The account-switched restart card had two near-synonymous dismissive
buttons (Dismiss / Collapse) whose distinct behaviors weren't legible,
and body copy that attributed "keep working with the current account"
to Collapse — which only minimizes the notice.
Remap to intent-named actions:
- Dismiss -> "Keep old account" (clears the notice; pane stays on the
old account)
- Collapse -> "Not now" (minimize to the pill; decision deferred)
- Body no longer claims the minimize action commits you to an account.
Restart, the title, and the heading are unchanged. New English strings
get fresh auto keys so stale es/ja/ko/zh translations don't persist;
catalog re-synced.
* fix(updater): humanize update-error card and hide raw error behind Show details
Windows auto-update failures surfaced the raw electron-updater message as the
card headline — most visibly the PowerShell "Command failed: … Get-Authenticode
Signature …" dump when antivirus/EDR blocks the post-download signature check.
That reads as a crash, not an actionable state.
Classify each failure and lead with one plain-language sentence + the right
action, keeping the raw error one click away:
- New shared classifier (updater-windows-signature-check): distinguishes an
AV/EDR-blocked signature check (environment) from a genuine wrong-publisher
mismatch (security). The two are mutually exclusive so a real integrity
failure is never softened into "try again".
- UpdateCard: raw error moves behind a collapsed "Show details" toggle; adds a
security-stop variant (wrong publisher → no retry, "Open official releases")
and the AV-blocked variant ("Update Verification Blocked"). HTTP/2 and generic
paths keep their existing actions, now with the same details disclosure.
- Main process records a windows_signature_check_blocked lifecycle event so we
can size the affected Windows cohort in the field.
Verified each error scenario in a running Electron build (signature-check
blocked, wrong-publisher security stop, HTTP/2, generic, and the expanded
details view).
* fix(updater): make Show details a caret disclosure above the action row
Move the raw-error toggle directly above the Retry/Download buttons and give it
a rotating chevron; the Last error block now expands in place beneath the caret
instead of appearing above the summary, with the action buttons pinned below.
* fix(updater): prevent signature-check bypass
* fix(updater): surface retry start immediately
* feat(agents): pi session resume support
* fix(pi): require persisted session files for resume
* test(sleeping-agent): use non-resumable sentinel in malformed-record fixture
The 'drops malformed sleeping agent resume records' test used agent:'pi' as
its example of an unknown/non-resumable agent, expecting the record to be
dropped. This PR added 'pi' to RESUMABLE_TUI_AGENTS, making that fixture
valid and retained, so the toBeUndefined assertion broke. Switch the
malformed-case fixture to a genuinely non-resumable sentinel
('definitely-not-an-agent') so the drop-malformed path is still exercised;
no other assertions changed.
* Add durable resume identity for Pi sessions without fabricating turn sta
Pi's `session_start` hook now carries the session file needed to resume
a sleeping pane, but until now Orca either discarded it or treated it
as a fake status transition. Thread a `providerSessionOnly` envelope
through the hook listener, relay, main-process server, and renderer
store so resume identity (and its session-file-scoped equality/claim
key) can be persisted and replayed without emitting prompt telemetry
or a visible working/done row.
* Add durable resume identity for completed Pi sessions
Pi's agent_end hook marks a turn done, but the underlying TUI session
stays alive and resumable. Previously a `done` status wiped sleeping
records and launch config as if the session ended, so hibernation,
manual worktree sleep, and quit-capture all lost Pi's resume identity.
- Track a "live recovery" record for done-but-still-resumable Pi
sessions, exempting it from the usual done-state cleanup paths in
agent-status.ts and agent-hibernation-planner.ts
- Gate providerSessionOnly rows and sleeping-agent schema records on
actual resumability (getAgentResumeArgv) instead of trusting the
presence of a provider session
- Wait for Pi to persist its session file before advertising resume
metadata, and treat `/reload` as a non-terminal event so it doesn't
clobber visible status
- Extend SSH relay envelopes to carry providerSessionOnly so remote
hosts get the same behavior
* Add explicit periodic/quit mode to sleeping-agent session capture
Split captureAllSleepingAgentSessions into 'periodic' and 'quit' modes
so a background checkpoint can no longer downgrade a confirmed-quit
record or promote a completed Pi session without an authoritative
transcript path. Updates all call sites and tests accordingly.
* Use normalizeAgentStatusPayload for default pi status
Remove unnecessary JSON.stringify wrapper and call the appropriate normalization function directly.
---------
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
Translate 105 ko.json entries that were still identical to their English
source after the machine bootstrap — dialog bodies, settings descriptions,
shortcut recorder labels, Windows firewall notices, Jira connection copy,
crash-report notices, and source-control panel text.
Intentionally left in English: brand/product names, command and config
examples, CSS/selector strings used by animated visuals, and lowercase
concatenation fragments whose word order cannot be reordered safely.
* refactor(native-chat): extract option appliers and improve toggle UI
- Extract setOption apply logic to native-chat-session-option-apply.ts
- Add invokeAction method for toggle-only options without tracked baseline
- Replace checkbox UI with On/Off radio groups for boolean options
- Display option values only in pills (remove redundant label prefixes)
- Serialize concurrent applies so later dispatches win in order
* Remove focus-visible border from composer container
The container uses a steady hairline border (no focus/click flash), with focus-visible styling delegated to the inner textarea. The container is a layout wrapper, not a focus target.
* Prevent option commits on model switch during dispatch
When a user changes an option and the model switches before dispatch
completes, committing the stale option would overwrite state under
the new model. Guard by capturing the baseline option state before
dispatch and validating it hasn't changed post-dispatch.
* Use invokeAction for model option in NativeChatComposer test
Codex model is an agent picker mid-session, so setOption rejects.
Update test to use invokeAction to match the actual UI behavior.
