26 Commits

Include renames

Author	SHA1	Message	Date
Ruben FiszelandClaude Opus 5	45102c8265	fix: let the hub_sync job read the uid and hub_base_url settings (#11106) ... Claude-Session: https://claude.ai/code/session_01Q6triDksvGJ4YK2gA1acEc Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-09-13 17:10:33 +00:00
3 people	8aab5034a6	feat: guest JWT entry for embedded apps (#10954) ... * feat: guest JWT entry for embedded apps (jwt_guest_) A second way in for a guest, alongside the signed-in guest session: a JWT the embedding customer's backend mints and signs, verified per request against a per-workspace key (a PEM public key or a JWKS URL), resolving to the same seatless guest identity confined to the one app its app_path claim names. Bearer prefix jwt_guest_, stateless (no token row). See PR #10954. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * feat: surface guest JWT as the embed method in the app deploy drawer The deploy drawer explained the secret-URL embed but not the guest JWT path, so the primary way to embed an app for a customer's own authenticated users was undiscoverable. For a guest-mode app with guests enabled, show how to mint a `jwt_guest_` token and append `guest.<jwt>` to the app URL, with a copyable iframe template pre-filled with this app's workspace_id and app_path, and a note that new guest emails are refused past the instance's free allowance (the live count is shown just above). Also log a guest JWT allowance refusal at warn, not info: the caller gets a bare 401 (the reason must not leak to an unauthenticated caller), so the log is the admin's signal that the instance hit its guest cap. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix: correct the guest JWT minting instructions in the embed block The block said "sign it with the workspace's guest JWT key", but that setting holds the public verification key. Clarify the keypair relationship (configure the public key or a JWKS URL in the workspace; sign with the matching private key), name the accepted algorithms (RS/PS/ES; HS* refused), and keep the required claims, so an embedder knows how to actually mint the token. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * feat: fall back to the instance JWT issuer for guest verification (off on cloud) A workspace with no guest key of its own now verifies guest JWTs against the instance issuer (JWT_EXT_JWKS_URL, already used by jwt_ext_), so an operator running one issuer configures it once. Verification and the guest grant are CE; granting a full login from that issuer stays EE (jwt_ext_, unchanged). Disabled under CLOUD_HOSTED, where one instance issuer must not be trusted to mint guests in every tenant's workspace — there the per-workspace key is the only source, which also stays the override everywhere. The workspace settings note (hidden on cloud) explains the fallback. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix: embed instructions cover both the workspace key and instance issuer The embed block said to set the workspace's guest JWT key; now it says Windmill verifies against the workspace key or, off cloud, the instance issuer (JWT_EXT_JWKS_URL) when no workspace key is set. The instance clause is hidden under isCloudHosted(). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix: show the guest JWT embed block only when Embed is toggled It belongs with the iframe snippet, not the plain-URL view, so gate it on embedMode alongside the guest-mode / guests-enabled checks. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix: trust the instance issuer in the guest fallback; refresh stale docs P1 (CI review): the fallback wrapped JWT_EXT_JWKS_URL as a workspace JwksUrl, so it hit validate_guest_jwks_url and was refused for http/private issuers unless ALLOW_PRIVATE_GUEST_JWKS_URLS was also set — a self-hosted internal issuer that works for jwt_ext_ failed for guests, though the UI says setting the env var is enough. fetch_jwks now fetches the instance issuer without the https/private restriction (matching the jwt_ext_ loader; it stays operator-trusted), while a workspace-admin URL is validated and pinned as before. All the size/key/URL bounds still apply to both. P2 (CI review): refresh the stale docs that said a missing workspace key always refuses a guest JWT — the module, bearer, key-source, and EditGuestJwtKey field docs now describe the workspace key with the off-cloud instance-issuer fallback. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix: fetch the trusted instance issuer like the jwt_ext_ loader P1 (CI review): the instance-issuer fetch skipped SSRF validation but still disabled redirects and default cert validation, so an instance issuer that works for jwt_ext_ through a redirect or an operator-approved self-signed cert failed the guest fallback. Fetch it with HTTP_CLIENT_PERMISSIVE (follows redirects, honors ACCEPT_INVALID_CERTS) — the same behavior jwt_ext_ has — while a workspace-admin URL stays validated, DNS-pinned and redirect-free. The body size cap still bounds both. P2 (CI review): the WorkspaceSettings field doc still said None/None means no JWT guests; it now names the off-cloud instance-issuer fallback. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * docs: schema summary + OpenAPI cover the guest JWT columns and fallback P2 (CI review): summarized_schema.txt was missing guest_activity.jwt_entry and the two workspace_settings guest-JWT key columns (required by docs/validation.md after a schema change). The edit_guest_jwt_key OpenAPI description now notes that clearing the workspace key falls back to the instance issuer (JWT_EXT_JWKS_URL) off cloud rather than necessarily stopping guest JWTs. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix: keep JWKS single-flight locks in a self-cleaning map, not a bounded cache P1 (CI review): JWKS_FETCH_LOCKS was a 200-entry quick_cache. Past 200 cold URLs it can evict a lock whose fetch is still in flight; the next request for that URL then mints a fresh lock and starts a second fetch, so cycling configured workspaces defeats single-flight and can storm the issuers. Replace it with a plain map guarded by a JwksFetchLock RAII handle that removes each entry once its last holder drops, so the map only ever holds the fetches in flight and never evicts an in-flight lock. Add a unit test pinning the shared-lock and self-cleaning invariants. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * chore: update ee-repo-ref to c2270eb5fe2d9f0968253e6b460c33186363f4e7 This commit updates the EE repository reference after PR #773 was merged in windmill-ee-private. Previous ee-repo-ref: 5a1d9dee34159512c0823fddcd3d096490edbcce New ee-repo-ref: c2270eb5fe2d9f0968253e6b460c33186363f4e7 Automated by sync-ee-ref workflow. --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com> Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>	2026-09-05 10:23:37 +00:00
3 people	fce635d3c4	feat: guest app execution mode, a role that takes no seat (#10929) ... * feat: guest app execution mode, a fourth role that takes no seat Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: make the guest grant a server-minted label, not a declarable scope Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * chore: pin ee-repo-ref to the guest session companion branch Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: close the relabel hole, guest embed tokens, read-path switch, custom-path entry Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: guest tokens are not rescopable and guest embed tokens keep the sentinel Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: guest-derived tokens share one constraint set; gate sign-in on guest discovery Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: the label alone governs a guest; refuse guests with accounts; unserialize discovery Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: guest discovery fails closed; SAML aborts if the guest cookie write fails Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * refactor: enforce the guest switch once at the auth door; sign-in for a guest of another app Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: guest app-mode decided once at the on-behalf resolver; clear a stale guest session before offering another app's sign-in Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: a guest may use anonymous apps; await the stale-session logout; trim comments Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: a guest's path confinement waits for the app's mode, so anonymous apps stay open to it Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: guest target survives http (Lax cookie), rides SAML RelayState; tell account holders on arrival * fix: a guest uses an anonymous app as itself; S3 uploads confined by app mode * fix: a guest upload needs an app policy; a missing app does not skip the confinement * fix: guests are gated on the Enterprise plan server-side; pin ee-repo-ref * fix: the guest plan gate fails closed on non-enterprise builds; settings report the effective switch * fix: guest controls read the plan, not the key; gate the guest tests on the features they need * docs: tighten the guest session invariant comments * feat: 100 free guests per 30 days, then a quarter seat each on Enterprise and a hard cap elsewhere; superadmin guest list; refusals reach the page Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: the cap is exact, an account ends a guest session at the door, popups close, and guest mode survives the CLI round trip Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * feat: a superadmin switch over guests for the whole instance; the pre-existing-user flag keeps its meaning Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: drop the dead guest-access helper, name the instance setting once, guests tab states, CE save order Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: a guest app path is refused at the mint if it could widen the scope; the instance toggle waits for its reload Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: guests stop at the launched-by-me job grant; canonical app paths at the mint and discovery; the toggle ends on the stored value Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: only the scope grammar's own characters bar an app path from guests, refused at deploy as well as at the mint Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: the deploy-time guest path guard checks the destination of a rename and refuses a leading slash Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: a workspace rename keeps the guest switch; the rename guard reads the deployed mode under the row lock Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * fix: guest_activity follows a workspace rename and goes with a workspace delete Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * chore: pin ee-repo-ref to the state-bound guest target Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * chore: pin ee-repo-ref; the guest cookie is never cleared by a callback Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * docs: the workspace-scoped guest_activity delete moves an instance-wide count; assert the mint records the guest Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * test: the seeded allowance is a day old, so only the mint can write today's guest_activity row Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BayTppRCstWX6qTf3LMco5 * chore: update ee-repo-ref to 1a10132e4f3cb442c7d0c2cf6e5d92d150bf6e07 This commit updates the EE repository reference after PR #769 was merged in windmill-ee-private. Previous ee-repo-ref: 32841072aa396bff91d30bd91854fa348cb3c439 New ee-repo-ref: 1a10132e4f3cb442c7d0c2cf6e5d92d150bf6e07 Automated by sync-ee-ref workflow. --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>	2026-09-04 22:47:28 +02:00
Ruben Fiszelandwindmill-internal-app[bot]	cfcfe298dd	feat(ai-chat): make reusable skills ai_skill resources you select per workspace (#10914) ... * feat(ai-chat): make reusable skills ai_skill resources you select per workspace * chore: pin the ee ref to the skill telemetry counters * fix: address review findings on skill authoring, import and migration * fix: enforce skill selection in read_skill and stop imports clobbering resources * feat: carry format_extension from the hub into synced resource types * fix: let an edit set or clear a resource type's format_extension * fix: regenerate the sqlx cache and close the review round findings * fix: close the round-2 findings on folder ACLs, cached sync and truncation * refactor: make the skills migration non-destructive and use design-system inputs * fix: close the round-4 findings on folder owners, startup sync and truncation * fix: clear obsolete extensions, guard folder owners, and report skipped skills * fix: honor explicit-null extensions and report same-type migration conflicts * fix: scope skill actions to the committed workspace and paginate the listing * fix: keep the drawer scoped to the live workspace and surface truncation * fix: discard a skills refresh for a workspace the chat has left * chore: update ee-repo-ref to 6efe7a73c745c2e1377a34498523c00d89010a3d This commit updates the EE repository reference after PR #764 was merged in windmill-ee-private. Previous ee-repo-ref: 55998c142bc72edd08532748af1974b16035658d New ee-repo-ref: 6efe7a73c745c2e1377a34498523c00d89010a3d Automated by sync-ee-ref workflow. --------- Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>	2026-09-01 12:51:27 +00:00
hugocasaandClaude Opus 5	c2279db8a9	fix: allow job tokens to read the automate_username_creation setting (#10869) ... * fix: let a job token read the automate_username_creation setting Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test: use an ungated global setting as the confinement control Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-27 11:37:37 +02:00
hugocasaandClaude Fable 5	9c557859c5	feat: AI agent evals: datasets, scored runs and comparison (#10633) ... * feat: eval datasets and standalone runs for reusable AI agents Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat: agent eval drawer with case editor, runs and capture entry points Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * docs: document AI agent eval datasets and standalone runs Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: say how many eval cases the list is not showing Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: address review findings on eval datasets - keep an edited case's conversation and tool inputs: serde(flatten) silently drops Box<RawValue> fields, so the update payload is spelled out - remount the case editor per case so one case's turns cannot leak into another - require jobs:read / flow_conversations:read on the capture endpoints, which UserDB does not gate by token scope - take the dataset lock in create and update so a delete cannot be undone by a concurrent metadata write, and delete cases before metadata - load more cases beyond the first page, and stop capping the agent picker - record that the version stamp is taken at enqueue, not at resolution Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: address round-2 review findings on eval datasets - block operators from dataset and case writes - pass the editor's operating workspace through the drawer and the capture request, instead of assuming the navigation workspace - discard superseded case-list responses so switching datasets cannot land the previous dataset's cases - reject a dataset without a case_id (or vice versa) rather than running an inline case under a dangling association - run unsaved edits inline instead of silently running the stored case - surface the API error body on a failed run - fetch dataset metadata concurrently when listing - $bindable() without a default on the optional open prop - correct the permission and enqueue-time-version wording in the docs Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: run an untouched saved case by reference again The editor writes back keys the stored case omits, so comparing the raw objects reported every unedited case as edited: the run went inline and lost the dataset/case stamp its history depends on. Compare a normalized form, and pin it with a test. Also scope the history query to the drawer's workspace and drop superseded responses. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat: show a dataset's cases as a table, and fix round-4 review findings The case list showed one case at a time with no overview. It is now a table with the case, where it was captured from, and its last run — the last-run column is a single jobs query on the path stamp rather than a request per row. Review fixes in the same file: - keep the edit baseline on the selected case rather than looking it up in the loaded page, so a case beyond page 1 is not treated as unedited and run stale - release the loading state when a superseded case load returns early - reload every loaded page after a write instead of collapsing to page 1 - last remaining 'resolved to' wording in the version tooltip Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat: run a dataset as an experiment, with scorers as runnables An experiment runs every case of a dataset against one subject and records the exact case set it executed, so a result set stays reproducible while the dataset keeps changing. Each case runs as its own small flow — the agent, then a step per scorer — so a case keeps the run stamp, history query and trajectory view a single run already has, and scorers need no orchestration of their own. Results are read back per step by node id rather than by walking a nested loop's status. A scorer is any runnable taking (input, output, expected): a script, a flow, or a reusable agent used as a judge. A judge is prompted with the case and the answer as one JSON message; a script or flow receives them as named arguments. Scores accept a bare number, a boolean or {score}. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat: results table for an experiment, with scorer columns One row per case: status, the agent's answer, and a column per scorer, with the mean per scorer above the table and a link into each case's run for its trajectory. Averages skip cases a scorer produced no number for — counting a missing score as zero would read as a regression. The drawer's left pane becomes Cases / Results, and Results carries the scorer picker and Run dataset. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat: compare an experiment against a baseline Per-scorer deltas on each row and on the mean, and a filter down to the rows that regressed. Rows join by case id, so a case added after the baseline ran has no delta instead of counting as a change. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: address round-5 review findings on experiments - match scorers by label when diffing two experiments; joining by array position subtracted one scorer from another whenever the scorer sets differed - report a row's status from the case job, not the agent step, so a case whose scorer failed no longer reads as a success - delete a dataset's experiments with it: they hold copies of its cases, and a recreated dataset of the same path would have exposed them - select the experiment that Run dataset just started instead of leaving the table on the previous one - expected is scored now, so stop describing it as having no consumer Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: address round-6 review findings on experiments - hold the dataset lock across an experiment launch, so a delete landing between reading the cases and writing the experiment cannot recreate the deleted dataset's inputs - match scorers between experiments on kind and path, not on label: labels default to a path's last segment, so f/a/quality and f/b/quality compared against each other - average mean deltas over the cases both runs scored; comparing each run's own average reported a regression from a case the baseline never ran, with no regressed row to point at - openapi: the row status is the job's, which is also canceled/skipped; runEval takes scorers; the update-case body no longer advertises source, which the handler deliberately ignores - record why the experiment prefix cannot reach a sibling dataset Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: address round-7 review findings on experiments - release the dataset lock for the push loop and retake it for the write, re-checking the dataset still exists: holding it across the whole launch made every capture and case edit on that dataset 409 until the last job queued - assemble experiment results with bounded concurrency; a 100-case, 3-scorer experiment was 400 sequential lookups, each itself several queries - clear the baseline when it becomes the selected experiment, which was comparing a run against itself and reporting zero deltas - take the header mean over the same cases as its delta while comparing, so the two numbers beside each other describe the same set - a canceled or skipped case is no longer the same grey dot as a running one Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: address round-8 review findings on experiments - verify the dataset's identity, not just its existence, before recording an experiment: the path can be deleted and recreated during the push loop, and the experiment holds copies of the old dataset's cases - give the recording lock a longer budget than a case edit, since its jobs are already queued and giving up strands them, and say so when it fails - keep score lookups sequential within a case: nesting two bounded streams multiplied into 32 in-flight queries against a 50-connection pool - clear a baseline that no longer belongs to the loaded experiments, so switching datasets does not leave comparison mode on with nothing to compare - keep a scorer's own mean when the baseline never ran it, instead of blanking a column full of numbers - EvalCaseDraft.expected no longer claims nothing scores it Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: do not trust an experiment's job ids, and require write to record one Experiment objects live in workspace object storage, which a script can write directly, and results are read on the unrestricted pool — so a forged experiment naming another flow job returned output the jobs API would have refused. Only jobs this server stamped with that experiment's id are read now. Also from round 9: - recording an experiment requires write on the dataset, not read: it persists into the dataset's namespace and its shared list - clear the results table when the selection changes and surface a failed load, instead of labelling the previous experiment's numbers as the new one's - a storage fault is no longer reported as a deleted dataset - the lock-timeout message at the recording site no longer says to retry, which would run the whole dataset again on top of the jobs already queued - ExperimentRow.status documents canceled and skipped Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: bind the experiment trust check to the requested dataset The previous check matched jobs on the experiment id alone, which the stored object supplies — so copying another dataset's experiment JSON under a readable key carried its jobs' output along with it. A job is now only read if it was stamped for this experiment *and* for the dataset the caller's read access was checked against, and an experiment that names a different dataset is not served from this key at all. Also from round 10: - add the .sqlx entry for that query; without it every SQLX_OFFLINE build failed - serve results over GET: as POST the route-scope middleware classified a read as ai_evals:write, locking read-only tokens out of their own results - clear the selected and baseline experiments synchronously when the dataset changes, so the previous dataset's id is not requested under the new one Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: address round-11 review findings on experiments and scorers - give scorers the whole case input, not just the message: an answer that came from attachments or a replayed conversation could not be judged on it - accept a judge's boolean and structured {score} answers, including stringified ones, and pin every documented scorer shape with a test - record an experiment for the cases that did launch when a later push fails, instead of leaving those jobs running with nothing to attribute them to - do not capture a preview parent's synthetic runnable_path as a host flow; the saved case could not be rerun - clear the case table before loading a dataset and surface a failed load, so a failure cannot leave the previous dataset's cases under the new name - keep the results table through a refresh of the same experiment - exclude flow-step jobs from the per-case last-run lookup - drop case sets from the experiment list, which is only used to pick a run - report a database failure at the recording lock as itself, not as contention Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: address round-12 review findings on capture and run history - load flow_node.flow for flownode parents: an agent inside a deployed branch or loop captured without its agent, host flow or tool bindings - decide host_flow_path by whether the path resolves to a flow, not by job kind: excluding previews wholesale also dropped the flow editor's step test, whose path is real - page the per-case last-run lookup by created_before until the loaded cases are covered; one page of 200 reported older cases as never run - do not record an experiment when nothing launched - only attach the case input to a job when a scorer will read it - keep the case table through a save; only a different dataset clears it - drop the superseded duplicate comment on the score parser Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: stop refetching run history on every case write Reading the case list before the first await made the whole job-history query a dependency of it, so every save, delete and Load more refetched up to 1000 job rows and blanked the column. Read untracked instead. - an empty Last run cell now distinguishes never-ran from not-found-within the page bound, which the comment already claimed and the cell did not - reloading a dataset no longer replaces a populated table with a skeleton - keep the score-parser comment that describes every shape it handles Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * refactor: keep eval datasets in Postgres instead of object storage Datasets, cases and experiments become rows (`eval_dataset`, `eval_case`, `eval_experiment`, `eval_experiment_case`) rather than objects under a `wmill_eval_datasets/` prefix. What a run produced is still the job's: only case inputs and an experiment's case snapshot are stored. This removes the machinery the object store needed: - The advisory lock and the read-modify-write of a per-dataset JSONL. A case is a row, so there is nothing to serialize. - The launch-time identity check on the dataset. The foreign key makes a concurrent delete fail the transaction instead. - The trust guard on an experiment's job ids, which existed because a script can write workspace object storage directly and could forge an experiment naming somebody else's job. An experiment now chooses every job id and records itself before pushing anything, so a launch that dies partway leaves a recorded case whose job is missing rather than a running job nothing accounts for; cases that never reached the queue are removed again. Row-level security on `eval_dataset` is the authority on who may read or write a dataset, so `extra_perms` grants work and the rule is not mirrored in Rust. Cases and experiments carry a read policy derived from their dataset and no write policy: they are written on the unrestricted pool after the dataset row itself has been asked, with `SELECT ... FOR UPDATE`, whether the caller may write it. Cases are capped at 256 KiB each and 10 000 per dataset, refused rather than truncated. Attachments are S3 references, not inline bytes, so a case that approaches either cap is a mistake rather than a use case. Evals no longer need the `parquet` feature or a configured workspace object storage. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * style: align the eval drawer with the design system - Scorer chips are `Badge`s rather than a hand-rolled bordered span, and the section header is a `Label` with its tooltip, as are the case editor's fields (which also gets the label colour right). - The results table showed status as a coloured bullet, which says nothing to a colour-blind reader. It now carries the same icons the runs table uses, with the status as its accessible name. - Feedback colours move to the `-500` shades the brand guidelines name. - The conversation JSON error uses `TextInput`'s `error` prop for the border and the caption style for the message, as elsewhere. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: author an expected answer, tags and attachments on a case Every scorer is handed `(input, output, expected)`, but nothing could produce an `expected` except a conversation capture: the case editor had no field for it and a captured run left it empty. So: - The editor gains Expected, Tags and a read-only list of the attachments a captured case carries. Expected is plain text, or JSON when the answer has structure. - Capturing from an AI agent run keeps what that run answered, which is the only moment a reference answer exists for free. The results table also laid itself out by content, so a long answer pushed the scores — the numbers the table exists for — off the edge of the pane. It is fixed-layout now, with the text columns bounded. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs: expected is captured from a run and can be authored Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: link a saved agent when inserting an ai agent step "AI Agent" in the step picker was a leaf that always created a blank step, so reusing a saved agent meant inserting a blank one, opening its step input and linking it there. It is a category now, like Flow and AI Sandbox, listing the workspace's `ai_agent` resources next to a blank option, filtered by the picker's own search. A picked agent produces a step that is already linked rather than one linked afterwards: `agent` set, no tools, and only the flow-local `user_message`/`user_attachments` transforms. Seeding the brain keys there would leave transforms a linked step never reads and that `AgentResourceBar` strips on its next link change. Each `on:new` forwarder rebuilds the insert detail field by field instead of spreading it, so a new field is dropped unless the forwarder names it. `agentPath` is typed on both `GraphEventHandlers.insert` and `FlowGraphV2`'s `onInsert` so the next one to forget it fails the check. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: restore the link on cancel and simplify the agent bar Cancel on an agent edit forked the step into a standalone copy, which is the opposite of what the word means and needed a paragraph under the card to explain. It discards the edits and re-links the step now, leaving the agent untouched; diverging from an agent is Unlink's job, on the linked card. This flow's `tool_inputs` survive the round trip as overrides, so Cancel no longer folds them into the tools the way Unlink does. Linking a step to a saved agent happens in the step picker at insert time, so the bar's own resource picker is gone and "Save as agent" is the one action left. Its `+` button was a trap besides: it opened the generic resource form, where an agent would have to be written as raw JSON. The card itself was `surface-secondary`, the sections token, so in dark mode it was darker than the pane and read as a sunken well rather than an elevated card. It uses `surface-tertiary` as the brand table prescribes, its tool chips are `Badge`s, and the editing card no longer overflows the pane and clips its own buttons. The remaining tooltip follows the inline `Label` convention rather than sitting in a flex row whose gap stacked on the trigger's own margin. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: rework the AI agent evals surface into one table Evals become a single pane: a dataset of cases, one column per scorer, one row per case, with the run being looked at chosen from the toolbar. Runs are permanent. Running the whole dataset opens one; running a single case records nothing at all — it is a job, and looking at what it did is not a claim that it belongs in the history. Its result and its scores sit over the row until they are saved as a run, which carries the cases that were not rerun and the scoring jobs themselves, so the number that is saved is the number that was looked at. A scorer is a runnable: a judge agent or a script, created in one click and edited in place. Scores carry a reason and per-assertion checks, shown on hover with a rescore button. What ran is always named. A run records the agent version, or — for a configuration that is not deployed — a hash of it, so a table can say that its numbers describe an agent that no longer exists: those rows dim and the table offers to rerun. An agent's draft can be run directly instead of the deployed value, and once those edits are deployed the runs that made them are recognised as that version. A step with no agent of its own is evaluable too, and saving it as an agent moves its history onto it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: keep an agent's in-progress edits on the agent Editing a linked agent forks it into the step, which is what makes the edits runnable there — but the agent is what is being edited, so that is where the unsaved state belongs. The edit is mirrored into the agent's own resource draft as it is made. It then survives leaving the flow, shows the agent as drafted wherever it appears, and is what evals run when asked to run the draft rather than what is deployed. Deploying or cancelling clears it; opening Edit without changing anything does not mark the agent as drafted. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: shape the evals surface around a saved agent Evals hang off an `ai_agent` resource, so the surface is now only ever about one: the `draft` subject kind, the standalone-step subject and the move that carried a step's history onto a newly saved agent are gone. - A run is permanent and numbered per agent. Running a single case is a trial: it answers in the panel and never touches the table. - "Run scorers only" opens a run of its own that reuses the answers of the run you are looking at, so a scorer added later measures what already ran without calling the agent again. - A draft run whose configuration is later deployed is stamped, once, to the version it became, so its label stops reading `v23 + edits` forever. - A scorer can carry a pass threshold, read off the scores already recorded. - The table is the case, its answer and one number per scorer; datasets are created and edited in a drawer; a run that executed an earlier state of the current draft says so above the table, in one line. - Which agent a step is, whether it is being edited, and which version it is on is a strip above the step's tabs, because it is true of every tab. - Capturing a case from a step test or a conversation is dropped, and with it the `memory` override on a linked step that nothing set. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: run past versions of an agent, and number versions per resource The evals home becomes one table of every run of the agent, whichever dataset each is of, with one badge per scorer. A list spanning datasets cannot hold every dataset's scorers to look a name up, so a score carries its name and kind with its number, and thresholds are joined in per run and column. Run now asks what to run: the latest agent, resolved when the run executes as a flow step does, any past version, or the unsaved edits. Pinning is a subject kind of its own, since a linked step resolves the resource live and inlining is the only way to run a version that is no longer current. Scorers move into the edit-dataset drawer. The column header over a run reports and nothing else: a run is permanent, and a control there that changed the columns would edit the past from the one place that must not. Adding one offers four ways rather than two, writing and reusing being different jobs, and both new kinds open with a summary filled in. Versions are numbered per resource. `resource_version.id` is one identity sequence for the whole table, so an agent saved nine times read v4 ... v24, and the gaps counted writes in workspaces the reader cannot see. The id stays how a version is addressed; the new number is what it is called, in the resource history drawer as well as here. It is assigned on write rather than counted on read because trimming past the cap and clearing a history both take the oldest rows, and counting the survivors would renumber a version a run already names. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: read the dataset a remembered selection names Reopening the evals modal restored the last dataset from storage as a bare path, without reading the row it names. Every "is this already the one?" test compared against that selection, so all of them short-circuited and the dataset was never loaded: editing it opened a drawer with no summary, no scorers and no cases. The remembered path is now brought into context the same way any other choice is, and the tests compare against the dataset that is loaded rather than the one that is selected, so a selection can no longer stand for a read that did not happen. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: give dialogs a trail in their header A dialog deep enough to navigate had nowhere to say where you were: the header held a fixed title, and the way back was a control each body placed for itself, somewhere in a toolbar that moves with everything else the toolbar holds. The header is the one part of the surface that does not move, which is where the trail belongs. `Modal` takes an optional `trail` of levels below its title, rendered as a breadcrumb whose ancestors are the way back. Declarative on purpose: callers of this depth already hold the state that says where they are, so the dialog reads it rather than owning a stack they would have to push and pop in step with it. Escape follows the trail. Leaving a level is what someone deep in a dialog means by it, and closing the whole surface throws away the navigating they did to get there; at the root it closes as before. That only works if a dialog can tell it is the surface being addressed, so `Disposable` now answers `isTopmost()` and the dialog asks before acting: it keeps Escape for itself, so nothing else was arbitrating between it and a drawer opened from inside it, and both were acting on one key press. Evals is the first caller: its runs list is the root, a run is a level in it, and the back button that used to sit above the table is gone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: portal dialogs out of wherever they were opened from A dialog rendered in place inherits whatever the calling component happens to sit inside. One `transform`, `filter` or `overflow` anywhere above it makes its `fixed` positioning resolve against that ancestor instead of the viewport, and a surface meant to cover the app is then confined to a box it never asked for: the nav rail paints over it and its own edges are clipped. Drawers have always portalled for this reason. Dialogs only did so when an enclosing pane claimed them, and rendered in place otherwise, so the same screen could show a drawer over everything and a dialog trapped behind the nav. They now portal the same way: to the pane when one claims it, to `body` otherwise. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: make the dialog's title the first step of its trail The trail listed levels below the title, so a dialog one level deep read "Evals > All runs > Run 20 · v6": three steps for two places, the first two of them the same place under different names. The title is the root, so it is the root's own segment, and the trail a dialog is given is now the whole path with that segment at its head. Its height stopped moving too. A heading carries a line-height of its own, so a header holding only an h3 stood six pixels shorter than one holding segments as well, and the dialog's whole top edge stepped as you navigated. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: sharpen the evals controls around where you are standing Each screen now offers what belongs to it. The list starts runs; a run is a record, so it offers only the one thing that acts on the record itself, which is measuring the answers it already stored. Starting a fresh run from inside one asked which agent and which dataset from the screen least about either, and scoring an existing run was offered from the list, where there is no run to score. Which run and what it is read against are one question asked twice, so they sit together rather than at opposite ends of a row. Choosing what to run is now a toggle over the two states worth naming, the draft and the saved agent, with every earlier version one click further: running an old version is deliberate, and a list made all three look alike. The draft is read when the dialog opens rather than taken from the caller's polled copy, which could be seconds behind an agent edited a moment ago and would leave the option out exactly when it is the reason for opening the dialog. The dataset field carries its path under it and its edit button on hover, as a resource picker does, so the closed field says what the open list said. Edits waiting on an agent are a "draft" here as everywhere else in Windmill, rather than "+ edits". The dialog runs an evaluation rather than "the agent", which is what it was already called everywhere it is recorded. An agent being edited keeps its evals button on a line of its own, clear of the decision to save or discard. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: settle the evals controls on the patterns Windmill already has The version choice uses ToggleButtonMore, as the AI provider picker does: the two states worth naming stay in the group, the rest are behind the overflow menu, and the one you pick joins the group rather than appearing in a second control below it. The deployed one says which version it resolves to. A run offers nothing to start. Scoring an existing run again was the last thing left there, and it was one button explaining a distinction that the run and the dataset already make between them. The warning that a run executed an earlier draft is about the run on screen, so it goes when the run does rather than following you back to the list, and it sits against the table instead of inside a frame of its own. A dataset just created stays open for its scorers and cases: those are what a dataset is, they can only be added to one that exists, and closing on create sent you to find it again to add them. Scorer settings are a cog rather than a word, now that the row holds three actions. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: close the gap in the version toggle and say what naming a dataset does The overflow trigger is not a pill, so the room it reserves showed as a gap between it and the button before it; it is pulled in by that much. The dataset field gets its clear button, which is also the slot the edit button is positioned against, so the two now sit where a resource picker puts them. Naming a new dataset said nothing about what happens next, and the drawer looked like it was missing the rest of itself. It says so instead: a scorer and a case both belong to a dataset, so there is nothing to attach either to until this one exists, and creating it leaves the drawer open on them. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: choose a dataset's scorers while naming it A scorer is a reference to a runnable, not a child of the dataset, so it needs the dataset's name but not its row. The list is collected in the drawer while the dataset is being named and sent with the create, which already accepts one, so a dataset arrives holding the columns that were chosen for it rather than being made empty and then edited to hold them. Cases stay where they were: a case *is* a row of the dataset, so there is nothing for it to be a row of until one exists. The drawer says which of the two is which instead of leaving the screen looking like it is missing the rest of itself. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: level the version toggle and name the dataset in its own field The overflow trigger stands a row taller than a toggle button, so the group grew to its height and left the sunken background showing under every pill beside it. Every child of the group is the same height now, which is why the AI provider picker never had the band: it sizes them all alike. The dataset field says the summary with the path after it rather than carrying the path on a line below. The list stacks the two, which a one-line field cannot do, so it says both the other way round. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: tidy the evals forms and the run's own controls Picking a scorer that exists chooses between two sources rather than showing both: the ones already measuring something, and everything else in the workspace. The first list says what each is called with its path under it and what it already measures on the right, instead of three columns that were the same path truncated three ways whenever a scorer had no name of its own. A dataset's drawer says what it is for on the page rather than under an icon, and its summary is sized like the field beneath it. The run's own row lines up with the table under it, the warning above that table is spaced off the rule rather than sitting on it, and adding a case is gone from a run: a run is a record of cases that were answered, so curating them from it is editing what it measured. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: create a dataset holding the cases written for it Creating a dataset takes the cases to create it with, so one can be assembled in a single act instead of made empty and then filled in. The drawer holds them while the dataset is being named, gives them ids of its own to be edited by, and sends them with the create. Every case is checked before the dataset is written. `eval_case` grants users no write, so the rows cannot be inserted in the transaction that creates the dataset under the caller's own policies; validating first is what keeps "created holding these cases" from becoming "created, holding some of them", and the rows that do follow go in one transaction of their own. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: name the button for what it opens, and say what each version is Starting an evaluation asks which state of the agent and which dataset, and both cost a provider bill, so a button that read as spending one on the way past was lying about the click. It opens something, and says so. Running one case from the panel keeps its own name and its play icon, because that one does run on click. The version options say what they are rather than what they are not: what a flow step would or would not run is a fact about somewhere else, and someone choosing what to evaluate is not standing in it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: give the editing card two rows and mark evals as beta At the width of a step panel the card's one row wrapped: the line naming the agent, the line saying what saving does, and the two buttons deciding the edits' fate all fought for it. Deciding gets a row of its own, and evals sits against the line it is about, since evals of an agent being edited run the edits. Evals is named wherever it is offered. It read as a word in one state of the card and as an icon in the other, which is two things to recognise for one door. The dialog carries a beta badge against its own name, before any level below it: every way in lands there, so it is said once and stays put as you navigate. The version toggle spells out which is which. Both are the agent at v2 and the difference between them is the whole choice, so it is worth the width. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: name a new dataset, and lay the scorer's settings out like a step's inputs A new dataset arrives called "Dataset 1", which the path follows as it follows any summary: a dataset with none was one every table could only call by its path, and the two seeds are what the summary rule already produces. Scorer settings put each field's description between its label and its input, where a step's inputs put theirs, and its inputs are the size the rest of the drawer uses. The runnable behind the column is a link to it with its kind's icon, since it is a resource of its own and the one thing about it these fields cannot change. The line explaining that a pass line re-reads recorded scores went: the threshold is a number to set, and how it is applied is not a decision being made here. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: curate a dataset in the drawer and save it in one act The drawer holds the cases while they are edited and writes them when it is saved: added, changed and dropped, whichever it is. Typing no longer writes, so a set is never half saved while someone is still deciding what is in it, and Save means the same thing whether the dataset exists yet or not. A case panel offers reading rather than acting. Running one case now and editing one from a run were the last two ways to change a record from the screen showing it, and the machinery behind the first went with it. The answer is rendered as the prose it is, under what it is: the case's result, whichever run is selected above it. The rest is what the run's table was doing to its own edges: a column name is clipped to its column rather than running into the next, the table squares off against an open panel, and that panel closes with the run it belonged to. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: one border above a table, and a link to the run's job The row above the table drew a bottom border and the table draws its own top edge, so every table sat under two lines. The row keeps its spacing and the table keeps its edge. A column header no longer spins while its scores arrive: the cells under it are where the numbers are missing, and they say so themselves. The beta badge is the height of the word beside it rather than of the line it sits on. A run is one flow and therefore one job, so the run says where that job is: what it is doing, what it cost and what it logged are all there rather than reconstructed from the table. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: stream scores as each scorer finishes, and show them per case A scorer runs after the agent inside the case's own iteration, so its verdict can be read as soon as its step is done. Waiting for the iteration to end held every column of a case back until the last of them finished, which is why answers arrived one at a time and scores all at once. Reading a job that is still running needs one guard: a module with nothing in it is a step that has not run, not one that produced nothing, and recording the second makes a failure that never goes away. The panel beside the table shows what each column made of the case and why. The reason a judge gave was stored and never shown, which is the half of a score that says anything. It stops repeating the question the header already asks, and a case still running reads as waiting rather than as an answer that says "Running". A run is a number beside a dataset, so the list puts the two together. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: score a case with every scorer at once The scorers of a case read the answer and never each other, so they ran one after another for no reason: measuring a case now takes as long as its slowest column rather than as long as all of them. Each is a branch of its own, kept from failing the others, so a judge that errors costs its own column and no more. An iteration is three steps again — answer, payload, scores — rather than one per scorer, and each branch is named for the column it produces, so the graph of a run says which scorer did what instead of spelling out an id. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: read a judge's score out of the JSON it nearly wrote A judge quoting the agent inside its own reason writes those quotes unescaped, which is invalid JSON and also the most ordinary sentence for it to produce. The whole verdict was being thrown away over it, so a column that had a number reported having none. The number and the reason are now read straight out of such text. Deliberately not a second JSON parser: it finds the two keys and takes what follows, which is what survives a quote in the middle of a sentence. A case still running says so with a spinner rather than with the word "Running" sitting where its answer goes. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: ask a judge for a shape instead of trusting it to write one A new judge carries an output schema, so the provider holds it to `{score, reason}` rather than the prompt asking it to. Windmill already delivers a schema whichever way the model takes it, a tool for Claude and Bedrock and the native parameter elsewhere, so there is no list of models to keep here. An agent with no runs offers its first one where the first row would be, rather than from a toolbar above a table that has nothing in it. Starting a run no longer picks a dataset for you. It fell back to whichever came first, which on an agent that has never run means offering another agent's set as though it were the obvious one; and with no dataset at all it says so and offers the one move there is. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat: report a column that failed throughout, and hold the run dialog The runs overview dropped any column that produced no number, so a judge that failed on every case of a run vanished from the row and read as a column nobody had asked for. The aggregate now reports every column that has cells, with the count of the ones it failed on, and the badge says "failed" where there is nothing to average. A column with no cells at all is still left out: that one was added after the run and has nothing to say about it. Creating a dataset closes the drawer rather than turning it into an edit of what it just made: scorers and cases already ship with the create, so there is nothing left to stay open for. Reached from the run dialog, it gives the screen back with the new dataset selected, and the dialog keeps the version you had already chosen. Also: - the case panel's job link moves to the panel's own header, where its scope is: the job is the whole iteration, not the answer it sat over - one action in the scorer drawer's header, as its neighbours have. The reuse list picks rather than adds, and says which dataset each column already measures - adding a case is the last row of the list it lands in - the pane shows what it has read rather than an empty state it has not earned yet, and its rows say they open - the linked agent card loses a border it had inside another one * fix: keep the linked agent card's outline The card is a thing inside the step's inputs rather than a section of them, and the outline is what says so. Only the rule inside it goes: the detail it separates is already set apart by being detail. * refactor: fit the eval surface to the shipped design * feat: give a nested dialog a back control and the runs list its own moves Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: put a dialog's description under its title Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * refactor: fold a dialog's back control into the crumb it returns to Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: edit a dataset's cases as a table rather than a list beside a form Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: edit a dataset's cases in the grid the data tables are edited in Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: edit a grid cell of prose in place, and cap a dataset at one page Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * refactor: keep the cell editor's styles beside it, not in the vendored theme Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: keep an empty cell empty and cap the editor's growth Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: name the step that assembles a run for the scorers Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: run the payload step natively, and say so when nothing serves that tag Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: report an answer as answered while its scorers are still running Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: let a scorer say a case is not one it measures Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: score the answer, and leave a case with no expected answer unmeasured Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * refactor: split the evals backend into modules Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: record what a run produced so it outlives its jobs * fix: read only the agent step's own tool jobs into the payload * fix: pin a run's configuration and give the judge the attachments * feat: write a dataset's cases in one transaction * chore: refresh the sqlx cache for the eval queries * fix: drop results a newer selection has superseded * fix: keep a draft the agent editor never opened on * feat: let a run record what it produced instead of waiting to be read * fix: serialize the replacements of a dataset's cases * fix: stop the poller from superseding a read slower than its interval * chore: refresh the sqlx cache * fix: keep a failed read from settling a cell as a case with no answer * fix: hold the case grid while its save is in flight * fix: keep a failed collect step from failing the run it recorded * chore: refresh the sqlx cache * fix: commit an open cell into the save that reads it * refactor: size the eval buttons with unifiedSize * docs: describe a run as the one flow it is * fix: show a run's recorded rows when part of it cannot be collected * refactor: size the remaining PR-added buttons with unifiedSize * fix: save the dataset name that was submitted, not the one typed after * fix: force an open cell into the save that was pressed for it * fix: refuse to score a run whose evidence could not be read * fix: hold one lock over a dataset's case count and its writes * fix: keep one unreadable run from costing the whole runs list * refactor: drop the banned bindable-default from the eval props * fix: hold the scorer controls while the dataset is written * fix: read only the caller's own draft of an agent * docs: say in the contract that a run pins its configuration * fix: say a scorer did not run rather than blaming a missing answer * feat: resume the agent draft you already had when you press Edit * refactor: build the trail and dataset controls from Button * fix: clear the open-cell flag when the drawer reopens * chore: refresh the sqlx cache * fix: read a run's configuration and its version from one snapshot * fix: refuse a dataset path or summary the column cannot hold * refactor: handle the agent draft the way the resource editor does * fix: run only a configuration the launch actually read * docs: bound dataset path and summary where they are submitted * fix: surface a stalled agent draft instead of claiming it is kept * fix: stop claiming a draft holds edits a failed write never sent * fix: word a missing score only once the run says whether the case answered * fix: let a breadcrumb crumb shrink so its truncation applies * docs: describe where an agent's unsaved edits live and what drops them * fix: keep harvesting scores when the run cannot yet word a missing one * fix: report a refused draft write the card was reading as a save * fix: drop the refused draft write when the server copy is taken instead * refactor: build the scorer and dataset pickers from the design system * fix: say what removing a scorer column actually does * fix: drop a refused draft write wherever the server copy is read * fix: let a picker row be as tall as the two lines it holds * docs: record what removing a scorer column does to recorded runs * fix: send a queued draft write before reopening, and drop only what it refuses * refactor: write the agent draft at commit points instead of mirroring keystrokes Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * refactor: run an agent's edits from the step instead of keeping them as a draft Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: make the diff badge keyboard operable and refuse an edits run without its edits Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: drop the dataset icon from the scorer picker rows Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: size the evals buttons like the rest of windmill and call a run of edits edits Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: count a brain expression as an edit of the linked agent * fix: cap scorers per dataset and report a launched run as launched * fix: harvest scores in one read, refuse duplicate case ids, allow group paths * fix: mint scorer ids server-side, save a dataset edit in one request, check attachments * fix: write a dataset edit and its cases in one transaction * fix: atomic dataset create/edit, reset eval pane per agent, stable pending scorer ids * refactor: govern eval_case writes by RLS so a dataset edit is one transaction * fix: pin launch snapshot, order case locks, cap dataset size, guard stale load * fix: cap dataset bytes on single-case writes, reset run-dialog flag on load failure * feat: migrate eval datasets on username change, settle unspawned cases, drop unused case endpoints * fix: resolve scorer scripts as the caller and pin their hash; migrate scorer paths on rename * fix: bound a failed tool call's error to the payload truncation cap * fix: pin scorer hash as a hex string, reject missing judges, migrate eval authorship * fix: record an out-of-range scorer result as an error, not a score * fix: resolve judges in one caller-scoped read, pin deployed scripts, bound pass_if * fix: settle unspawned cases only when the run completes, and their score cells too * feat: reassign eval datasets and their path references when offboarding a user * fix: use the regex backreference in offboarding eval path rewrites * fix: register eval datasets in offboarding registries, keep resource-version param name * refactor: name the resource-version path param id, since it is the row id not the version * fix: validate dataset paths canonically, clone eval data on fork, surface eval load and launch failures * docs: note MCP tool results are not yet surfaced to eval scorers * fix: show the eval error state on any load failure, not only an empty dataset list * fix: preserve eval case order across a batched save Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LVjewUvXFEjNLLw7kxz41h * docs: scope the eval launch delete-safety guarantee to the assembly window Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LVjewUvXFEjNLLw7kxz41h * fix: only offer deployed scripts as eval scorers, drop unbuilt rescore claim Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LVjewUvXFEjNLLw7kxz41h * fix: enforce 0-1 scorer threshold in the settings drawer and clear stale eval load errors Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LVjewUvXFEjNLLw7kxz41h * fix: scope subject version/hash reads to the caller and keep a 0 pass threshold Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LVjewUvXFEjNLLw7kxz41h * fix: select the saved dataset when creating or renaming from the Run dialog Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LVjewUvXFEjNLLw7kxz41h * fix: gate eval dataset rename on path ownership, not just write access Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LVjewUvXFEjNLLw7kxz41h * fix: tolerate a malformed agent config when resolving the deployed label Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LVjewUvXFEjNLLw7kxz41h * refactor: trim eval code and comments, fix shared select and modal paths Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: drop the rename warning when editing an eval dataset path Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: add eval dataset delete, keep summary on partial edits, settle resultless scorer cells Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test: cover parseThreshold and subjectLabel Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: hold dataset Save during a scorer write, derive draft_hash only from the carried draft Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-24 11:08:56 +02:00
3 people	dc27db68de	fix: require item read scope on workspace tarball export (#10797) ... * fix: require item read scope on workspace tarball export Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: accept a wildcard path grant for whole-domain scope checks Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix: let a wildcard path grant delegate the unqualified scope Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com> Co-authored-by: Ruben Fiszel <ruben@windmill.dev>	2026-08-22 12:06:54 +02:00
hugocasaandClaude Opus 5	9022dc9d44	fix: confine job tokens to workspace-scoped API routes (#10631) ... * fix: confine job tokens to workspace-scoped API routes Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: keep the object-storage connection test reachable from a job token Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: keep the workspace-exists check the CLI makes reachable from a job Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test: reconcile the job-token caps after #10124 The workspace-confinement middleware answers a workspace-less route before the privilege gate behind it runs, so the cases #10124 added on those routes now see 403 rather than 401. Rejection is what they assert, but two of them needed more than a status change: - `list_worker_groups` asserted only that the response body omits the static env value, which an error body satisfies for the wrong reason. It now asserts the status, keeping the secret check as a second assertion. - `require_super_admin` lost its only unshadowed route. `GET /api/w/{workspace}/users/list_addable` is gated solely by that call and names a workspace, so it reaches the gate and pins it at 401. The module doc states the two-layer rule once; the file covers both caps, so it is no longer named for either one alone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs: correct the workspaces/exists rationale and the parquet gate note `workspace` carries no row-level security, so `exists_workspace` running through `user_db` does not filter by membership as the comment claimed. State what the route actually discloses — whether a workspace id is taken. The object-storage case explained why a 404 would satisfy the assertion for the wrong reason, which described the earlier `assert_ne!(403)`; against the 422 it now asserts, a 404 fails. Say instead why the case is gated on the feature. * fix: let a job token keep the workspace-less routes that carry no workspace Confinement refused every route outside the allowlist, including ones that answer purely from the caller's own account or from the request body. Those cross no workspace boundary, so refusing them buys nothing: - `users/email` returns a value already inside the token, and `workspaces/allowed_domain_auto_invite` tests the caller's own address against a static list. Neither opens a transaction. - `users/usage` reads the caller's own row; `users/tutorial_progress` reads and upserts a UI bitfield keyed on the same email. - `schedules/preview` takes no `ApiAuthed` at all — it computes the occurrences of the cron expression in the body and returns nothing the caller did not send. The rule, not the list, is what the doc comment states: answers from the caller's own account, the request body, or content identical for every workspace; never naming another workspace, never instance configuration. The candidates it excludes are written down with their reasons, since `users/list_invites` reads as caller-scoped until you notice the response carries a workspace id per invite. Regression covers both directions — the new entries answer, and the rejected caller-scoped reads stay refused. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test: reach the privilege gates confinement hides `require_devops_role` and `require_instance_admin` gate only workspace-less routes, so confinement answers every request that would reach them: no HTTP case can tell whether they still cap job tokens, and both could lose that check with this suite green. `require_super_admin` has a workspace-scoped route to reach it; these two have none, so call them directly instead. The identity used is the fixture's real superadmin, so the passing half proves the rejection keys off `job_id` rather than off the user. Also separate two claims the write-allowlist doc had merged into one sentence: no entry writes outside the caller's own account, but what each may read differs, and the workspace-existence check answers for any id. * docs: say that the object-storage probe writes The write-allowlist lead claimed no entry writes state outside the caller's own account. `test_s3_bucket` puts an object into the store the body names and deletes it again, so it does write; a failure between the two leaves the object behind. The invariant that holds is about Windmill state. Say so in the lead, and describe the put/delete in the entry itself rather than leaving "acts only on the store the request body describes" to imply a read. * test: cover the last two job-token gates confinement hides Seven guards key on `ApiAuthed::job_id`. Three keep a workspace-scoped route and are exercised over HTTP; the other four are reachable only through workspace-less routes, which confinement now answers first, so nothing observed whether they still cap job tokens. `require_devops_role` and `require_instance_admin` were already called directly. Add the two that were not: `forbid_superadmin_job_token`, and `forbid_elevated_job_token`, whose call sites are `create_token`, `update_token_scopes` and `set_password` — all workspace-less. Both key on two conditions rather than one, so all three combinations are pinned: neither fires without job provenance, and neither fires for an unelevated identity. The second matters — collapsing either into a blanket job-token refusal would stop ordinary users creating tokens, and no other case would catch it. The doc comment records which of the seven each route covers. * docs: correct which job-token gates have no observable route The previous commit put `forbid_elevated_job_token` among the guards reachable only through workspace-less routes, and its message named three call sites. It has six, and two are workspaced: `mint_app_embed_token` and `mint_raw_app_sdk_token`. Its superadmin branch is therefore already exercised over HTTP — the 401 the embed-token case asserts is this gate. So three of the seven lack an observable route, not four. Its direct assertions stay: the embed-token case only ever reaches it with an elevated identity, and the unelevated-negative case is what would catch the gate being collapsed into a blanket job-token refusal. * test: pin is_instance_admin, and stop enumerating gates in prose `is_instance_admin` is `authed.is_admin && authed.job_id.is_none()`, so a census built by searching for `job_id.is_some()` could not see it. Both its call sites are workspace-less, and it returns a bool that selects obfuscation rather than refusing — a job token reading `true` leaks `env_vars_static` instead of being turned away. Pin both directions. The doc comment tried to account for every job-token guard and which route exercised it. It was wrong three times running: the count, the call sites of `forbid_elevated_job_token`, and the claim that the CUSTOM_INSTANCE_DB case covers `is_super_admin_authed` when that path tests `job_id` inline. A table that has to be rederived from six crates to stay true does not belong in a comment, so it now states only why these calls are direct. * docs: name the right is_instance_admin caller The comment credited "the concurrency-group listing" with obfuscating rows. The second caller is `prune_concurrency_group`, which returns PermissionDenied; the obfuscating one is `list_worker_groups`. Keep the claim to that single caller, which is what makes this guard fail by leaking `env_vars_static` rather than by admitting a request. --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-21 11:55:42 +02:00
hugocasaandClaude Opus 5	ee533273dd	fix: confine jobs:run tokens to the jobs of the runnables they may start (#10635) ... * fix: confine path-scoped jobs:run tokens to their runnable's jobs Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: project singlestepflow onto its runnable and confine kind-only run scopes Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: keep every by-id job read reachable by a jobs:run token Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: whitelist the dbt and wac-approval by-id job reads for run tokens Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: let an apps:run scope satisfy job-read confinement for that app's runs Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: apply run-scope confinement on top of the approval-token read bypass Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: confine the resume-secret job reads to the run scope as well Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-14 18:57:57 +02:00
Ruben FiszelandClaude Opus 5	633d7bcb2e	feat: add trigger_history table with source tracking (#10696) ... * feat: add trigger_history table with source tracking Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: gate trigger history reads on scopes and harden its writers Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: filter trigger history scopes in SQL and match the cleared-handler diff Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: record a trigger restore from the trashbin in its history Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: record bulk http trigger creates and document the recording boundary Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: lock the trigger row when capturing its history preimage Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: only record an auto-disable that actually flipped the schedule Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * chore: state the auto-disable invariant once instead of at four call sites Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat: render trigger history changes as a structured field diff Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: make a server-initiated disable atomic with its history row Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * docs: note that the auto-disable savepoint takes no pool connection Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * docs: note the flow fallback is the last chance to disable Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: never leave a trigger enabled because its history row failed Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: retry the disable history row instead of dropping it on first failure Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: use the design-system Button for the change-value expander Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: hold the trigger row lock across its disable history row Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: keep the history-loss alert out of the listener cancellation race Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: read the history workspace through the trigger-workspace seam Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-14 17:57:11 +02:00
3 people	7e1c1fa3a4	feat(apps): use the windmill-client SDK from raw app frontend code (#10377) ... * feat(apps): use the windmill-client SDK from raw app frontend code Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): bound the raw app SDK token to deployed runnables Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): deny dependency jobs and survive a failed SDK mint Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): confine the SDK token's users scope to the viewer's identity Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * docs: describe the full raw-app SDK sentinel narrowing Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): deny workflow-as-code replay for raw app SDK tokens Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): deny preview-flow restart replay for raw app SDK tokens Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): re-prompt when an app widens its SDK scopes mid-consent Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * feat(apps): support the frontend SDK in sandboxed raw apps Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): hand the sandboxed SDK token over only once per loaded document Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): bind the sandboxed SDK handoff to the document we loaded Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): use an unguessable nonce for the sandboxed SDK handoff Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): reply to the sandboxed SDK handshake over its own port Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): answer the raw app handshake only over a transferred port Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): set frontend_sdk_scopes in the S3-gated policy literals Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * docs: describe the sandboxed wrapper's credential as it now works Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * ui nit * feat(apps): make the frontend SDK work in the raw app editor preview Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): guard the preview token mint and drop superseded responses Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): refuse job tokens on every raw app SDK mint path Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * docs: correct the mint caller list and the preview retry rationale Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): apply the consent response's render mode before rendering Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): restart the viewer when a redeploy changes the render mode Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): restart on every render-mode change, not just the first Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): clear the preview's SDK credential when scopes go away Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): remove window.process in the preview instead of blanking it Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * docs: cut the raw app SDK comments down to the invariant Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * refactor(apps): use randomUUID for the raw app handshake nonce Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): re-read the render mode before rendering without a token Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * fix(apps): make the raw app handshake nonce unguessable again Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * chore: pin the EE ref to a commit that builds against this OSS tree Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018Gmsk9kAG7p9t2Qy6ADRJz * refactor(apps): authenticate the raw app preview by session instead of a token The editor preview is same-origin and unsandboxed, so app code there already holds the editing user's session cookie. Minting a scoped bearer for it added an endpoint and a portable 12h credential without containing anything. Inject only BASE_URL and WM_WORKSPACE: `windmill-client` falls back to credentialed same-origin requests when it finds no token, so the SDK runs as the editing user. Drops POST /apps/preview_sdk_token and the mint/race handling in the editor. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * fix(sdk): send credentials only outside the browser The API answers `Access-Control-Allow-Origin: *` and never sets `allow_credentials`, so a credentialed cross-origin request fails before the bearer is read — which is what a sandboxed raw app issues. Keying this on the browser rather than on `WM_TOKEN` leaves non-browser callers byte-identical, and browsers keep sending cookies same-origin through fetch's own default. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * remove windmill-client from templates * fix(sdk): drop credentials only for raw app bundles A sandboxed raw app calls the API from an opaque origin, and the API answers `Access-Control-Allow-Origin: *`, which a credentialed request can never pair with. Gate on WM_RAW_APP, set by the two places that build a raw app's `window.process.env`, so every other windmill-client consumer is untouched. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * feat(apps): make frontend SDK access sandbox-only An unsandboxed bundle runs same-origin with the viewer's full session, so a consent prompt there implies a boundary that does not exist and the token adds nothing it could not already do. Advertise scopes and mint only when isolation is on; turning the toggle off clears the declared scopes with it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * Revert "refactor(apps): authenticate the raw app preview by session instead of a token" This reverts commit 81905e455b, restoring POST /apps/preview_sdk_token. Session auth gave the preview the editing user's full permissions and worked regardless of policy, so an app that would 403 for a viewer — or that declares no scopes at all — ran fine in the preview and broke only once deployed. The preview now takes the same credential as a deployed app, gated the same way: sandbox off or no scopes means no env at all. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * docs(apps): state the sandbox-only SDK contract in the public schema The Policy and EmbedTokenResponse descriptions still promised a token to any raw app with non-empty scopes, and said raw apps skip tokens entirely. Point authors at adding windmill-client themselves too, since the starter templates no longer carry it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * fix(apps): drop the preview token before minting its replacement A mint is asynchronous, so clearing the env only on the empty-scope path left the running preview — and any build fed meanwhile — holding scopes the policy had just removed, or a token for the workspace just left, for as long as the request took. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * fix(apps): restart the preview realm when its credential changes Re-feeding the build resets the preview's DOM but keeps its JavaScript realm, so the previous bundle's timers, listeners and pending callbacks went on using the client they imported — and the token it captured at module load — after the policy dropped it. Reload both shells instead; each replays the build on its way back, so only the new realm survives. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * fix(apps): start the preview once per credential change Restarting the realm made its shell replay the build immediately, so a delayed mint ran the app once tokenless and again tokenful — mount-time side effects twice per scope or workspace change. Hold the build back until the mint settles: the shell comes back blank and whichever finishes last starts the app. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * fix(apps): wait for the detached preview shell before replaying Its reload was only initiated, never awaited — unlike the inline iframe it had no readiness flag — so a mint settling first posted the build to the retiring document, which then ran alongside the replacement shell's own replay. Track readiness from both paths that announce it: `load` for a freshly opened window, `appPreviewReady` for a reload. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KPCW1WB5QeYrgJmgwcywNA * chore: update ee-repo-ref to 99e143fa1e2e6c33b3525366a5afe48f7a4f020e This commit updates the EE repository reference after PR #688 was merged in windmill-ee-private. Previous ee-repo-ref: 609e197fbc08f1ce83dd86f816748cc19d213f77 New ee-repo-ref: 99e143fa1e2e6c33b3525366a5afe48f7a4f020e Automated by sync-ee-ref workflow. * nit better description --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>	2026-08-03 19:31:57 +00:00
Ruben FiszelandClaude Opus 5	c69f08073a	fix: add apps:run to the token scope picker and confine path-scoped app tokens (#10428) ... * fix: expose apps:run in the token scope picker and let apps:write grant it Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: confine path-scoped app run/write tokens to the app they name Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: let apps:run read back its own app's S3 files, condense scope comments Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: let apps:write mint apps:run and extend run read-back to app S3 display routes Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * docs: drop stale embed-token wording from the app S3 helper summary Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-07-31 11:38:05 +02:00
