorktree, - expandedDirs: nextExpandedDirs, - gitStatusHugeByWorktree: nextGitStatusHugeByWorktree, - gitStatusByWorktree: nextGitStatusByWorktree, - gitStatusHeadByWorktree: nextGitStatusHeadByWorktree, - gitBranchLineTotalByWorktree: nextGitBranchLineTotalByWorktree, - gitIgnoredPathsByWorktree: nextGitIgnoredPathsByWorktree, - gitConflictOperationByWorktree: nextGitConflictOperationByWorktree, - trackedConflictPathsByWorktree: nextTrackedConflictPathsByWorktree, - gitBranchChangesByWorktree: nextGitBranchChangesByWorktree, - gitBranchCompareSummaryByWorktree: nextGitBranchCompareSummaryByWorktree, - gitBranchCompareRequestKeyByWorktree: nextGitBranchCompareRequestKeyByWorktree, - gitBranchCompareRequestStatusHeadByWorktree: nextGitBranchCompareRequestStatusHeadByWorktree, + openFiles: nextOpenFiles, + browserTabsByWorktree: omitByWorktree(s.browserTabsByWorktree), + // Why: closeBrowserTab records a Cmd+Shift+T undo snapshot, but a deleted worktree's tabs can't be restored; purge it. + recentlyClosedBrowserTabsByWorktree: omitByWorktree(s.recentlyClosedBrowserTabsByWorktree), + activeFileIdByWorktree: omitByWorktree(s.activeFileIdByWorktree), + activeBrowserTabIdByWorktree: omitByWorktree(s.activeBrowserTabIdByWorktree), + activeTabTypeByWorktree: omitByWorktree(s.activeTabTypeByWorktree), + rightSidebarExplorerViewByWorktree: omitByWorktree(s.rightSidebarExplorerViewByWorktree), + activeTabIdByWorktree: omitByWorktree(s.activeTabIdByWorktree), + // Why: the tab strip persists visual order per worktree; drop the entry so stale tab IDs aren't retained. + tabBarOrderByWorktree: omitByWorktree(s.tabBarOrderByWorktree), + pendingReconnectTabByWorktree: omitByWorktree(s.pendingReconnectTabByWorktree), + // Why: split-tab layout/group state is worktree-owned; leaving it makes a deleted worktree look restorable. + unifiedTabsByWorktree: omitByWorktree(s.unifiedTabsByWorktree), + groupsByWorktree: omitByWorktree(s.groupsByWorktree), + layoutByWorktree: omitByWorktree(s.layoutByWorktree), + activeGroupIdByWorktree: omitByWorktree(s.activeGroupIdByWorktree), + editorDrafts: omitByFileId(s.editorDrafts), + markdownViewMode: omitByFileId(s.markdownViewMode), + markdownRichModeSizeOverride: omitByFileId(s.markdownRichModeSizeOverride), + editorViewMode: omitByFileId(s.editorViewMode), + markdownFrontmatterVisible: omitByFileId(s.markdownFrontmatterVisible), + // Why: editorCursorLine is keyed by fileId; clear it with the other per-file state so it doesn't leak. + editorCursorLine: omitByFileId(s.editorCursorLine), + showDotfilesByWorktree: omitByWorktree(s.showDotfilesByWorktree), + expandedDirs: omitByWorktree(s.expandedDirs), + // Why: clear the huge-status marker so it doesn't linger after the worktree is gone. + gitStatusHugeByWorktree: omitByWorktree(s.gitStatusHugeByWorktree), + // Why: git status/compare caches stop refreshing once the worktree is deleted; remove them so no stale badges/diffs linger. + gitStatusByWorktree: omitByWorktree(s.gitStatusByWorktree), + gitStatusHeadByWorktree: omitByWorktree(s.gitStatusHeadByWorktree), + gitBranchLineTotalByWorktree: omitByWorktree(s.gitBranchLineTotalByWorktree), + gitIgnoredPathsByWorktree: omitByWorktree(s.gitIgnoredPathsByWorktree), + gitConflictOperationByWorktree: omitByWorktree(s.gitConflictOperationByWorktree), + trackedConflictPathsByWorktree: omitByWorktree(s.trackedConflictPathsByWorktree), + gitBranchChangesByWorktree: omitByWorktree(s.gitBranchChangesByWorktree), + gitBranchCompareSummaryByWorktree: omitByWorktree(s.gitBranchCompareSummaryByWorktree), + gitBranchCompareRequestKeyByWorktree: omitByWorktree(s.gitBranchCompareRequestKeyByWorktree), + gitBranchCompareRequestStatusHeadByWorktree: omitByWorktree( + s.gitBranchCompareRequestStatusHeadByWorktree + ), activeFileId: activeFileCleared ? null : s.activeFileId, activeBrowserTabId: removedActiveWorktree ? null : s.activeBrowserTabId, activeTabType: removedActiveWorktree || activeFileCleared ? 'terminal' : s.activeTabType, - everActivatedWorktreeIds: nextEverActivatedWorktreeIds, - lastVisitedAtByWorktreeId: nextLastVisitedAtByWorktreeId, + everActivatedWorktreeIds: s.everActivatedWorktreeIds.has(worktreeId) + ? new Set([...s.everActivatedWorktreeIds].filter((id) => id !== worktreeId)) + : s.everActivatedWorktreeIds, + lastVisitedAtByWorktreeId: removeWorktreeVisitEntries( + s.lastVisitedAtByWorktreeId, + new Set(worktreeIds), + executionHostId + ), sortEpoch: s.sortEpoch + 1 } }) diff --git a/src/renderer/src/store/slices/worktrees/teardown/worktree-purge-omitters.ts b/src/renderer/src/store/slices/worktrees/teardown/worktree-purge-omitters.ts index 52a76539499..d8e3b7b9e3b 100644 --- a/src/renderer/src/store/slices/worktrees/teardown/worktree-purge-omitters.ts +++ b/src/renderer/src/store/slices/worktrees/teardown/worktree-purge-omitters.ts @@ -3,6 +3,7 @@ import type { WorkspaceLineage } from '../../../../../../shared/worktree/lineage import { isWorkspaceKey, worktreeWorkspaceKey } from '../../../../../../shared/workspace-scope' import { normalizeRightSidebarRoute } from '../../../right-sidebar-route' import type { WorktreePurgeDoomedIds } from './worktree-purge-doomed-ids' +import { omitRecordKeys } from './record-key-omission' export function createWorktreePurgeOmitters( s: AppState, @@ -11,31 +12,15 @@ export function createWorktreePurgeOmitters( ) { const { doomedTabIds, doomedPtyIds, doomedBrowserWorkspaceIds, doomedPageIds, removedFileIds } = doomed - const omitByWorktree = (obj: Record): Record => { - let changed = false - const out = { ...obj } - for (const id of worktreeIdSet) { - if (id in out) { - delete out[id] - changed = true - } - } - return changed ? out : obj - } + const omitByWorktree = (obj: Record): Record => + omitRecordKeys(obj, worktreeIdSet) const omitWorkspaceLineageByWorktree = ( obj: Record - ): Record => { - let changed = false - const out = { ...obj } - for (const id of worktreeIdSet) { - const childKey = isWorkspaceKey(id) ? id : worktreeWorkspaceKey(id) - if (childKey in out) { - delete out[childKey] - changed = true - } - } - return changed ? out : obj - } + ): Record => + omitRecordKeys( + obj, + [...worktreeIdSet].map((id) => (isWorkspaceKey(id) ? id : worktreeWorkspaceKey(id))) + ) const pruneRightSidebarTabByWorktree = (): AppState['rightSidebarTabByWorktree'] => { const omitted = omitByWorktree(s.rightSidebarTabByWorktree) let changed = omitted !== s.rightSidebarTabByWorktree @@ -50,94 +35,40 @@ export function createWorktreePurgeOmitters( } return changed ? out : omitted } - const omitByTabId = (obj: Record): Record => { - let changed = false - const out = { ...obj } - for (const tabId of doomedTabIds) { - if (tabId in out) { - delete out[tabId] - changed = true - } - } - return changed ? out : obj - } + const omitByTabId = (obj: Record): Record => + omitRecordKeys(obj, doomedTabIds) const survivingTabIds = new Set( Object.entries(s.tabsByWorktree) .filter(([worktreeId]) => !worktreeIdSet.has(worktreeId)) .flatMap(([, tabs]) => tabs.map((tab) => tab.id)) ) - const omitRetiredDirectSshLedgerByTabId = (obj: Record): Record => { - let changed = false - const out = { ...obj } - for (const tabId of doomedTabIds) { - if (!survivingTabIds.has(tabId) && tabId in out) { - delete out[tabId] - changed = true - } - } - return changed ? out : obj - } - const omitByPtyId = (obj: Record): Record => { - let changed = false - const out = { ...obj } - for (const ptyId of doomedPtyIds) { - if (ptyId in out) { - delete out[ptyId] - changed = true - } - } - return changed ? out : obj - } + const omitRetiredDirectSshLedgerByTabId = (obj: Record): Record => + omitRecordKeys( + obj, + [...doomedTabIds].filter((tabId) => !survivingTabIds.has(tabId)) + ) + const omitByPtyId = (obj: Record): Record => + omitRecordKeys(obj, doomedPtyIds) // Pane-scoped maps are keyed `${tabId}:${leafId}`; tabId never contains ":", so the prefix before the first ":" is the owning tab. const omitByPaneKeyTabPrefix = (obj: Record): Record => { // Null-tolerant like omitByTabId: some worktree-isolation callers omit these slices (production store always inits to {}). if (!obj) { return obj } - let changed = false - const out = { ...obj } - for (const paneKey of Object.keys(obj)) { - const sep = paneKey.indexOf(':') - if (sep > 0 && doomedTabIds.has(paneKey.slice(0, sep))) { - delete out[paneKey] - changed = true - } - } - return changed ? out : obj - } - const omitByBrowserWorkspaceId = (obj: Record): Record => { - let changed = false - const out = { ...obj } - for (const workspaceId of doomedBrowserWorkspaceIds) { - if (workspaceId in out) { - delete out[workspaceId] - changed = true - } - } - return changed ? out : obj - } - const omitByPageId = (obj: Record): Record => { - let changed = false - const out = { ...obj } - for (const pageId of doomedPageIds) { - if (pageId in out) { - delete out[pageId] - changed = true - } - } - return changed ? out : obj - } - const omitByFileId = (obj: Record): Record => { - let changed = false - const out = { ...obj } - for (const fileId of removedFileIds) { - if (fileId in out) { - delete out[fileId] - changed = true - } - } - return changed ? out : obj + return omitRecordKeys( + obj, + Object.keys(obj).filter((paneKey) => { + const sep = paneKey.indexOf(':') + return sep > 0 && doomedTabIds.has(paneKey.slice(0, sep)) + }) + ) } + const omitByBrowserWorkspaceId = (obj: Record): Record => + omitRecordKeys(obj, doomedBrowserWorkspaceIds) + const omitByPageId = (obj: Record): Record => + omitRecordKeys(obj, doomedPageIds) + const omitByFileId = (obj: Record): Record => + omitRecordKeys(obj, removedFileIds) return { omitByWorktree, diff --git a/src/renderer/src/store/slices/worktrees/teardown/worktree-teardown-array-identity.test.ts b/src/renderer/src/store/slices/worktrees/teardown/worktree-teardown-array-identity.test.ts new file mode 100644 index 00000000000..5e122d9dbef --- /dev/null +++ b/src/renderer/src/store/slices/worktrees/teardown/worktree-teardown-array-identity.test.ts @@ -0,0 +1,58 @@ +import { describe, expect, it } from 'vitest' +import type { AppState } from '../../../types' +import { applyRemoveWorktreeSuccessState } from './remove-worktree-store-cleanup' + +type OpenFile = AppState['openFiles'][number] + +const REMOVED = 'repo-1::/repos/one/removed' +const KEPT = 'repo-1::/repos/one/kept' + +function fileFor(worktreeId: string, id: string): OpenFile { + return { id, worktreeId, path: `${worktreeId}/f.ts`, name: 'f.ts' } as unknown as OpenFile +} + +function buildState(openFiles: OpenFile[]): AppState { + return { + worktreesByRepo: { 'repo-1': [] }, + tabsByWorktree: { [KEPT]: [] }, + openFiles, + everActivatedWorktreeIds: new Set(), + lastVisitedAtByWorktreeId: {}, + deleteStateByWorktreeId: {}, + sortEpoch: 0 + } as unknown as AppState +} + +function removeWorktree(state: AppState): AppState { + let current = state + applyRemoveWorktreeSuccessState( + (update) => { + const patch = typeof update === 'function' ? update(current) : update + current = { ...current, ...patch } + }, + REMOVED, + new Set() + ) + return current +} + +describe('worktree removal openFiles identity', () => { + it('keeps the openFiles reference when the removed worktree had no open file', () => { + // openFiles is selected whole by the editor panel, file explorer and git-status + // polling, so a fresh array here rerenders all of them for no data change. + const before = buildState([fileFor(KEPT, 'kept-file')]) + + const after = removeWorktree(before) + + expect(after.openFiles).toBe(before.openFiles) + }) + + it('still drops the removed worktree files', () => { + const before = buildState([fileFor(KEPT, 'kept-file'), fileFor(REMOVED, 'gone-file')]) + + const after = removeWorktree(before) + + expect(after.openFiles).not.toBe(before.openFiles) + expect(after.openFiles.map((f) => f.id)).toEqual(['kept-file']) + }) +}) diff --git a/src/renderer/src/store/store-identity-churn-probe.test.ts b/src/renderer/src/store/store-identity-churn-probe.test.ts new file mode 100644 index 00000000000..04ae08e12d3 --- /dev/null +++ b/src/renderer/src/store/store-identity-churn-probe.test.ts @@ -0,0 +1,214 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' +import { create, type StoreApi } from 'zustand' +import { withReactCommitCascadeWriteProbe } from './react-commit-cascade-write-probe' +import { + armStoreIdentityChurnProbe, + disarmStoreIdentityChurnProbe, + readStoreIdentityChurnReport, + withStoreIdentityChurnProbe +} from './store-identity-churn-probe' + +type ProbeState = { + rows: { id: string; label: string }[] + entries: Record + counter: number + refresh: (rows: { id: string; label: string }[]) => void + touch: (id: string, status: string) => void + bump: () => void +} + +function createProbeStore() { + return create()( + withStoreIdentityChurnProbe((set) => ({ + rows: [{ id: 'a', label: 'A' }], + entries: { a: { status: 'idle' } }, + counter: 0, + refresh: (rows) => set({ rows }), + touch: (id, status) => set((state) => ({ entries: { ...state.entries, [id]: { status } } })), + bump: () => set((state) => ({ counter: state.counter + 1 })) + })) + ) +} + +function churnFor(field: string): number { + return readStoreIdentityChurnReport().find((row) => row.field === field)?.churnedWrites ?? 0 +} + +describe('store identity churn probe', () => { + beforeEach(() => { + // Arming resets the counters; disarming immediately leaves a clean, off probe. + armStoreIdentityChurnProbe() + disarmStoreIdentityChurnProbe() + }) + + it('flags a refresh that rebuilds an array with unchanged contents', () => { + const store = createProbeStore() + armStoreIdentityChurnProbe() + + store.getState().refresh([{ id: 'a', label: 'A' }]) + + expect(churnFor('rows')).toBe(1) + expect(readStoreIdentityChurnReport()[0]).toMatchObject({ + field: 'rows', + churnedWrites: 1, + replacedWrites: 1, + sites: [] + }) + }) + + it('flags a keyed update that rewrites an entry with the same value', () => { + const store = createProbeStore() + armStoreIdentityChurnProbe() + + store.getState().touch('a', 'idle') + + expect(churnFor('entries')).toBe(1) + }) + + it('does not flag writes that change the value', () => { + const store = createProbeStore() + armStoreIdentityChurnProbe() + + store.getState().refresh([{ id: 'a', label: 'B' }]) + store.getState().touch('a', 'running') + store.getState().bump() + + expect(readStoreIdentityChurnReport()).toEqual([]) + }) + + it('does not flag a refresh that returns the original reference', () => { + const store = createProbeStore() + const original = store.getState().rows + armStoreIdentityChurnProbe() + + store.getState().refresh(original) + + expect(readStoreIdentityChurnReport()).toEqual([]) + }) + + it('names the write site when capture is requested', () => { + const store = createProbeStore() + armStoreIdentityChurnProbe({ captureSites: true }) + + store.getState().refresh([{ id: 'a', label: 'A' }]) + + const [row] = readStoreIdentityChurnReport() + expect(row.sites).toHaveLength(1) + expect(row.sites[0]).toMatchObject({ churnedWrites: 1 }) + expect(row.sites[0].site).toContain('store-identity-churn-probe.test') + }) + + it('leaves a functional updater to zustand: called once, with the live state', () => { + const store = createProbeStore() + const seen: unknown[] = [] + armStoreIdentityChurnProbe() + + store.setState((state) => { + seen.push(state) + return { counter: state.counter + 1 } + }) + store.setState((state) => { + seen.push(state) + return { rows: [{ ...state.rows[0] }] } + }) + + expect(seen).toHaveLength(2) + expect(seen[1]).toMatchObject({ counter: 1 }) + expect(store.getState().counter).toBe(1) + expect(churnFor('rows')).toBe(1) + }) + + it('ignores a write zustand itself drops as identical', () => { + const store = createProbeStore() + armStoreIdentityChurnProbe() + + store.setState((state) => state) + + expect(readStoreIdentityChurnReport()).toEqual([]) + }) + + it('passes disarmed writes straight through without reading state', () => { + const innerSet = vi.fn() + const innerGet = vi.fn(() => ({ counter: 0 })) + const api = { setState: innerSet, getState: innerGet } as unknown as StoreApi<{ + counter: number + }> + const creator = withStoreIdentityChurnProbe<{ counter: number }>(() => ({ counter: 0 })) + creator(innerSet, innerGet, api) + const updater = (state: { counter: number }) => ({ counter: state.counter + 1 }) + + api.setState(updater, true) + + // The disarmed path forwards the exact arguments and never calls get(). + expect(innerSet).toHaveBeenCalledTimes(1) + expect(innerSet.mock.calls[0]).toEqual([updater, true]) + expect(innerGet).not.toHaveBeenCalled() + }) + + it('composes with the cascade probe without dropping or doubling a write', () => { + // Mirrors store/index.ts: churn probe outermost, cascade probe inside it. + const store = create()( + withStoreIdentityChurnProbe( + withReactCommitCascadeWriteProbe((set) => ({ + rows: [{ id: 'a', label: 'A' }], + entries: { a: { status: 'idle' } }, + counter: 0, + refresh: (rows) => set({ rows }), + touch: (id, status) => + set((state) => ({ entries: { ...state.entries, [id]: { status } } })), + bump: () => set((state) => ({ counter: state.counter + 1 })) + })) + ) + ) + let updaterCalls = 0 + armStoreIdentityChurnProbe({ captureSites: true }) + + store.getState().bump() + store.setState((state) => { + updaterCalls += 1 + return { counter: state.counter + 10 } + }) + store.getState().refresh([{ id: 'a', label: 'A' }]) + store.setState({ ...store.getState(), counter: 100 }, true) + + expect(updaterCalls).toBe(1) + expect(store.getState().counter).toBe(100) + expect(store.getState().rows).toEqual([{ id: 'a', label: 'A' }]) + // The named site is this test, not the sibling probe's wrapper frame. + const [row] = readStoreIdentityChurnReport() + expect(row).toMatchObject({ field: 'rows', churnedWrites: 1 }) + expect(row.sites[0].site).toContain('store-identity-churn-probe.test') + }) + + it('still sees churn on a replace write', () => { + const store = createProbeStore() + const rows = store.getState().rows + armStoreIdentityChurnProbe() + + store.setState({ ...store.getState(), rows: [{ ...rows[0] }] }, true) + + expect(churnFor('rows')).toBe(1) + }) + + it('records nothing while disarmed', () => { + const store = createProbeStore() + + store.getState().refresh([{ id: 'a', label: 'A' }]) + + expect(readStoreIdentityChurnReport()).toEqual([]) + }) + + it('treats distinct class instances as changed rather than equal', () => { + const store = create<{ value: unknown; put: (value: unknown) => void }>()( + withStoreIdentityChurnProbe((set) => ({ + value: new Map([['a', 1]]), + put: (value) => set({ value }) + })) + ) + armStoreIdentityChurnProbe() + + store.getState().put(new Map([['a', 1]])) + + expect(readStoreIdentityChurnReport()).toEqual([]) + }) +}) diff --git a/src/renderer/src/store/store-identity-churn-probe.ts b/src/renderer/src/store/store-identity-churn-probe.ts new file mode 100644 index 00000000000..32b289c13b9 --- /dev/null +++ b/src/renderer/src/store/store-identity-churn-probe.ts @@ -0,0 +1,206 @@ +/** + * Counts store writes that hand out a NEW reference for a field whose value did + * not change — the write-side half of Zustand rerender churn. + * + * Why a runtime probe and not a lint rule: the read side is statically decidable + * (app-store-performance flags selectors that allocate), but whether a `set()` + * reallocated for nothing depends on the payload, so only an executed write can + * answer it. A field that churns re-renders every component selecting it, with + * no data change to show for it. + * + * Cost when disarmed: one boolean field load per write, matching + * react-commit-cascade-write-probe. Comparison work only happens while armed. + * Nothing in the app arms it, so store/index.ts installs it only in dev and + * store-exposing builds; a shipped build never runs the wrapper at all. + * + * The wrapper never resolves a functional updater itself: zustand keeps sole + * ownership of when and with what argument an updater runs, so the probe cannot + * double-invoke it or hand it a stale state. + */ +import type { StateCreator } from 'zustand' + +export const storeIdentityChurnProbe = { armed: false, captureSites: false } + +export type StoreIdentityChurnRow = { + field: string + /** Writes that replaced the reference while the value stayed equal. */ + churnedWrites: number + /** Writes that replaced the reference at all. */ + replacedWrites: number + /** Write sites that churned, worst first; empty unless capture was requested. */ + sites: { site: string; churnedWrites: number }[] +} + +// Why bounded: an unbounded deep compare over a fully populated store would +// dominate the measurement it is trying to take. +const NODE_BUDGET = 20_000 +const MAX_DEPTH = 12 + +type CompareBudget = { nodesLeft: number } + +// Why plain-only: Map/Set/Date/class instances expose no own enumerable keys, so a +// key-wise compare would call two different instances equal. +function isPlainRecord(value: unknown): value is Record { + if (typeof value !== 'object' || value === null || Array.isArray(value)) { + return false + } + const prototype = Object.getPrototypeOf(value) + return prototype === Object.prototype || prototype === null +} + +/** Value equality with a node budget; an exhausted budget reports "changed". */ +function valuesEqual(left: unknown, right: unknown, depth: number, budget: CompareBudget): boolean { + if (Object.is(left, right)) { + return true + } + budget.nodesLeft -= 1 + if (budget.nodesLeft <= 0 || depth > MAX_DEPTH) { + return false + } + if (Array.isArray(left) || Array.isArray(right)) { + if (!Array.isArray(left) || !Array.isArray(right) || left.length !== right.length) { + return false + } + return left.every((entry, index) => valuesEqual(entry, right[index], depth + 1, budget)) + } + if (!isPlainRecord(left) || !isPlainRecord(right)) { + return false + } + const leftKeys = Object.keys(left) + if (leftKeys.length !== Object.keys(right).length) { + return false + } + return leftKeys.every( + (key) => Object.hasOwn(right, key) && valuesEqual(left[key], right[key], depth + 1, budget) + ) +} + +const churnedWritesByField = new Map() +const replacedWritesByField = new Map() +const churnedWritesByFieldSite = new Map>() + +function increment(counts: Map, field: string): void { + counts.set(field, (counts.get(field) ?? 0) + 1) +} + +export function armStoreIdentityChurnProbe(options?: { captureSites?: boolean }): void { + churnedWritesByField.clear() + replacedWritesByField.clear() + churnedWritesByFieldSite.clear() + storeIdentityChurnProbe.captureSites = options?.captureSites === true + storeIdentityChurnProbe.armed = true +} + +// Why the first non-probe, non-zustand frame: the caller that built the partial is +// the code to fix; the frames above it are the shared write plumbing. Every store +// write middleware is named *-probe.ts, so a sibling wrapper's frame is skipped too. +const SOURCE_FRAME = /:\d+:\d+\)?