* feat(keybindings): swap tab-switch chords to the common convention for new users
New installs now get the widespread mapping — Mod+Shift+[ / ] cycles across
all tabs, Mod+Alt+[ / ] cycles within the active tab type. Pre-existing installs
keep today's mapping: a one-time cohort seed (frozen on first launch via the
fileExistedOnLoad signal, mirroring the telemetry migration) pins the legacy
chords into keybindings.json, skipping any action the user already customized.
- shared registry: swap the four tab.*SameType / tab.*AllTypes defaults; export
LEGACY_TAB_SWITCH_BINDINGS for the seed
- persistence: migrateTabSwitchKeybindings freezes the existing-vs-fresh cohort
(tabSwitchKeybindingSeed = pending | done)
- keybinding-file: seedLegacyTabSwitchBindings writes the legacy pins into the
active-platform section so Settings reset still works
- refresh stale default-chord comments
- tests for the swapped defaults, the seed (fresh/existing/customized/idempotent),
and the cohort migration
* test(keybindings): prove existing-user parity + make the seed strictly per-action
seedLegacyTabSwitchBindings now pins each un-customized action individually
instead of skipping all four when any one is customized. A partially-customized
existing user keeps their rebound action AND the pre-swap default on the rest;
no existing user's behavior changes. The skip check keys on this platform's
effective overrides so a foreign-platform-only override can't leave the active
platform on a new default.
Adds keybinding-service.test.ts: constructs a real KeybindingService and asserts
effective bindings + real keystroke matching for both cohorts across darwin/
linux/win32, plus partial-customization, idempotency, and seed-failure retry.
* fix(keybindings): preserve legacy files during tab shortcut seed
* fix(keybindings): preserve valid pre-swap overrides
* feat(terminal): add flag-gated render-desync sentinel for WebGL panes
Detects the buffer-clean/render-stale glyph garble class in the field: per
visible WebGL pane, compare the cells the xterm buffer says hold glyphs
against the ink actually present on the canvas, sampled in the same task as
a forced synchronous redraw so a divergence proves the render model/atlas is
wrong rather than a missed present. A trip requires the same screen cells to
stay divergent across three samples (real desync is pinned; scroll lag moves),
then records a webgl-render-desync breadcrumb, stashes evidence (canvas PNG +
buffer text) for bug reports, and runs the same shared-atlas recovery a tab
reveal performs, so a stuck-garbled pane self-heals within seconds.
Off by default; arm on any build via
localStorage.setItem('orca:render-desync-sentinel', '1') and reload.
* fix(terminal): invalidate glyph cache on atlas replacement
Reproduce the WebGL atlas identity mismatch with two live terminals and force cached geometry to rebuild whenever a different shared atlas is attached. Persist flag-gated render-desync evidence and retain the investigation tooling used to validate the field signature.
* fix(terminal): harden render desync diagnostics
* docs(reliability): clarify Linux WebGL evidence gap
* feat(settings): unify skill card status and add Linear agent-skill section
Show the freshness-aware status pill ("Up to date" / "Update available")
on the Orchestration skill card, matching the Computer Use card — it was
the only installed-skill card still limited to a presence-only "Installed"
pill because it omitted the freshnessSkillName prop.
Add a Linear capability section under Settings → Agents that only appears
once the Linear provider is connected. It mirrors the other skill panes:
an orca-linear skill card (Update / Re-check + freshness pill) and a "How
to use it" grid of copyable /orca-linear prompts.
The Orchestration and Linear "How to use it" surfaces now share extracted
SkillUsageExamplesSection / SkillUsageExampleDialog components (the old
orchestration-only dialog is renamed/generalized with a slashCommand prop).
The nav registry and Settings page gate the Linear section on one shared
useLinearProviderConnected() selector so the sidebar entry and the section
appear and disappear together.
* fix(settings): tighten Linear skill section behavior
* fix(settings): preserve Linear freshness and dialog translations
* fix(linear): migrate legacy skill installs
* Revert "fix(linear): migrate legacy skill installs"
This reverts commit 3ff45f9245.
* fix(settings): show Linear freshness in sidebar
* fix(settings): type Linear navigation target
* Replace assistant-prose heuristic with explicit turn lifecycle markers
Extract provider-authored turn boundaries (completion, interruption) directly
from Claude/Codex transcripts so the chat view knows when work ends without
guessing from message presence. Reconciles live hook state with transcript
lifecycle: when a terminal boundary lands, it settles a dropped Stop hook
instead of letting prose mislead the UI into showing 'working' after done.
* fix(review): cover Claude terminal stop_reasons and RPC lifecycle frames
Treat max_tokens/stop_sequence/refusal as completed markers so capable hosts
do not stay working after a dropped Stop, and assert lifecycle payloads on
runtime subscribe/read frames plus mid-turn non-terminal stop_reason cases.
* test(native-chat): clarify that lifecycle field is optional
Add type assertion and comment documenting that lifecycle field is optional and can be omitted in truncation-gating test fixtures.
* fix(native-chat): settle status on interruption despite working subagent
When Claude's turn is explicitly interrupted, the session should show ready
immediately — even if background subagents are still running. Add an
interruption check before consulting the hook's working-subagents flag so
interruptions take precedence. Also normalize omitted lifecycle timestamps
to null instead of leaving them undefined, and add test coverage for both
cases.
* fix(native-chat): settle loading spinner on explicit turn boundaries
Explicit transcript turn-lifecycle markers now fully replace the prose-fallback
settlement path. Remove the now-unused `turnLifecycleCapable` flag and wire
lifecycle to suppress spinner even when hook status lingers. Refine Claude
lifecycle detection to distinguish terminal stops from mid-turn tool_use rows,
exclude harness noise from new-generation detection, and apply clock-skew slack
over SSH/relay. Serialize PTY sends per line to prevent rapid prompts from gluing
before Enter, clearing unsubmitted input on cancel. Update working suppression to
detect epoch rollovers so interrupt+next-turn without a ready gap resets the
spinner correctly.