$/ +const PROBE_FRAME = /-probe\.[cm]?[jt]s\b/ + +function callingSite(): string { + const stack = new Error('store identity churn site').stack?.split('\n') ?? [] + for (const line of stack.slice(2)) { + const frame = line.trim() + if (SOURCE_FRAME.test(frame) && !PROBE_FRAME.test(frame) && !frame.includes('node_modules')) { + return frame + } + } + return 'unknown' +} + +export function disarmStoreIdentityChurnProbe(): void { + storeIdentityChurnProbe.armed = false +} + +/** Fields that churned at least once, worst first. */ +export function readStoreIdentityChurnReport(): StoreIdentityChurnRow[] { + return [...churnedWritesByField.entries()] + .map(([field, churnedWrites]) => ({ + field, + churnedWrites, + replacedWrites: replacedWritesByField.get(field) ?? 0, + sites: [...(churnedWritesByFieldSite.get(field)?.entries() ?? [])] + .map(([site, count]) => ({ site, churnedWrites: count })) + .sort((left, right) => right.churnedWrites - left.churnedWrites) + })) + .sort((left, right) => right.churnedWrites - left.churnedWrites) +} + +function recordSite(field: string): void { + const site = callingSite() + let sites = churnedWritesByFieldSite.get(field) + if (!sites) { + sites = new Map() + churnedWritesByFieldSite.set(field, sites) + } + sites.set(site, (sites.get(site) ?? 0) + 1) +} + +/** + * `fields` is the write's own keys when they are knowable: `set(partial)` merges, + * so no field outside the partial can have changed. A functional updater or a + * replace write falls back to every field; the extra cost there is one Object.is + * per untouched field, since the deep compare only runs on replaced references. + */ +function recordWrite( + previous: Record, + next: Record, + fields: readonly string[] +): void { + const budget: CompareBudget = { nodesLeft: NODE_BUDGET } + for (const field of fields) { + const before = previous[field] + const after = next[field] + if (Object.is(before, after)) { + continue + } + increment(replacedWritesByField, field) + // Primitives cannot churn: a different primitive is a real change. + if (typeof after !== 'object' || after === null) { + continue + } + if (valuesEqual(before, after, 0, budget)) { + increment(churnedWritesByField, field) + if (storeIdentityChurnProbe.captureSites) { + recordSite(field) + } + } + } +} + +/** + * Wraps the state creator rather than patching setState, for the same reason as + * react-commit-cascade-write-probe: slices capture the `set` closure built before + * `api` exists, and slice-internal writes are the ones that churn. + */ +export function withStoreIdentityChurnProbe( + createState: StateCreator +): StateCreator { + return (set, get, api) => { + const wrapped = ((partial: unknown, replace?: unknown): void => { + if (!storeIdentityChurnProbe.armed) { + ;(set as (nextPartial: unknown, nextReplace?: unknown) => void)(partial, replace) + return + } + const previous = get() as Record + // Why the write is passed through untouched: zustand owns when and how an + // updater runs. The probe only compares the states on either side of it. + ;(set as (nextPartial: unknown, nextReplace?: unknown) => void)(partial, replace) + try { + const next = get() as Record + if (next === previous) { + return + } + const fields = + replace !== true && partial !== null && typeof partial === 'object' + ? Object.keys(partial) + : Object.keys(next) + recordWrite(previous, next, fields) + } catch { + // A diagnostic on the app's universal write path must never break writes. + } + }) as typeof set + api.setState = wrapped as typeof api.setState + return createState(wrapped, get, api) + } +} diff --git a/src/renderer/src/store/terminals/terminal-shutdown-guards-identity.test.ts b/src/renderer/src/store/terminals/terminal-shutdown-guards-identity.test.ts new file mode 100644 index 00000000000..c353991c2c1 --- /dev/null +++ b/src/renderer/src/store/terminals/terminal-shutdown-guards-identity.test.ts @@ -0,0 +1,72 @@ +import { describe, expect, it, vi } from 'vitest' +import type { AppState } from '../types' +import { createTerminalShutdownGuardController } from './terminal-shutdown-guards' + +vi.mock('@/components/terminal-pane/pty-transport', () => ({ + restorePtyDataHandlersAfterFailedShutdown: vi.fn(), + unregisterPtyDataHandlers: vi.fn(() => []) +})) +vi.mock('@/components/terminal-pane/terminal-parked-watcher-registry', () => ({ + disposeParkedTerminalWatchersForPtyIds: vi.fn() +})) +vi.mock('@/components/terminal-pane/pty-shutdown-exit-deferral', () => ({ + clearCommittedPtyShutdownSettlements: vi.fn(), + hasCommittedPtyShutdownSettlement: vi.fn(() => false), + markCommittedPtyShutdowns: vi.fn(), + noteCommittedPtyShutdownSettlements: vi.fn(), + settleDeferredPtyShutdownExits: vi.fn() +})) + +function harness(initial: Partial, exitGuardPtyIds: readonly string[]) { + let current = initial as AppState + const set = vi.fn((update: unknown) => { + const patch = + typeof update === 'function' ? (update as (s: AppState) => object)(current) : update + current = { ...current, ...(patch as object) } + }) + const guards = createTerminalShutdownGuardController({ + exitGuardPtyIds, + get: (() => current) as never, + keepIdentifiers: false, + rendererShutdownPtyIds: exitGuardPtyIds, + runtimeEnvironmentId: null, + set: set as never, + tabs: [] + }) + return { guards, set, state: () => current } +} + +describe('markShutdownPending identity', () => { + it('does not write the store when there is nothing to guard', () => { + const { guards, set } = harness({ suppressedPtyExitIds: {}, pendingPtyShutdownIds: {} }, []) + + guards.markShutdownPending() + + expect(set).not.toHaveBeenCalled() + }) + + it('still counts a pending owner when every id is already suppressed', () => { + const suppressedPtyExitIds: Record = { 'pty-1': true } + const { guards, state } = harness( + { suppressedPtyExitIds, pendingPtyShutdownIds: { 'pty-1': 1 } }, + ['pty-1'] + ) + + guards.markShutdownPending() + + expect(state().suppressedPtyExitIds).toBe(suppressedPtyExitIds) + expect(state().pendingPtyShutdownIds).toEqual({ 'pty-1': 2 }) + }) + + it('suppresses the ids that were not yet suppressed', () => { + const { guards, state } = harness( + { suppressedPtyExitIds: { 'pty-1': true }, pendingPtyShutdownIds: {} }, + ['pty-1', 'pty-2'] + ) + + guards.markShutdownPending() + + expect(state().suppressedPtyExitIds).toEqual({ 'pty-1': true, 'pty-2': true }) + expect(state().pendingPtyShutdownIds).toEqual({ 'pty-1': 1, 'pty-2': 1 }) + }) +}) diff --git a/src/renderer/src/store/terminals/terminal-shutdown-guards.ts b/src/renderer/src/store/terminals/terminal-shutdown-guards.ts index 83342bd5041..0eba0fda927 100644 --- a/src/renderer/src/store/terminals/terminal-shutdown-guards.ts +++ b/src/renderer/src/store/terminals/terminal-shutdown-guards.ts @@ -13,6 +13,7 @@ import { settleDeferredPtyShutdownExits } from '@/components/terminal-pane/pty-shutdown-exit-deferral' import type { TerminalStoreGet, TerminalStoreSet } from './terminal-state' +import { copyOnWriteRecord } from '../copy-on-write-record' export type TerminalShutdownGuardController = { commitHandlerSnapshots: () => void @@ -48,18 +49,22 @@ export function createTerminalShutdownGuardController({ let partialRendererStopSettled = false const markShutdownPending = (): void => { + // Why the early return: tearing down a worktree whose panes already exited passes + // no guard ids, and the spreads below would still hand both maps a new identity. + if (exitGuardPtyIds.length === 0) { + return + } set((state) => { const pendingPtyShutdownIds = { ...state.pendingPtyShutdownIds } + // Why copy-on-write: re-guarding an already-suppressed pty writes the same `true`. + const suppressedPtyExitIds = copyOnWriteRecord(state.suppressedPtyExitIds) for (const ptyId of exitGuardPtyIds) { pendingPtyShutdownIds[ptyId] = (pendingPtyShutdownIds[ptyId] ?? 0) + 1 + if (state.suppressedPtyExitIds[ptyId] !== true) { + suppressedPtyExitIds.set(ptyId, true) + } } - return { - suppressedPtyExitIds: { - ...state.suppressedPtyExitIds, - ...Object.fromEntries(exitGuardPtyIds.map((ptyId) => [ptyId, true] as const)) - }, - pendingPtyShutdownIds - } + return { suppressedPtyExitIds: suppressedPtyExitIds.read(), pendingPtyShutdownIds } }) } diff --git a/src/renderer/src/store/terminals/terminal-shutdown-map-identity.test.ts b/src/renderer/src/store/terminals/terminal-shutdown-map-identity.test.ts new file mode 100644 index 00000000000..1b6cff948f1 --- /dev/null +++ b/src/renderer/src/store/terminals/terminal-shutdown-map-identity.test.ts @@ -0,0 +1,109 @@ +import { describe, expect, it } from 'vitest' +import type { AppState } from '../types' +import type { TerminalTab } from '../../../../shared/terminal-tab-types' +import { commitTerminalShutdownState } from './terminal-shutdown-state' + +const WORKTREE = 'repo::/tmp/app' +const TAB_ID = 'tab-a' + +const tab = { id: TAB_ID, worktreeId: WORKTREE } as unknown as TerminalTab + +/** Maps a shutdown with nothing left to clear must not re-reference. */ +const UNTOUCHED_FIELDS = [ + 'ptyIdsByTabId', + 'suppressedPtyExitIds', + 'pendingPtyShutdownIds', + 'pendingCodexPaneRestartIds', + 'codexRestartNoticeByPtyId', + 'pendingSetupSplitByTabId', + 'pendingIssueCommandSplitByTabId', + 'terminalLayoutsByTabId', + 'runtimePaneTitlesByTabId', + 'lastKnownRelayPtyIdByTabId' +] as const + +function buildState(overrides: Partial = {}): AppState { + return { + tabsByWorktree: { [WORKTREE]: [tab] }, + // The tab already exited: its pty list is present and empty. + ptyIdsByTabId: { [TAB_ID]: [] }, + suppressedPtyExitIds: {}, + pendingPtyShutdownIds: {}, + pendingCodexPaneRestartIds: {}, + codexRestartNoticeByPtyId: {}, + pendingSetupSplitByTabId: {}, + pendingIssueCommandSplitByTabId: {}, + terminalLayoutsByTabId: {}, + runtimePaneTitlesByTabId: {}, + lastKnownRelayPtyIdByTabId: {}, + unreadTerminalTabs: {}, + unreadTerminalPanes: {}, + unreadAgentCompletionPanes: {}, + lastTerminalInputAtByPaneKey: {}, + sleepingAgentSessionsByPaneKey: {}, + // Post-write actions this helper calls; irrelevant to the identity contract. + dropAgentStatusByWorktree: () => undefined, + clearPaneForegroundAgentByWorktree: () => undefined, + clearSleepingAgentSessionsByWorktree: () => undefined, + isPtyShutdownPending: () => false, + ...overrides + } as unknown as AppState +} + +function commit(state: AppState, exitGuardPtyIds: readonly string[] = []): AppState { + let current = state + commitTerminalShutdownState({ + exitGuardPtyIds, + get: (() => current) as never, + keepIdentifiers: true, + retainedCompletionEvidence: [], + set: ((update: unknown) => { + const patch = + typeof update === 'function' ? (update as (s: AppState) => object)(current) : update + current = { ...current, ...(patch as object) } + }) as never, + shutdownReason: 'manual-sleep', + sleepingAgentSessionRecords: {}, + tabs: [tab], + worktreeId: WORKTREE + }) + return current +} + +describe('terminal shutdown map identity', () => { + it('keeps every map reference when the panes already exited', () => { + const before = buildState() + + const after = commit(before) + + for (const field of UNTOUCHED_FIELDS) { + expect(after[field], field).toBe(before[field]) + } + }) + + it('creates an absent pty-id entry rather than skipping it', () => { + // An absent key is not an empty array: the spread this replaces created the key. + const before = buildState({ ptyIdsByTabId: {} } as Partial) + + const after = commit(before) + + expect(after.ptyIdsByTabId).not.toBe(before.ptyIdsByTabId) + expect(TAB_ID in after.ptyIdsByTabId).toBe(true) + expect(after.ptyIdsByTabId[TAB_ID]).toEqual([]) + }) + + it('still clears a live pty list and drops the exit-guard bookkeeping', () => { + const before = buildState({ + ptyIdsByTabId: { [TAB_ID]: ['pty-1'] }, + pendingPtyShutdownIds: { 'pty-1': 1 }, + codexRestartNoticeByPtyId: { 'pty-1': { reason: 'x' } } + } as unknown as Partial) + + const after = commit(before, ['pty-1']) + + expect(after.ptyIdsByTabId[TAB_ID]).toEqual([]) + expect(after.suppressedPtyExitIds['pty-1']).toBe(true) + expect('pty-1' in after.pendingPtyShutdownIds).toBe(false) + expect('pty-1' in after.codexRestartNoticeByPtyId).toBe(false) + }) +}) diff --git a/src/renderer/src/store/terminals/terminal-shutdown-state.ts b/src/renderer/src/store/terminals/terminal-shutdown-state.ts index a4cb7979978..389c8166a18 100644 --- a/src/renderer/src/store/terminals/terminal-shutdown-state.ts +++ b/src/renderer/src/store/terminals/terminal-shutdown-state.ts @@ -12,6 +12,7 @@ import { type RetainedAgentEntry } from '../slices/agent-status' import type { TerminalStoreGet, TerminalStoreSet } from './terminal-state' +import { copyOnWriteRecord } from '../copy-on-write-record' export function commitTerminalShutdownState({ exitGuardPtyIds, @@ -45,120 +46,102 @@ export function commitTerminalShutdownState({ clearTransientTerminalState(tab, index) ) } - const ptyIdsByTabId = { - ...state.ptyIdsByTabId, - ...Object.fromEntries(tabs.map((tab) => [tab.id, [] as string[]] as const)) - } - const runtimePaneTitlesByTabId = keepIdentifiers - ? state.runtimePaneTitlesByTabId - : { ...state.runtimePaneTitlesByTabId } - const suppressedPtyExitIds = { - ...state.suppressedPtyExitIds, - ...Object.fromEntries(exitGuardPtyIds.map((ptyId) => [ptyId, true] as const)) - } - const pendingPtyShutdownIds = { ...state.pendingPtyShutdownIds } - for (const ptyId of exitGuardPtyIds) { - const remainingOwners = (pendingPtyShutdownIds[ptyId] ?? 0) - 1 - if (remainingOwners > 0) { - pendingPtyShutdownIds[ptyId] = remainingOwners - } else { - delete pendingPtyShutdownIds[ptyId] + // Why copy-on-write everywhere below: a worktree whose panes already exited hits + // this with nothing to clear, and unconditional spreads then hand every map a new + // identity for no data change. ptyIdsByTabId is the costly one — six components + // select it whole, and selectLivePtyIdsForWorktree memoizes per sidebar card on + // its identity, so churning it rebuilds that record once per card. + const ptyIdsByTabId = copyOnWriteRecord(state.ptyIdsByTabId) + for (const tab of tabs) { + // Why `!== undefined`: an absent key is not an empty array, and the spread this + // replaces created the key. Only an already-empty entry can be skipped. + const current = state.ptyIdsByTabId[tab.id] + if (current === undefined || current.length > 0) { + ptyIdsByTabId.set(tab.id, []) } } - - // Sleeping terminals retain restart intent, but a wake can receive a different live PTY id. - const pendingCodexPaneRestartIds = keepIdentifiers - ? state.pendingCodexPaneRestartIds - : { ...state.pendingCodexPaneRestartIds } - const codexRestartNoticeByPtyId = { ...state.codexRestartNoticeByPtyId } + const suppressedPtyExitIds = copyOnWriteRecord(state.suppressedPtyExitIds) + const pendingPtyShutdownIds = copyOnWriteRecord(state.pendingPtyShutdownIds) + const pendingCodexPaneRestartIds = copyOnWriteRecord(state.pendingCodexPaneRestartIds) + const codexRestartNoticeByPtyId = copyOnWriteRecord(state.codexRestartNoticeByPtyId) for (const ptyId of exitGuardPtyIds) { + if (state.suppressedPtyExitIds[ptyId] !== true) { + suppressedPtyExitIds.set(ptyId, true) + } + // An absent owner count meant `delete` of a missing key, which changed nothing. + if (ptyId in state.pendingPtyShutdownIds) { + const remainingOwners = (state.pendingPtyShutdownIds[ptyId] ?? 0) - 1 + if (remainingOwners > 0) { + pendingPtyShutdownIds.set(ptyId, remainingOwners) + } else { + pendingPtyShutdownIds.delete(ptyId) + } + } + // Sleeping terminals retain restart intent, but a wake can receive a different live PTY id. if (!keepIdentifiers) { - delete pendingCodexPaneRestartIds[ptyId] + pendingCodexPaneRestartIds.delete(ptyId) } - delete codexRestartNoticeByPtyId[ptyId] + codexRestartNoticeByPtyId.delete(ptyId) } - const pendingSetupSplitByTabId = { ...state.pendingSetupSplitByTabId } - const pendingIssueCommandSplitByTabId = { ...state.pendingIssueCommandSplitByTabId } - const terminalLayoutsByTabId = { ...state.terminalLayoutsByTabId } - let unreadTerminalTabs = state.unreadTerminalTabs - let unreadTerminalPanes = state.unreadTerminalPanes - let unreadAgentCompletionPanes = state.unreadAgentCompletionPanes - let lastTerminalInputAtByPaneKey = state.lastTerminalInputAtByPaneKey + const runtimePaneTitlesByTabId = copyOnWriteRecord(state.runtimePaneTitlesByTabId) + const pendingSetupSplitByTabId = copyOnWriteRecord(state.pendingSetupSplitByTabId) + const pendingIssueCommandSplitByTabId = copyOnWriteRecord(state.pendingIssueCommandSplitByTabId) + const terminalLayoutsByTabId = copyOnWriteRecord(state.terminalLayoutsByTabId) + const lastKnownRelayPtyIdByTabId = copyOnWriteRecord(state.lastKnownRelayPtyIdByTabId) + const unreadTerminalTabs = copyOnWriteRecord(state.unreadTerminalTabs) + const unreadTerminalPanes = copyOnWriteRecord(state.unreadTerminalPanes) + const unreadAgentCompletionPanes = copyOnWriteRecord(state.unreadAgentCompletionPanes) + const lastTerminalInputAtByPaneKey = copyOnWriteRecord(state.lastTerminalInputAtByPaneKey) for (const tab of tabs) { - if (!keepIdentifiers) { - delete runtimePaneTitlesByTabId[tab.id] - } - delete pendingSetupSplitByTabId[tab.id] - delete pendingIssueCommandSplitByTabId[tab.id] - if (unreadTerminalTabs[tab.id]) { - if (unreadTerminalTabs === state.unreadTerminalTabs) { - unreadTerminalTabs = { ...state.unreadTerminalTabs } - } - delete unreadTerminalTabs[tab.id] - } - for (const paneKey of Object.keys(unreadTerminalPanes)) { - if (paneKey.startsWith(`${tab.id}:`)) { - if (unreadTerminalPanes === state.unreadTerminalPanes) { - unreadTerminalPanes = { ...unreadTerminalPanes } - } - delete unreadTerminalPanes[paneKey] + pendingSetupSplitByTabId.delete(tab.id) + pendingIssueCommandSplitByTabId.delete(tab.id) + unreadTerminalTabs.delete(tab.id) + const panePrefix = `${tab.id}:` + for (const paneKey of Object.keys(state.unreadTerminalPanes)) { + if (paneKey.startsWith(panePrefix)) { + unreadTerminalPanes.delete(paneKey) } } - for (const paneKey of Object.keys(unreadAgentCompletionPanes)) { - if (paneKey.startsWith(`${tab.id}:`)) { - if (unreadAgentCompletionPanes === state.unreadAgentCompletionPanes) { - unreadAgentCompletionPanes = { ...unreadAgentCompletionPanes } - } - delete unreadAgentCompletionPanes[paneKey] + for (const paneKey of Object.keys(state.unreadAgentCompletionPanes)) { + if (paneKey.startsWith(panePrefix)) { + unreadAgentCompletionPanes.delete(paneKey) } } - for (const paneKey of Object.keys(lastTerminalInputAtByPaneKey)) { - if (paneKey.startsWith(`${tab.id}:`)) { - if (lastTerminalInputAtByPaneKey === state.lastTerminalInputAtByPaneKey) { - lastTerminalInputAtByPaneKey = { ...lastTerminalInputAtByPaneKey } - } - delete lastTerminalInputAtByPaneKey[paneKey] + for (const paneKey of Object.keys(state.lastTerminalInputAtByPaneKey)) { + if (paneKey.startsWith(panePrefix)) { + lastTerminalInputAtByPaneKey.delete(paneKey) } } if (!keepIdentifiers) { - const layout = terminalLayoutsByTabId[tab.id] - if (layout?.ptyIdsByLeafId) { - terminalLayoutsByTabId[tab.id] = { ...layout, ptyIdsByLeafId: {} } + runtimePaneTitlesByTabId.delete(tab.id) + lastKnownRelayPtyIdByTabId.delete(tab.id) + const layout = state.terminalLayoutsByTabId[tab.id] + // Why the emptiness check: replacing an already-empty map with a fresh {} is + // the same value with a new identity. + if (layout?.ptyIdsByLeafId && Object.keys(layout.ptyIdsByLeafId).length > 0) { + terminalLayoutsByTabId.set(tab.id, { ...layout, ptyIdsByLeafId: {} }) } } } - const lastKnownRelayPtyIdByTabId = keepIdentifiers - ? state.lastKnownRelayPtyIdByTabId - : { ...state.lastKnownRelayPtyIdByTabId } - if (!keepIdentifiers) { - for (const tab of tabs) { - delete lastKnownRelayPtyIdByTabId[tab.id] - } - } - return { tabsByWorktree, - ptyIdsByTabId, - lastKnownRelayPtyIdByTabId, - runtimePaneTitlesByTabId, - suppressedPtyExitIds, - pendingPtyShutdownIds, - pendingCodexPaneRestartIds, - codexRestartNoticeByPtyId, - pendingSetupSplitByTabId, - pendingIssueCommandSplitByTabId, - terminalLayoutsByTabId, - ...(unreadTerminalTabs !== state.unreadTerminalTabs ? { unreadTerminalTabs } : {}), - ...(unreadTerminalPanes !== state.unreadTerminalPanes ? { unreadTerminalPanes } : {}), - ...(unreadAgentCompletionPanes !== state.unreadAgentCompletionPanes - ? { unreadAgentCompletionPanes } - : {}), - ...(lastTerminalInputAtByPaneKey !== state.lastTerminalInputAtByPaneKey - ? { lastTerminalInputAtByPaneKey } - : {}) + ptyIdsByTabId: ptyIdsByTabId.read(), + lastKnownRelayPtyIdByTabId: lastKnownRelayPtyIdByTabId.read(), + runtimePaneTitlesByTabId: runtimePaneTitlesByTabId.read(), + suppressedPtyExitIds: suppressedPtyExitIds.read(), + pendingPtyShutdownIds: pendingPtyShutdownIds.read(), + pendingCodexPaneRestartIds: pendingCodexPaneRestartIds.read(), + codexRestartNoticeByPtyId: codexRestartNoticeByPtyId.read(), + pendingSetupSplitByTabId: pendingSetupSplitByTabId.read(), + pendingIssueCommandSplitByTabId: pendingIssueCommandSplitByTabId.read(), + terminalLayoutsByTabId: terminalLayoutsByTabId.read(), + unreadTerminalTabs: unreadTerminalTabs.read(), + unreadTerminalPanes: unreadTerminalPanes.read(), + unreadAgentCompletionPanes: unreadAgentCompletionPanes.read(), + lastTerminalInputAtByPaneKey: lastTerminalInputAtByPaneKey.read() } }) @@ -174,7 +157,10 @@ export function commitTerminalShutdownState({ ).records : state.sleepingAgentSessionsByPaneKey return { - sleepingAgentSessionsByPaneKey: { ...base, ...sleepingAgentSessionRecords } + sleepingAgentSessionsByPaneKey: { + ...base, + ...sleepingAgentSessionRecords + } } }) } else { diff --git a/src/renderer/src/store/terminals/terminal-tab-close-map-identity.test.ts b/src/renderer/src/store/terminals/terminal-tab-close-map-identity.test.ts new file mode 100644 index 00000000000..1b44147ca30 --- /dev/null +++ b/src/renderer/src/store/terminals/terminal-tab-close-map-identity.test.ts @@ -0,0 +1,110 @@ +import { describe, it, expect, vi, beforeEach } from 'vitest' +import type * as AgentStatusModule from '@/lib/agent-status' +import { createTestStore, makeTab, makeWorktree, seedStore } from '../slices/store-test-helpers' +import { createStoreCascadesMockApi } from '../slices/store-cascades-test-harness' + +vi.mock('sonner', () => ({ + toast: { info: vi.fn(), success: vi.fn(), error: vi.fn(), warning: vi.fn() } +})) + +vi.mock('@/components/terminal-pane/pty-dispatcher', () => ({ + restorePtyDataHandlersAfterFailedShutdown: vi.fn(), + unregisterPtyDataHandlers: vi.fn<() => unknown[]>(() => []) +})) + +vi.mock('@/lib/agent-status', async (importOriginal) => ({ + ...(await importOriginal()), + detectAgentStatusFromTitle: vi.fn().mockReturnValue(null) +})) + +const mockApi = createStoreCascadesMockApi() + +const WORKTREE = 'repo::/tmp/app' + +/** Maps a closing tab has no entry in; closing must not give them a new reference. */ +const UNTOUCHED_FIELDS = [ + 'terminalLayoutsByTabId', + 'ptyIdsByTabId', + 'runtimePaneTitlesByTabId', + 'lastKnownRelayPtyIdByTabId', + 'deferredSshSessionIdsByTabId', + 'pendingReconnectPtyIdByTabId', + 'directSshPaneRetryByTabId', + 'directSshLivePtyBindingByTabId', + 'pendingStartupByTabId', + 'automaticAgentResumeClaimsByTabId', + 'nativeChatLaunchPromptByTabId', + 'nativeChatLaunchDraftByTabId', + 'pendingInitialCwdByTabId', + 'pendingSetupSplitByTabId', + 'pendingIssueCommandSplitByTabId', + 'expandedPaneByTabId', + 'canExpandPaneByTabId', + 'cacheTimerByKey', + 'lastTerminalInputAtByPaneKey', + 'unreadTerminalTabs', + 'unreadTerminalPanes', + 'unreadAgentCompletionPanes', + 'tabBarOrderByWorktree' +] as const + +function storeWithTwoTabs(): ReturnType { + const store = createTestStore() + seedStore(store, { + repos: [{ id: 'repo', path: '/tmp/app', name: 'app' }] as never, + worktreesByRepo: { + repo: [makeWorktree({ id: WORKTREE, repoId: 'repo', path: '/tmp/app' })] + }, + tabsByWorktree: { + [WORKTREE]: [ + makeTab({ id: 'tab-a', worktreeId: WORKTREE }), + makeTab({ id: 'tab-b', worktreeId: WORKTREE }) + ] + } + }) + return store +} + +describe('closeTab map identity', () => { + beforeEach(() => { + vi.clearAllMocks() + mockApi.worktrees.updateMeta.mockResolvedValue({}) + }) + + it('keeps the reference of every per-tab map the closing tab had no entry in', () => { + const store = storeWithTwoTabs() + const before = store.getState() + const snapshot = Object.fromEntries( + UNTOUCHED_FIELDS.map((field) => [field, before[field]]) + ) as Record + + store.getState().closeTab('tab-a') + + const after = store.getState() + // The tab really closed — otherwise the identity assertions below are vacuous. + expect(after.tabsByWorktree[WORKTREE].map((tab) => tab.id)).toEqual(['tab-b']) + for (const field of UNTOUCHED_FIELDS) { + expect(after[field], field).toBe(snapshot[field]) + } + }) + + it('still drops the closing tab from a map that did hold it', () => { + const store = storeWithTwoTabs() + store.setState({ + expandedPaneByTabId: { 'tab-a': true, 'tab-b': false }, + pendingStartupByTabId: { 'tab-a': true }, + cacheTimerByKey: { 'tab-a:leaf': 1, 'tab-b:leaf': 2 }, + unreadTerminalPanes: { 'tab-a:leaf': true } + } as never) + const before = store.getState() + + store.getState().closeTab('tab-a') + + const after = store.getState() + expect(after.expandedPaneByTabId).not.toBe(before.expandedPaneByTabId) + expect(after.expandedPaneByTabId).toEqual({ 'tab-b': false }) + expect(after.pendingStartupByTabId).toEqual({}) + expect(after.cacheTimerByKey).toEqual({ 'tab-b:leaf': 2 }) + expect(after.unreadTerminalPanes).toEqual({}) + }) +}) diff --git a/src/renderer/src/store/terminals/terminal-tab-close.ts b/src/renderer/src/store/terminals/terminal-tab-close.ts index d469bc99510..2e1d45127a5 100644 --- a/src/renderer/src/store/terminals/terminal-tab-close.ts +++ b/src/renderer/src/store/terminals/terminal-tab-close.ts @@ -16,6 +16,8 @@ import { import type { TerminalSlice, TerminalStoreGet, TerminalStoreSet } from './terminal-state' import { startTerminalTabProviderRetirement } from './terminal-tab-close-providers' import { omitUnverifiedPtyLossTabIds } from './terminal-unverified-pty-loss' +import { removePaneKeysByTabPrefix } from '../slices/agent-status-pane-keyed-records' +import { omitRecordKeys } from '../slices/worktrees/teardown/record-key-omission' export function createTerminalTabCloseActions( set: TerminalStoreSet, @@ -42,6 +44,11 @@ export function createTerminalTabCloseActions( }) } set((s) => { + // Why hoisted: omitRecordKeys takes an iterable, and this closes over one + // array instead of allocating a fresh [tabId] at each of the call sites below. + const closingTabIds = [tabId] + const omitByTabId = (record: Record): Record => + omitRecordKeys(record, closingTabIds) const next = { ...s.tabsByWorktree } let closedTab: TerminalTab | null = null let closedWorktreeId: string | null = null @@ -96,105 +103,67 @@ export function createTerminalTabCloseActions( ...(closedPosition ? { position: closedPosition } : {}) } : null - const nextExpanded = { ...s.expandedPaneByTabId } - delete nextExpanded[tabId] - const nextCanExpand = { ...s.canExpandPaneByTabId } - delete nextCanExpand[tabId] - const nextLayouts = { ...s.terminalLayoutsByTabId } - delete nextLayouts[tabId] - const nextPtyIdsByTabId = { ...s.ptyIdsByTabId } - delete nextPtyIdsByTabId[tabId] - const nextLastKnownRelay = { ...s.lastKnownRelayPtyIdByTabId } - delete nextLastKnownRelay[tabId] - const nextDeferredSshSessionIdsByTabId = { ...s.deferredSshSessionIdsByTabId } - delete nextDeferredSshSessionIdsByTabId[tabId] - const nextPendingReconnectPtyIdByTabId = { ...s.pendingReconnectPtyIdByTabId } - delete nextPendingReconnectPtyIdByTabId[tabId] - const nextRuntimePaneTitlesByTabId = { ...s.runtimePaneTitlesByTabId } - delete nextRuntimePaneTitlesByTabId[tabId] - const nextDirectSshPaneRetryByTabId = { ...s.directSshPaneRetryByTabId } - delete nextDirectSshPaneRetryByTabId[tabId] - const nextDirectSshLivePtyBindingByTabId = { - ...s.directSshLivePtyBindingByTabId - } - delete nextDirectSshLivePtyBindingByTabId[tabId] - const nextDirectSshPaneRetryHistoryByTabId = { - ...s.directSshPaneRetryHistoryByTabId - } - delete nextDirectSshPaneRetryHistoryByTabId[tabId] + const nextExpanded = omitByTabId(s.expandedPaneByTabId) + const nextCanExpand = omitByTabId(s.canExpandPaneByTabId) + const nextLayouts = omitByTabId(s.terminalLayoutsByTabId) + const nextPtyIdsByTabId = omitByTabId(s.ptyIdsByTabId) + const nextLastKnownRelay = omitByTabId(s.lastKnownRelayPtyIdByTabId) + const nextDeferredSshSessionIdsByTabId = omitByTabId(s.deferredSshSessionIdsByTabId) + const nextPendingReconnectPtyIdByTabId = omitByTabId(s.pendingReconnectPtyIdByTabId) + const nextRuntimePaneTitlesByTabId = omitByTabId(s.runtimePaneTitlesByTabId) + const nextDirectSshPaneRetryByTabId = omitByTabId(s.directSshPaneRetryByTabId) + const nextDirectSshLivePtyBindingByTabId = omitByTabId(s.directSshLivePtyBindingByTabId) + const nextDirectSshPaneRetryHistoryByTabId = omitByTabId(s.directSshPaneRetryHistoryByTabId) const nextUnverifiedPtyLossTabIds = omitUnverifiedPtyLossTabIds(s.unverifiedPtyLossTabIds, [ tabId ]) // Why: keep the same reference when the closing tab had no unread flag, so unrelated closes don't force full-state selector re-eval. - let nextUnreadTerminalTabs = s.unreadTerminalTabs - if (s.unreadTerminalTabs[tabId]) { - nextUnreadTerminalTabs = { ...s.unreadTerminalTabs } - delete nextUnreadTerminalTabs[tabId] - } - let nextUnreadTerminalPanes = s.unreadTerminalPanes - for (const paneKey of Object.keys(s.unreadTerminalPanes)) { - if (paneKey.startsWith(`${tabId}:`)) { - if (nextUnreadTerminalPanes === s.unreadTerminalPanes) { - nextUnreadTerminalPanes = { ...s.unreadTerminalPanes } - } - delete nextUnreadTerminalPanes[paneKey] - } - } - let nextUnreadAgentCompletionPanes = s.unreadAgentCompletionPanes - for (const paneKey of Object.keys(s.unreadAgentCompletionPanes)) { - if (paneKey.startsWith(`${tabId}:`)) { - if (nextUnreadAgentCompletionPanes === s.unreadAgentCompletionPanes) { - nextUnreadAgentCompletionPanes = { ...s.unreadAgentCompletionPanes } - } - delete nextUnreadAgentCompletionPanes[paneKey] - } - } - const nextLastTerminalInputAtByPaneKey = { ...s.lastTerminalInputAtByPaneKey } - for (const paneKey of Object.keys(nextLastTerminalInputAtByPaneKey)) { - if (paneKey.startsWith(`${tabId}:`)) { - delete nextLastTerminalInputAtByPaneKey[paneKey] - } - } + const nextUnreadTerminalTabs = omitByTabId(s.unreadTerminalTabs) + const nextUnreadTerminalPanes = removePaneKeysByTabPrefix(s.unreadTerminalPanes, tabId) + const nextUnreadAgentCompletionPanes = removePaneKeysByTabPrefix( + s.unreadAgentCompletionPanes, + tabId + ) + const nextLastTerminalInputAtByPaneKey = removePaneKeysByTabPrefix( + s.lastTerminalInputAtByPaneKey, + tabId + ) const nextSleepingAgentSessionsByPaneKey = retiresSession ? removeSleepingAgentSessionsForTab(s.sleepingAgentSessionsByPaneKey, tabId) : s.sleepingAgentSessionsByPaneKey - const nextPendingStartupByTabId = { ...s.pendingStartupByTabId } - delete nextPendingStartupByTabId[tabId] - const nextAutomaticAgentResumeClaimsByTabId = { ...s.automaticAgentResumeClaimsByTabId } - delete nextAutomaticAgentResumeClaimsByTabId[tabId] - const nextNativeChatLaunchPromptByTabId = { ...s.nativeChatLaunchPromptByTabId } - delete nextNativeChatLaunchPromptByTabId[tabId] - const nextNativeChatLaunchDraftByTabId = { ...s.nativeChatLaunchDraftByTabId } - delete nextNativeChatLaunchDraftByTabId[tabId] - const nextPendingInitialCwdByTabId = { ...s.pendingInitialCwdByTabId } - delete nextPendingInitialCwdByTabId[tabId] - const nextPendingSetupSplitByTabId = { ...s.pendingSetupSplitByTabId } - delete nextPendingSetupSplitByTabId[tabId] - const nextPendingIssueCommandSplitByTabId = { ...s.pendingIssueCommandSplitByTabId } - delete nextPendingIssueCommandSplitByTabId[tabId] - const nextCacheTimer = { ...s.cacheTimerByKey } + const nextPendingStartupByTabId = omitByTabId(s.pendingStartupByTabId) + const nextAutomaticAgentResumeClaimsByTabId = omitByTabId( + s.automaticAgentResumeClaimsByTabId + ) + const nextNativeChatLaunchPromptByTabId = omitByTabId(s.nativeChatLaunchPromptByTabId) + const nextNativeChatLaunchDraftByTabId = omitByTabId(s.nativeChatLaunchDraftByTabId) + const nextPendingInitialCwdByTabId = omitByTabId(s.pendingInitialCwdByTabId) + const nextPendingSetupSplitByTabId = omitByTabId(s.pendingSetupSplitByTabId) + const nextPendingIssueCommandSplitByTabId = omitByTabId(s.pendingIssueCommandSplitByTabId) // Why: cache timer keys are `${tabId}:${leafId}` composites; remove all entries for the closing tab. - for (const key of Object.keys(nextCacheTimer)) { - if (key.startsWith(`${tabId}:`)) { - delete nextCacheTimer[key] - } - } + const nextCacheTimer = removePaneKeysByTabPrefix(s.cacheTimerByKey, tabId) // Why: keep activeTabIdByWorktree in sync when closing a background-worktree tab, else the stale remembered tab falls back to tabs[0] on switch. - const nextActiveTabIdByWorktree = { ...s.activeTabIdByWorktree } + let nextActiveTabIdByWorktree = s.activeTabIdByWorktree for (const [wId, tabs] of Object.entries(next)) { - if (nextActiveTabIdByWorktree[wId] === tabId) { - nextActiveTabIdByWorktree[wId] = tabs[0]?.id ?? null + if (nextActiveTabIdByWorktree[wId] !== tabId) { + continue } + if (nextActiveTabIdByWorktree === s.activeTabIdByWorktree) { + nextActiveTabIdByWorktree = { ...s.activeTabIdByWorktree } + } + nextActiveTabIdByWorktree[wId] = tabs[0]?.id ?? null } // Why: keep tabBarOrderByWorktree in sync so stale terminal IDs don't linger and shift positions on later tab operations. - const nextTabBarOrderByWorktree: Record = { - ...s.tabBarOrderByWorktree - } - for (const wId of Object.keys(nextTabBarOrderByWorktree)) { - const order = nextTabBarOrderByWorktree[wId] - if (order?.includes(tabId)) { - nextTabBarOrderByWorktree[wId] = order.filter((entryId) => entryId !== tabId) + let nextTabBarOrderByWorktree: Record = s.tabBarOrderByWorktree + for (const wId of Object.keys(s.tabBarOrderByWorktree)) { + const order = s.tabBarOrderByWorktree[wId] + if (!order?.includes(tabId)) { + continue } + if (nextTabBarOrderByWorktree === s.tabBarOrderByWorktree) { + nextTabBarOrderByWorktree = { ...s.tabBarOrderByWorktree } + } + nextTabBarOrderByWorktree[wId] = order.filter((entryId) => entryId !== tabId) } // Why: clean up unconsumed snapshot/cold-restore data (e.g. tab closed before TerminalPane mounted) to prevent unbounded store growth across restarts. let nextSnapshots = s.pendingSnapshotByPtyId diff --git a/src/renderer/src/store/terminals/workspace-terminal-placeholders.test.ts b/src/renderer/src/store/terminals/workspace-terminal-placeholders.test.ts new file mode 100644 index 00000000000..f52898dce26 --- /dev/null +++ b/src/renderer/src/store/terminals/workspace-terminal-placeholders.test.ts @@ -0,0 +1,121 @@ +import { describe, expect, it } from 'vitest' +import type { Repo } from '../../../../shared/repo-types' +import type { Worktree } from '../../../../shared/worktree/types' +import { DEFAULT_REPO_BADGE_COLOR } from '../../../../shared/constants' +import { buildRuntimeSessionPlaceholders } from './workspace-terminal-placeholders' +import { addHydratedSshWorktreePlaceholders } from './workspace-terminal-ssh-placeholders' + +const repo: Repo = { + id: 'repo-1', + path: '/repos/one', + displayName: 'one', + badgeColor: DEFAULT_REPO_BADGE_COLOR, + addedAt: 0, + connectionId: null, + executionHostId: 'local' +} + +const worktree: Worktree = { + id: 'repo-1::/repos/one', + repoId: 'repo-1', + hostId: 'local', + displayName: 'main', + comment: '', + linkedIssue: null, + linkedPR: null, + linkedLinearIssue: null, + linkedGitLabMR: null, + linkedGitLabIssue: null, + isArchived: false, + isUnread: false, + isPinned: false, + sortOrder: 0, + lastActivityAt: 0, + path: '/repos/one', + head: '', + branch: '', + isBare: false, + isMainWorktree: true +} + +describe('buildRuntimeSessionPlaceholders', () => { + it('returns the original repos array when no placeholder repo is needed', () => { + const repos = [repo] + const worktreesByRepo = { 'repo-1': [worktree] } + + const result = buildRuntimeSessionPlaceholders({ + repos, + runtimeHostIdByWorkspaceSessionKey: {}, + worktreesByRepo + }) + + // Hydration writes these straight to the store; a fresh array would rerender + // every component selecting the whole repo list for no data change. + expect(result.repos).toBe(repos) + expect(result.worktreesByRepo).toBe(worktreesByRepo) + }) + + it('keeps the original repos array when the session only references known repos', () => { + const repos = [repo] + + const result = buildRuntimeSessionPlaceholders({ + repos, + runtimeHostIdByWorkspaceSessionKey: { 'repo-1::/repos/one': 'runtime:host-1' }, + worktreesByRepo: { 'repo-1': [worktree] } + }) + + expect(result.repos).toBe(repos) + }) + + it('still appends a placeholder repo for an unknown runtime workspace', () => { + const repos = [repo] + + const result = buildRuntimeSessionPlaceholders({ + repos, + runtimeHostIdByWorkspaceSessionKey: { 'repo-2::/repos/two': 'runtime:host-1' }, + worktreesByRepo: { 'repo-1': [worktree] } + }) + + expect(result.repos).not.toBe(repos) + expect(result.repos.map((entry) => entry.id)).toEqual(['repo-1', 'repo-2']) + // The caller's array must not be mutated in place. + expect(repos).toHaveLength(1) + }) +}) + +describe('addHydratedSshWorktreePlaceholders', () => { + it('returns the original map when no SSH placeholder is needed', () => { + const worktreesByRepo = { 'repo-1': [worktree] } + + const result = addHydratedSshWorktreePlaceholders([repo], worktreesByRepo, { + 'repo-1::/repos/one': [] + }) + + expect(result).toBe(worktreesByRepo) + }) + + it('returns the original map when the SSH worktree is already present', () => { + const sshRepo: Repo = { ...repo, id: 'ssh-repo', connectionId: 'ssh-1' } + const sshWorktree: Worktree = { ...worktree, id: 'ssh-repo::/repos/ssh', repoId: 'ssh-repo' } + const worktreesByRepo = { 'ssh-repo': [sshWorktree] } + + const result = addHydratedSshWorktreePlaceholders([sshRepo], worktreesByRepo, { + 'ssh-repo::/repos/ssh': [] + }) + + expect(result).toBe(worktreesByRepo) + }) + + it('still adds a placeholder for an SSH worktree with no row, without mutating the caller', () => { + const sshRepo: Repo = { ...repo, id: 'ssh-repo', connectionId: 'ssh-1' } + const worktreesByRepo = { 'ssh-repo': [] as Worktree[] } + + const result = addHydratedSshWorktreePlaceholders([sshRepo], worktreesByRepo, { + 'ssh-repo::/repos/ssh': [] + }) + + expect(result).not.toBe(worktreesByRepo) + expect(result['ssh-repo'].map((entry) => entry.id)).toEqual(['ssh-repo::/repos/ssh']) + expect(worktreesByRepo['ssh-repo']).toHaveLength(0) + }) +}) diff --git a/src/renderer/src/store/terminals/workspace-terminal-placeholders.ts b/src/renderer/src/store/terminals/workspace-terminal-placeholders.ts index e83adf8b937..94a520993a0 100644 --- a/src/renderer/src/store/terminals/workspace-terminal-placeholders.ts +++ b/src/renderer/src/store/terminals/workspace-terminal-placeholders.ts @@ -20,10 +20,12 @@ export function buildRuntimeSessionPlaceholders({ runtimeHostIdByWorkspaceSessionKey: Record worktreesByRepo: Record }): { - repos: Repo[] + repos: readonly Repo[] worktreesByRepo: Record } { - let nextRepos = repos.slice() + // Why copy-on-write: hydration writes both straight to the store, and an unconditional copy + // rerendered every whole-array/map selector on every hydration with no data change. + let nextRepos: readonly Repo[] = repos let nextWorktreesByRepo = worktreesByRepo for (const workspaceSessionKey of Object.keys(runtimeHostIdByWorkspaceSessionKey)) { const hostId = runtimeHostIdByWorkspaceSessionKey[workspaceSessionKey] diff --git a/src/renderer/src/store/terminals/workspace-terminal-ssh-placeholders.ts b/src/renderer/src/store/terminals/workspace-terminal-ssh-placeholders.ts index 3012b017e15..b2cdc5581e2 100644 --- a/src/renderer/src/store/terminals/workspace-terminal-ssh-placeholders.ts +++ b/src/renderer/src/store/terminals/workspace-terminal-ssh-placeholders.ts @@ -12,7 +12,9 @@ export function addHydratedSshWorktreePlaceholders( tabsByWorktree: Record ): Record { const sshRepoIds = new Set(repos.filter((repo) => repo.connectionId).map((repo) => repo.id)) - const worktreesByRepo = { ...sourceWorktreesByRepo } + // Why copy-on-write: hydration writes this map straight to the store; an unconditional copy + // rerendered every whole-map selector on every hydration with no data change. + let worktreesByRepo = sourceWorktreesByRepo for (const worktreeId of Object.keys(tabsByWorktree)) { const repoId = getRepoIdFromWorktreeId(worktreeId) if (!sshRepoIds.has(repoId)) { @@ -45,6 +47,9 @@ export function addHydratedSshWorktreePlaceholders( isBare: false, isMainWorktree: false } + if (worktreesByRepo === sourceWorktreesByRepo) { + worktreesByRepo = { ...sourceWorktreesByRepo } + } worktreesByRepo[repoId] = [...(worktreesByRepo[repoId] ?? []), placeholder] } return worktreesByRepo diff --git a/src/renderer/src/web/preload-api/web-agent-accounts-api.ts b/src/renderer/src/web/preload-api/web-agent-accounts-api.ts index 921c8ca7a5d..26b11ad68dc 100644 --- a/src/renderer/src/web/preload-api/web-agent-accounts-api.ts +++ b/src/renderer/src/web/preload-api/web-agent-accounts-api.ts @@ -3,12 +3,14 @@ import type { PreloadApi } from '../../../../preload/api-types' export function createMiniMaxCredentialsApi(): NonNullable< Partial['minimaxCredentials'] > { - const notConfigured = { configured: false } + const notConfigured = { configured: false, cookieConfigured: false, apiKeyConfigured: false } const unsupportedError = new Error('MiniMax cookie storage is only available in the desktop app.') return { getStatus: () => Promise.resolve(notConfigured), saveCookie: () => Promise.reject(unsupportedError), - clearCookie: () => Promise.resolve(notConfigured) + clearCookie: () => Promise.resolve(notConfigured), + saveApiKey: () => Promise.reject(unsupportedError), + clearApiKey: () => Promise.resolve(notConfigured) } } diff --git a/src/renderer/src/web/preload-api/web-preferences-store.ts b/src/renderer/src/web/preload-api/web-preferences-store.ts index 13c755b1d87..f42c31d628a 100644 --- a/src/renderer/src/web/preload-api/web-preferences-store.ts +++ b/src/renderer/src/web/preload-api/web-preferences-store.ts @@ -139,6 +139,12 @@ export async function getRuntimeBackedStoredSettings(): Promise if (typeof result.settings.minimaxUsageModels === 'string') { runtimeSettings.minimaxUsageModels = result.settings.minimaxUsageModels } + if ( + result.settings.minimaxEndpoint === 'overseas' || + result.settings.minimaxEndpoint === 'cn' + ) { + runtimeSettings.minimaxEndpoint = result.settings.minimaxEndpoint + } if (Array.isArray(result.settings.prBotAuthorOverrides)) { runtimeSettings.prBotAuthorOverrides = normalizePRBotAuthorOverrides( result.settings.prBotAuthorOverrides @@ -204,6 +210,9 @@ export async function syncRuntimeBackedSettings( if (typeof updates.minimaxUsageModels === 'string') { runtimeUpdates.minimaxUsageModels = updates.minimaxUsageModels } + if (updates.minimaxEndpoint === 'overseas' || updates.minimaxEndpoint === 'cn') { + runtimeUpdates.minimaxEndpoint = updates.minimaxEndpoint + } if (Array.isArray(updates.prBotAuthorOverrides)) { runtimeUpdates.prBotAuthorOverrides = normalizePRBotAuthorOverrides( updates.prBotAuthorOverrides diff --git a/src/renderer/src/web/preload-api/web-rate-limits-api.ts b/src/renderer/src/web/preload-api/web-rate-limits-api.ts index 7587a838660..023b7e3fd3a 100644 --- a/src/renderer/src/web/preload-api/web-rate-limits-api.ts +++ b/src/renderer/src/web/preload-api/web-rate-limits-api.ts @@ -13,6 +13,7 @@ export function createRateLimitsApi(): NonNullable['rateLimi minimax: null, grok: null, minimaxCookieConfigured: false, + minimaxApiKeyConfigured: false, grokAuthConfigured: false, claudeTarget: { runtime: 'host', wslDistro: null }, codexTarget: { runtime: 'host', wslDistro: null }, diff --git a/src/renderer/src/web/web-preload-api-agent-providers.test.ts b/src/renderer/src/web/web-preload-api-agent-providers.test.ts index 35324baaa13..c5bfca8af71 100644 --- a/src/renderer/src/web/web-preload-api-agent-providers.test.ts +++ b/src/renderer/src/web/web-preload-api-agent-providers.test.ts @@ -158,9 +158,23 @@ describe('web MiniMax preload API', () => { it('exposes desktop-only MiniMax credential reads as unconfigured and rejects saves', async () => { const { api } = await installApi('Linux') - await expect(api.minimaxCredentials.getStatus()).resolves.toEqual({ configured: false }) + await expect(api.minimaxCredentials.getStatus()).resolves.toEqual({ + configured: false, + cookieConfigured: false, + apiKeyConfigured: false + }) await expect(api.minimaxCredentials.saveCookie('_token=abc')).rejects.toThrow(/desktop app/i) - await expect(api.minimaxCredentials.clearCookie()).resolves.toEqual({ configured: false }) + await expect(api.minimaxCredentials.clearCookie()).resolves.toEqual({ + configured: false, + cookieConfigured: false, + apiKeyConfigured: false + }) + await expect(api.minimaxCredentials.saveApiKey('sk-test')).rejects.toThrow(/desktop app/i) + await expect(api.minimaxCredentials.clearApiKey()).resolves.toEqual({ + configured: false, + cookieConfigured: false, + apiKeyConfigured: false + }) }) }) diff --git a/src/renderer/src/web/web-preload-api-settings.test.ts b/src/renderer/src/web/web-preload-api-settings.test.ts index 3ac1a49aa59..cbaff1f70c9 100644 --- a/src/renderer/src/web/web-preload-api-settings.test.ts +++ b/src/renderer/src/web/web-preload-api-settings.test.ts @@ -597,7 +597,8 @@ describe('web settings preload API', () => { result: { settings: { minimaxGroupId: 'group-42', - minimaxUsageModels: 'general,abab6.5' + minimaxUsageModels: 'general,abab6.5', + minimaxEndpoint: 'cn' } }, _meta: { runtimeId: 'runtime-1' } @@ -617,12 +618,15 @@ describe('web settings preload API', () => { const stored = JSON.parse(globals.storage.getItem('orca.web.settings.v1') ?? '{}') as { minimaxGroupId?: string minimaxUsageModels?: string + minimaxEndpoint?: string } expect(settings.minimaxGroupId).toBe('group-42') expect(settings.minimaxUsageModels).toBe('general,abab6.5') + expect(settings.minimaxEndpoint).toBe('cn') expect(stored.minimaxGroupId).toBe('group-42') expect(stored.minimaxUsageModels).toBe('general,abab6.5') + expect(stored.minimaxEndpoint).toBe('cn') expect(runtimeCalls).toEqual([{ method: 'settings.get', params: undefined }]) }) @@ -743,7 +747,8 @@ describe('web settings preload API', () => { result: { settings: { minimaxGroupId: 'group-42', - minimaxUsageModels: 'general,abab6.5' + minimaxUsageModels: 'general,abab6.5', + minimaxEndpoint: 'cn' } }, _meta: { runtimeId: 'runtime-1' } @@ -761,24 +766,29 @@ describe('web settings preload API', () => { const settings = await globals.window.api.settings.set({ minimaxGroupId: 'group-42', - minimaxUsageModels: 'general,abab6.5' + minimaxUsageModels: 'general,abab6.5', + minimaxEndpoint: 'cn' }) const stored = JSON.parse(globals.storage.getItem('orca.web.settings.v1') ?? '{}') as { minimaxGroupId?: string minimaxUsageModels?: string + minimaxEndpoint?: string } expect(settings.minimaxGroupId).toBe('group-42') expect(settings.minimaxUsageModels).toBe('general,abab6.5') + expect(settings.minimaxEndpoint).toBe('cn') expect(stored.minimaxGroupId).toBe('group-42') expect(stored.minimaxUsageModels).toBe('general,abab6.5') + expect(stored.minimaxEndpoint).toBe('cn') expect(runtimeCalls).toEqual([ { method: 'settings.update', params: { minimaxGroupId: 'group-42', - minimaxUsageModels: 'general,abab6.5' + minimaxUsageModels: 'general,abab6.5', + minimaxEndpoint: 'cn' } } ]) diff --git a/src/shared/agent-process-recognition.test.ts b/src/shared/agent-process-recognition.test.ts index 015fb2964a1..d97d334d6d3 100644 --- a/src/shared/agent-process-recognition.test.ts +++ b/src/shared/agent-process-recognition.test.ts @@ -178,6 +178,19 @@ describe('agent process recognition', () => { expect(isRecognizedAgentType('vibe')).toBe(true) }) + it('recognizes Kimi Code by the kimi-code process its launcher becomes', () => { + expect(recognizeAgentProcess('/home/dev/.kimi-code/bin/kimi')).toEqual({ + agent: 'kimi', + processName: 'kimi' + }) + expect(recognizeAgentProcess('kimi-code')).toEqual({ + agent: 'kimi', + processName: 'kimi-code' + }) + expect(isExpectedAgentProcess('/home/dev/.kimi-code/bin/kimi', 'kimi')).toBe(true) + expect(isRecognizedAgentType('kimi-code')).toBe(true) + }) + it('recognizes Qwen Code by its installed qwen executable', () => { expect(recognizeAgentProcess('/home/dev/.local/bin/qwen')).toEqual({ agent: 'qwen-code', diff --git a/src/shared/agent-session-wire.ts b/src/shared/agent-session-wire.ts index b7360bd0e62..bb222528532 100644 --- a/src/shared/agent-session-wire.ts +++ b/src/shared/agent-session-wire.ts @@ -68,6 +68,11 @@ export type AgentSessionBackgroundTaskState = { supportsTaskStop?: boolean } +export type AgentSessionTurnActivity = { + turnId: string + text: string +} + /** Backward paging is the client's normal read; 40 matches the page size the * mobile list renders without a visible fill-in. */ export const AGENT_SESSION_HISTORY_DEFAULT_LIMIT = 40 @@ -145,6 +150,10 @@ export type AgentSessionSubscribeEvent = fence: number handoff?: AgentSessionHandoffStatus backgroundTasks?: AgentSessionBackgroundTaskState | null + /** Omitted when unchanged; null clears a previous provider catalog. */ + commands?: AgentSessionSlashCommand[] | null + /** Latest provider-authored turn activity; optional for mixed-version hosts. */ + activity?: AgentSessionTurnActivity | null } | { type: 'batch' @@ -154,6 +163,10 @@ export type AgentSessionSubscribeEvent = fence?: number handoff?: AgentSessionHandoffStatus backgroundTasks?: AgentSessionBackgroundTaskState | null + /** Omitted when unchanged; null clears a previous provider catalog. */ + commands?: AgentSessionSlashCommand[] | null + /** Additive ephemeral state; it never creates or advances journal rows. */ + activity?: AgentSessionTurnActivity | null } | { type: 'reset' @@ -163,6 +176,9 @@ export type AgentSessionSubscribeEvent = fence: number handoff?: AgentSessionHandoffStatus backgroundTasks?: AgentSessionBackgroundTaskState | null + /** Omitted when unchanged; null clears a previous provider catalog. */ + commands?: AgentSessionSlashCommand[] | null + activity?: AgentSessionTurnActivity | null } | { type: 'end' } @@ -178,6 +194,13 @@ export type AgentSessionStatusSummary = { /** Null until the journal holds a persisted user or assistant message. */ status: StructuredAgentSessionProjectedStatus | null latestPrompt: string + /** Provider model in force for the next turn; absent until the host has read the options. */ + model?: string + /** The tool the running turn is inside. Absent unless `status` is 'working'. */ + toolName?: string + toolInput?: string + /** Preview of the newest assistant prose, so a settled row says what the agent said. */ + lastAssistantMessage?: string providerSession?: AgentProviderSessionMetadata updatedAt: number } @@ -305,6 +328,23 @@ export type AgentSessionModelOption = { efforts: AgentSessionOptionChoice[] } +/** One entry of the `/` menu the running provider reports for itself. `skill` + * marks a name the session loaded as a skill rather than a built-in command; + * commands the provider reserves for a terminal UI are already removed. */ +export type AgentSessionSlashCommand = { + name: string + kind: 'command' | 'skill' + /** Membership is authoritative, but this provider report did not classify the name. */ + kindUnspecified?: true +} + +/** The provider's own command surface, read per session. Additive read-only + * surface: a host that predates it answers `method_not_found`, and the client + * keeps rendering its curated catalog. */ +export type AgentSessionCommandsResult = { + commands?: AgentSessionSlashCommand[] +} + /** Provider-reported choices and effective next-turn values. Additive read-only * surface so older hosts can reject it without changing structured v1 writes. */ export type AgentSessionOptionsResult = { diff --git a/src/shared/constants.test.ts b/src/shared/constants.test.ts index ea41c8a6af9..9bf262ef45a 100644 --- a/src/shared/constants.test.ts +++ b/src/shared/constants.test.ts @@ -181,4 +181,9 @@ describe('MiniMax defaults', () => { expect(settings.minimaxGroupId).toBe('') expect(settings.minimaxUsageModels).toBe('general') }) + + it('defaults the MiniMax endpoint to overseas', () => { + const settings = getDefaultSettings('/tmp') + expect(settings.minimaxEndpoint).toBe('overseas') + }) }) diff --git a/src/shared/default-global-settings.ts b/src/shared/default-global-settings.ts index 313e9fc6a9a..c9b7a08928a 100644 --- a/src/shared/default-global-settings.ts +++ b/src/shared/default-global-settings.ts @@ -199,6 +199,7 @@ export function buildDefaultSettings(args: { opencodeWorkspaceId: '', minimaxGroupId: '', minimaxUsageModels: 'general', + minimaxEndpoint: 'overseas', geminiCliOAuthEnabled: false, agentCmdOverrides: {}, agentDefaultArgs: { ...DEFAULT_TUI_AGENT_ARGS }, diff --git a/src/shared/global-settings-types.ts b/src/shared/global-settings-types.ts index bc590a19a2e..bef153ba8c9 100644 --- a/src/shared/global-settings-types.ts +++ b/src/shared/global-settings-types.ts @@ -42,6 +42,9 @@ import type { WorktreeVisibilitySourcePreferences } from './repo-types' +/** MiniMax account region used to select the quota endpoint. */ +export type MiniMaxEndpoint = 'overseas' | 'cn' + export type WorktreeVisibilityDefaults = { /** Default for worktrees outside a recognized source. */ external?: ExternalWorktreeVisibility @@ -199,7 +202,7 @@ export type GlobalSettings = { openLinksInAppPreferencePrompted: boolean /** Opt-in: Shift+modifier click inverts openLinksInApp instead of always forcing the system browser. Off keeps the historical one-way escape hatch. */ openLinksInAppModifierInverts?: boolean - /** Show terminal link actions on plain click; off restores modifier-click-only terminal links. */ + /** Show link actions on plain click in the terminal and chat; off restores modifier-click-only terminal links. */ terminalLinkActionPopoverEnabled?: boolean /** Opt-in: open new coding-agent tabs in native chat instead of the raw terminal; optional for legacy settings. */ openAgentTabsInChatByDefault?: boolean @@ -360,6 +363,8 @@ export type GlobalSettings = { minimaxGroupId: string /** Comma-separated MiniMax model names to show in the status bar usage window. */ minimaxUsageModels: string + /** MiniMax account region; defaults to overseas for existing users. */ + minimaxEndpoint: MiniMaxEndpoint /** Extract OAuth credentials from the local Gemini CLI for rate-limit fetching. Off by default (explicit opt-in). */ geminiCliOAuthEnabled: boolean /** Per-agent CLI command overrides. A missing key means use the catalog default binary name. */ diff --git a/src/shared/mobile-notification-policy.test.ts b/src/shared/mobile-notification-policy.test.ts new file mode 100644 index 00000000000..a7ecff1ba70 --- /dev/null +++ b/src/shared/mobile-notification-policy.test.ts @@ -0,0 +1,47 @@ +import { describe, expect, it } from 'vitest' +import { allowsMobileNotification } from './mobile-notification-policy' +import { + MOBILE_PUSH_SOURCES, + MOBILE_PUSH_AGENT_STATES, + parseMobilePushRegistration +} from './mobile-push-contract' + +describe('notification delivery preferences', () => { + const filter = { sources: MOBILE_PUSH_SOURCES, agentStates: MOBILE_PUSH_AGENT_STATES } + it.each(['agent-task-complete', 'terminal-bell', 'plugin'])( + 'mirrors desktop settings for %s, but permits an explicit override', + (source) => { + const event = { source, desktopAllowed: false } + expect(allowsMobileNotification(filter, event)).toBe(false) + expect(allowsMobileNotification({ ...filter, followDesktop: true }, event)).toBe(false) + expect(allowsMobileNotification({ ...filter, followDesktop: false }, event)).toBe(true) + expect(allowsMobileNotification(filter, { source })).toBe(true) + } + ) + it('keeps bells independent of agent states and supports disabling them', () => { + expect( + allowsMobileNotification({ ...filter, agentStates: [] }, { source: 'terminal-bell' }) + ).toBe(true) + expect( + allowsMobileNotification( + { ...filter, sources: ['agent-task-complete'] }, + { source: 'terminal-bell' } + ) + ).toBe(false) + }) + it.each(['working', 'unknown'])('never presents %s agent activity', (agentState) => { + expect(allowsMobileNotification(filter, { source: 'agent-task-complete', agentState })).toBe( + false + ) + }) + it('preserves independent mode and silence through a desktop restart', () => { + expect( + parseMobilePushRegistration({ + registrationId: 'r', + platform: 'ios', + registeredAt: 1, + filter: { ...filter, followDesktop: false, sound: false } + })?.filter + ).toEqual({ ...filter, followDesktop: false, sound: false }) + }) +}) diff --git a/src/shared/mobile-notification-policy.ts b/src/shared/mobile-notification-policy.ts new file mode 100644 index 00000000000..d1c8a51475f --- /dev/null +++ b/src/shared/mobile-notification-policy.ts @@ -0,0 +1,34 @@ +import type { MobilePushAgentState, MobilePushFilter } from './mobile-push-contract' + +export type MobileNotificationPolicyEvent = { + source: string + agentState?: string + desktopAllowed?: boolean +} + +export function mapPushAgentState( + source: string, + state: string | undefined +): MobilePushAgentState | null | undefined { + if (source !== 'agent-task-complete') { + return null + } + if (state === 'blocked' || state === 'waiting' || state === 'needs-input') { + return 'needs-input' + } + return state === undefined || state === 'done' || state === 'finished' ? 'finished' : undefined +} + +export function allowsMobileNotification( + filter: MobilePushFilter, + event: MobileNotificationPolicyEvent +): boolean { + if (filter.followDesktop !== false && event.desktopAllowed === false) { + return false + } + if (!filter.sources.some((source) => source === event.source)) { + return false + } + const state = mapPushAgentState(event.source, event.agentState) + return state !== undefined && (state === null || filter.agentStates.includes(state)) +} diff --git a/src/shared/mobile-push-contract.ts b/src/shared/mobile-push-contract.ts new file mode 100644 index 00000000000..0e52a217571 --- /dev/null +++ b/src/shared/mobile-push-contract.ts @@ -0,0 +1,106 @@ +// Why: the desktop host, the push gateway, and the phone must agree on these +// exact strings. See docs/reference/mobile-push-contract.md. + +export const MOBILE_PUSH_SOURCES = ['agent-task-complete', 'terminal-bell', 'plugin'] as const +export type MobilePushSource = (typeof MOBILE_PUSH_SOURCES)[number] + +// The only two states a phone can be told about; the host maps its richer +// agent status onto them before it ever reaches the gateway. +export const MOBILE_PUSH_AGENT_STATES = ['needs-input', 'finished'] as const +export type MobilePushAgentState = (typeof MOBILE_PUSH_AGENT_STATES)[number] + +export const MOBILE_PUSH_PLATFORMS = ['ios', 'android'] as const +export type MobilePushPlatform = (typeof MOBILE_PUSH_PLATFORMS)[number] + +export const MOBILE_PUSH_APNS_ENVIRONMENTS = ['sandbox', 'production'] as const +export type MobilePushApnsEnvironment = (typeof MOBILE_PUSH_APNS_ENVIRONMENTS)[number] + +export type MobilePushFilter = { + followDesktop?: boolean + sound?: boolean + sources: readonly MobilePushSource[] + agentStates: readonly MobilePushAgentState[] +} + +/** Persisted on the paired DeviceEntry so a host restart can push without the phone re-registering. */ +export type MobilePushRegistration = { + registrationId: string + platform: MobilePushPlatform + filter: MobilePushFilter + registeredAt: number +} + +export type MobilePushRegisterInput = { + deviceId: string + platform: MobilePushPlatform + token: string + apnsEnvironment?: MobilePushApnsEnvironment + filter: MobilePushFilter +} + +export type MobilePushRegisterResult = + | { registered: true; registrationId: string } + | { + registered: false + // `registration_storage_failed`: the gateway accepted the token but the host + // could not persist it, so the phone must register again rather than believe + // a push route that does not exist. `throttled`: this device registered too + // often in the last minute; whatever it registered before still stands. + reason: + | 'gateway_unreachable' + | 'gateway_rejected' + | 'not_mobile' + | 'registration_storage_failed' + | 'throttled' + } + +function isStringMember(value: unknown, members: readonly T[]): value is T { + return typeof value === 'string' && (members as readonly string[]).includes(value) +} + +function parseFilter(value: unknown): MobilePushFilter | null { + if (!value || typeof value !== 'object') { + return null + } + const filter = value as Partial + if (!Array.isArray(filter.sources) || !Array.isArray(filter.agentStates)) { + return null + } + return { + ...(typeof filter.sound === 'boolean' ? { sound: filter.sound } : {}), + ...(typeof filter.followDesktop === 'boolean' ? { followDesktop: filter.followDesktop } : {}), + sources: filter.sources.filter((entry) => isStringMember(entry, MOBILE_PUSH_SOURCES)), + agentStates: filter.agentStates.filter((entry) => + isStringMember(entry, MOBILE_PUSH_AGENT_STATES) + ) + } +} + +/** + * Reads a persisted registration back. Returns undefined for anything an older or + * corrupted registry may hold, so a bad row degrades to "this device has no push" + * instead of failing the whole registry load. + */ +export function parseMobilePushRegistration(value: unknown): MobilePushRegistration | undefined { + if (!value || typeof value !== 'object') { + return undefined + } + const registration = value as Partial + const filter = parseFilter(registration.filter) + if ( + typeof registration.registrationId !== 'string' || + registration.registrationId.length === 0 || + !isStringMember(registration.platform, MOBILE_PUSH_PLATFORMS) || + !filter || + typeof registration.registeredAt !== 'number' || + !Number.isFinite(registration.registeredAt) + ) { + return undefined + } + return { + registrationId: registration.registrationId, + platform: registration.platform, + filter, + registeredAt: registration.registeredAt + } +} diff --git a/src/shared/native-chat-session-option-defaults.ts b/src/shared/native-chat-session-option-defaults.ts index 41f607ca12a..b913a0070a2 100644 --- a/src/shared/native-chat-session-option-defaults.ts +++ b/src/shared/native-chat-session-option-defaults.ts @@ -1,6 +1,7 @@ import type { AgentType } from './agent-status-types' import { sessionOptionValueIsValid } from './agent-session-option-catalog' import type { + NativeChatSessionOptionSettingsMutation, PersistedNativeChatSessionOptions, SessionOptionValue } from './native-chat-session-options' @@ -33,7 +34,7 @@ export function resolveNativeChatSessionOptionDefaults( * strings. Claude's `fastMode` is a boolean the durable `Record` * record cannot carry, and the providers' remaining keys are settable only * mid-session, never seeded at launch. */ -const STRUCTURED_LAUNCH_SEED_OPTION_IDS = ['model', 'effort'] as const +export const STRUCTURED_LAUNCH_SEED_OPTION_IDS = ['model', 'effort'] as const /** The saved selection a structured create seeds into its reservation, narrowed * to the wire-safe string subset the durable record and both providers accept. */ @@ -55,6 +56,41 @@ export function resolveStructuredLaunchSeedOptions( return Object.keys(seeded).length > 0 ? seeded : undefined } +/** Fold a settled batch of picks onto the durable record. A surface that must send the + * whole object back — rather than merging key by key — applies them in one pass so a + * later pick in the batch cannot drop an earlier one. */ +export function applyNativeChatSessionOptionPicks(args: { + persisted: PersistedNativeChatSessionOptions | null | undefined + agent: AgentType + picks: Extract['picks'] +}): PersistedNativeChatSessionOptions { + let persisted = args.persisted ?? {} + for (const pick of args.picks) { + persisted = updateNativeChatSessionOptionDefaults({ persisted, agent: args.agent, ...pick }) + } + return persisted +} + +/** Applies one host-owned delta to the latest record. Returning null means the + * authoritative model list found nothing to retire. */ +export function applyNativeChatSessionOptionSettingsMutation( + persisted: PersistedNativeChatSessionOptions | null | undefined, + mutation: NativeChatSessionOptionSettingsMutation +): PersistedNativeChatSessionOptions | null { + if (mutation.type === 'apply-picks') { + return applyNativeChatSessionOptionPicks({ + persisted, + agent: mutation.agent, + picks: mutation.picks + }) + } + const modelId = persisted?.[mutation.agent]?.model + if (!modelId || mutation.availableModelIds.includes(modelId)) { + return null + } + return clearNativeChatSessionOptionModel(persisted, mutation.agent) +} + /** Why: an authoritative probe proved this id gone, and a stale `model` is emitted * verbatim as a launch flag — grok exits fatally on an unknown one. Dropping only * `model` keeps the per-model option values for a later reselect. */ diff --git a/src/shared/native-chat-session-options.ts b/src/shared/native-chat-session-options.ts index 33567d39131..d2e94caaa7a 100644 --- a/src/shared/native-chat-session-options.ts +++ b/src/shared/native-chat-session-options.ts @@ -1,3 +1,5 @@ +import type { AgentType } from './agent-status-types' + export type SessionOptionValue = string | boolean export type SessionOptionSelectChoice = { @@ -75,6 +77,19 @@ export type PersistedNativeChatSessionOptions = Partial< > > +export type NativeChatSessionOptionSettingsMutation = + | { + type: 'apply-picks' + agent: AgentType + picks: readonly { + modelId: string + optionId: string + value: SessionOptionValue + adoptModelAsLaunchDefault?: boolean + }[] + } + | { type: 'clear-model-if-missing'; agent: AgentType; availableModelIds: readonly string[] } + export type SessionOptionsSurface = { getSnapshot(): SessionOptionDescriptor[] /** Apply an absolute target; known flip-only options use their tracked baseline. */ diff --git a/src/shared/native-chat-slash-commands.test.ts b/src/shared/native-chat-slash-commands.test.ts index bd9f7984551..32d3d2d8876 100644 --- a/src/shared/native-chat-slash-commands.test.ts +++ b/src/shared/native-chat-slash-commands.test.ts @@ -4,6 +4,8 @@ import { filterSlashCommands, getAgentSlashCommands, isSlashCommandDraft, + sessionReportedSkillNames, + sessionSlashCommandSuggestions, slashCommandDispatchText } from './native-chat-slash-commands' @@ -64,3 +66,27 @@ describe('dispatch vs completion text', () => { expect(applySlashSuggestion({ name: 'model' })).toBe('/model ') }) }) + +describe('a session that reports its own command surface', () => { + const reported = [ + { name: 'clear', kind: 'command' as const }, + { name: 'opsx:apply', kind: 'command' as const }, + { name: 'ref-oss', kind: 'skill' as const } + ] + + it('offers exactly the reported commands, described from the curated catalog', () => { + expect(sessionSlashCommandSuggestions('claude', reported)).toEqual([ + { name: 'clear', description: 'Clear conversation history' }, + { name: 'opsx:apply' } + ]) + }) + + it('does not resurrect a curated command the session never reported', () => { + const names = sessionSlashCommandSuggestions('claude', reported).map((c) => c.name) + expect(names).not.toContain('compact') + }) + + it('splits skills out for the picker to group on its own', () => { + expect(sessionReportedSkillNames(reported)).toEqual(['ref-oss']) + }) +}) diff --git a/src/shared/native-chat-slash-commands.ts b/src/shared/native-chat-slash-commands.ts index 99c0152577f..9337e9c78f7 100644 --- a/src/shared/native-chat-slash-commands.ts +++ b/src/shared/native-chat-slash-commands.ts @@ -4,6 +4,7 @@ // mirrored copy to drift, unlike the agent-specific parsers in src/shared that // Metro forces us to duplicate. +import type { AgentSessionSlashCommand } from './agent-session-wire' import type { AgentType } from './agent-status-types' export type SlashCommandSuggestion = { @@ -11,6 +12,7 @@ export type SlashCommandSuggestion = { name: string /** Optional one-line description for the suggestion row. */ description?: string + kindUnspecified?: true } // Best-effort, curated per-agent catalogs. The CLIs ship no machine-readable @@ -90,6 +92,36 @@ export function getAgentSlashCommands(agent: AgentType): readonly SlashCommandSu return COMMANDS_BY_AGENT[agent] ?? COMMON_COMMANDS } +/** The command rows for a session that reports its own `/` surface. The report + * is the authority on WHICH commands exist; the curated catalog above is kept + * only as the description source for the names both know about. Skills are + * excluded — they render in the picker's own skills group. */ +export function sessionSlashCommandSuggestions( + agent: AgentType, + reported: readonly AgentSessionSlashCommand[] +): readonly SlashCommandSuggestion[] { + const described = new Map( + getAgentSlashCommands(agent).map((command) => [command.name, command.description]) + ) + return reported + .filter((entry) => entry.kind === 'command') + .map((entry) => { + const description = described.get(entry.name) + return { + name: entry.name, + ...