* fix(github): attribute GitHub API outages instead of blank/"failed" states
When GitHub's API is unreachable (5xx outage, network, or rate limit), Orca
showed no PR data with no explanation, so it read as an Orca bug rather than a
GitHub-side problem.
- Add a shared classifier (classifyGitHubUnavailable) reused by the main
process and the renderer so every surface attributes an outage identically.
A live outage returns HTTP 5xx, which the PR-refresh classifier previously
had no branch for (fell through to the un-attributed "refresh failed").
- Right-sidebar Checks panel: show GitHub-attributed copy in the error
empty-state, plus an inline banner over stale cached PR data so an outage
doesn't look like a normal (silently out-of-date) panel.
- Tasks/PR-list page: replace the vague "N of M projects failed to load" with
a GitHub-attributed banner when the failure is a reachability problem.
Copy names GitHub as the source and reassures it isn't an Orca problem, with no
status-page link. Stays GitHub-scoped so GitLab/other providers aren't
mislabeled.
* fix(github): keep outage attribution accurate
* fix(github): preserve outage attribution edge cases
* fix(github): preserve Tasks outage attribution
* fix(github): avoid false outage attribution
* fix(runtime): tolerate absent browser certificate state
* fix(ui): preserve exhaustive optional state handling
* fix(github): preserve outage attribution for combined queries
* fix(github): preserve runtime failure attribution
* chore: restore unrelated UI files to main (out of scope)
native-chat-session-option-labels.ts and skill-freshness-group.tsx switch
tweaks were unrelated to GitHub API outage attribution — they fix pre-existing
switch-exhaustiveness lint on main, which this PR's CI (oxlint) doesn't gate on.
Restore them to origin/main so this PR's diff stays focused; the exhaustiveness
cleanup belongs in its own change. (sync-runtime-graph.ts is already identical
to main, so no diff there to revert.)
* fix(github): drop Orca self-reference from outage copy
* fix(github): drop em-dashes from outage copy
* fix(agent-status): clear answered Claude question waits at answer time
An answered AskUserQuestion left the amber "waiting" indicator on sidebar
rows and tabs until the agent's next tool hook or turn end — unbounded
linger while the model thinks or streams after the answer (measured 17s
for a 1000-word reply, 44s for 3000 words).
Root cause is an event-shape change: newer Claude reports the
AskUserQuestion wait as PermissionRequest (not the PreToolUse shape #7852
special-cased), so the wait inherited real-permission stickiness and
shouldKeepClaudePermissionVisible swallowed the answer-time
PostToolUse(AskUserQuestion) working event — the identity match can never
succeed because the question's PermissionRequest carries no inheritable
tool_use_id. That silently undid #8311 for questions.
Two scoped changes, both keyed on the tool name rather than the hook
event name:
- Sticky permission hold now exempts AskUserQuestion waits, so the real
answer-time hook (when Claude sends one) clears the wait as #8311
intended.
- New guarded inference for the hook Claude may never send: the submit
keystroke (Enter or digit quick-select) into a pane whose fresh status
is a waiting AskUserQuestion synthesizes the post-answer state, exactly
mirroring the existing interrupt inference (renderer baseline capture,
main-process re-validation, listener lead-state sync so child-driven
refreshes cannot resurrect the dismissed question).
Real permission waits (other tools) keep their sticky semantics; batched
input and pastes never match the submit classifier.
Verified live against a real claude CLI: waiting -> working within ~50ms
of both Enter and digit answers, question card dropped, unanswered
questions still hold amber, permission stickiness covered by tests.
* fix(agent-status): guard question answer inference
Keep multi-question, multi-select, and free-text selector interactions waiting until the full prompt is submitted. Wire native-chat answers into the same guarded inference only after every paced runtime write succeeds, with cancellation and delivery-failure coverage.
* chore(skills): refresh manifest for rc.2
* fix(agent-status): verify native chat answer delivery
* fix(agent-status): await verified question delivery
* fix(agent-status): pin native-chat answer baseline before delivery
The native-chat question-answered inference read the live pane status at
settle time (after the paced send + remote acceptance, which can span
seconds on SSH). If a replacement AskUserQuestion became current in that
window, the settle callback minted a fresh baseline from the new question
and the server cleared *its* wait — dismissing a question the user never
answered.
Capture the answered question's baseline before delivery and have the
inference getter return it, so the server re-validates against the pinned
baseline and rejects a changed status — the same capture-then-revalidate
contract the terminal keystroke path already uses. Also hoist the
shouldStepNativeChatAskAnswer predicate to a single evaluation.
Regression test swaps the live status between sendAnswer and settle and
asserts the answered question's baseline is used (fails against the prior
live-read getter).
Follow-up polish to #8730: key @keyframes on the resolved row/frames (so a same-row fallback like hover on a jumping-less pet doesn't restart idle), mirror the importer's 60s per-frame cap at render, and trim comments to the 1-2 line convention.
Imported .codex-pet bundles played every animation at a flat 8 fps (~9x too
fast). Mirror Codex's exact per-frame duration tables, render uneven holds as
step-end keyframes, upgrade legacy-persisted pets at render, and add the Codex
mascot's pointer interactions (hover, grab-and-hold on frame 0, horizontal-drag
running). Hardened over several adversarial review rounds: honor explicit fps,
start each row/pet from frame 0, scope drag to its pointer, and guard untrusted
persisted data.
Fixes#8729
Co-authored-by: nasagong <zinho2000@gachon.ac.kr>
#9085 made resolveNativeChatSessionOptionDefaults fall back to the catalog
default model (sonnet) and effort (high) whenever the user had not explicitly
picked one, injecting `--model sonnet --effort high` (and codex `-m`/`-c`
equivalents) onto every agent spawn: composer, worktree activation,
source-control agents, onboarding, work-item launch, floating terminal, and
new-tab launch. That silently overrode each user's own CLI-configured default
model.