(description ? { description } : {}), + ...(entry.kindUnspecified ? { kindUnspecified: true as const } : {}) + } + }) +} + +/** Names the session reported as skills, in the order it reported them. */ +export function sessionReportedSkillNames( + reported: readonly AgentSessionSlashCommand[] +): readonly string[] { + return reported.filter((entry) => entry.kind === 'skill').map((entry) => entry.name) +} + /** Whether the draft is a slash command (leading `/`, ignoring leading space). * Slash drafts dispatch to the agent's own TUI and must NOT render an optimistic * user bubble — they are control actions, not chat turns. */ diff --git a/src/shared/notification-burst-cooldown.ts b/src/shared/notification-burst-cooldown.ts new file mode 100644 index 00000000000..e7616c57746 --- /dev/null +++ b/src/shared/notification-burst-cooldown.ts @@ -0,0 +1,37 @@ +const NOTIFICATION_COOLDOWN_MS = 5000 +const MAX_RECENT_NOTIFICATION_KEYS = 50 + +function pruneRecentNotifications(recentNotifications: Map, now: number): void { + if (recentNotifications.size <= MAX_RECENT_NOTIFICATION_KEYS) { + return + } + + for (const [key, ts] of recentNotifications) { + if (now - ts >= NOTIFICATION_COOLDOWN_MS) { + recentNotifications.delete(key) + } + } + + while (recentNotifications.size > MAX_RECENT_NOTIFICATION_KEYS) { + const oldest = recentNotifications.keys().next() + if (oldest.done) { + break + } + recentNotifications.delete(oldest.value) + } +} + +export function reserveNotificationCooldown( + recentNotifications: Map, + dedupeKey: string, + now: number +): boolean { + const lastSentAt = recentNotifications.get(dedupeKey) ?? 0 + if (now - lastSentAt < NOTIFICATION_COOLDOWN_MS) { + return false + } + recentNotifications.delete(dedupeKey) + recentNotifications.set(dedupeKey, now) + pruneRecentNotifications(recentNotifications, now) + return true +} diff --git a/src/shared/process-table-snapshot-reader.ts b/src/shared/process-table-snapshot-reader.ts index 962a24c7f48..06d3407f3b9 100644 --- a/src/shared/process-table-snapshot-reader.ts +++ b/src/shared/process-table-snapshot-reader.ts @@ -6,7 +6,9 @@ import { PS_ARGS, PS_MAX_BUFFER_BYTES, ProcessTableCaptureError, + SHELL_FOREGROUND_PS_ARGS, parseProcessTableRows, + parseShellForegroundRows, parseStrictProcessTableRows, type ProcessTableRow } from './process-table-snapshot' @@ -250,29 +252,47 @@ async function readLinuxProcessStartTimes( return result } +async function captureProcessTable(args: readonly string[]): Promise { + let stdout: string + try { + ;({ stdout } = await execFile('ps', [...args], { + encoding: 'utf-8', + timeout: PS_TIMEOUT_MS, + maxBuffer: PS_MAX_BUFFER_BYTES + })) + } catch (error) { + // A ceiling hit is truncation, not absence: name it in the domain vocabulary. + if ((error as { code?: unknown } | null)?.code === 'ERR_CHILD_PROCESS_STDIO_MAXBUFFER') { + throw new ProcessTableCaptureError('capture_truncated') + } + throw error + } + return assertWholeCapture(stdout) +} + const processTableReader = createProcessTableSnapshotReader({ runPs: async () => { - let stdout: string - try { - ;({ stdout } = await execFile('ps', [...PS_ARGS], { - encoding: 'utf-8', - timeout: PS_TIMEOUT_MS, - maxBuffer: PS_MAX_BUFFER_BYTES - })) - } catch (error) { - // A ceiling hit is truncation, not absence: name it in the domain vocabulary. - if ((error as { code?: unknown } | null)?.code === 'ERR_CHILD_PROCESS_STDIO_MAXBUFFER') { - throw new ProcessTableCaptureError('capture_truncated') - } - throw error - } - const baseCapture = createProcessTableCapture(assertWholeCapture(stdout)) + const stdout = await captureProcessTable(PS_ARGS) + const baseCapture = createProcessTableCapture(stdout) const startTimesByPid = await readLinuxProcessStartTimes(baseCapture.lenient()) return createProcessTableCapture(stdout, startTimesByPid, process.platform === 'linux') }, now: () => Date.now() }) +// Its own reader, not a column-set flag on the shared one: terminal-name resolution dominates +// macOS capture time, and a shell proof must not queue behind a full capture it cannot use. +const shellForegroundReader = createProcessTableSnapshotReader({ + runPs: async () => parseShellForegroundRows(await captureProcessTable(SHELL_FOREGROUND_PS_ARGS)), + now: () => Date.now() +}) + +export async function getFreshShellForegroundSnapshot(): Promise { + return process.platform === 'darwin' + ? shellForegroundReader.getFreshSnapshot() + : getFreshProcessTableSnapshot() +} + export async function getProcessTableSnapshot(): Promise { return (await processTableReader.getSnapshot()).lenient() } @@ -334,4 +354,5 @@ export async function getStrictProcessTableSnapshotWithAge(): Promise<{ export function resetProcessTableSnapshotForTests(): void { processTableReader.reset() + shellForegroundReader.reset() } diff --git a/src/shared/process-table-snapshot.ts b/src/shared/process-table-snapshot.ts index 3b3236079c4..81a669e6d97 100644 --- a/src/shared/process-table-snapshot.ts +++ b/src/shared/process-table-snapshot.ts @@ -38,6 +38,47 @@ export const CHEAP_PS_ARGS = ( : ['-axo', 'pid=,ppid=,pgid=,tpgid=,stat='] ) as readonly string[] +/** + * Shell-proof tier: job control plus argv, dropping only the columns the shell predicate never + * reads — macOS `tty=` (0.29s of the 0.34s on a 1,900-process Mac) and the start marker. Enough + * to name a pane's foreground process; never enough to correlate a pid across captures. + */ +export const SHELL_FOREGROUND_PS_ARGS = [ + '-axo', + 'pid=,ppid=,pgid=,tpgid=,stat=,command=' +] as readonly string[] + +/** + * Parse a {@link SHELL_FOREGROUND_PS_ARGS} capture, anchored to exactly those columns. + * Not {@link parseProcessTableRows}: with no `tty=` to absorb it, that parser's optional + * tty/start pair eats the head of an argv shaped `python 3 app.py`, and a command-less zombie + * row parses into a garbage pid/stat pair. + * + * Lenient per row like its siblings, but a capture yielding none is unreadable rather than a + * machine with no processes: the shell proof must not read that as "the shell is gone". + */ +export function parseShellForegroundRows(stdout: string): ProcessTableRow[] { + const rows: ProcessTableRow[] = [] + for (const rawLine of stdout.split(/\r?\n/)) { + const match = rawLine.trim().match(/^(\d+)\s+(\d+)\s+(-?\d+)\s+(-?\d+)\s+(\S+)\s+(.+)$/) + const pid = match ? Number(match[1]) : 0 + if (match && Number.isSafeInteger(pid) && pid > 0) { + rows.push({ + pid, + ppid: Number(match[2]), + pgid: Number(match[3]), + tpgid: Number(match[4]), + stat: match[5], + command: match[6] + }) + } + } + if (rows.length === 0) { + throw new ProcessTableCaptureError('empty_capture') + } + return rows +} + export type CheapProcessTableRow = { pid: number ppid: number diff --git a/src/shared/protocol-version.ts b/src/shared/protocol-version.ts index ef342d55d6a..fcdbfc44fad 100644 --- a/src/shared/protocol-version.ts +++ b/src/shared/protocol-version.ts @@ -180,6 +180,12 @@ export const AUTOMATION_OWNER_FENCING_UPDATE_REQUIRED_MESSAGE = 'Editing automations on this host requires a newer Orca server. Update the HUB and try again.' export const AUTOMATION_CREATE_IDEMPOTENCY_RUNTIME_CAPABILITY = 'automation.create-idempotency.v1' as const +// Why: registered on every build, so it is a STATIC capability. Mobile hides its +// background-notification settings entirely unless a paired host advertises it — +// an older host has no notifications.registerPush to call. +export const NOTIFICATION_DELIVERY_PREFERENCES_CAPABILITY = + 'notifications.delivery-preferences.v1' as const +export const NOTIFICATIONS_REMOTE_PUSH_RUNTIME_CAPABILITY = 'notifications.remote-push.v1' as const // Generic native clients include the CLI and must not claim Electron-only page // placement support. @@ -271,7 +277,9 @@ export const RUNTIME_CAPABILITIES = [ SKILL_DELETE_CAPABILITY, AUTOMATION_LIST_HOST_SCOPE_RUNTIME_CAPABILITY, AUTOMATION_OWNER_FENCING_RUNTIME_CAPABILITY, - AUTOMATION_CREATE_IDEMPOTENCY_RUNTIME_CAPABILITY + AUTOMATION_CREATE_IDEMPOTENCY_RUNTIME_CAPABILITY, + NOTIFICATIONS_REMOTE_PUSH_RUNTIME_CAPABILITY, + NOTIFICATION_DELIVERY_PREFERENCES_CAPABILITY ] as const export type RuntimeCapability = (typeof RUNTIME_CAPABILITIES)[number] | (string & {}) diff --git a/src/shared/rate-limit-types.test.ts b/src/shared/rate-limit-types.test.ts index 6d35c2d22fc..f11dc638d41 100644 --- a/src/shared/rate-limit-types.test.ts +++ b/src/shared/rate-limit-types.test.ts @@ -17,6 +17,7 @@ describe('RateLimitState', () => { minimax: null, grok: null, minimaxCookieConfigured: false, + minimaxApiKeyConfigured: false, grokAuthConfigured: false, claudeTarget: { runtime: 'host', wslDistro: null }, codexTarget: { runtime: 'host', wslDistro: null }, @@ -27,5 +28,6 @@ describe('RateLimitState', () => { expect(state.antigravity).toBeNull() expect(state.minimax).toBeNull() expect(state.minimaxCookieConfigured).toBe(false) + expect(state.minimaxApiKeyConfigured).toBe(false) }) }) diff --git a/src/shared/rate-limit-types.ts b/src/shared/rate-limit-types.ts index 83210fba2cc..5744e3e6749 100644 --- a/src/shared/rate-limit-types.ts +++ b/src/shared/rate-limit-types.ts @@ -131,6 +131,13 @@ export type RateLimitState = { * between snapshot refreshes. */ minimaxCookieConfigured: boolean + /** + * True when a MiniMax API key is persisted on disk. The key value itself + * never leaves main, so the renderer only sees this boolean. The status bar + * ORs it with the cookie flag to decide whether to keep the MiniMax bar + * visible across reloads. + */ + minimaxApiKeyConfigured: boolean /** True when main finds a Grok CLI session file (~/.grok/auth.json or GROK_HOME). */ grokAuthConfigured: boolean claudeTarget: RateLimitRuntimeTarget diff --git a/src/shared/shell-foreground-snapshot.test.ts b/src/shared/shell-foreground-snapshot.test.ts new file mode 100644 index 00000000000..e4cfa5f53a5 --- /dev/null +++ b/src/shared/shell-foreground-snapshot.test.ts @@ -0,0 +1,77 @@ +import { afterEach, beforeEach, expect, it, vi } from 'vitest' + +const { execFileMock } = vi.hoisted(() => ({ execFileMock: vi.fn() })) +vi.mock('node:child_process', () => ({ execFile: execFileMock })) + +import { + getFreshShellForegroundSnapshot, + getProcessTableSnapshot, + resetProcessTableSnapshotForTests +} from './process-table-snapshot-reader' +import { parseShellForegroundRows } from './process-table-snapshot' + +type Callback = (error: Error | null, result: { stdout: string; stderr: string }) => void +const platform = Object.getOwnPropertyDescriptor(process, 'platform')! +const shell = '100 99 100 100 Ss+ /bin/zsh -l' + +beforeEach(() => { + Object.defineProperty(process, 'platform', { value: 'darwin' }) + execFileMock.mockReset() + resetProcessTableSnapshotForTests() +}) +afterEach(() => Object.defineProperty(process, 'platform', platform)) + +it('answers concurrent shell proofs without waiting for a pending full capture', async () => { + let finishFull!: Callback + execFileMock.mockImplementation((_program, args: string[], _options, callback: Callback) => { + if (args[1]?.includes('tty=')) { + finishFull = callback + } else { + expect(args).toEqual(['-axo', 'pid=,ppid=,pgid=,tpgid=,stat=,command=']) + callback(null, { stdout: shell, stderr: '' }) + } + }) + const full = getProcessTableSnapshot() + const [first, second] = await Promise.all([ + getFreshShellForegroundSnapshot(), + getFreshShellForegroundSnapshot() + ]) + expect(first).toEqual([ + { pid: 100, ppid: 99, pgid: 100, tpgid: 100, stat: 'Ss+', command: '/bin/zsh -l' } + ]) + expect(second).toBe(first) + expect(execFileMock).toHaveBeenCalledTimes(2) + finishFull(null, { stdout: shell, stderr: '' }) + await full + await getFreshShellForegroundSnapshot() + expect(execFileMock).toHaveBeenCalledTimes(3) +}) + +it('requires a new capture after an earlier shell proof has started', async () => { + const callbacks: Callback[] = [] + execFileMock.mockImplementation((_program, _args, _options, callback: Callback) => { + callbacks.push(callback) + }) + const first = getFreshShellForegroundSnapshot() + await vi.waitFor(() => expect(callbacks).toHaveLength(1)) + const second = getFreshShellForegroundSnapshot() + callbacks[0]!(null, { stdout: shell, stderr: '' }) + await first + await vi.waitFor(() => expect(callbacks).toHaveLength(2)) + callbacks[1]!(null, { stdout: shell.replace('Ss+', 'Ss'), stderr: '' }) + expect((await second)[0]?.stat).toBe('Ss') +}) + +// With no `tty=` column to absorb them, the shared parser read `python`/`3` as tty/start. +it('keeps an argv whose second token is numeric', () => { + expect(parseShellForegroundRows('101 100 101 101 S+ /usr/bin/python 3 app.py')).toEqual([ + { pid: 101, ppid: 100, pgid: 101, tpgid: 101, stat: 'S+', command: '/usr/bin/python 3 app.py' } + ]) +}) + +it('rejects an unreadable shell capture', async () => { + execFileMock.mockImplementation((_program, _args, _options, callback: Callback) => { + callback(null, { stdout: '', stderr: '' }) + }) + await expect(getFreshShellForegroundSnapshot()).rejects.toThrow('empty_capture') +}) diff --git a/src/shared/structured-agent-session-coalescer.test.ts b/src/shared/structured-agent-session-coalescer.test.ts index 770b08308af..f5b9dbdec86 100644 --- a/src/shared/structured-agent-session-coalescer.test.ts +++ b/src/shared/structured-agent-session-coalescer.test.ts @@ -4,7 +4,8 @@ import { createStructuredAgentSessionEventCoalescer } from './structured-agent-s function batch( sequence: number, - backgroundTasks?: Extract['backgroundTasks'] + backgroundTasks?: Extract['backgroundTasks'], + activity?: Extract['activity'] ): Extract { return { type: 'batch', @@ -15,7 +16,8 @@ function batch( removedItemIds: [], submissions: [] }, - ...(backgroundTasks !== undefined ? { backgroundTasks } : {}) + ...(backgroundTasks !== undefined ? { backgroundTasks } : {}), + ...(activity !== undefined ? { activity } : {}) } } @@ -53,4 +55,17 @@ describe('structured agent session event coalescer', () => { expect(events).toHaveLength(1) expect(events[0]).toMatchObject({ backgroundTasks: null }) }) + + it('keeps only the latest ephemeral activity value', () => { + const events: AgentSessionSubscribeEvent[] = [] + const coalescer = createStructuredAgentSessionEventCoalescer((event) => events.push(event)) + + coalescer.push(batch(1, undefined, { turnId: 'turn-1', text: 'Thinking' })) + coalescer.push(batch(1, undefined, { turnId: 'turn-1', text: 'Checking the result' })) + coalescer.push(batch(1, undefined, null)) + coalescer.flush() + + expect(events).toHaveLength(1) + expect(events[0]).toMatchObject({ activity: null }) + }) }) diff --git a/src/shared/structured-agent-session-coalescer.ts b/src/shared/structured-agent-session-coalescer.ts index fe982d67a69..d3c8cacffb7 100644 --- a/src/shared/structured-agent-session-coalescer.ts +++ b/src/shared/structured-agent-session-coalescer.ts @@ -25,6 +25,9 @@ function mergeBatch( } return { type: 'batch', + ...(right.commands !== undefined || left.commands !== undefined + ? { commands: right.commands !== undefined ? right.commands : left.commands } + : {}), sessionId: right.sessionId, batch: { cursor: right.batch.cursor, @@ -43,6 +46,9 @@ function mergeBatch( ? right.backgroundTasks : (left.backgroundTasks ?? null) } + : {}), + ...(right.activity !== undefined || left.activity !== undefined + ? { activity: right.activity !== undefined ? right.activity : (left.activity ?? null) } : {}) } } diff --git a/src/shared/structured-agent-session-option-picks.test.ts b/src/shared/structured-agent-session-option-picks.test.ts new file mode 100644 index 00000000000..5a46cd4ebc5 --- /dev/null +++ b/src/shared/structured-agent-session-option-picks.test.ts @@ -0,0 +1,167 @@ +import { describe, expect, it } from 'vitest' +import { CODEX_SESSION_OPTION_CATALOG } from './agent-session-option-catalog-claude-codex' +import { + applyNativeChatSessionOptionPicks, + resolveStructuredLaunchSeedOptions, + updateNativeChatSessionOptionDefaults +} from './native-chat-session-option-defaults' +import type { PersistedNativeChatSessionOptions } from './native-chat-session-options' +import { + applyStructuredAgentSessionOptions, + createStructuredAgentSessionOptionState, + structuredAgentSessionOptionPicks +} from './structured-agent-session-options' + +function liveState(current: { model: string; effort?: string }) { + return applyStructuredAgentSessionOptions( + createStructuredAgentSessionOptionState('codex'), + CODEX_SESSION_OPTION_CATALOG, + { + models: [ + { + id: 'account-model', + label: 'Account Model', + isDefault: true, + defaultEffort: 'medium', + efforts: [ + { value: 'medium', label: 'Medium' }, + { value: 'high', label: 'High' } + ] + }, + { + id: 'other-model', + label: 'Other Model', + isDefault: false, + defaultEffort: 'low', + efforts: [ + { value: 'low', label: 'Low' }, + { value: 'medium', label: 'Medium' } + ] + } + ], + current + } + ) +} + +function persist( + picks: readonly { modelId: string; optionId: string; value: string }[] +): PersistedNativeChatSessionOptions { + return applyNativeChatSessionOptionPicks({ persisted: undefined, agent: 'codex', picks }) +} + +describe('structuredAgentSessionOptionPicks', () => { + it('pins the model an effort-only pick was chosen against', () => { + const picks = structuredAgentSessionOptionPicks(liveState({ model: 'account-model' }), { + effort: 'high' + }) + expect(picks).toEqual([{ modelId: 'account-model', optionId: 'effort', value: 'high' }]) + // Without the model the launch resolves nothing at all, so the effort would be dead. + expect(resolveStructuredLaunchSeedOptions(persist(picks), 'codex')).toEqual({ + model: 'account-model', + effort: 'high' + }) + }) + + it('remembers the effort the provider reconciled, not the one in force before', () => { + const state = liveState({ model: 'account-model', effort: 'high' }) + const picks = structuredAgentSessionOptionPicks(state, { + model: 'other-model', + effort: 'low' + }) + expect(picks).toEqual([ + { modelId: 'other-model', optionId: 'model', value: 'other-model' }, + { modelId: 'other-model', optionId: 'effort', value: 'low' } + ]) + expect(resolveStructuredLaunchSeedOptions(persist(picks), 'codex')).toEqual({ + model: 'other-model', + effort: 'low' + }) + }) + + it('reads the committed model rather than the record a deferred commit has not settled', () => { + // The caller passes pre-commit state: the record still tracks the old model. + const state = liveState({ model: 'account-model', effort: 'medium' }) + expect(structuredAgentSessionOptionPicks(state, { model: 'other-model' })).toEqual([ + { modelId: 'other-model', optionId: 'model', value: 'other-model' } + ]) + }) + + it('keeps a per-model effort so reselecting the old model restores its level', () => { + const persisted = persist([ + ...structuredAgentSessionOptionPicks(liveState({ model: 'account-model' }), { + effort: 'high' + }), + ...structuredAgentSessionOptionPicks(liveState({ model: 'account-model', effort: 'high' }), { + model: 'other-model', + effort: 'low' + }) + ]) + const reselected = updateNativeChatSessionOptionDefaults({ + persisted, + agent: 'codex', + modelId: 'account-model', + optionId: 'model', + value: 'account-model' + }) + expect(resolveStructuredLaunchSeedOptions(reselected, 'codex')).toEqual({ + model: 'account-model', + effort: 'high' + }) + }) + + it('writes nothing before the provider catalog lands', () => { + expect( + structuredAgentSessionOptionPicks(createStructuredAgentSessionOptionState('codex'), { + effort: 'high' + }) + ).toEqual([]) + }) + + it('drops ids a launch cannot seed back', () => { + expect( + structuredAgentSessionOptionPicks(liveState({ model: 'account-model' }), { + permissionMode: 'plan' + }) + ).toEqual([]) + }) +}) + +describe('applyNativeChatSessionOptionPicks', () => { + it('keeps a later pick in the batch from dropping an earlier one', () => { + const persisted = applyNativeChatSessionOptionPicks({ + persisted: undefined, + agent: 'codex', + picks: [ + { modelId: 'gpt-fast', optionId: 'model', value: 'gpt-fast' }, + { modelId: 'gpt-fast', optionId: 'effort', value: 'low' } + ] + }) + expect(resolveStructuredLaunchSeedOptions(persisted, 'codex')).toEqual({ + model: 'gpt-fast', + effort: 'low' + }) + }) + + it('leaves every other agent untouched', () => { + const persisted = applyNativeChatSessionOptionPicks({ + persisted: { claude: { model: 'opus', valuesByModel: { opus: { effort: 'high' } } } }, + agent: 'codex', + picks: [{ modelId: 'gpt-fast', optionId: 'effort', value: 'low' }] + }) + expect(resolveStructuredLaunchSeedOptions(persisted, 'claude')).toEqual({ + model: 'opus', + effort: 'high' + }) + expect(resolveStructuredLaunchSeedOptions(persisted, 'codex')).toEqual({ + model: 'gpt-fast', + effort: 'low' + }) + }) + + it('returns the record unchanged for an empty batch', () => { + expect( + applyNativeChatSessionOptionPicks({ persisted: undefined, agent: 'codex', picks: [] }) + ).toEqual({}) + }) +}) diff --git a/src/shared/structured-agent-session-options.ts b/src/shared/structured-agent-session-options.ts index d746a52025c..87c7e1ccfd2 100644 --- a/src/shared/structured-agent-session-options.ts +++ b/src/shared/structured-agent-session-options.ts @@ -13,6 +13,7 @@ import { setTrackedSessionOption, type NativeChatSessionOptionRecord } from './native-chat-session-option-state' +import { STRUCTURED_LAUNCH_SEED_OPTION_IDS } from './native-chat-session-option-defaults' import type { SessionOptionDescriptor, SessionOptionValue } from './native-chat-session-options' import type { AgentSessionOptionsResult } from './agent-session-wire' @@ -148,3 +149,44 @@ export function commitStructuredAgentSessionOptionValues( } return next } + +export type StructuredSessionOptionPick = { + modelId: string + optionId: string + value: string +} + +/** + * The picks a mutation must remember so the next launch starts where the user left off. + * Keyed off the same ids the launch seed reads back, so a pick this surface cannot + * re-seed is never written. + * + * Model and effort travel as a pair: a launch resolves a stored effort only under a + * stored model, so an effort-only pick adopts the model it was chosen against. Values + * come from what the provider committed, not