Return undefined (spawn the agent exactly as its own CLI would) unless the user
has explicitly selected a model, and carry only explicitly stored option
values. An explicit model selection still applies that model and its catalog
option defaults via resolveAgentSessionOptionLaunch.
* feat(editor): bindable shortcut to add a markdown review note
Adds editor.addReviewNote (default Mod+Alt+N) to the shared keybinding
registry and wires it into all three markdown surfaces: the rich editor
key handler invokes the annotation popover opener, the Monaco editor
installs a keydown listener that opens the composer for the tracked
selection target, and the preview maps the DOM selection to its
annotation block. openAnnotationPopover now prefers the live selection
target over synced state so the shortcut works even before the sync
render lands.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XYtipbTz8N4woN1sxTK1ia
* fix(editor): cover list items and Monaco path for add-review-note shortcut
Tag the preview's list-item annotation blocks with data-annotation-block-key
so the shortcut resolves selections inside li blocks (review feedback), and
extend the e2e spec to drive the Monaco source-editor wiring.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XYtipbTz8N4woN1sxTK1ia
* docs(e2e): explain store-driven view-mode switch in add-review-note spec
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XYtipbTz8N4woN1sxTK1ia
* refactor(editor): extract add-review-note + selection-flush modules to satisfy max-lines after rebase
* refactor(editor): spread key-handler params and extract TOC hook to satisfy max-lines
* test(editor): move add-review-note installer test into its own describe
* fix(editor): pass add-review-note chord through when Monaco cannot act; cover preview surface e2e
* fix(editor): unify add-review-note chord consumption — consume only when a composer opens
* fix(editor): gate list-item annotation block key on composer availability
* fix(editor): require live selection for keyboard add-review-note
* chore: retrigger CI against current main (merge ref built during transient main breakage at 6e91ca6c0)
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Brennan Benson <79079362+brennanb2025@users.noreply.github.com>
* fix(agents): recognize OpenCode native OC | tab titles
OpenCode's native OSC titles use `OC | <task>` without an `opencode`
token, so title classifiers left tabs as Claude/unknown. Map the native
marker (optional mux prefix) to OpenCode identity in both title
classifiers, exclude it from isClaudeAgent, and cover lookalikes plus
stale Claude launch reclaim.
Builds on and supersedes #8590 (credit @gatsby74). Fixes#8478.
* fix(agents): drop renderer import from #8478 shared repro
tsconfig.node includes src/shared tests; importing agent-status pulled
renderer modules outside the node project and failed typecheck. Assert
opencode identity via shared title classifiers only (OpenCode TUI sets
"OpenCode" and `OC | ${title}`).
* fix(runtime): fill browser cert failure map in mobile snapshot fixtures
Main's #9104 reads browserCertificateFailuresByPageId in buildMobileBrowserTab
but left partial AppState test helpers without the field, breaking PR Checks
merge commits. Default the map in fixtures and use optional chaining so partial
state cannot throw.
* fix(settings): show "Up to date" for current skills in sidebar
Orchestration and Computer Use showed a green "Installed" badge in the
settings sidebar even when freshness tracking knows the copy is current.
Add an "up-to-date" nav status so a present, no-update skill reads
"Up to date" (pairing with the existing "Update available" state), while
Voice — which tracks model readiness, not skill freshness — keeps
"Installed".
* fix(settings): require freshness proof for current badge
Only label a skill Up to date when every discovered placement is current. Keep Installed for unloaded, blocked, inaccessible, or unrecognized inventories, and share the classifier with the existing freshness pill. Regenerate the current skill manifest so its app version matches the package release and the required verification gate passes.
* Add mobile image-diff previews via shared data-URI builder
- Extracts a `buildImageDataUri` helper (src/shared/image-data-uri.ts) shared by
the desktop ImageViewer and mobile, so both trim whitespace-wrapped base64 and
skip non-previewable mimes (e.g. application/pdf) the same way.
- Adds mobile-diff-image-preview.ts to render binary git.diff results (add/modify/
delete) as images instead of falling back to "Binary preview unavailable".
- Extracts resolveMobileFileTabDoc to consolidate the session file-tab loading
logic (diff/image/html/text) out of the route file for testability.
* Fix stale binary image fallback for empty modified diffs and relay reads
- mobileDiffImageDataUri now distinguishes a true deletion (modified
side absent) from a modify whose binary bytes arrived empty
(relay/size-cap cases), returning null instead of the stale
pre-change image
- readWorkingDiffFile passes the file path to bufferToBlob so relay
working-tree reads can detect previewable image extensions instead
of always reporting empty binary content
- add mobile-file-tab-doc.test.ts covering diff/image/binary/text
resolution paths
* Regenerate skill bundle manifest for 1.4.144-rc.2
Co-authored-by: Orca <help@stably.ai>
* fix(review): trim comments to AGENTS.md's one/two-line why-only rule
Comments in mobile-diff-image-preview.ts and mobile-file-tab-doc.ts ran
3-6 lines and narrated mechanism instead of stating only the non-obvious
reason, per AGENTS.md's "Code Comments: Document the Why, Briefly" rule.
Co-authored-by: Orca <help@stably.ai>
* Distinguish read failures from true deletions in binary diff results
- Working-tree stat/readFile errors and relay reads previously collapsed
onto the same empty-content signal as a genuine deletion, letting
previewers fall back to stale original bytes on a failed read.
- Add modifiedDeleted/missing flags through status.ts, git-handler-ops,
and git-working-file-read so only proven deletions trigger the
original-bytes fallback; failed reads now return null.
- Tighten buildImageDataUri to accept only image/* mimes instead of
special-casing application/pdf.
* fix(relay): expect missing:false on index blob maxBuffer overflow
readBlobAtIndex now returns a missing flag so staged deletions are
distinct from size-capped binary reads; update the overflow test.
* Allow opening deleted files to show pre-delete text or image diffs
Deleted files can now be opened to view their pre-delete content via
git.diff (including images via modifiedDeleted). Only unresolved conflicts
remain unopenable. Centralizes the canOpen rule in canOpenMobileGitStatusEntry()
to keep opener guards consistent across the mobile source control UI.
---------
Co-authored-by: Orca <help@stably.ai>
* fix(skills): hide split icon and stop update-terminal respawn on refocus
The "Update skills" modal embeds a one-off command terminal via
OnboardingInlineCommandTerminal. Two issues:
1. The pane header showed the regular terminal's split ("Split Terminal
Right") icon, which makes no sense for an ephemeral single-command
terminal. Add an `allowSplit` prop (default true) threaded
TerminalPane -> TerminalPaneHeaderOverlay, and have all onboarding
inline command terminals pass `allowSplit={false}` to omit it.
2. Alt-tabbing out of Orca and back respawned the terminal. On window
focus, useSkillFreshness republishes {inventory: null, loading: true}
to retract stale update authority; the dialog reacts by nulling the
draft command, then re-sets it after the scan, and the
`key={terminalCommand}` remount spawns a fresh PTY. Add a ref-counted
suspendSkillFreshnessFocusRescan(); the dialog holds focus rescans
while a live terminal is showing so refocus no longer tears it down.
Explicit Re-check and install-changed refreshes still run.
* fix(skills): preserve freshness checks without respawning terminal
* fix(skills): keep revalidated terminal safe and focused
* Add per-model session-option pickers (model/effort/fast-mode) to native
Introduces a shared agent session-option catalog (Claude/Codex/Gemini/Cursor)
with model-scoped options, launch-command composition, mid-session dispatch
via slash commands, and per-model persisted defaults. Wires the new
NativeChatSessionOptionPickers UI into the composer, threads sessionOptions
through every startup-plan builder (worktree creation, onboarding, source
control actions, folder workspace, direct work-item launches), and adds
localized strings and tests across the affected surfaces.
* Add per-model session-option pickers with a verified Claude model switch
Native chat's model/effort picker now dispatches option commands through a
body-then-verified-Enter write path and, for Claude, arms a PTY observer that
classifies the cached-history confirmation prompt as applied/rejected/needs
interaction before the picker returns — falling back to the terminal only
when Claude genuinely requires manual input, and clearing stale truth
otherwise. Also reorders the composer's model/effort pills, disambiguates
their tooltip labels, and switches disabled-reason strings to a closed enum
so producer and localized copy can't drift.
* Add live model/effort detection from Claude's TUI header
- Reads the mounted xterm's main-buffer snapshot (falling back to the
rendered screen when the alternate screen owns the buffer) to parse
Claude's header for the currently active model and effort, so the
native chat picker reflects reality instead of only dispatched state
- Extracts command-apply recording and reported-value application into
dedicated modules, and factors file-link click handling into a hook,
to keep native-chat-pty-session-options.ts focused
* fix(terminal): forward Windows Ctrl Alt chords
* fix(terminal): route rescued Windows Ctrl+Alt chords through xterm's own key encoders
Replace the hand-rolled Alt-prefix encoder in the custom key handler with a
narrow repair of xterm's third-level-shift classification. xterm's keyboard
service already computes the correct bytes for every input protocol (legacy
ESC-prefixed, kitty CSI-u, win32-input-mode) before _isThirdLevelShift
discards them on Windows Ctrl+Alt; rescuing only provably-genuine chords
(Chromium's layout-wide AltGraph simulation, crbug 762557) lets those
encoders deliver protocol-correct, layout-aware bytes with no duplicated
encoding knowledge in Orca.
Fixes vs the previous approach: kitty-mode TUIs now receive CSI-u instead of
legacy bytes, digits/punctuation no longer alias to plain Alt chords,
letters follow the logical layout (Dvorak/Colemak), Ctrl+Alt+Shift and
Ctrl+Alt+F-keys gain Linux parity, and handled keys get xterm's stock
preventDefault/stopPropagation. Firefox web clients keep stock behavior; a
real-Terminal contract test fails loudly if an xterm upgrade removes the
seam, degrading at runtime to the historical dropped-chord behavior.
Co-authored-by: Orca <help@stably.ai>
* Refactor Windows Ctrl+Alt chord test helpers and clarify AltGraph commen
- Extract a shared getCore() helper in the test file to dedupe repeated
`_core` casts across third-level-shift and keyboard-service lookups.
- Correct the AltGraph comment: Chromium simulates AltGraph per composing
keypress, not for the whole Ctrl+Alt press duration.
- Warn via console when xterm no longer exposes `_core._isThirdLevelShift`,
so a silent classification-repair failure is diagnosable in the wild.
* Clarify comment explaining why Windows Ctrl+Alt chords bypass AltGraph c
The comment previously implied Chromium always sets AltGraph=true for
composable chords; the revised wording states the actual mechanism
(Alt+Ctrl modifiers get replaced by AltGraph) so the inverse case is
unambiguous.
---------
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
Co-authored-by: Orca <help@stably.ai>
OMP runs as a `shell → omp → pi` process tree and Orca recognizes both
`omp` and `pi` as distinct agents, so the foreground-process reader
alternates between reporting the two across command boundaries. In
`resolveTabAgentFromSignals` every identity signal is re-owned onto the
pane's durable owner within its title-identity group (pi and omp share
one) — except the process signal, which was consumed raw. Because it
ranks above launchAgent, each `pi` read repainted an OMP-owned tab's icon
to the Pi glyph and the next read flipped it back.