what was requested — it reconciles an + * effort the newly selected model cannot run before reporting back. + * + * `state` may still be pre-commit: a changed model arrives in `committed`, and an + * unchanged one is already what the record tracks, so neither reading depends on the + * commit having landed. + */ +export function structuredAgentSessionOptionPicks( + state: StructuredAgentSessionOptionState, + committed: Readonly> +): StructuredSessionOptionPick[] { + if (!state.catalog) { + return [] + } + const committedModel = committed.model + const modelId = + typeof committedModel === 'string' && committedModel.trim() + ? committedModel + : resolveEffectiveNativeChatModelId(state.catalog, state.catalog.models, state.record) + if (!modelId) { + return [] + } + return STRUCTURED_LAUNCH_SEED_OPTION_IDS.flatMap((optionId) => { + const value = committed[optionId] + return typeof value === 'string' && value.trim() ? [{ modelId, optionId, value }] : [] + }) +} diff --git a/src/shared/structured-agent-session-projection.test.ts b/src/shared/structured-agent-session-projection.test.ts index 8bdce30577e..08d4de2fa0c 100644 --- a/src/shared/structured-agent-session-projection.test.ts +++ b/src/shared/structured-agent-session-projection.test.ts @@ -80,6 +80,144 @@ describe('structured agent session status projection', () => { }) }) + it('carries the running tool and the newest assistant prose the sidebar row shows', () => { + const ask = item('ask', 1, { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: 'look at the sidebar' }] + }) + const running = item('running', 2, { + kind: 'status', + text: 'Working', + turnLifecycle: { turnId: 'turn-1', state: 'running' } + }) + const said = item('said', 3, { + kind: 'message', + role: 'assistant', + blocks: [{ type: 'text', text: 'Reading the card first.' }] + }) + const tool = item('tool', 4, { + kind: 'tool-call', + name: 'Read', + input: { file_path: '/repo/src/WorktreeCard.tsx' }, + state: 'running' + }) + + expect(projectStructuredAgentSessionStatusSummary([ask, running, said, tool])).toEqual({ + status: 'working', + latestPrompt: 'look at the sidebar', + toolName: 'Read', + toolInput: '/repo/src/WorktreeCard.tsx', + lastAssistantMessage: 'Reading the card first.' + }) + }) + + it('clears the previous answer as soon as the next prompt is persisted', () => { + const firstAsk = item('first-ask', 1, { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: 'first task' }] + }) + const previousAnswer = item('previous-answer', 2, { + kind: 'message', + role: 'assistant', + blocks: [{ type: 'text', text: 'The first task is done.' }] + }) + const nextAsk = item('next-ask', 3, { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: 'second task' }] + }) + expect(projectStructuredAgentSessionStatusSummary([firstAsk, previousAnswer, nextAsk])).toEqual( + { + status: 'idle', + latestPrompt: 'second task' + } + ) + }) + + it('reports no tool line once the turn settles, even with an abandoned running call', () => { + const ask = item('ask', 1, { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: 'go' }] + }) + const abandoned = item('abandoned', 2, { + kind: 'tool-call', + name: 'Bash', + input: { command: 'sleep 600' }, + state: 'running' + }) + + expect(projectStructuredAgentSessionStatusSummary([ask, abandoned])).toEqual({ + status: 'idle', + latestPrompt: 'go' + }) + }) + + it('never adopts a running call from a turn older than the live one', () => { + const ask = item('ask', 1, { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: 'go' }] + }) + const abandoned = item('abandoned', 2, { + kind: 'tool-call', + name: 'Bash', + input: { command: 'sleep 600' }, + state: 'running' + }) + const running = item('running', 3, { + kind: 'status', + text: 'Working', + turnLifecycle: { turnId: 'turn-2', state: 'running' } + }) + + expect(projectStructuredAgentSessionStatusSummary([ask, abandoned, running])).toEqual({ + status: 'working', + latestPrompt: 'go' + }) + }) + + it('skips a tool-only assistant item to reach the newest prose', () => { + const ask = item('ask', 1, { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: 'go' }] + }) + const said = item('said', 2, { + kind: 'message', + role: 'assistant', + blocks: [{ type: 'text', text: 'Done — the card now aligns.' }] + }) + const wordless = item('wordless', 3, { + kind: 'message', + role: 'assistant', + blocks: [{ type: 'tool-call', name: 'Read', input: {} }] + }) + + expect( + projectStructuredAgentSessionStatusSummary([ask, said, wordless]).lastAssistantMessage + ).toBe('Done — the card now aligns.') + }) + + it('bounds the assistant preview at the shared agent-status preview cap', () => { + const ask = item('ask', 1, { + kind: 'message', + role: 'user', + blocks: [{ type: 'text', text: 'go' }] + }) + const rambled = item('rambled', 2, { + kind: 'message', + role: 'assistant', + blocks: [{ type: 'text', text: 'y'.repeat(AGENT_STATUS_MAX_FIELD_LENGTH * 40) }] + }) + + expect( + projectStructuredAgentSessionStatusSummary([ask, rambled]).lastAssistantMessage + ).toHaveLength(AGENT_STATUS_MAX_FIELD_LENGTH) + }) + it('bounds the wire prompt at the shared agent-status preview cap', () => { const pasted = item('pasted', 1, { kind: 'message', diff --git a/src/shared/structured-agent-session-projection.ts b/src/shared/structured-agent-session-projection.ts index 6a5f01ba9ea..7c55f2b8379 100644 --- a/src/shared/structured-agent-session-projection.ts +++ b/src/shared/structured-agent-session-projection.ts @@ -1,5 +1,17 @@ -import { normalizePromptField } from './agent-status-field-normalization' -import type { AgentJournalRenderItem } from './agent-session-journal-types' +import { + AGENT_STATUS_MAX_FIELD_LENGTH, + normalizeOptionalField, + normalizePromptField +} from './agent-status-field-normalization' +import type { + AgentJournalRenderItem, + AgentJournalToolCallItem +} from './agent-session-journal-types' +import { + AGENT_STATUS_TOOL_INPUT_MAX_LENGTH, + AGENT_STATUS_TOOL_NAME_MAX_LENGTH +} from './agent-status-types' +import { describeToolInput } from './native-chat-tool-summary' import type { NativeChatBlock, NativeChatMessage } from './native-chat-types' import { sha256 } from './sha256' @@ -163,6 +175,10 @@ export function projectStructuredAgentSessionStatus( return activeStructuredAgentSessionTurnId(items) ? 'working' : 'idle' } +function messageProse(blocks: readonly NativeChatBlock[]): string { + return blocks.flatMap((block) => (block.type === 'text' ? [block.text] : [])).join('\n') +} + /** The newest user prompt, as the sidebar quotes it. */ export function latestStructuredAgentSessionPrompt( items: readonly AgentJournalRenderItem[] @@ -170,24 +186,95 @@ export function latestStructuredAgentSessionPrompt( for (let index = items.length - 1; index >= 0; index -= 1) { const body = items[index]?.body if (body?.kind === 'message' && body.role === 'user') { - return body.blocks.flatMap((block) => (block.type === 'text' ? [block.text] : [])).join('\n') + return messageProse(body.blocks) } } return '' } +/** The newest assistant prose in the latest user turn. Tool-only assistant items + * are skipped; the user boundary clears prose from the preceding turn. */ +export function latestStructuredAgentSessionAssistantMessage( + items: readonly AgentJournalRenderItem[] +): string { + for (let index = items.length - 1; index >= 0; index -= 1) { + const body = items[index]?.body + if (body?.kind === 'message' && body.role === 'user') { + return '' + } + if (body?.kind === 'message' && body.role === 'assistant') { + const prose = messageProse(body.blocks) + if (prose.trim()) { + return prose + } + } + } + return '' +} + +/** The tool call the newest turn is still inside, or null when nothing is running. + * Scanning stops at the turn's own lifecycle row so an abandoned `running` call + * from an earlier crashed turn can never be reported as live work. */ +export function activeStructuredAgentSessionToolCall( + items: readonly AgentJournalRenderItem[] +): AgentJournalToolCallItem | null { + for (let index = items.length - 1; index >= 0; index -= 1) { + const body = items[index]?.body + if (body?.kind === 'status' && body.turnLifecycle) { + return null + } + if (body?.kind === 'tool-call' && body.state === 'running') { + return body + } + } + return null +} + +/** The activity fields a sidebar row shows beside the prompt, named as the agent-status + * entry names them so the client can hand them straight to a row. */ +export type StructuredAgentSessionStatusProjection = { + status: StructuredAgentSessionProjectedStatus | null + latestPrompt: string + /** Present only while a turn is running — see showsAgentToolPreview, which reads + * these on any state that carries them. */ + toolName?: string + toolInput?: string + lastAssistantMessage?: string +} + /** One projection shared by host and client: null status means "no turn yet", not idle. - * The prompt is bounded to the same preview every other agent-status row carries — a send - * admits 256 KB, and one status frame carries every retained session at once. */ + * Every text field is bounded to the same preview an agent-status row carries — a send + * admits 256 KB, and one status frame carries every retained session at once. The + * assistant line is bounded harder than the hook field it stands in for (a preview, not + * the 8 KB body): a streamed reply re-projects on every journal checkpoint, so the frame + * has to stay small even though the row only ever renders one line of it. */ export function projectStructuredAgentSessionStatusSummary( items: readonly AgentJournalRenderItem[] -): { status: StructuredAgentSessionProjectedStatus | null; latestPrompt: string } { +): StructuredAgentSessionStatusProjection { if (!hasPersistedStructuredAgentSessionTurn(items)) { return { status: null, latestPrompt: '' } } + const status = projectStructuredAgentSessionStatus(items) + const activeToolCall = status === 'working' ? activeStructuredAgentSessionToolCall(items) : null + const toolName = activeToolCall + ? normalizeOptionalField(activeToolCall.name, AGENT_STATUS_TOOL_NAME_MAX_LENGTH) + : undefined + const toolInput = activeToolCall + ? normalizeOptionalField( + describeToolInput(activeToolCall.input), + AGENT_STATUS_TOOL_INPUT_MAX_LENGTH + ) + : undefined + const lastAssistantMessage = normalizeOptionalField( + latestStructuredAgentSessionAssistantMessage(items), + AGENT_STATUS_MAX_FIELD_LENGTH + ) return { - status: projectStructuredAgentSessionStatus(items), - latestPrompt: normalizePromptField(latestStructuredAgentSessionPrompt(items)) + status, + latestPrompt: normalizePromptField(latestStructuredAgentSessionPrompt(items)), + ...(toolName ? { toolName } : {}), + ...(toolInput ? { toolInput } : {}), + ...(lastAssistantMessage ? { lastAssistantMessage } : {}) } } diff --git a/src/shared/structured-agent-session-reducer.test.ts b/src/shared/structured-agent-session-reducer.test.ts index d36b6717758..9aec45c8bf1 100644 --- a/src/shared/structured-agent-session-reducer.test.ts +++ b/src/shared/structured-agent-session-reducer.test.ts @@ -408,4 +408,98 @@ describe('structured agent session reducer', () => { expect(withoutCapability.backgroundTasks).toBeUndefined() }) + + it('projects ephemeral activity without changing transcript identity and clears it', () => { + const initial = reduceStructuredAgentSession(EMPTY_STRUCTURED_AGENT_SESSION, { + type: 'event', + event: { + type: 'snapshot', + sessionId: 'session-a', + fence: 1, + page: hydrationPage([item('message', 1)]) + } + }) + const active = reduceStructuredAgentSession(initial, { + type: 'event', + event: { + type: 'batch', + sessionId: 'session-a', + batch: { + cursor: initial.cursor!, + items: [], + removedItemIds: [], + submissions: [] + }, + activity: { turnId: 'turn-1', text: 'Checking the renderer' } + } + }) + + expect(active.activity).toEqual({ turnId: 'turn-1', text: 'Checking the renderer' }) + expect(active.items).toBe(initial.items) + + const cleared = reduceStructuredAgentSession(active, { + type: 'event', + event: { + type: 'batch', + sessionId: 'session-a', + batch: { + cursor: active.cursor!, + items: [], + removedItemIds: [], + submissions: [] + }, + activity: null + } + }) + + expect(cleared.activity).toBeNull() + expect(cleared.items).toBe(active.items) + }) + + it('retains same-epoch activity across a newer journal tail refresh', () => { + const active = reduceStructuredAgentSession(EMPTY_STRUCTURED_AGENT_SESSION, { + type: 'event', + event: { + type: 'snapshot', + sessionId: 'session-a', + fence: 1, + page: hydrationPage([item('first', 1)]), + activity: { turnId: 'turn-1', text: 'Checking the renderer' } + } + }) + const refreshed = reduceStructuredAgentSession(active, { + type: 'tail-page', + page: hydrationPage([item('latest', 2)]) + }) + + expect(refreshed.activity).toEqual({ turnId: 'turn-1', text: 'Checking the renderer' }) + }) +}) + +it('applies catalog-only checkpoints without replacing transcript or submission state', () => { + const state = reduceStructuredAgentSession(EMPTY_STRUCTURED_AGENT_SESSION, { + type: 'event', + event: { + type: 'snapshot', + sessionId: 'session-a', + fence: 1, + page: hydrationPage([item('one', 1)], [submission(1)]), + commands: [] + } + }) + const event = { + type: 'batch' as const, + sessionId: 'session-a', + fence: 1, + commands: [{ name: 'loaded', kind: 'skill' as const }], + batch: { cursor: state.cursor!, items: [], removedItemIds: [], submissions: [] } + } + const updated = reduceStructuredAgentSession(state, { type: 'event', event }) + expect(updated.commands).toEqual(event.commands) + expect(updated.items).toBe(state.items) + expect(updated.submissions).toBe(state.submissions) + expect(updated.cursor).toBe(state.cursor) + expect(reduceStructuredAgentSession(updated, { type: 'event', event })).toBe(updated) + const { commands: _commands, ...oldEvent } = event + expect(reduceStructuredAgentSession(updated, { type: 'event', event: oldEvent })).toBe(updated) }) diff --git a/src/shared/structured-agent-session-reducer.ts b/src/shared/structured-agent-session-reducer.ts index 88d41b2f8e5..6283df7d0a3 100644 --- a/src/shared/structured-agent-session-reducer.ts +++ b/src/shared/structured-agent-session-reducer.ts @@ -5,9 +5,11 @@ import type { } from './agent-session-journal-types' import type { AgentSessionBackgroundTaskState, + AgentSessionSlashCommand, AgentSessionHandoffStatus, AgentSessionHistoryPage, - AgentSessionSubscribeEvent + AgentSessionSubscribeEvent, + AgentSessionTurnActivity } from './agent-session-wire' export type StructuredAgentSessionState = { @@ -21,6 +23,8 @@ export type StructuredAgentSessionState = { error?: string handoff: AgentSessionHandoffStatus | null backgroundTasks?: AgentSessionBackgroundTaskState | null + commands?: AgentSessionSlashCommand[] | null + activity?: AgentSessionTurnActivity | null } export type StructuredAgentSessionAction = @@ -77,7 +81,8 @@ function replacePage( page: AgentSessionHistoryPage, fence: number, handoff?: AgentSessionHandoffStatus, - backgroundTasks?: AgentSessionBackgroundTaskState | null + backgroundTasks?: AgentSessionBackgroundTaskState | null, + activity?: AgentSessionTurnActivity | null ): StructuredAgentSessionState { return { epoch: page.epoch, @@ -88,6 +93,7 @@ function replacePage( hasOlder: page.hasOlder, status: 'ready', handoff: handoff ?? null, + activity: activity ?? null, ...(backgroundTasks !== undefined ? { backgroundTasks } : page.backgroundTasks !== undefined @@ -182,6 +188,8 @@ export function reduceStructuredAgentSession( hasOlder: action.page.hasOlder, status: 'ready', handoff: state.handoff, + ...(sameEpoch ? { commands: state.commands } : {}), + ...(sameEpoch && state.activity !== undefined ? { activity: state.activity } : {}), ...(action.page.backgroundTasks !== undefined ? { backgroundTasks: action.page.backgroundTasks } : state.backgroundTasks !== undefined @@ -205,7 +213,10 @@ export function reduceStructuredAgentSession( return state } if (event.type === 'snapshot' || event.type === 'reset') { - return replacePage(event.page, event.fence, event.handoff, event.backgroundTasks) + return { + ...replacePage(event.page, event.fence, event.handoff, event.backgroundTasks, event.activity), + commands: event.commands + } } if (state.epoch !== event.batch.cursor.epoch) { return state @@ -215,6 +226,7 @@ export function reduceStructuredAgentSession( } const backgroundTasks = event.backgroundTasks !== undefined ? event.backgroundTasks : state.backgroundTasks + const activity = event.activity !== undefined ? event.activity : state.activity const journalUnchanged = event.batch.items.length === 0 && event.batch.removedItemIds.length === 0 && @@ -224,7 +236,10 @@ export function reduceStructuredAgentSession( journalUnchanged && (event.fence === undefined || event.fence === state.fence) && (event.handoff === undefined || event.handoff === state.handoff) && + (event.commands === undefined || event.commands === state.commands) && backgroundTaskStatesEqual(backgroundTasks, state.backgroundTasks) && + activity?.turnId === state.activity?.turnId && + activity?.text === state.activity?.text && state.status === 'ready' && state.error === undefined ) { @@ -243,7 +258,9 @@ export function reduceStructuredAgentSession( status: 'ready', error: undefined, handoff: event.handoff ?? state.handoff, - ...(backgroundTasks !== undefined ? { backgroundTasks } : {}) + commands: event.commands !== undefined ? event.commands : state.commands, + ...(backgroundTasks !== undefined ? { backgroundTasks } : {}), + ...(activity !== undefined ? { activity } : {}) } } diff --git a/src/shared/structured-native-chat-launch-route.test.ts b/src/shared/structured-native-chat-launch-route.test.ts new file mode 100644 index 00000000000..48cb117fdf5 --- /dev/null +++ b/src/shared/structured-native-chat-launch-route.test.ts @@ -0,0 +1,115 @@ +/** + * The shared half of the launch route: the renderer's `resolveAgentLaunchRoute` and orchestration's + * worker-mode decision both answer from these, so a change here moves both surfaces at once. + */ + +import { describe, expect, it } from 'vitest' +import { STRUCTURED_AGENT_SESSION_RUNTIME_CAPABILITY } from './protocol-version' +import { + agentTabsDefaultToNativeChat, + prefersStructuredNativeChatByDefault, + resolveStructuredNativeChatSupport, + type StructuredNativeChatSupportInput +} from './structured-native-chat-launch-route' + +const ON = { + experimentalNativeChat: true, + openAgentTabsInChatByDefault: true, + experimentalStructuredNativeChat: true +} + +function support(overrides: Partial = {}) { + return resolveStructuredNativeChatSupport({ + agent: 'claude', + executionHostId: 'local', + platform: 'darwin', + hostCapabilities: [STRUCTURED_AGENT_SESSION_RUNTIME_CAPABILITY], + workspaceKind: 'git-worktree', + ...overrides + }) +} + +describe('the settings default', () => { + it('needs all three toggles for structured, and the first two for native chat', () => { + expect(prefersStructuredNativeChatByDefault(ON)).toBe(true) + expect(prefersStructuredNativeChatByDefault({ ...ON, experimentalNativeChat: false })).toBe( + false + ) + expect( + prefersStructuredNativeChatByDefault({ ...ON, openAgentTabsInChatByDefault: false }) + ).toBe(false) + expect( + prefersStructuredNativeChatByDefault({ ...ON, experimentalStructuredNativeChat: false }) + ).toBe(false) + expect(agentTabsDefaultToNativeChat({ ...ON, experimentalStructuredNativeChat: false })).toBe( + true + ) + }) + + it.each([null, undefined, {}])('reads %s as no preference', (settings) => { + expect(prefersStructuredNativeChatByDefault(settings)).toBe(false) + expect(agentTabsDefaultToNativeChat(settings)).toBe(false) + }) +}) + +describe('per-launch structured feasibility', () => { + it.each(['claude', 'codex'] as const)('supports a local %s launch', (agent) => { + expect(support({ agent })).toEqual({ supported: true }) + }) + + it.each([ + ['grok', { agent: 'grok' }, 'agent-without-structured-session'], + ['openclaude', { agent: 'openclaude' }, 'agent-without-structured-session'], + ['a draft prompt', { isDraftPrompt: true }, 'draft-prompt'], + ['a floating workspace', { workspaceKind: 'floating' }, 'floating-workspace'], + ['a custom TUI launch', { requiresTuiLaunchCustomization: true }, 'tui-launch-customization'], + ['an SSH host', { executionHostId: 'ssh:host-a' }, 'remote-execution-host'], + ['Codex on Windows', { agent: 'codex', platform: 'win32' }, 'codex-on-windows'], + ['a missing capability', { hostCapabilities: [] }, 'runtime-capability'] + ] as [string, Partial, string][])( + 'names %s as the blocker', + (_name, overrides, blocker) => { + expect(support(overrides)).toEqual({ supported: false, blocker }) + } + ) + + it('leaves a Windows Claude launch to the executing host', () => { + expect(support({ agent: 'claude', platform: 'win32' })).toEqual({ supported: true }) + }) + + it('blocks a WSL or repair-required project runtime', () => { + expect( + support({ + projectRuntime: { + status: 'resolved', + runtime: { + kind: 'wsl', + hostPlatform: 'wsl', + projectId: 'repo-1', + distro: 'Ubuntu', + reason: 'project-override', + cacheKey: 'wsl' + } + } + }) + ).toEqual({ supported: false, blocker: 'project-runtime' }) + expect( + support({ + projectRuntime: { + status: 'repair-required', + repair: { + projectId: 'repo-1', + preferredRuntime: { kind: 'wsl', distro: null }, + reason: 'wsl-distro-required', + source: 'project-override', + cacheKey: 'repair' + } + } + }) + ).toEqual({ supported: false, blocker: 'project-runtime' }) + }) + + it('supports a folder workspace without widening floating scope', () => { + expect(support({ workspaceKind: 'folder' })).toEqual({ supported: true }) + }) +}) diff --git a/src/shared/structured-native-chat-launch-route.ts b/src/shared/structured-native-chat-launch-route.ts new file mode 100644 index 00000000000..b97ffcc0dac --- /dev/null +++ b/src/shared/structured-native-chat-launch-route.ts @@ -0,0 +1,99 @@ +/** + * The one place that answers "should this launch be a structured native chat session?". + * + * Both launch surfaces call it. The renderer asks when a user opens an agent tab + * (`resolveAgentLaunchRoute`); orchestration asks when it dispatches a worker, because the mode is + * the user's own default rather than a per-call flag. Keeping the two halves — the settings default + * and the per-launch feasibility — here is what stops the second caller from growing a copy that + * drifts. + */ + +import { isAgentSessionHandleProvider } from './agent-session-provider-handle' +import type { GlobalSettings } from './global-settings-types' +import type { ProjectExecutionRuntimeResolution } from './project-execution-runtime' +import { STRUCTURED_AGENT_SESSION_RUNTIME_CAPABILITY } from './protocol-version' +import type { TuiAgent } from './tui-agent' + +export type NativeChatDefaultSettings = Pick< + GlobalSettings, + 'experimentalNativeChat' | 'experimentalStructuredNativeChat' | 'openAgentTabsInChatByDefault' +> + +/** Why a launch that the user's default asked to be structured cannot be. */ +export type StructuredNativeChatBlocker = + | 'agent-without-structured-session' + | 'draft-prompt' + | 'floating-workspace' + | 'tui-launch-customization' + | 'remote-execution-host' + | 'codex-on-windows' + | 'project-runtime' + | 'runtime-capability' + +export type StructuredNativeChatSupport = + | { supported: true } + | { supported: false; blocker: StructuredNativeChatBlocker } + +export type StructuredNativeChatSupportInput = { + agent: TuiAgent + executionHostId: string + platform: NodeJS.Platform + hostCapabilities: readonly string[] + workspaceKind?: 'git-worktree' | 'folder' | 'floating' + projectRuntime?: ProjectExecutionRuntimeResolution | null + /** A draft stays terminal-backed: the composer, not a turn, owns unsent text. */ + isDraftPrompt?: boolean + requiresTuiLaunchCustomization?: boolean +} + +/** The user's default for a new agent tab: native chat rather than the raw TUI. */ +export function agentTabsDefaultToNativeChat( + settings: Partial | null | undefined +): boolean { + return ( + settings?.experimentalNativeChat === true && settings?.openAgentTabsInChatByDefault === true + ) +} + +/** ...and specifically a structured native chat session rather than a terminal rendered as chat. */ +export function prefersStructuredNativeChatByDefault( + settings: Partial | null | undefined +): boolean { + return ( + agentTabsDefaultToNativeChat(settings) && settings?.experimentalStructuredNativeChat === true + ) +} + +export function resolveStructuredNativeChatSupport( + input: StructuredNativeChatSupportInput +): StructuredNativeChatSupport { + if (!isAgentSessionHandleProvider(input.agent)) { + return { supported: false, blocker: 'agent-without-structured-session' } + } + if (input.isDraftPrompt === true) { + return { supported: false, blocker: 'draft-prompt' } + } + if (input.workspaceKind === 'floating') { + return { supported: false, blocker: 'floating-workspace' } + } + if (input.requiresTuiLaunchCustomization === true) { + return { supported: false, blocker: 'tui-launch-customization' } + } + if (input.executionHostId !