Re-own the process signal the same way the hook and title signals already
are, so a same-group read collapses onto the owner while a genuine
cross-group process (e.g. Codex) still stands and can reclaim a reused
pane. The tab title text was unaffected — its owner resolution already
excludes the process signal — so this is icon-only.
Adds regression coverage for the oscillation, the mirrored/restored pane
(launchAgent gone, durable hook record), and the cross-group scope guard.
* fix(terminal): restore clickable links after returning to a worktree
Clicking a terminal link (file path, URL, term_* handle, OSC-8) does nothing after switching to another worktree and back, until the terminal is scrolled a little.
xterm's linkifier only re-runs link providers on mousemove when the hovered buffer cell changes vs its cached `_lastBufferCell`. Hiding the terminal surface fires `mouseleave`, which clears the current link but leaves that cache. On reveal the pointer returns to the same cell, so the mousemove handler short-circuits and the link is never re-established; a scroll shifts the buffer position and re-linkifies, which is the "scroll to fix" symptom.
Reset the linkifier hover-cell cache when a terminal surface is revealed (resumeTerminalVisibility) so the next mousemove re-evaluates providers. Covers all link types, including file-path links whose geometry click fallback does not compensate after reveal.
Verified in real Electron via a new e2e that fails without the reset and passes with it, for both a file-path link and a URL.
* test(terminal): prove restored links activate after reveal
* Add safe skill freshness detection
* Accept observed copy-mode rail outcomes
* chore(skills): regenerate snapshot artifacts for the merged guide content
The rebase onto main picked up the reviewed guide fixes (#8624), so the
current manifest hashes and a new appended snapshot generation must
match those bytes; the registry keeps all prior snapshots so existing
installs classify as outdated rather than unrecognized.
* fix(skills): canonicalize snapshot file order and guard released history
Historical snapshots kept git ls-tree byte-order while the working-tree
walk and runtime observation use the sorted depth-first order, so any
future multi-file skill would misclassify older installs as unrecognized
and churn spurious registry revisions; all producers now share one
canonical order (no digest changes for today's single-file packages).
Also rejects executable files from shipped skills (Windows observation
cannot see execute bits, which would misclassify pristine Windows
installs) and adds an explicit append-only invariant for released
snapshots so a generation-logic change cannot rewrite them silently.
* fix(skills): throttle focus rescans and correct self-blocked placement copy
Every window focus re-read and re-hashed all installed packages, and the
nudge and panel each forced their own trailing rescan for one event; a
15s cooldown plus a shared invalidation latch keep one bounded scan per
event while install-change events stay immediate. Bundle artifacts are
now loaded once per run instead of re-parsed on every scan. A read-only
or otherwise unsupported outdated placement now explains that it blocks
itself instead of blaming a phantom sibling placement; the supported
topology set moved to shared so eligibility and copy cannot drift.
* feat(skills): move freshness surfacing to a lingering toast and update modal
The Skills page has been unreachable since its toolbox menu entry was
removed (#4535), so surfacing freshness there buried the feature behind
its own nudge. The nudge now lingers until acted on (ignoring it records
nothing; only the explicit close persists dismissal keys) and opens an
update modal hosting the pre-filled editable terminal, an honest
current/blocked summary, and the per-placement rows in a collapsed
Details section. A compact 'Check for skill updates' row in CLI settings
is the manual re-entry point. Skills page restored to main; design-doc
surfacing section records the venue decision.
* fix(skills): mount update dialog inside the link-routing provider and fold freshness into the setup rails
The dialog hosts a live terminal pane that requires the link-routing
preference context; mounted outside the provider it crashed the renderer
the moment an eligible update existed (caught by live QA — unit tests
mock the terminal). It now mounts inside the provider behind its own
recoverable boundary.
The separate 'Check for skill updates' settings button is gone: the
setup rails' own pill now carries freshness (Update available / Up to
date, falling back to Installed for blocked or unrecognized copies and
for non-local runtimes the local-only scan cannot vouch for), and
Re-check refreshes both installation detection and the freshness
inventory. Wired for the CLI, Orchestration, Computer Use, and
Per-Workspace Environments rails.
* fix(skills): use the sleek scrollbar style in the update dialog
* chore(skills): regenerate manifest for merged main (v1.4.142-rc.1)
Main advanced to 1.4.142-rc.1 with a v1.4.141 release, so the embedded
appVersion and release mapping were stale on the PR's merged tree. Only
appVersion and the new release entry change; no snapshot digests move
(released history preserved).
* fix(skills): bound and batch freshness work
* fix(skills): harden freshness integrity checks
* fix(skills): accept observed copy topology outcomes
* chore(skills): regenerate manifest for current main
* fix(skills): preserve update terminal lifecycle
* chore(skills): regenerate manifest for current main
* fix(skills): fail closed on stale freshness scans
* chore(skills): regenerate manifest for current main
* fix(skills): preserve freshness safety under focus churn
* feat(skills): group the update modal by skill with plain-language status
The Update skills modal now lists only skills that will update or that can't
(with why), grouped by skill with their install locations nested underneath —
no more one row per placement.
- Statuses collapse to "Update available" / "Can't update" at the skill level.
- A location's problem is a chip (Duplicate, Unrecognized, Inaccessible, Read
only, In a repo, External/Broken link, Plugin cache) with a hover tooltip
that explains what it means for the user and what to do.
- Up-to-date, unrecognized-only, and unreadable-only skills are hidden; a
current/unrecognized/etc. location only appears when it explains a shown skill.
- Copy is de-jargoned (drops "copy"/"placement"/"snapshot"/"official copy") and
names the mechanism as the npx skills update command, not "Orca's update".
- Rename the section to "Update details"; drop the unreachable newer-known state.
Renderer-only: derivation is a pure module (groupSkillFreshness) with unit tests;
no IPC or main-process change. Locales updated for all five languages.