== 'local') { + return { supported: false, blocker: 'remote-execution-host' } + } + // Codex's Windows refusal is deliberate and settled elsewhere, so it stays a client-side answer. + // Claude's is measured by the executing host at create time (agentSession.createSupport) because + // only that host knows whether it can read a provider child's start time. + if (input.agent === 'codex' && input.platform === 'win32') { + return { supported: false, blocker: 'codex-on-windows' } + } + const projectRuntime = input.projectRuntime + if (projectRuntime?.status === 'repair-required' || projectRuntime?.runtime.kind === 'wsl') { + return { supported: false, blocker: 'project-runtime' } + } + if (!input.hostCapabilities.includes(STRUCTURED_AGENT_SESSION_RUNTIME_CAPABILITY)) { + return { supported: false, blocker: 'runtime-capability' } + } + return { supported: true } +} diff --git a/src/shared/structured-session-marker.ts b/src/shared/structured-session-marker.ts new file mode 100644 index 00000000000..36c90f432ee --- /dev/null +++ b/src/shared/structured-session-marker.ts @@ -0,0 +1,13 @@ +/** + * The marker a structured chat session's child carries when it has NO orchestration identity. + * + * It names nothing on purpose — no handle, no pane key, no session id, no token — so it grants no + * authority and cannot be replayed or impersonated. Its only job is to let a CLI verb that would + * otherwise GUESS an implicit terminal refuse instead: a structured session has no pane, so every + * guess resolves to a sibling, and `orchestration check` is destructive by default. + */ +export const ORCA_STRUCTURED_SESSION_ENV = 'ORCA_STRUCTURED_SESSION' + +export function isStructuredSessionWithoutIdentity(env: NodeJS.ProcessEnv = process.env): boolean { + return (env[ORCA_STRUCTURED_SESSION_ENV] ?? '').length > 0 +} diff --git a/src/shared/tui-agent-config.ts b/src/shared/tui-agent-config.ts index 664c0e39106..0bb2c35a040 100644 --- a/src/shared/tui-agent-config.ts +++ b/src/shared/tui-agent-config.ts @@ -233,7 +233,10 @@ const TUI_AGENT_CONFIG_SOURCE: Record = { ctrlEnterEncoding: 'csi-u' }, kimi: { + // Why: the `kimi` launcher runs as `kimi-code`, so foreground-process recognition never + // matches the agent without the alias — terminal reuse and `dispatch --inject` fail. detectCmd: 'kimi', + detectCmdAliases: ['kimi-code'], promptInjectionMode: 'stdin-after-start' }, 'mistral-vibe': { diff --git a/src/shared/tui-agent-launch-customization.ts b/src/shared/tui-agent-launch-customization.ts new file mode 100644 index 00000000000..b31edff8a01 --- /dev/null +++ b/src/shared/tui-agent-launch-customization.ts @@ -0,0 +1,43 @@ +import type { GlobalSettings } from './global-settings-types' +import type { TuiAgent } from './tui-agent' +import { getTuiAgentDefaultArgs, getTuiAgentDefaultEnv } from './tui-agent-launch-defaults' + +/** + * Whether the user configured a TUI launch this agent would lose outside a terminal. + * + * Shared rather than renderer-local because both launch surfaces have to answer it: the renderer + * routes such a launch back to the TUI, and orchestration falls a worker back to a PTY so the + * custom command, arguments and environment still apply. + */ +export function hasExplicitTuiLaunchCustomization( + settings: + | Partial> + | null + | undefined, + agent: TuiAgent +): boolean { + const configuredArgs = settings?.agentDefaultArgs?.[agent] + const configuredEnv = settings?.agentDefaultEnv?.[agent] + const defaultEnv = getTuiAgentDefaultEnv(agent) + const envIsCustomized = + configuredEnv !== undefined && + (Object.keys(configuredEnv).length !== Object.keys(defaultEnv).length || + Object.entries(configuredEnv).some(([key, value]) => defaultEnv[key] !== value)) + return ( + Boolean(settings?.agentCmdOverrides?.[agent]?.trim()) || + hasExplicitTuiAgentArgs(agent, configuredArgs) || + envIsCustomized + ) +} + +export function hasSemanticallyNonEmptyAgentArgs(value: string | null | undefined): boolean { + return Boolean(value?.trim()) +} + +export function hasExplicitTuiAgentArgs( + agent: TuiAgent, + value: string | null | undefined +): boolean { + const trimmed = value?.trim() ?? '' + return trimmed.length > 0 && trimmed !== getTuiAgentDefaultArgs(agent).trim() +} diff --git a/src/shared/worker-transcript-text.ts b/src/shared/worker-transcript-text.ts new file mode 100644 index 00000000000..6176a8b9b97 --- /dev/null +++ b/src/shared/worker-transcript-text.ts @@ -0,0 +1,60 @@ +/** + * The one plain-text rendering of a worker transcript message. + * + * The CLI prints `worker-read --source transcript` with it, and `terminal read` serves a structured + * worker's recent output through it, so a peer sees the same text either way. Shared rather than + * copied: two renderings would let the two surfaces disagree about what a tool call looked like. + */ + +import { + isSubagentGroupFallbackText, + subagentGroupFallbackText +} from './native-chat-subagent-summary' +import type { NativeChatMessage } from './native-chat-types' + +export function formatWorkerTranscriptMessage(message: NativeChatMessage): string { + // Every roster block is written beside a plain-text twin carrying the same + // sentence, for clients that cannot draw the block. Text surfaces are those + // clients, so they print the twin and drop the block — the mirror of the + // renderer, which draws the block and drops the twin. Either way the sentence + // prints once. + const hasTwin = message.blocks.some( + (block) => block.type === 'text' && isSubagentGroupFallbackText(block.text) + ) + const blocks = message.blocks.map((block) => { + if (block.type === 'text') { + return block.text + } + if (block.type === 'tool-call') { + return `[tool ${block.name}] ${safeJson(block.input)}` + } + if (block.type === 'tool-result') { + return `[tool result${block.isError ? ' error' : ''}] ${block.output}` + } + if (block.type === 'image-ref') { + return block.url ? `[image] ${block.url}` : `[image omitted]` + } + if (block.type === 'subagent-group') { + // Stand in for the block only when no twin is being printed: the wire + // admits a roster that arrived without one, and dropping that + // unconditionally would lose the sentence altogether. Presence, not a + // byte compare against a recomputed sentence — a roster from a newer + // build holds a state this build reads as `unverifiable`, so recomputing + // yields a different sentence and both would print. + return hasTwin ? null : `[subagents] ${subagentGroupFallbackText(block.agents)}` + } + // The journal deliberately admits block types this build does not know, and + // a newer remote host can send one over the wire. Degrade to a marker rather + // than reading fields off a shape that has none. + return '[unsupported block]' + }) + return `[${message.role}] ${blocks.filter((line) => line !== null).join('\n')}`.trimEnd() +} + +function safeJson(value: unknown): string { + try { + return JSON.stringify(value) + } catch { + return '[unserializable input]' + } +} diff --git a/src/shared/worktree/removal.ts b/src/shared/worktree/removal.ts index 8f5a6c4a4d2..59e5803eddb 100644 --- a/src/shared/worktree/removal.ts +++ b/src/shared/worktree/removal.ts @@ -15,6 +15,7 @@ export type WorktreeForceDeleteReason = | 'orphan-directory' | 'missing-registration' | 'unstopped-pty' + | 'running-agent-session' // Why: everything before this separator is the worktree id — a user-chosen filesystem path. // Only the detail after it is Orca's own wording, so verdict matchers anchor on the boundary @@ -32,6 +33,17 @@ export const UNSTOPPED_PTY_LIVE_DETAIL_PREFIX = 'still live:' // its own matcher the force affordance stayed hidden for the very case it was added for. export const WORKTREE_TEARDOWN_TIMEOUT_PREFIX = 'Timed out waiting for physical PTY teardown:' +// Why (#11960 again): a running agent SESSION blocks removal for the same reason an unstopped PTY +// does, and it needs its own prefix for the same reason the timeout above needed one — the desktop +// force affordance comes only from the classifier below, so a refusal with no matcher shows raw +// CLI wording and hides the Force Delete button. Matcher and hint stay in this file together. +export const RUNNING_AGENT_SESSION_REMOVAL_PREFIX = + 'Refusing to remove worktree with running agent sessions:' + +export function isRunningAgentSessionRemovalError(error: string): boolean { + return error.includes(RUNNING_AGENT_SESSION_REMOVAL_PREFIX) +} + export function isUnstoppedPtyRemovalError(error: string): boolean { return ( error.includes(UNSTOPPED_PTY_REMOVAL_PREFIX) || error.includes(WORKTREE_TEARDOWN_TIMEOUT_PREFIX) @@ -103,6 +115,12 @@ export function classifyWorktreeForceDeleteReason( if (isUnstoppedPtyRemovalError(error)) { return allowUnverifiedPtyStop ? null : 'unstopped-pty' } + // Same placement and the same reason: decided BEFORE the `force` guard, because an ordinary + // desktop delete already passes force:true to skip the dirty-file prompt and that says nothing + // about whether the user has waived closing a live agent session. Only the waiver itself does. + if (isRunningAgentSessionRemovalError(error)) { + return allowUnverifiedPtyStop ? null : 'running-agent-session' + } if (force) { return null } diff --git a/tests/e2e/cross-version-wire/cross-version-agent-session-wire.unit.test.ts b/tests/e2e/cross-version-wire/cross-version-agent-session-wire.unit.test.ts index 8a407a29d15..7f73f74c149 100644 --- a/tests/e2e/cross-version-wire/cross-version-agent-session-wire.unit.test.ts +++ b/tests/e2e/cross-version-wire/cross-version-agent-session-wire.unit.test.ts @@ -101,6 +101,11 @@ const STRUCTURED_CALLS: { hostMethod: 'readOptions', result: { current: { model: 'gpt-live' } } }, + { + method: 'agentSession.commands', + hostMethod: 'readCommands', + result: { commands: [{ name: 'clear', kind: 'command' }] } + }, { method: 'agentSession.reveal', hostMethod: 'revealSession', @@ -339,6 +344,7 @@ function structuredHostStub(): Record> { requestHandoff: vi.fn(async () => ({ status: { owner: 'native' } })), handoffStatus: vi.fn(async () => ({ owner: 'native' })), readOptions: vi.fn(async () => ({ models: [], current: { model: 'gpt-live' } })), + readCommands: vi.fn(() => ({ commands: [{ name: 'clear', kind: 'command' as const }] })), history: vi.fn(() => ({ ok: true, page: { items: [] } })), subscribe: vi.fn(() => () => undefined), subscribeStatus: vi.fn((subscriber: { emit: (event: unknown) => void }) => { diff --git a/tests/e2e/floating-workspace-reopen-webgl-recovery.spec.ts b/tests/e2e/floating-workspace-reopen-webgl-recovery.spec.ts index a1f48f1982d..cbd7da7d805 100644 --- a/tests/e2e/floating-workspace-reopen-webgl-recovery.spec.ts +++ b/tests/e2e/floating-workspace-reopen-webgl-recovery.spec.ts @@ -420,8 +420,9 @@ test.describe('floating workspace reopen WebGL recovery @headful', () => { expect(afterReopen.equals(baseline), 'reopened terminal should render clean glyphs').toBe(true) }) - test('window focus regain recovers the corrupted atlas (harness control)', async ({ - orcaPage + test('system resume recovers the corrupted atlas (harness control)', async ({ + orcaPage, + electronApp }) => { // Why: control proving the injected corruption is exactly the class the // existing recovery machinery heals — isolating the reopen gap above as a @@ -431,13 +432,17 @@ test.describe('floating workspace reopen WebGL recovery @headful', () => { const { baseline, corrupted } = shots! expect(corrupted.equals(baseline)).toBe(false) - await orcaPage.evaluate(() => { - window.dispatchEvent(new Event('focus')) + await electronApp.evaluate(({ BrowserWindow }) => { + const mainWindow = BrowserWindow.getAllWindows()[0] + if (!mainWindow) { + throw new Error('Orca window unavailable for system resume') + } + mainWindow.webContents.send('system:resumed') }) await settleRecoveryWindows(orcaPage) - const afterFocus = await screenshotFloatingTerminal(orcaPage) - console.log(`[floating-control] healedByFocus=${afterFocus.equals(baseline)}`) - expect(afterFocus.equals(baseline), 'window focus should heal the atlas').toBe(true) + const afterResume = await screenshotFloatingTerminal(orcaPage) + console.log(`[floating-control] healedByResume=${afterResume.equals(baseline)}`) + expect(afterResume.equals(baseline), 'system resume should heal the atlas').toBe(true) }) }) diff --git a/tests/e2e/helpers/electron-launch-args.ts b/tests/e2e/helpers/electron-launch-args.ts index 9128a5fb551..6868f48b083 100644 --- a/tests/e2e/helpers/electron-launch-args.ts +++ b/tests/e2e/helpers/electron-launch-args.ts @@ -11,6 +11,16 @@ export function getOrcaElectronLaunchArgs(mainPath: string, headful: boolean): s // Crash tests must not block later launches on AppKit's saved-window recovery dialog. return [...keychainArgs, appPath, '-ApplePersistenceIgnoreState', 'YES'] } + if (headful && process.platform === 'linux' && process.env.CI) { + // Hosted runners have no GPU; SwiftShader keeps WebGL assertions from silently skipping. + return [ + '--use-gl=angle', + '--use-angle=swiftshader', + '--enable-unsafe-swiftshader', + '--disable-gpu-sandbox', + appPath + ] + } if (headful || process.platform !== 'linux') { return [...keychainArgs, appPath] } diff --git a/tests/e2e/helpers/electron-launch-args.unit.test.ts b/tests/e2e/helpers/electron-launch-args.unit.test.ts index 636299fbc4e..c538d6f9a85 100644 --- a/tests/e2e/helpers/electron-launch-args.unit.test.ts +++ b/tests/e2e/helpers/electron-launch-args.unit.test.ts @@ -1,8 +1,31 @@ import { join } from 'node:path' -import { describe, expect, it } from 'vitest' +import { afterEach, describe, expect, it, vi } from 'vitest' import { getOrcaElectronLaunchArgs } from './electron-launch-args' describe('getOrcaElectronLaunchArgs', () => { + afterEach(() => vi.unstubAllGlobals()) + + it.each([ + ['linux', 'true', true, true], + ['linux', undefined, true, false], + ['linux', 'true', false, false], + ['darwin', 'true', true, false], + ['win32', 'true', true, false] + ] as const)( + 'scopes software WebGL to Linux CI headful launches: %s/%s/%s', + (platform, ci, headful, enabled) => { + vi.stubGlobal('process', { ...process, platform, env: { ...process.env, CI: ci } }) + const args = getOrcaElectronLaunchArgs(join('orca', 'out', 'main', 'index.js'), headful) + expect(args.includes('--use-gl=angle')).toBe(enabled) + expect(args.includes('--use-angle=swiftshader')).toBe(enabled) + expect(args.includes('--enable-unsafe-swiftshader')).toBe(enabled) + if (enabled) { + expect(args).toContain('--disable-gpu-sandbox') + expect(args).not.toContain('--disable-gpu') + } + } + ) + it('launches the package root that owns the compiled main entry', () => { const root = join('workspace', 'orca') const mainPath = join(root, 'out', 'main', 'index.js') diff --git a/tests/e2e/helpers/sidebar-project-visibility.ts b/tests/e2e/helpers/sidebar-project-visibility.ts new file mode 100644 index 00000000000..92843ef9461 --- /dev/null +++ b/tests/e2e/helpers/sidebar-project-visibility.ts @@ -0,0 +1,27 @@ +import { expect, type Page } from '@stablyai/playwright-test' + +export async function expectSidebarProjectVisible(page: Page, projectName: string): Promise { + const sidebar = page.getByRole('listbox', { name: 'Worktrees', exact: true }) + const label = sidebar.getByText(projectName, { exact: false }).first() + await sidebar.evaluate((element) => { + element.scrollTop = 0 + element.dispatchEvent(new Event('scroll', { bubbles: true })) + }) + await expect + .poll( + async () => { + if (await label.isVisible()) { + return true + } + // Virtualized project headers mount only as their scroll range enters the viewport. + await sidebar.evaluate((element) => { + element.scrollTop += Math.max(1, Math.floor(element.clientHeight * 0.8)) + element.dispatchEvent(new Event('scroll', { bubbles: true })) + }) + return false + }, + { message: `sidebar never rendered project ${projectName}`, intervals: [100] } + ) + .toBe(true) + await expect(label).toBeVisible() +} diff --git a/tests/e2e/orchestration-worker-settlement-release-cli.spec.ts b/tests/e2e/orchestration-worker-settlement-release-cli.spec.ts index f71668306b2..43b1f6a196f 100644 --- a/tests/e2e/orchestration-worker-settlement-release-cli.spec.ts +++ b/tests/e2e/orchestration-worker-settlement-release-cli.spec.ts @@ -284,6 +284,42 @@ test('compiled CLI rejects false completion then reconciles the dead retained wo db.close() } + const retained = invokeCompiledCli(userDataDir, [ + 'orchestration', + 'worker-release', + '--dispatch', + dispatch.result.dispatch!.id, + '--json' + ]) + expect(retained.status).toBe(0) + expect(JSON.parse(retained.stdout)).toMatchObject({ + ok: true, + result: { state: 'retained', reason: 'external_terminal', processAction: 'none' } + }) + const recovery = new Database(path.join(userDataDir, 'orchestration.db')) + try { + expect( + recovery + .prepare( + 'SELECT ownership_state, release_state FROM worker_terminal_resources WHERE owner_dispatch_id = ?' + ) + .get(dispatch.result.dispatch!.id) + ).toEqual({ ownership_state: 'external', release_state: 'retained' }) + // Seed the owned, abandoned recovery state after separately proving completion and external retention. + recovery + .prepare( + "UPDATE worker_terminal_resources SET ownership_state = 'owned', retained_reason = 'user_requested' WHERE owner_dispatch_id = ?" + ) + .run(dispatch.result.dispatch!.id) + recovery + .prepare( + "UPDATE worker_dispatches SET state = 'abandoned', stage = 'abandoned' WHERE dispatch_id = ?" + ) + .run(dispatch.result.dispatch!.id) + } finally { + recovery.close() + } + const released = invokeCompiledCli(userDataDir, [ 'orchestration', 'worker-release', diff --git a/tests/e2e/paired-client-hosted-browser.spec.ts b/tests/e2e/paired-client-hosted-browser.spec.ts index 46c2abac567..5470e59c0b3 100644 --- a/tests/e2e/paired-client-hosted-browser.spec.ts +++ b/tests/e2e/paired-client-hosted-browser.spec.ts @@ -151,13 +151,19 @@ async function waitForMirroredBrowserPage( worktreeId: string, url: string ): Promise { + let mirrored: MirroredBrowserPage | null = null await expect - .poll(() => findMirroredBrowserPage(page, worktreeId, url), { - timeout: 20_000, - message: `paired client never materialized ${url}` - }) + .poll( + async () => { + mirrored = await findMirroredBrowserPage(page, worktreeId, url) + return mirrored + }, + { + timeout: 20_000, + message: `paired client never materialized ${url}` + } + ) .not.toBeNull() - const mirrored = await findMirroredBrowserPage(page, worktreeId, url) if (!mirrored) { throw new Error(`Mirrored browser page disappeared for ${url}`) } diff --git a/tests/e2e/paired-cmd-j-host-qualified-tabs.spec.ts b/tests/e2e/paired-cmd-j-host-qualified-tabs.spec.ts index f0e62b048b1..6ac5b227350 100644 --- a/tests/e2e/paired-cmd-j-host-qualified-tabs.spec.ts +++ b/tests/e2e/paired-cmd-j-host-qualified-tabs.spec.ts @@ -1,4 +1,5 @@ import { errors } from '@stablyai/playwright-test' +import { encodePaletteIdentity } from '../../src/renderer/src/lib/palette-match/palette-ranking' import { expect, test } from './helpers/orca-app' import { createRuntimeDesktopPairingOffer, @@ -382,19 +383,45 @@ test('routes same-id browser and simulator Cmd-J rows to their owning paired hos worktreeId: seeded.sharedWorktreeId } ) + const remoteBrowserIdentity = encodePaletteIdentity([ + 'browser-page', + remoteHostId, + seeded.sharedWorktreeId, + seeded.remoteWorkspaceId, + seeded.remotePageId + ]) + const localBrowserIdentity = encodePaletteIdentity([ + 'browser-page', + 'local', + seeded.sharedWorktreeId, + 'browser-local', + 'page-local' + ]) + const remoteSimulatorIdentity = encodePaletteIdentity([ + 'simulator-tab', + remoteHostId, + seeded.sharedWorktreeId, + 'simulator-remote' + ]) + const localSimulatorIdentity = encodePaletteIdentity([ + 'simulator-tab', + 'local', + seeded.sharedWorktreeId, + 'simulator-local' + ]) expect(remoteBrowserAfterOpen.browserCount).toBe(2) expect(remoteBrowserAfterOpen.owner).toBe(remoteHostId) await input.fill('New Tab') - await expect( - palette.locator(`[cmdk-item][data-value="browser-page:${seeded.remotePageId}"]`) - ).toHaveCount(1) + await expect(palette.locator(`[cmdk-item][data-value="${remoteBrowserIdentity}"]`)).toHaveCount( + 1 + ) await expect(palette.getByText('Local browser proof', { exact: true })).toHaveCount(0) await testInfo.attach('cmd-j-host-qualified-browser.png', { body: await page.screenshot(), contentType: 'image/png' }) await expectSameIdCollisionIntact('remote browser page click') - await palette.locator(`[cmdk-item][data-value="browser-page:${seeded.remotePageId}"]`).click() + await