* chore(skills): regenerate manifest for current main (v1.4.143-rc.0)
* feat(skills): don't let a duplicate block the update; clearer skipped copy
- Eligibility: a clean standalone duplicate no longer poisons the whole
name — the canonical copy still updates and the duplicate is flagged;
a duplicate-only skill stays unoffered.
- Update modal: "Can't update" -> "Skipped" with a reason-specific
sentence (edited/read-only/in-a-repo/plugin/link); chips describe only
the location state; footer "Check now" -> "Re-check".
- Settings sidebar nav pills go amber "Update available" when a skill is
updatable, matching the setup cards.
- Localized new strings across en/es/ja/ko/zh.
* chore(skills): regenerate manifest for merged main (v1.4.144-rc.1)
* feat(native-chat): add native chat view across mobile
* fix(native-chat): address review findings and CodeRabbit threads
Correctness:
- Restore an independent initial readSession seed and surface initial-drain
errors as snapshot frames so the chat view can never strand on 'loading'
- Pair mobile tool results to calls by ordinal FIFO (parallel calls no longer
misgraft results); clear a pending ask only when its own call resolves
- Show a new streaming reply immediately (same-turn suppression, not length)
- Delegate mobile noise filtering to the shared harness-injected classifier
- Admit soft-leaving mobile clients in beginMobileInputFloor (parity with
mobileTookFloor) so grace-window writes aren't dropped
- Self-heal a stale 'working' status once this turn's reply lands
- Catch RPC rejections in mobile file-open helpers; guard sanitizeToolInput
key collisions; settle web/runtime transports on unrecognized first frames
and forward snapshot errors
Perf:
- Throttle the mobile streaming bubble (50ms) so per-part status frames stop
re-parsing the whole accumulated markdown
- Short-circuit markdown path detection on dot-less or oversized runs
(quadratic backtracking guard)
UX/minor:
- Wire hold-mode dictation through the native chat composer
- Allow scoped-package (@) paths in file-path detection
- Move caret after mid-text autocomplete insertion; index-prefixed ask option
keys; single scroll-to-end effect; bounded wait + toast when image attach
races a resubscribe; count-based pending reconciliation; cache-hit search
cancels stale debounce; chat-tab toggle wins over in-flight preference load
- Share shouldStepNativeChatAskAnswer between desktop and mobile; import
block guards/source priority from shared instead of local copies
- Defensive non-positive transcript limits; test strengthening (TTL expiry,
post-unsubscribe stale frame, lease readiness, filtered console.error)
* refactor(native-chat): share desktop/mobile chat logic in src/shared
Extract the parity-mirrored native-chat modules into shared implementations
both surfaces re-export: ask parsing (registry, parseAskFromStatus,
extractPendingAsk, formatAskAnswer), answer stepping offsets/scheduler, diff
detection/parsing, harness-noise filtering, tool fold/pair/split, and tool
summaries. Removes the hand-synced copies and their stale Metro comments.
Divergence reconciliations take the safer side of each: diffs truncate at
120 lines/32KB everywhere (desktop previously unbounded), tool-run summaries
cap at 3 parts with bounded-depth previews, nameless tool calls are skipped,
and basenames split on both separators.
Also: settle and kill every sibling quick-open pass when one reaches
maxResults (main rg/git and relay git; relay rg already did) so a capped
search cannot leave a scan walking a huge tree; fold window-bounding into
the shared merger's applyAppend; localize the web 'Pair a host' snapshot
error.
* fix(native-chat): address CodeRabbit follow-ups on shared modules
- Attachment lease gate re-checks connection/target/tab after the bounded
wait, so a tab/host switch or disconnect mid-wait can't send into a stale
terminal; a moved-away target drops silently like the pre-wait guard and
only an unrecovered lease surfaces the toast. Adds hook tests.
- extractPendingAsk parses transcript tool-calls through the same
registered-parser + canonical-shape fallback as live status, so a custom
question tool that rendered live survives reconnect/replay.
- Direct unit tests for the shared ask parser (FIFO ordering, fallback,
malformed payloads) and tool-summary bounded preview (depth/collection
caps, circular refs, basename/command branches).
* fix(native-chat): treat initialLimit 0 as a valid empty window
Both engine guards used truthiness, so an explicit zero limit skipped the
bounded tail reader and fell back to an unbounded incremental read. Latent
only (every caller clamps positive), hardened for consistency with the
tail reader's non-positive-limit handling.
* fix(mobile): native-chat composer lock UX + send-failure feedback
- Distinguish input-lock reasons: transport 'disconnected' shows Reconnecting…
instead of mislabeling a reconnect as locked-by-another-client
- Guard the composer lock behind a 600ms hold so connState blips / lease
hand-offs don't flicker the placeholder; unlock stays instant
- Surface a rejected send inline above the composer (a bottom toast hides
behind the keyboard); auto-dismisses after 4s
- waiting-session hint invites the first message instead of implying the
agent is still starting
* test(mobile): sync answer-send pacing test to the 500ms advance buffer
Missed in merge 8fe3c391c, which carried main's NATIVE_CHAT_ADVANCE_BUFFER_MS
300->500 (#8568) into the shared stepping module that mobile derives from.
* fix(mobile): restore terminal stream after chat cold start
* fix(native-chat): harden retries, optimistic sends, and file scans
* fix(mobile): deliver AskUserQuestion answers by option number (STA-1860)
Port #8840's fix to the mobile native chat: the Ask card now tracks
per-question option INDICES (+ free text) and the answer-send hook drives
Claude's arrow-navigate selector with buildAskAnswerKeys keystroke groups —
option numbers, next-tab arrows, Enter — paced one selector step apart, instead
of pasting label text that the selector ignores (which silently committed the
default option). Non-Claude agents keep the pasted-label path via the
selection-based formatAskAnswer.