palette.locator(`[cmdk-item][data-value="${remoteBrowserIdentity}"]`).click() await expect .poll(() => page.evaluate((worktreeId) => { @@ -433,11 +460,11 @@ test('routes same-id browser and simulator Cmd-J rows to their owning paired hos palette = page.getByRole('dialog', { name: 'Jump to...' }) input = palette.getByPlaceholder('Search chats, terminals, worktrees, settings, and actions...') await input.fill('local.example.test') - await expect(palette.locator('[cmdk-item][data-value="browser-page:page-local"]')).toHaveCount( + await expect(palette.locator(`[cmdk-item][data-value="${localBrowserIdentity}"]`)).toHaveCount( 1 ) await expectSameIdCollisionIntact('local browser page click') - await palette.locator('[cmdk-item][data-value="browser-page:page-local"]').click() + await palette.locator(`[cmdk-item][data-value="${localBrowserIdentity}"]`).click() await expect .poll(() => page.evaluate((worktreeId) => { @@ -469,7 +496,7 @@ test('routes same-id browser and simulator Cmd-J rows to their owning paired hos contentType: 'image/png' }) await expectSameIdCollisionIntact('remote simulator click') - await palette.locator('[cmdk-item][data-value="simulator-tab:simulator-remote"]').click() + await palette.locator(`[cmdk-item][data-value="${remoteSimulatorIdentity}"]`).click() await expect .poll(() => page.evaluate((worktreeId) => { @@ -496,9 +523,7 @@ test('routes same-id browser and simulator Cmd-J rows to their owning paired hos palette = page.getByRole('dialog', { name: 'Jump to...' }) input = palette.getByPlaceholder('Search chats, terminals, worktrees, settings, and actions...') await input.fill('Local emulator proof') - const localSimulatorRow = palette.locator( - '[cmdk-item][data-value="simulator-tab:simulator-local"]' - ) + const localSimulatorRow = palette.locator(`[cmdk-item][data-value="${localSimulatorIdentity}"]`) await expect(localSimulatorRow).toHaveCount(1) await expectSameIdCollisionIntact('local simulator click') await localSimulatorRow.click() diff --git a/tests/e2e/paired-remote-html-preview-local-render.spec.ts b/tests/e2e/paired-remote-html-preview-local-render.spec.ts index dd849091294..7351c5e0978 100644 --- a/tests/e2e/paired-remote-html-preview-local-render.spec.ts +++ b/tests/e2e/paired-remote-html-preview-local-render.spec.ts @@ -582,7 +582,10 @@ test('renders a paired HTML doc as a document browser tab while the host gains n return { before, after: document.activeElement?.tagName ?? null } }) console.log(`[preview-e2e] before-focus ${JSON.stringify(guestFocus)}`) - const confirmationTitle = page.getByRole('heading', { name: 'Open link to example.com?' }) + const confirmation = page.getByRole('dialog', { name: 'Open link to example.com?' }) + const confirmationTitle = confirmation.getByRole('heading', { + name: 'Open link to example.com?' + }) await expect .poll( async () => { @@ -601,8 +604,8 @@ test('renders a paired HTML doc as a document browser tab while the host gains n } ) .toBe(true) - await expect(page.getByText(EXTERNAL_LINK_URL, { exact: true })).toBeVisible() - await page.getByRole('button', { name: 'Cancel', exact: true }).click() + await expect(confirmation.getByText(EXTERNAL_LINK_URL, { exact: true })).toBeVisible() + await confirmation.getByRole('button', { name: 'Cancel', exact: true }).click() await expect(confirmationTitle).not.toBeVisible() const afterCancel = await readPairedHtmlPreviewInventory(page, inventoryArgs) expect({ @@ -620,7 +623,7 @@ test('renders a paired HTML doc as a document browser tab while the host gains n } await page.mouse.click(point.x, point.y) await expect(confirmationTitle).toBeVisible({ timeout: 30_000 }) - await page.getByRole('button', { name: 'Open link', exact: true }).click() + await confirmation.getByRole('button', { name: 'Open link', exact: true }).click() await expect .poll( async () => { diff --git a/tests/e2e/pr11346-selected-runtime-add.spec.ts b/tests/e2e/pr11346-selected-runtime-add.spec.ts index 6225f13c623..09630d277b3 100644 --- a/tests/e2e/pr11346-selected-runtime-add.spec.ts +++ b/tests/e2e/pr11346-selected-runtime-add.spec.ts @@ -1,3 +1,4 @@ +import { expectSidebarProjectVisible } from './helpers/sidebar-project-visibility' import { openSidebarProjectDialog } from './helpers/sidebar-project-dialog' import { rmSync } from 'node:fs' import path from 'node:path' @@ -727,7 +728,7 @@ async function runSelectedRuntimeAddJourney( ...fixture.nestedRepoPaths.map((repoPath) => path.basename(repoPath)) ]) { // Why: duplicate checkout names are disambiguated with a parent path. - await expect(client.page.getByText(projectName, { exact: false }).first()).toBeVisible() + await expectSidebarProjectVisible(client.page, projectName) } expect(await client.getDirectSshAttemptTargetIds()).toEqual([]) // Why: revealing the client must not leak into the HUB's window visibility. diff --git a/tests/e2e/remote-agent-session-focus-authority.spec.ts b/tests/e2e/remote-agent-session-focus-authority.spec.ts index 4fb4c3e7ae8..f8f9facd446 100644 --- a/tests/e2e/remote-agent-session-focus-authority.spec.ts +++ b/tests/e2e/remote-agent-session-focus-authority.spec.ts @@ -374,7 +374,19 @@ test('headed paired host keeps structured agent focus viewer-local @headful', as afterTabId: toWebTerminalSurfaceTabId(`${predecessorHostTabId}::${predecessorHostLeafId}`) }) const legacyWebTabId = toWebTerminalSurfaceTabId(legacy.terminal.tabId) - const mirroredLegacyGroup = legacy.mirror.tabGroups.find((group) => group.id === legacyGroup.id) + await expect + .poll( + async () => { + const order = await readRenderedTabOrder(client.page) + const anchorIndex = order.indexOf(predecessorWebTabId) + return anchorIndex === -1 ? [] : order.slice(anchorIndex, anchorIndex + 3) + }, + { timeout: 15_000, message: 'Legacy placement did not reach the rendered tab order' } + ) + .toEqual([predecessorWebTabId, legacyWebTabId, successorWebTabId]) + const mirroredLegacyGroup = ( + await readClientMirror(client.page, session.worktreeId) + ).tabGroups.find((group) => group.id === legacyGroup.id) if (!mirroredLegacyGroup) { throw new Error('Legacy placement mirrored group is missing') } diff --git a/tests/e2e/ssh-cold-hydration-gap-tab-seeding.spec.ts b/tests/e2e/ssh-cold-hydration-gap-tab-seeding.spec.ts index cdcdf38fa5a..93b16df0a11 100644 --- a/tests/e2e/ssh-cold-hydration-gap-tab-seeding.spec.ts +++ b/tests/e2e/ssh-cold-hydration-gap-tab-seeding.spec.ts @@ -128,8 +128,16 @@ function unblockRemoteWorkspaceGet( snapshotPath: string, saved: string ): void { - // Detached: a FIFO write blocks until the reader drains it, which must not stall the test. - spawnSync('docker', [ + const replacementPath = `${snapshotPath}.release` + const releaseScript = [ + `printf '%s' ${shellQuote(saved)} > ${shellQuote(replacementPath)}`, + `exec 3> ${shellQuote(snapshotPath)}`, + // Publish the complete file before the held reader can issue another snapshot read. + `mv -f ${shellQuote(replacementPath)} ${shellQuote(snapshotPath)}`, + `printf '%s' ${shellQuote(saved)} >&3`, + 'exec 3>&-' + ].join(' && ') + const release = spawnSync('docker', [ 'exec', '-d', target.containerName, @@ -137,8 +145,10 @@ function unblockRemoteWorkspaceGet( '--noprofile', '--norc', '-c', - `printf '%s' ${shellQuote(saved)} > ${snapshotPath} && rm -f ${snapshotPath} && printf '%s' ${shellQuote(saved)} > ${snapshotPath}` + releaseScript ]) + expect(release.error, 'failed to launch the snapshot release writer').toBeUndefined() + expect(release.status, release.stderr?.toString()).toBe(0) } async function connectAndSeedTabs( diff --git a/tests/e2e/ssh-docker-transport-drop-recovery.spec.ts b/tests/e2e/ssh-docker-transport-drop-recovery.spec.ts index c64761ede80..83a7ae66f65 100644 --- a/tests/e2e/ssh-docker-transport-drop-recovery.spec.ts +++ b/tests/e2e/ssh-docker-transport-drop-recovery.spec.ts @@ -163,10 +163,7 @@ test.describe('SSH transport drop recovery', () => { } }) - // #18018: local authority-aware recovery still loses the flooded pane's relay channel. - test.fixme('stays bounded when a disconnected shell floods its pty', async ({ - orcaPage - }, testInfo) => { + test('stays bounded when a disconnected shell floods its pty', async ({ orcaPage }, testInfo) => { test.slow() // Timeouts here are deliberately generous: this guards memory, not latency. A 48MB flood plus a // reconnect lands near 60s wall-clock end to end, so a 60s bind timeout was marginal and made diff --git a/tests/e2e/tabs.spec.ts b/tests/e2e/tabs.spec.ts index 2faabafc1cc..50d02e507a4 100644 --- a/tests/e2e/tabs.spec.ts +++ b/tests/e2e/tabs.spec.ts @@ -39,6 +39,22 @@ function tabLocator(page: Page, tabId: string) { return page.locator(`${SORTABLE_TAB}[data-tab-id="${tabId}"]`).first() } +async function closeTabFromTabBar(page: Page, tabId: string): Promise { + const tab = tabLocator(page, tabId) + await tab.hover() + await tab.getByRole('button', { name: /^Close tab /i }).click() + const confirmation = page.getByRole('dialog', { name: 'Stop running command?' }) + // A shell still starting under load may require the running-command confirmation. + await expect + .poll(async () => (await confirmation.isVisible()) || (await tab.count()) === 0, { + timeout: 5_000 + }) + .toBe(true) + if (await confirmation.isVisible()) { + await confirmation.getByRole('button', { name: 'Stop and Close', exact: true }).click() + } +} + /** Count rendered tabs in the tab bar (user-visible, not store-level). */ async function countRenderedTabs(page: Page): Promise { return page.locator(SORTABLE_TAB).count() @@ -73,6 +89,8 @@ test.describe('Tabs', () => { await waitForStartupWorktreeRefresh(orcaPage) await waitForActiveWorktree(orcaPage) await ensureTerminalVisible(orcaPage) + const initialTabId = (await getActiveTabId(orcaPage))! + await expect(tabLocator(orcaPage, initialTabId)).toBeVisible() }) /** @@ -94,7 +112,7 @@ test.describe('Tabs', () => { // Why: the "+" dropdown uses Radix , which exposes the // label text as the accessible name once the menu is open. const newTerminalMenuItem = orcaPage.getByRole('menuitem', { name: /New Terminal/i }).first() - await newTerminalMenuItem.click({ force: true }) + await newTerminalMenuItem.click() await expect(newTerminalMenuItem).toBeHidden({ timeout: 3_000 }) // Final assertion is on the rendered tab count — the tab bar itself must @@ -138,8 +156,7 @@ test.describe('Tabs', () => { await orcaPage.getByRole('button', { name: 'New tab' }).click({ force: true }) const newMarkdownMenuItem = orcaPage.getByRole('menuitem', { name: /New Markdown/i }).first() - await newMarkdownMenuItem.click({ force: true }) - await expect(newMarkdownMenuItem).toBeHidden({ timeout: 3_000 }) + await newMarkdownMenuItem.click() // Why: require an id that did not exist before the click, so an already-open // Markdown file can't satisfy the assertions (or be deleted by cleanup), and @@ -161,6 +178,7 @@ test.describe('Tabs', () => { const editor = orcaPage.locator('.rich-markdown-editor') await expect(editor).toBeVisible({ timeout: 25_000 }) + await expect(newMarkdownMenuItem).toBeHidden({ timeout: 3_000 }) await expect .poll(() => editor.evaluate((element) => document.activeElement === element), { @@ -519,12 +537,7 @@ test.describe('Tabs', () => { const tabsBefore = await countRenderedTabs(orcaPage) const activeId = await getActiveTabId(orcaPage) expect(activeId).not.toBeNull() - const activeTab = tabLocator(orcaPage, activeId!) - // Why: hover the tab first so the close button reveals its hover style. - // The button is interactive regardless but hovering matches real user - // behaviour and keeps click coordinates stable. - await activeTab.hover() - await activeTab.getByRole('button', { name: /^Close tab /i }).click() + await closeTabFromTabBar(orcaPage, activeId!) await expect .poll(() => countRenderedTabs(orcaPage), { @@ -562,9 +575,7 @@ test.describe('Tabs', () => { const activeTabBefore = await getActiveTabId(orcaPage) expect(activeTabBefore).not.toBeNull() - const activeTab = tabLocator(orcaPage, activeTabBefore!) - await activeTab.hover() - await activeTab.getByRole('button', { name: /^Close tab /i }).click() + await closeTabFromTabBar(orcaPage, activeTabBefore!) // Final DOM assertion: some *other* tab element now carries data-active. await expect diff --git a/tests/e2e/terminal-document-visibility-webgl-recovery.spec.ts b/tests/e2e/terminal-document-visibility-webgl-recovery.spec.ts index 1aa355b317e..d189df075c5 100644 --- a/tests/e2e/terminal-document-visibility-webgl-recovery.spec.ts +++ b/tests/e2e/terminal-document-visibility-webgl-recovery.spec.ts @@ -281,7 +281,7 @@ async function dispatchDocumentVisibilityCycle(page: Page): Promise { } test.describe('terminal document visibility WebGL recovery', () => { - test('preserves the WebGL atlas and keeps terminal text painted after document visibility resumes', async ({ + test('@headful preserves the WebGL atlas and keeps terminal text painted after document visibility resumes', async ({ electronApp, orcaPage }, testInfo) => { diff --git a/tests/e2e/terminal-foreground-redraw-freeze.spec.ts b/tests/e2e/terminal-foreground-redraw-freeze.spec.ts index fd6fad6f42d..05630042fae 100644 --- a/tests/e2e/terminal-foreground-redraw-freeze.spec.ts +++ b/tests/e2e/terminal-foreground-redraw-freeze.spec.ts @@ -339,7 +339,7 @@ function annotateMeasurement( } test.describe('Terminal foreground redraw freeze repro', () => { - test('Codex-style line rewrites request a visible row refresh', async ({ + test('@headful Codex-style line rewrites request a visible row refresh', async ({ orcaPage }, testInfo) => { await waitForSessionReady(orcaPage) diff --git a/tests/e2e/terminal-hangul-terminating-digit-native.spec.ts b/tests/e2e/terminal-hangul-terminating-digit-native.spec.ts index 4344f94adaf..9d0127d4df4 100644 --- a/tests/e2e/terminal-hangul-terminating-digit-native.spec.ts +++ b/tests/e2e/terminal-hangul-terminating-digit-native.spec.ts @@ -190,6 +190,8 @@ test.describe('Hangul terminating digit @headful', () => { })) console.log(`[digit-diag] ${JSON.stringify(launchDiagnostics)}`) if (INJECTOR === 'nested') { + expect(launchDiagnostics.ozonePlatform).toBe('wayland') + expect(launchDiagnostics.waylandDisplay).toBeTruthy() // Under Wayland the app's ready-to-show never fires here, so the window // stays hidden and the compositor has nothing to give keyboard focus to. await electronApp.evaluate(({ BrowserWindow }) => { diff --git a/tests/e2e/terminal-tab-switch-visual-restore.spec.ts b/tests/e2e/terminal-tab-switch-visual-restore.spec.ts index 938dc852179..f12e7a7a60a 100644 --- a/tests/e2e/terminal-tab-switch-visual-restore.spec.ts +++ b/tests/e2e/terminal-tab-switch-visual-restore.spec.ts @@ -794,7 +794,7 @@ test.describe('Terminal tab switch visual restore', () => { .toContain(marker) }) - test('keeps returned tab glyphs intact across tab switches', async ({ orcaPage }, testInfo) => { + test('@headful keeps returned tab glyphs intact across tab switches', async ({ orcaPage }, testInfo) => { // Why: screenshot equality catches WebGL atlas corruption on the tab being // resumed, not just stale cols/rows geometry checks. await waitForSessionReady(orcaPage) diff --git a/tests/e2e/terminal-webgl-atlas-budget.spec.ts b/tests/e2e/terminal-webgl-atlas-budget.spec.ts index 70b9a87c9d6..229140a3d35 100644 --- a/tests/e2e/terminal-webgl-atlas-budget.spec.ts +++ b/tests/e2e/terminal-webgl-atlas-budget.spec.ts @@ -431,7 +431,7 @@ async function runAtlasReplacementScenario(page: Page): Promise { test.describe.configure({ timeout: 120_000 }) - test('keeps shared glyph pages bindable through overflow and recovery @terminal-rendering-golden', async ({ + test('@headful keeps shared glyph pages bindable through overflow and recovery @terminal-rendering-golden', async ({ orcaPage }) => { await waitForActiveTerminalManager(orcaPage) @@ -451,7 +451,7 @@ test.describe('terminal WebGL atlas budget', () => { expect(result.pixelDiffAfterWipe).toBe(0) }) - test('rebuilds cached vertices after attaching a different shared atlas @terminal-rendering-golden', async ({ + test('@headful rebuilds cached vertices after attaching a different shared atlas @terminal-rendering-golden', async ({ orcaPage }) => { await waitForActiveTerminalManager(orcaPage) diff --git a/tests/e2e/worktree-jump-palette-filter.spec.ts b/tests/e2e/worktree-jump-palette-filter.spec.ts index 9c5d6146ba1..5b9a2cf9e29 100644 --- a/tests/e2e/worktree-jump-palette-filter.spec.ts +++ b/tests/e2e/worktree-jump-palette-filter.spec.ts @@ -1,4 +1,7 @@ import type { Locator, Page } from '@stablyai/playwright-test' +import type { ExecutionHostId } from '../../src/shared/execution-host' +import { getPaletteWorktreeIdentity } from '../../src/renderer/src/lib/palette-repo-resolution' +import { encodePaletteIdentity } from '../../src/renderer/src/lib/palette-match/palette-ranking' import { expect, test } from './helpers/orca-app' import { waitForActiveWorktree, waitForSessionReady } from './helpers/store' @@ -8,18 +11,29 @@ const REMOTE_WORKSPACE = 'E2E Palette Remote Workspace' const REMOTE_HOST = 'E2E Palette Builder' const SEARCH_PLACEHOLDER = 'Search chats, terminals, worktrees, settings, and actions...' -type PaletteFilterFixture = { localWorktreeId: string; remoteWorktreeId: string } +type PaletteFilterFixture = { + localRepoId: string + localWorktreeId: string + remoteWorktreeId: string + remoteHostId: ExecutionHostId +} async function seedPaletteFilterFixture(page: Page): Promise { return page.evaluate( - ({ localProject, remoteHost, remoteProject, remoteWorkspace }) => { + async ({ localProject, remoteHost, remoteProject, remoteWorkspace }) => { const store = window.__store if (!store) { throw new Error('window.__store is unavailable') } + const sourceRepo = store.getState().repos[0] + if ( + !sourceRepo || + !(await store.getState().updateRepo(sourceRepo.id, { displayName: localProject })) + ) { + throw new Error('Failed to persist the local palette fixture name') + } const state = store.getState() - const sourceRepo = state.repos[0] const sourceWorktree = Object.values(state.worktreesByRepo) .flat() .find((worktree) => worktree.repoId === sourceRepo?.id && !worktree.isArchived) @@ -31,13 +45,14 @@ async function seedPaletteFilterFixture(page: Page): Promise - project.sourceRepoIds.includes(sourceRepo.id) - ? { ...project, displayName: localProject } - : project - ) store.setState({ - repos: [ - ...state.repos.map((repo) => - repo.id === sourceRepo.id ? { ...repo, displayName: localProject } : repo - ), - remoteRepo - ], - projects, + repos: [...state.repos, remoteRepo], sshTargetLabels, worktreesByRepo: { ...state.worktreesByRepo, @@ -79,7 +81,12 @@ async function seedPaletteFilterFixture(page: Page): Promise { @@ -154,8 +165,15 @@ test.describe('Worktree jump-palette filters', () => { await waitForSessionReady(orcaPage) await waitForActiveWorktree(orcaPage) }) + test.afterEach(async ({ orcaPage }) => { + await orcaPage.evaluate(() => { + const store = window.__store?.getState() + store?.setFilterRepoIds([]) + store?.closeModal() + }) + }) - test('filters workspace results by host, intersects project selection, and resets on close', async ({ + test('filters results, intersects fields, and reseeds from the sidebar on reopen', async ({ orcaPage }) => { const fixture = await seedPaletteFilterFixture(orcaPage) @@ -166,10 +184,12 @@ test.describe('Worktree jump-palette filters', () => { await selectRemoteHost(orcaPage, true) await expect(filterTrigger(orcaPage)).toContainText('1') await expect(palette(orcaPage).getByLabel(`Remove filter ${REMOTE_HOST}`)).toBeVisible() - await expect(worktreeRow(orcaPage, fixture.remoteWorktreeId)).toBeVisible() + await expect( + worktreeRow(orcaPage, fixture.remoteWorktreeId, fixture.remoteHostId) + ).toBeVisible() await expect(worktreeRow(orcaPage, fixture.localWorktreeId)).toHaveCount(0) - // P2: host and project fields intersect, with the filter-specific empty state. + // P2: host and repository fields intersect, with the filter-specific empty state. await palette(orcaPage).getByPlaceholder(SEARCH_PLACEHOLDER).fill('') await filterTrigger(orcaPage).click() await palette(orcaPage).getByText('Projects', { exact: true }).click() @@ -183,7 +203,7 @@ test.describe('Worktree jump-palette filters', () => { palette(orcaPage).getByText('Clear the filter above, or widen it to more hosts and projects.') ).toBeVisible() - // P3: clear restores both rows; closing drops the ephemeral filter. + // P3: clear restores both rows; reopening replaces ephemeral state with the sidebar scope. await filterTrigger(orcaPage).click() await palette(orcaPage).getByRole('button', { name: 'Clear all' }).last().click() await filterTrigger(orcaPage).click() @@ -191,11 +211,36 @@ test.describe('Worktree jump-palette filters', () => { await searchFixtureWorkspaces(orcaPage, fixture) await selectRemoteHost(orcaPage) - await orcaPage.evaluate(() => window.__store?.getState().closeModal()) + await orcaPage.evaluate((repoId) => { + const store = window.__store?.getState() + store?.closeModal() + store?.setFilterRepoIds([repoId]) + }, fixture.localRepoId) await expect(palette(orcaPage)).toBeHidden() await openPalette(orcaPage) - await searchFixtureWorkspaces(orcaPage, fixture) - await expect(filterTrigger(orcaPage)).not.toContainText('1') + await palette(orcaPage).getByPlaceholder(SEARCH_PLACEHOLDER).fill('E2E Palette') + await expect(filterTrigger(orcaPage)).toContainText('1') + await expect(worktreeRow(orcaPage, fixture.localWorktreeId)).toBeVisible() + await expect(worktreeRow(orcaPage, fixture.remoteWorktreeId, fixture.remoteHostId)).toHaveCount( + 0 + ) + }) + + test('opens with the sidebar repository scope without widening it', async ({ orcaPage }) => { + const fixture = await seedPaletteFilterFixture(orcaPage) + await orcaPage.evaluate((repoId) => { + window.__store?.getState().setFilterRepoIds([repoId]) + }, fixture.localRepoId) + + await openPalette(orcaPage) + await palette(orcaPage).getByPlaceholder(SEARCH_PLACEHOLDER).fill('E2E Palette') + + await expect(filterTrigger(orcaPage)).toContainText('1') + await expect(palette(orcaPage).getByLabel(`Remove filter ${LOCAL_PROJECT}`)).toBeVisible() + await expect(worktreeRow(orcaPage, fixture.localWorktreeId)).toBeVisible() + await expect(worktreeRow(orcaPage, fixture.remoteWorktreeId, fixture.remoteHostId)).toHaveCount( + 0 + ) }) test('pressing Enter creates a worktree from a typed name', async ({ orcaPage }) => { @@ -221,11 +266,5 @@ test.describe('Worktree jump-palette filters', () => { await expect(createDialog).toBeHidden() // The page declined the press rather than consuming it, so it is still open. await expect(automationsHeading).toBeVisible() - - // Why a second press: with nothing layered above, the real page chrome must not - // trip the overlay check, or Escape would never close Automations again. - await orcaPage.keyboard.press('Escape') - - await expect(automationsHeading).toBeHidden() }) })