Backcompat: keystrokes are built client-side and written through the EXISTING
terminal.send passthrough with enter:false — the same contract the permission
card already uses — so an older desktop runtime (SSH/relay included) replays
them verbatim; no RPC/contract change in either update order. Free text is
newline-sanitized because terminal.send has no paste framing.
Drops the now-unused formatCompleteAskAnswer from the shared module.
* fix native chat send and runtime races
* fix mobile native chat formatting
* fix(native-chat): mobile empty state matches desktop copy
Mobile showed a single generic line ('Send a message to get started') where
desktop shows a titled two-line empty state naming the agent ('Start a chat with
Claude' + 'Ask Claude to inspect code, explain output, or make a change.'). Align
them from one source of truth so they can't drift again:
- Extract the agent-type label map + formatAgentTypeLabel to
src/shared/agent-type-label.ts (desktop re-exports; mobile imports).
- Add src/shared/native-chat-empty-state.ts with the canonical English copy;
desktop uses it as its i18n fallbacks (localization unchanged — en/es/ja/ko/zh
keys still win), mobile substitutes the agent label and renders it directly
(mobile ships English only).
- Mobile: render title + subtitle for waiting-session AND ready-but-empty (both
are 'start a chat'), error copy for errors; keep the loading spinner.
Live-verified on the iOS sim against a pn-dev of this branch. typecheck node/web
+ mobile tsc clean; 30 mobile + 428 desktop/shared native-chat tests green.
* style: oxfmt the empty-state parity test (line wrap)
---------
Co-authored-by: Brennan Benson <brennanbenson@Brennans-MacBook-Pro.local>
Co-authored-by: Brennan Benson <79079362+brennanb2025@users.noreply.github.com>
* Add native macOS menu bar status item with settings toggle
Extend the Windows-only system tray into a shared status-item module
so macOS gets a template menu bar icon (Open/Settings/Check for
Updates/Quit), a theme-aware attention dot, and a "Show Menu Bar
Icon" appearance setting. Also fixes a startup race where a tray
"Settings…" click could fire before the renderer's ui:openSettings
listener attached, by queuing a one-shot pending-open-settings intent
the renderer consumes on mount.
* Fix Retina blur, race conditions, and menu-label duplication in tray Set
- Rebuild the tray attention icon's @2x representation since toBitmap only
read 1x pixels, blurring the glyph on Retina displays
- Fix premultiplied-alpha math so light-glyph tinting uses per-pixel alpha
instead of a flat 0xff, keeping antialiased edges valid
- Always push ui:openSettings and leave a longer-lived pending flag, since
there was no reliable signal that a renderer's listener was attached
- Preserve tray attention state across macOS menu-bar hide/show toggles
instead of resetting it on tray destroy
- Route macOS tray creation through syncMacMenuBarIcon so startup and the
live toggle share one visibility policy
- Reuse app-menu translation keys for tray Settings/Check for Updates and
drop the now-duplicate tray-scoped locale strings
* Make menu bar icon default on and platform-independent in settings
Previously showMenuBarIcon was hardcoded to darwin-only in both the
default settings and the sanitize/load paths, so a profile written on
macOS lost its opt-out when touched from another OS. Store the raw
boolean everywhere and let darwin-specific consumers decide whether
to act on it, so the value round-trips unchanged across platforms.
* Fix Settings menu click being silently dropped after a slow cold rendere
Replace the 60s pendingOpenSettings TTL with an untimed intent, since a
cold renderer start can outrun any fixed timeout and cause the flag to
expire before the click is consumed. webContents-id scoping plus
consume-on-read still prevent the intent from leaking to an unrelated
renderer. Adds a test covering the queued-before-mount pull path.
* fix(ssh): reconnect persisted PTYs after shutdown races
* fix(ssh): harden shutdown reconnect-target derivation
- exclude runtime-owned (ephemeral-VM) targets: a renderer-driven startup
ssh.connect would dispose the runtime layer's live relay session
- exclude explicitly disconnected / auth-failed targets so startup never
dials a host the user chose to leave offline (sessions still restore
on tab focus via the deferred flow)
- avoid re-running buildTerminalSessionData via the default parameter
when the remote-session map is legitimately empty (null sentinel)
- recompute activeConnectionIdsAtShutdown alongside terminal-field
patches so a crash between patches cannot strand a stale target
- pin all of the above with tests, including the patch-writer path
Co-authored-by: Orca <help@stably.ai>
* fix(ssh): filter runtime-owned targets at startup reconnect consumption
Main's windowless-promotion path can persist runtime-owned ids into
activeConnectionIdsAtShutdown; renderer must never dial those.
Co-authored-by: Orca <help@stably.ai>
* test(ssh): pin runtime-owned exclusion independently of the status gate
Give the runtime-owned fixture a transient status so the status gate
alone cannot exclude it; verified by mutation (guard removal fails).
Co-authored-by: Orca <help@stably.ai>
* refactor(session): extract reconnect-target derivation to its own module
workspace-session.ts crossed the max-lines budget; the derivation is a
self-contained concern (session ids + live SSH state -> startup targets).
Parameter is now required (null = none) since every caller has the
terminal session data or must opt into the scan explicitly.
Co-authored-by: Orca <help@stably.ai>
* fix(ssh): exclude runtime-owned targets from the connected-states path too
A pane-level optimistic write can stamp a runtime-owned connection state;
the live 'connected' filter must apply the same exclusion as the
session-id union (CodeRabbit review finding). Pinned in the test on both
paths.
Co-authored-by: Orca <help@stably.ai>
---------
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
Co-authored-by: Orca <help@stably.ai>
Brennan Benson
slashdevcorpse
NeilandOrca
Jinwoo HongandOrca
eisen0419andJinjing
Asurada
hanjoonchoe
Trevin Chow
Jinho Choiandnasagong
fsdwenandJinjing