3 people	68debab877	feat(triggers): add AMQP (RabbitMQ) trigger via lapin (#10230) ... * feat(triggers): add AMQP (RabbitMQ) trigger using the lapin library Fixes WIN-2214 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(triggers): defer AMQP cross-workspace deploy pending utils-internal publish Revert the amqp_trigger additions to the shared windmill-utils-internal TriggerDeployKind and the frontend cross-workspace deploy adapter: the frontend installs the published npm package, which lacks the new kind until a release is cut. AMQP create/edit/delete/list/sync/capture are unaffected (they use local types); only cross-workspace deploy/merge of AMQP triggers waits on the package bump. Also document the at-most-once ack in the consumer loop. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(triggers): address AMQP review — at-least-once ack, workspace cascade, contracts - ack AMQP deliveries only after successful dispatch; nack+requeue on failure - add ON DELETE CASCADE workspace FK so amqp_trigger rows are cleaned on workspace deletion (and the listener stops) - fix the /amqp_triggers/test OpenAPI body and add amqp_trigger to WorkspaceDiffRow.kind - register AMQP in the generated workspace trigger tool (create_trigger) - drop banned $bindable defaults on optional props in the config section - add build_uri unit tests (encoding, ports, vhost) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(triggers): stop AMQP poison-message loop and reconnect on transient drops Chaos testing against a live RabbitMQ broker showed the previous nack(requeue) + immediate re-poll spun a tight redelivery loop (~1000 critical-error reports/sec) on a poison message, and any connection blip permanently disabled the trigger (lapin has no built-in reconnect). - on dispatch failure: nack+requeue then stop consuming; the listener framework re-lists the trigger after its ping goes stale (~15s), backing redelivery off to that cadence instead of a tight loop (verified: rate dropped from ~1000/s to ~1 per ~26s, message preserved) - on connection/stream error: stop and let the framework reconnect instead of disabling; persistent failures are still disabled via get_consumer (verified: a forced connection close now auto-reconnects and resumes) - finish the AI create-trigger action wiring for AMQP: add amqp to CreatedResourceTriggerKind, the action-card registry, and the drawer registry so the result card renders and its "Open" action works Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(triggers): complete AMQP frontend registries and defer merge rows - add amqp to capturableTriggerTypes (so AmqpCapture mounts), the Runs jobTriggerKinds filter, and CLOUD_DISABLED_TRIGGER_TYPES - wire AMQP into global AI chat mode: TRIGGER_KINDS, the request union, writeTriggerSchema, triggerServices, and the draft adapter - stop emitting actionable AMQP fork-comparison rows (revert amqp_trigger from TRIGGER_OR_SCHEDULE_TABLES) since cross-workspace deploy is deferred until windmill-utils-internal is published — avoids a deploy that fails with "Unknown kind: amqp_trigger" - use design-system TextInput instead of raw <input> in the config section Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(triggers): complete AMQP session/draft registries and constrain prefetch - add amqp to the session-deploy, draft-compare, preview-router, and copilot workspace-item registries so AMQP drafts/deploys/nav/path resolution work - include amqp_count in the MoveDrawer attached-trigger rename warning - replace the raw prefetch <input> with a design-system TextInput bounded to an integer 1-65535 (backend u16) and block save on invalid values Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(triggers): make AMQP disconnect/reconnect consistent with the Kafka trigger lapin, like rdkafka, has no transparent reconnect, so the AMQP listener now mirrors the Kafka trigger's explicit reconnect loop instead of relying on the framework re-list (which disabled the trigger once get_consumer failed on a sustained outage): - get_consumer returns cheaply; consume owns a (re)connect loop that retries with a 30s backoff, reports a critical error every 10 failed attempts, and reports a recovered critical error once it reconnects — never disabling the trigger on a connectivity failure - a consumer/stream error breaks out to reconnect rather than disabling - dispatch failure still nacks+requeues (at-least-once) with a short backoff to avoid a tight poison-message loop, keeping the connection alive Verified against a live RabbitMQ broker: killing the broker keeps the trigger enabled and retrying (attempt N), and restarting it auto-reconnects (logs "reconnected after N attempts") and resumes dispatch. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(triggers): complete AMQP capture registries and constrain prefetch contract - add the 'amqp' case to triggerKindToTriggerType so opening the AMQP editor from a capture button no longer throws "Unknown TriggerKind: amqp" - register AmqpIcon in CaptureTable's icon map and add an AMQP entry to the script/flow CaptureButton menu - bound the OpenAPI prefetch_count to an integer 1-65535 (matches the Rust u16) and regenerate clients/prompts - require a non-empty exchange name when the exchange binding is enabled - build_uri: fall back to "/" on a blank vhost and bracket IPv6 hosts (+ tests) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(triggers): wire AMQP into pipeline graph, git-sync, and preprocessor types - asset_graph: discover attached amqp_trigger rows and emit an AMQP TriggerEdge so AMQP triggers render (and can be opened/deleted) on the data-pipeline canvas - frontend pipeline graph: add amqp to NativeTriggerKind, the add-trigger menu, node presentation, event-trigger set, annotation keywords, and the editor/service registrations - git-sync: add the amqp_trigger include pattern (+ test) so an AMQP git-sync deployment stages only its .amqp_trigger.* file, not an unrelated same-path object - preprocessor starters: add the AMQP event to the generated TS/Python/PHP trigger event types (kind/payload/exchange/routing_key/queue_name/redelivered/ delivery_tag) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(triggers): finish AMQP pipeline/parser wiring, prefetch validation, source lists - fix a stray edit that corrupted the pre-existing MqttTriggerEditor import ($lib/... path) in PipelineTriggerEditors.svelte - reject prefetch_count = 0 server-side in validate_config (RabbitMQ treats 0 as unlimited) and defensively skip basic_qos(0) in build_consumer (covers the capture path that bypasses CRUD validation) - recognize `// on amqp` in the canonical parser (TriggerSpec::Amqp) and add amqp to the CLI non-autorun/event-trigger sets so a pipeline cascade never runs an AMQP-only node as a manual root without an event - add amqp to the preprocessor intro lists and both pipeline AI instructions Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(triggers): reject zero AMQP prefetch in all paths and finish guidance lists - extract a shared validate_amqp_options used by both CRUD validate_config and build_consumer, so capture configs (which bypass CRUD validation) also reject prefetch 0 instead of silently connecting with an unlimited buffer (+ unit tests for 0/1/65535/None) - add AMQP to the main script-writing preprocessor-sources prompt and the CLI triggers-skill guidance list Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * docs(triggers): de-duplicate AMQP prefetch comment and fix GET response text - keep the zero-prefetch rationale only on the shared validate_amqp_options doc; drop the redundant call-site comments - correct the getAmqpTrigger OpenAPI 200 description ("deleted" -> "retrieved") Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore: update ee-repo-ref to aaa6cb89b05b76139252c64f057e53b94d12ac60 This commit updates the EE repository reference after PR #680 was merged in windmill-ee-private. Previous ee-repo-ref: 5da5fd65aca9594b2611837a52e4677b544b0380 New ee-repo-ref: aaa6cb89b05b76139252c64f057e53b94d12ac60 Automated by sync-ee-ref workflow. * chore(migrations): consolidate the four AMQP migrations into one The table and the three enum ADD VALUE statements (trigger_kind, job_trigger_kind, draft_kind) are one atomic feature. ALTER TYPE ... ADD VALUE runs inside the migration transaction on PG >= 14 (Windmill's minimum) since the amqp_trigger table doesn't reference those enum types, so they can share a single migration instead of four. Verified applying cleanly in a single transaction on a fresh DB. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>	2026-07-21 15:10:29 +00:00
Ruben Fiszel	fd51d40f12	feat(pipelines): catalog declared measures and dimensions (#10190)	2026-07-21 07:34:44 +02:00
Ruben FiszelandClaude Opus 4.8	4917f79935	fix(mcp): let MCP tokens call preview run tools (jobs:run scope) — Fixes GIT-920 (#10107) ... * fix(mcp): let MCP tokens call preview run tools (jobs:run scope) The MCP proxy mints an internal JWT scoped to exactly `scope_for_route` for the endpoint it forwards to. For preview run routes (`run/preview`, `run/preview_bundle`, `run/preview_flow`, `run_wait_result/preview`, `run_wait_result/preview_flow`), `determine_kind_from_route` matched the `SCRIPT_JOBS` prefix `jobs/run_wait_result/p` (because "preview" starts with "p") and derived `jobs:run:scripts`. But the preview handlers run arbitrary request-supplied code with no deployed path and require the broad `jobs:run` scope, so `jobs:run:scripts` was rejected with 403 "Required scope: jobs:run". Preview/bundle routes now carry no runnable kind, so the derived scope is the broad `jobs:run` the handlers expect. This also aligns the route-level access check with the handler check for these routes. Fixes GIT-920 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(mcp): anchor preview-route match to endpoint segment Address CI review: `route_path.contains("preview")` also matched by-path runs of a deployed runnable whose path contains "preview" (e.g. `jobs/run_wait_result/p/f/team/preview_report`). Since determine_kind_from_route also feeds check_route_access, such a route would derive the broad `jobs:run` and reject a legitimately kind-scoped `jobs:run:scripts:*`/`jobs:run:flows:*` token with 403. Anchor the exception to the actual preview endpoints (`jobs/run/preview*`, `jobs/run_wait_result/preview*`) so by-path runs keep their kind. Add regression tests for preview-named by-path paths, and trim the comments per AGENTS.md. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-07-15 01:56:41 +02:00
3 people	9d61e4e59e	feat: self-host docs search for chat, mcp, cli; drop inkeep (#9772) ... * feat: self-host docs search for chat, mcp and cli; remove inkeep Embed a vendored docs snapshot (llms.txt/llms-full.txt) in the backend and serve ranking + page rendering from GET /api/docs/{search,page}. The AI chat, the MCP searchDocs/readDocsPage tools, and 'wmill docs' all consume it, so docs search works with no runtime egress and is no longer EE-gated. Removes the inkeep proxy. EE companion deletes inkeep_ee.rs (ee-repo-ref bumped). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * refactor: name read_docs_page param `url` instead of `path` search_docs returns each hit's `Source` URL, so the read tool now takes a `url` argument to match — the AI/MCP loop reads "search gives a Source URL, read takes that url" rather than copying a `Source:` URL into a `path` slot. A bare `/docs/...` path is still accepted and canonicalized before lookup. Regenerated openapi-deref, the MCP endpoint tools, and the frontend client. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * ci: add scheduled workflow to refresh the vendored docs snapshot The backend embeds docs_snapshot/*.gz at build time, so the in-product docs corpus is otherwise only as fresh as the last manual fetch.sh run. This adds a weekly (and manually dispatchable) job that re-runs fetch.sh, sanity-checks the result against truncation/garbage, and opens a PR via the internal app when the snapshot changed — so a human reviews the docs diff before it rides into the next release build. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * refactor: make docs tool-result strings caller-neutral The search/page endpoints back three differently-named consumers (the AI chat `read_docs_page` tool, the MCP `readDocsPage` tool, and the `wmill docs` CLI), so the shared rendered text shouldn't name one of them. Refer to "the docs page-reading tool" and its `url` argument instead, and add tests pinning the caller-neutral follow-up guidance. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * chore: point ee-repo-ref at inkeep-removal companion rebased on EE main The companion branch now carries only the inkeep_ee.rs deletion on top of EE main (was based on the native-job-retry EE line, which polluted the EE PR diff). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix(docs): expose docs:read in token catalog; precompute lowercased corpus Addresses two review nits on the self-hosted docs PR: - docs:read was enforced (ScopeDomain::Docs) but missing from the token scope catalog (token.rs ALL_SCOPES), so it couldn't be selected when creating a standard scoped token in the UI — leaving scope-restricted CLI/MCP docs use effectively ungrantable. Add a read-only "Documentation" group (no write surface) and a test asserting it is exposed. - search ran page.body.to_lowercase() on the whole corpus per query. Lowercase body/title/description once at parse time (into the OnceLock corpus) and scan the precomputed copies instead. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * chore: update ee-repo-ref to 27a4f41b8e5603d6e444efcfc420bd1c44a07eed This commit updates the EE repository reference after PR #630 was merged in windmill-ee-private. Previous ee-repo-ref: c7ec3a0c2fa38d4cb5e50bf0265eef4710de4860 New ee-repo-ref: 27a4f41b8e5603d6e444efcfc420bd1c44a07eed Automated by sync-ee-ref workflow. --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>	2026-06-25 16:32:07 +02:00
4 people	2879cbb65a	feat(apps): opt-in sandbox isolation for published & raw apps (alpha) (#9420) ... * feat(apps): sandbox published & raw apps with a scoped embed token Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore: point ee-repo-ref at embed-token EE commit Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(apps): allow top-navigation from the sandboxed app iframe Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(apps): share app localStorage across apps via the embedder Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(apps): publisher disable-sandbox option with per-version viewer consent Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(sqlx): cache for disable-sandbox queries Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore: bump ee-repo-ref to disable-sandbox EE commit Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): always sandbox the served raw-app wrapper + viewer fixes The raw-app wrapper served by get_raw_app_data now always carries `CSP: sandbox`. The publisher "disable sandbox isolation" opt-out is applied entirely on the viewer side, which (after per-version consent) builds its own same-origin blob wrapper — so the backend-served document stays isolated regardless of how it is reached, never via a relaxed real-origin URL. Also: - CORS on the global /apps_u mount so the opaque viewer can load custom-path public apps cross-origin. - Reject runnable-bridge messages unconditionally until the iframe is bound. - Relay the viewer's in-app hash up to the embedder address bar so deep links stay shareable (hash only; embedder keeps its own pathname). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * feat(apps): render public raw apps single-iframe (drop embed token) Public raw apps now render directly on the real origin with a single opaque bundle iframe and the page credential, instead of the opaque viewer + scoped-token indirection. The author bundle stays isolated in its own opaque iframe (CSP-sandboxed); low-code apps, whose code runs in the viewer frame, keep the opaque viewer + scoped token. embed_token now reports raw_app and skips minting a token for raw apps; the access check still gates visibility. Also set disable_sandbox: None in the remaining Policy constructors so the full feature build (all_sqlx_features, enterprise, license) compiles. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * chore: bump ee-repo-ref to single-iframe raw-app EE commit Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * feat(apps): grandfather existing apps as legacy-unsandboxed + authed-only consent Existing apps are stamped by migration as `legacy_unsandboxed` so they keep running same-origin on upgrade — no breakage and no consent prompt. New apps are sandboxed by default; re-deploying an app clears the flag. The publisher `disable_sandbox` consent prompt is now shown only to authenticated viewers — an anonymous viewer has no session to expose, so the prompt was meaningless friction. embed_token reports `legacy_unsandboxed` and `authed`. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * chore: bump ee-repo-ref to legacy-unsandboxed EE commit Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * feat(apps): deploy-time migration prompt for legacy-unsandboxed apps On the first re-deploy of a grandfathered (legacy-unsandboxed) app, the publisher must explicitly choose: enable sandbox isolation (the flag is cleared → the app becomes sandboxed) or keep running without isolation (→ disable_sandbox, with per-version viewer consent). updatePolicy() no longer carries the legacy flag through a deploy, so the choice is what sticks. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix(apps): disable the sandbox-isolation toggle until the app is deployed The Deploy-drawer "Disable sandbox isolation" toggle called setPublishState() — which updates the app by path — even before the app was first deployed, when the path is empty, throwing an error. Guard it with disabled={!savedApp}, matching the adjacent visibility toggle. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * feat(apps): sandbox the in-workspace low-code app viewer in an opaque iframe Extend the opaque-origin iframe isolation to the logged-in /apps/get viewer. /apps/get becomes an embedder that keeps the workspace chrome + Edit button and renders the app inside a cookieless, chrome-less /app_embed viewer route, handed a scoped embed token minted from the member's session. The app frame runs in an opaque origin (no allow-same-origin), so it cannot reach the member's session cookie or window.parent. - apps.rs: get_app_embed_token_for_path (authed, by-path, scope + RLS gated); mint_app_embed_token grants a path-scoped apps:read:{path} so the viewer can load its own app definition and no other - lib.rs: CORS on /apps (bearer-token only, no cookies) for the opaque viewer's by-path reads - new /app_embed/[workspace]/[...path] viewer route (private analog of /public) - PublicAppFrame: viewerUrl prop to point the opaque iframe at the viewer route Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(apps): unify in-workspace app viewers on the shared sandboxed path Route every in-workspace app display (low-code and raw) through the same PublicAppFrame -> PublicApp machinery as the public viewer, so the sandbox / legacy-unsandboxed / disable-sandbox-consent behavior is identical on every page. - new InWorkspaceAppViewer renders both app types via PublicAppFrame; /apps/get and /apps_raw/get become thin wrappers over it - /apps_raw/get previously rendered RawAppPreview directly (always isolated, with no legacy-grandfathering or consent handling); now consistent with the rest - retire the legacy same-origin raw viewer /apps/get_raw/[version] and re-point the apps-list row to /apps_raw/get; remove the dead /apps_raw/[ws]/[version] route - load the raw bundle secret in the shared viewer (getAppByPath doesn't return it) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): address PR review feedback (scope + policy hardening, nits) - require handler-level apps:read on list_apps / list_search_apps so a scoped embed token cannot read app definitions through the list endpoints. The route layer treats apps:run as satisfying read; the handler check (which does not) closes the gap. - treat legacy_unsandboxed as backend-owned: strip any client-provided value in create/update so it can only be set by the grandfather migration, not the API. - document mint_app_embed_token's caller-verifies-access contract. - use Button's declared onClick prop for the consent action (was onclick, which fell into the rest-spread and bypassed the component's click handling). - test: lock that the embed scopes cannot satisfy domain-level apps:read. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * docs(apps): document embed-token endpoints in openapi + fix doc nit Second-round review nits: - add the three app embed-token endpoints (apps/embed_token/p/{path}, apps_u/embed_token/{secret}, and the EE apps_u/embed_token_by_custom_path) plus the EmbedTokenResponse schema to openapi.yaml; note .html on get_data - mint_app_embed_token doc: "Both" -> "All" (it lists three call sites) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(apps): bound embed-token scopes to the caller's own The embed-token mint now enforces ensure_scopes_within_caller, so the minted scope set is always within the calling credential's own scopes (a no-op for regular unscoped sessions). Adds a unit test locking the boundary and documents the contract on mint_app_embed_token. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): raw-app ctx in external embeds + page credential in direct render - RawAppPreview: engage the storage relay only in opaque frames (probe Web Storage instead of just window.parent), so a public raw app embedded in an external iframe hydrates ctx/storage directly; add a relay-timeout fallback so an unresponsive parent can never stall the ctx handshake. - PublicAppFrame: in direct render, expose the page's own bearer credential through the AuthToken context (JWT public URLs), matching the previous route behavior; opaque-viewer mode keeps the embed token. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(apps): sandbox isolation UI polish + COI embed support for raw apps - Deploy drawer: move the sandbox toggle out of "Public URL" into its own "Sandbox isolation" section (the setting applies to every viewing surface, not just the public URL), with positive phrasing, visible helper text, and state-aware alerts (warning when disabled, info for pre-isolation apps). Toggling it now toasts its own message instead of the login-mode one. - Extract the deploy-time migration prompt into a shared LegacySandboxMigrationModal built on the common Modal component, and wire it into the raw app editor header too (it previously had no prompt, so re-deploying a pre-isolation raw app silently changed behavior). updateRawAppPolicy now also drops the backend-owned legacy flag, matching the low-code updatePolicy. - Viewer consent prompt: use the common ConfirmationModal and show the app path (new appPath prop) instead of the route pathname, falling back to "this app" when the path isn't known yet. - COI embeds: propagate the wm_coep opt-in to the raw-app wrapper document and have the backend assert COEP require-corp on it when the flag is present — required for the bundle iframe to load when the public app page is embedded inside a cross-origin-isolated page. Previously this only worked in dev because the Vite proxy injects the header; the production response lacked it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(apps): app navigation parity across sandboxed and direct viewers - Navbar component: same-app items relay query + hash to the embedder page (which mirrors them onto the root URL, keeping its own pathname and transport params), app items navigate the top page through a validated wm_embed_navigate relay instead of the cookieless viewer iframe, and external items keep opening a new tab. Selected-item detection now recognizes the /app_embed viewer route and ignores transport params. - Frontend-script `goto` and button `onSuccess: gotoUrl`: same-window navigation goes through a shared appNavigateSameWindow helper that relays to the embedder inside the opaque viewer (same-origin paths SPA-navigate, http(s) URLs do a full load, other schemes rejected) and keeps plain window.location everywhere else. - /apps/get and /apps_raw/get: key the viewer by workspace/path so in-route navigation fully remounts it — previously the URL changed but the app (and in sandbox mode its path-scoped token) did not follow. - wm_embed/wm_embedder_origin added to the reserved query params so they no longer leak into the app's ctx.query. - Raw apps: drop the sandbox attribute entirely for the unsandboxed (grandfathered/consented) blob path, matching the pre-isolation viewer exactly — the attribute added no isolation there and sandboxed popups (e.g. OAuth flows). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): preserve grandfathered policy across updates + in-workspace viewer parity Round of compatibility hardening so pre-existing apps behave exactly as before on every surface: - `legacy_unsandboxed` is now preserved across app updates unless the payload explicitly clears it (`false`, sent by the editor's migration prompt and the sandbox toggle). Unrelated update paths — CLI / git-sync redeploys, publish-mode toggles, cross-workspace promotion — no longer silently drop the grandfathering. Clients still can never SET the flag. - The embed-token endpoints (secret, path, EE custom-path) read only the sandbox-decision policy fields, leniently, and no longer mint a token for raw / legacy / disable_sandbox renders: the token is only consumed by the sandboxed low-code render, and minting for the others wrote a useless token row per view and could fail the render for scope-restricted callers. - In-workspace viewer parity with the pre-sandbox `/apps/get`: new `inWorkspace` mode on PublicApp (no "Powered by Windmill" badge / user overlay, no HTML-result approval gate, column flex wrapper, `hideRefreshBar` honored again), and the page's query/hash are forwarded into the opaque viewer so `ctx.query` / `ctx.hash` reach the app. - Raw apps: `window.ctx` is always `{ctx, workspace}` again (anonymous viewers of pre-existing bundles rely on `ctx.workspace`), and the runnable bridge's job-id scoping now applies only to sandboxed renders (`gateJobIds`) — an unsandboxed bundle holds the same credential as the bridge, so gating there only broke pre-existing apps polling persisted or runnable-returned job ids. - Document `disable_sandbox` / `legacy_unsandboxed` in the openapi Policy schema; add a unit test for the lenient policy read. - bump ee-repo-ref to the matching EE commit. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): keep share-link viewer credentials out of the isolated app context The JWT path segment of authenticated share URLs is an embedder-side credential, consumed only to mint the scoped embed token. Two transport channels still copied it into the isolated frame where app-authored code runs: - the opaque viewer iframe src defaulted to window.location.href — the public and custom-path routes now pass a sanitized viewerUrl (JWT segment stripped, query/hash preserved, captured once so the hash relay does not reload the iframe); - document.referrer on the same-origin iframe navigation carried the full embedder URL — both app iframes now set referrerpolicy="no-referrer" (sandboxed renders only for the raw bundle iframe, keeping exact legacy parity; nothing reads the referrer). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(frontend): drop unused import inherited from main merge `slide` import in AssistantMessage.svelte (from #9539) turns `npm run check` red on this branch. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): redirect the removed raw-app viewer path to the unified viewer The old same-origin raw-app viewer route (/apps/get_raw/{version}/{path}) was removed in favor of the sandboxed unified viewer. Re-add a thin client route at the old path that redirects stale bookmarks to /apps_raw/get/{path}, preserving query + hash (the pinned version is dropped — the unified viewer shows latest). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): narrow embed-token scopes and base consent on browser session - Embed token: resource access is metadata-only (list/type/exists) via a `resources:run` marker — resource values (get/get_value/get_value_interpolated/ list_search) are no longer reachable. Job reads are by-id only: an `app_embed` sentinel blocks the workspace-wide job enumeration/export routes (jobs/list, list_filtered_uuids, queue/list, completed/list, queue/export) while by-id result polling keeps working. - disable_sandbox consent now gates on whether the browser holds any Windmill session (cookie-only whoami) rather than workspace-scoped auth, so a viewer logged into a different workspace is still prompted before a same-origin render. - db-explorer: resolve the MySQL database name server-side (the metadata query already falls back to DATABASE()) instead of reading the resource value client-side; getTablesByResource derives the default db from the schema. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(apps): trim embed-scope and consent comments Reduce duplication — state the resource/job route exclusions and the workspace-session-vs-cookie rationale once at their source and reference them elsewhere; drop contrast/justification phrasing. No behavior change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(apps): make app sandbox isolation opt-in (alpha) Replace the disable_sandbox + legacy_unsandboxed policy pair and the per-version viewer consent with a single positive `sandbox` opt-in flag. Apps are unsandboxed by default (same-origin, full session — the pre-isolation behavior), so existing apps are unchanged and no migration is needed. Publishers opt an app into isolation from the deploy drawer, flagged alpha. - Policy.sandbox: Option<bool>; EmbedTokenResponse -> {token, expiration, raw_app, sandbox}; mint an embed token only for sandboxed low-code apps. - Drop the legacy-unsandboxed migration and the deploy-time migration prompt; remove the consent modal and the browser-session probe. - Deploy drawer: a single "Sandbox isolation" toggle (alpha), off by default, shared by the low-code and raw editors. - Bump ee-repo-ref to the companion EE commit. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): confine embed token to its intended user/folder/job routes The embed token's broad read scopes spanned whole domains while the matching routers are CORS-enabled for the opaque app iframe: - users:read / folders:read were domain-wide, so the token could reach users/list, users/list_usage, users/username_to_email/*, folders/list, etc. Restrict to an app_embed-sentinel allowlist: only users/whoami and folders/listnames; deny the rest of those domains. - jobs:read allowed jobs/completed/export, missed by the job denylist. Add it alongside jobs/queue/export. Extend the embed-scope allow/deny test matrix to cover all of these. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * docs(apps): align sandbox comments with the opt-in model The consent prompt, deploy-time migration, and legacy-unsandboxed grandfathering were removed when sandbox isolation became an opt-in policy flag; update the comments that still described them so they match the two-state (default-unsandboxed / opt-in-sandboxed) reality. Comments only, no behavior change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): confine embed-token job reads to runs the app launched App component jobs are stamped `created_by = the viewer`, so an embed token reads its own runs via the launched-by-viewer fast path. The token then also inherited the viewer's broader job access (share links, folder ACLs, admin RLS), letting user-authored app JS reuse it to read unrelated jobs by id. Stop embed tokens at the fast path: only jobs the viewer launched, never those merely visible to them. Return NotFound so the untrusted app can't probe existence. Regression test: an embed token reads its own launched job but is denied the foreign job (result/logs/getupdate) an admin viewer's normal token can read. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): allowlist embed-token apps/jobs routes + scope run to the app The embed token's apps:run/jobs:read reached more than a running app needs. Replace the job denylist with strict per-domain allowlists on the app_embed sentinel: - Apps: only the app's own definition (apps/get/p/<path>) and the public app-serving endpoints (apps_u/*). Denies workspace app inventory (exists, custom_path_exists, list, list_paths*). - Jobs: only the by-id poll routes the frontend JobLoader uses. Denies job counts and the job_signature/resume_urls capability-minting routes (the by-id reads remain confined to the app's own runs). Drop unqualified apps:run from APP_EMBED_SCOPES; mint apps:run:<path> instead and authorize apps:run:<requested path> first in execute_component, so the token can only run its own app's components, not another app's. Extend the embed-scope route matrix and add a path-scoped run unit test. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * docs(apps): clarify the sandbox toggle vs the on-behalf-of model The deploy-drawer sandbox copy leaned on "session" in a way that collided with the on-behalf-of permissioning right above it. Reword it to say the toggle governs what the app's browser-side code can reach in the viewer's browser — distinct from who its runnables execute as — and rename the label to "Isolate the app from the viewer's browser session". Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): path-scope embed-token S3 download to its own app The apps_u/* allowlist also admitted apps_u/download_s3_file/<path>, whose handler authorized any authenticated caller — so an embed token minted for app A could download app B's S3 files via B's on-behalf policy. Add the same path-scoped guard execute_component uses: download_s3_file_from_app now checks apps:read:<path> first, confining the token to its own app. Other path-taking apps_u routes are already covered (writes lack apps:write; embed_token/p path-checks; public_resource is type-constrained). Extend the path-scoping unit test to cover apps:read (download) alongside apps:run (execute). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): path-scope public-app-by-secret read to the embed token's app The apps_u/* allowlist admitted apps_u/public_app/<secret>, whose handler only checked the viewer's read access — so an embed token minted for app A could read app B's definition by secret (confused deputy via the viewer's identity). get_public_app_by_secret now binds a scoped caller to the resolved app with check_scopes(apps:read:<path>), confining it to its own app; unscoped sessions and anonymous access are unchanged. get_raw_app_data needs no binding (pure secret capability, no caller identity). Document the full set of app-resolving handlers the path-scoped read covers. Bump ee-repo-ref for the companion custom-path fix. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): preserve pre-sandbox behavior for db-explorer, edit link, jwt Three behavior-parity fixes for non-sandboxed (existing) apps that the sandbox-isolation refactor changed incidentally: - DB-explorer MySQL table picker: when the connection can see multiple non-system schemas, label the default db's tables unprefixed again. The resource-value read was removed globally, so identify the default db from the introspection script's `DATABASE() AS default_db_name` (carried on SQLSchema.defaultDb) instead of guessing "the single schema key". Equivalent to the prior resource.database match; editor-only (table picker). - In-workspace Edit button: restore `?nodraft=true` on both /apps/get and /apps_raw/get, so opening the editor from the viewer loads the deployed version, not a draft. - Custom-path (/a) viewer: restore the "could not authenticate user with jwt token" toast when a path JWT fails to resolve a user, instead of silently falling through. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): confine embed-token S3 downloads to the app's own keys/outputs download_s3_file_from_app authorized any authenticated caller for any S3 key (opt_authed.is_some() bypass). A sandboxed app's embed token carries the viewer's identity, so app-authored JS could fetch arbitrary S3 keys readable by the on-behalf identity, beyond the app's own declared keys or outputs. Route app embed tokens through the same allowlist as anonymous viewers — the app's declared allowed_s3_keys, or files produced by this app's own component runs — instead of the authed bypass. The produced-files check is parameterized by created_by (the embed viewer for a token, else anonymous) so a sandboxed app's own S3 outputs still render while arbitrary keys are denied. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): let embed tokens cancel their own jobs; gate cancel to launcher A sandboxed low-code app supersedes an in-flight component run on re-run by canceling it, but the embed token only had jobs:read, so cancellation silently failed and prior jobs ran to completion. - Permit the by-id jobs_u/queue/cancel POST for app_embed tokens at the route layer (the only write reachable through the existing by-id allowlist). - Gate cancel_job_api: an app_embed token may cancel ONLY jobs it launched (created_by == viewer). cancel_job_api had no other per-job ownership check, so this also confines the token instead of letting it cancel any job by id. - /app_embed now sets workspaceStore so cancellation targets the right workspace instead of an empty/stale one in the cookieless iframe. Add a shared has_app_embed_sentinel helper; cover cancel in the route matrix and the jobs_read_auth integration test (own job cancelable, foreign denied). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): drop get_root_job_id from the embed-token job allowlist Audit of the embed token's reachable job routes: get_root_job (jobs_u/ get_root_job_id) has no access check in its handler at all — it returns any job's root-job id by id — and the app runtime never calls it. Remove it from the by-id allowlist so the embed token can't probe a foreign job's flow lineage; add a denied-route assertion. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(apps): scope sandboxed-app localStorage per app Sandboxed apps shared one localStorage store (one key on the real origin), so an app could read or clobber another app's keys — and, with job ids stashed there, reuse its embed token to read another app's job. Scope the backing store per app. The embed-token endpoints now return the resolved app_path (EmbedTokenResponse; not a new disclosure — the viewer already receives the path when it loads the app). PublicAppFrame (low-code) and RawAppPreview (raw) key their backing store by it: wm_apps_localstorage:<app_path>. Same app shares one store across its public and in-workspace surfaces; different apps are isolated. Unsandboxed apps are unaffected (real same-origin localStorage, as before). Bump ee-repo-ref for the companion custom-path change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): scope embed access checks to embed tokens + key app storage by workspace - Apply the path-scoped read/run checks on the public-by-secret read and the component run path only when the caller is an app embed token, so other caller types keep their prior access. - Key the sandboxed app's backing client storage by workspace + path instead of path alone, and return the resolved workspace from the embed-token endpoints so the custom-path viewer can derive it. - Show a clear message instead of an indefinite loader when the viewer route is opened outside its embedder. Bumps ee-repo-ref to 5b8476b. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(apps): mint embed tokens only from the trusted embedder caller An app embed token must not reach the embed-token mint endpoints; refresh minting stays with the embedder session/JWT. Enforced at the scope route layer and at the mint chokepoint, with a route-matrix regression test. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix(apps): support S3 upload and frontend-script S3 download in sandboxed apps Sandboxed apps run with a scoped embed token (no cookie). Let the app's S3 file-input upload and the frontend-script download({s3}) helper work in that context: upload is reachable with apps:run and re-checked per-app at the handler; the script download routes through the app-scoped apps_u endpoint with the embed token instead of the cookie-authed job_helpers path. Default (unsandboxed) apps are unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * chore: update ee-repo-ref to b0cb761bf9852974e571b2978032d310cc998517 This commit updates the EE repository reference after PR #600 was merged in windmill-ee-private. Previous ee-repo-ref: e673c714a4618fdb72353a475f49c748e6016642 New ee-repo-ref: b0cb761bf9852974e571b2978032d310cc998517 Automated by sync-ee-ref workflow. --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com> Co-authored-by: Ruben Fiszel <ruben@windmill.dev>	2026-06-23 10:05:00 +02:00
centdixandClaude Opus 4.8	6f4017d694	feat(ai-chat): workspace AI chat skills (SKILL.md upload + read_skill tool) (#9648) ... * feat(ai-chat): workspace ai_skill table + CRUD API Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(ai-chat): AI Skills workspace settings tab with SKILL.md upload Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(ai-chat): advertise skills in global system prompt + read_skill tool Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * refactor(ai-chat): move custom skills into AI settings (paste or folder) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(ai-chat): cap folder import (depth<=3, max 50 skills, confirm dialog) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * style(ai-chat): give import folder its own labeled subsection Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(ai-chat): resolve svelte-check never-narrowing in skills preview Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix: address ai skills review issues * fix: validate ai skills and reload workspace list * fix(ai-chat): spec-align skill validation and cap skills per workspace Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(ai-chat): reject duplicate skill uploads, audit skill names Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(ai-chat): sync deref openapi specs with skill validation rules Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-06-23 00:16:13 +02:00
hugocasaandClaude Opus 4.8	e19594df2a	fix: re-enforce scoped API token boundaries across handlers (#9712) ... * fix: re-enforce per-path token scope on store rename, delete and interpolation Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix: enforce token scope on workspace export and resume-url minting Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix: enforce per-item and runnable scope on trigger create paths Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix: enforce app write scope before persistence and on rename Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix: enforce scope containment on mcp oauth approval Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix: scope mcp endpoint-proxy jwt to the proxied route Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix: treat resource-linked variables and resources as covered by resource scope Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix: only require variables:read for plaintext-secret workspace export Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix: handle singlestepflow resume, reject empty mcp grant, scope var-skipped tarball Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-06-22 23:55:19 +02:00
3 people	d666e8431c	feat: read-only flag on API tokens (#9144) ... * feat: read-only flag on API tokens, orthogonal to scopes Add a per-token `read_only` boolean set at creation time. When true, the token can only call HTTP methods classified as Read (GET/HEAD/OPTIONS). Mutating methods and job-run actions are rejected with 403, regardless of which scopes are attached. Surfaced as a prominent toggle in the standard token-creation flow and a discreet `2xs` toggle in MCP mode (where users often want write access, so we don't bias them toward enabling it). MCP enforcement: read-only tokens hide all script/flow/hub tools from `list_tools` and only see endpoint tools whose method is GET, and the runner rejects `call_tool` on anything mutating. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix: review fixes for read-only token flag - Exempt /api/mcp/* and /mcp/* paths from the read-only middleware check. MCP transport runs over POST (streamable HTTP / SSE), so otherwise the middleware would 403 every MCP request before the runner could enforce read-only at the tool-call level. - Tighten is_endpoint_read_only to GET only, matching the read_only_hint that create_endpoint_annotations actually emits. - Add unit test for check_read_only_for_route covering GET/HEAD/OPTIONS, mutating methods, and run paths. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore: bump ee-repo-ref to read-only-trigger-toggle Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(frontend): make read-only toggle discreet in both modes Match the MCP-mode treatment in standard mode: text-tertiary, 2xs, shared "Read-only" label. The tooltip switches per mode so the explanation still fits the context. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(frontend): gate read-only toggle behind Limit token permissions The read-only toggle now only shows when the user has limited the token's scopes (standard mode) or in MCP mode (which always picks an MCP scope). Turning the limit off also resets read-only so it doesn't silently stick. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(frontend): hide incompatible MCP tools when read-only is on When the read-only toggle is on in MCP mode: - Endpoint badges and the custom-mode endpoint MultiSelect filter to GET. - Already-selected non-GET endpoints are pruned from the scope. - The scripts/flows preview is replaced with a note explaining they're hidden (the runner already rejects script/flow runs for read-only). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(frontend): place read-only toggle at top of limited scope area The previous gate required at least one scope to be picked before the read-only toggle appeared, which made it look missing while the user was still building their scope list. Move the toggle inside ScopesPicker: - Standard mode: sits directly under the "Limit token permissions" toggle whenever Limit is on, before the scope selector. - MCP mode: sits at the top of the MCP scope block. readOnly is now $bindable on ScopesPicker so CreateToken still owns the value. The auto-reset on un-limit moves into ScopesPicker too. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(frontend): nest read-only toggle inside the scope list card Place the read-only toggle at the top of the scope list (between the Selected Scopes summary and the bordered domain list) via a new optional topSlot snippet on ScopeSelector. Keeps ScopeSelector decoupled from read-only specifics; ScopesPicker fills the slot. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore: update ee-repo-ref to 9bc8160be50b3e57a60daf4e1b71c389a6e02b8a This commit updates the EE repository reference after PR #571 was merged in windmill-ee-private. Previous ee-repo-ref: f53d26e6685dfd60bfa67686fbd7358169cfd130 New ee-repo-ref: 9bc8160be50b3e57a60daf4e1b71c389a6e02b8a Automated by sync-ee-ref workflow. * fix: address CI review for read-only token flag - P1 (Codex): narrow the MCP middleware exemption from "any /api/mcp/*" to just the streamable HTTP transport endpoints (/api/mcp/gateway, /api/mcp/w/{ws}/{mcp,sse,list_tools}). Without this, a read-only token could POST /api/mcp/gateway/oauth/server/approve and mint a follow-on non-read-only MCP token via the OAuth code/token exchange. - P2 (Claude/cubic): fix test comment/assertion mismatch — the run-path assertion now exercises GET (which is what the RUN_PATH_ACTIONS elevation comment describes) in addition to POST. Add a regression assertion for /api/mcp/gateway/oauth/server/approve. - P2 (cubic): short-circuit script/flow/hub-script/resource fetches in MCP list_tools when read_only is on — they would only be discarded below, so skipping the DB and resource fan-out is pure win. - P2 (cubic): when scopes are pre-supplied via the CreateToken prop, the ScopesPicker isn't rendered, which previously hid the read-only toggle entirely. Render it next to the pre-supplied scopes display. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>	2026-05-13 15:26:21 +00:00
hugocasaandClaude Opus 4.7	66db873651	fix: surface scope errors as 403 and show real message in CLI (#8953) ... * fix: surface scope errors as 403 and show real message in CLI Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: address review feedback on scope error PR - Backend: also patch handler-level check_scopes (lib.rs:223) — without this, endpoints using check_scopes (scripts, flows, jobs, …) still returned 401 for scope failures, which the CLI would render as the misleading auth message. - CLI: strip backend file refs and the duplicated "Permission denied:" / "Not authorized:" prefix from the surfaced error body. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>	2026-05-01 20:55:49 +00:00
3 people	d6c642b170	feat: add Azure Event Grid triggers (#8888) ... * feat: add Azure Event Grid triggers (EE) Introduces a new enterprise trigger kind `azure` that supports three modes via a single unified trigger type: - basic_push: Azure Event Grid basic — custom topics, system topics (Storage, Resource Manager, Key Vault, etc.), domains (push only) - namespace_push: Event Grid Namespace topics (CloudEvents over HTTP push) - namespace_pull: Event Grid Namespace topics (HTTP pull with lock-token ack/reject for dead-lettering) Auth uses a Service Principal resource (tenant_id, client_id, client_secret, subscription_id). Subscriptions are created in CloudEvents 1.0 schema so the push webhook handler and the pull listener share one payload parser. Backend - New crate `windmill-trigger-azure` (OSS stubs + EE impl symlinked from windmill-ee-private) - Migration `azure_trigger` table with CHECK constraints enforcing mode/columns coherence - `TriggerKind::Azure`, `JobTriggerKind::Azure`, `DeployedObject::AzureTrigger` variants - Push route `/api/azure/w/{workspace}/*path` handles classic Event Grid SubscriptionValidation handshake and CloudEvents 1.0 abuse-protection OPTIONS handshake - Optional inbound JWT validation (audience check only for v1) - Feature flag `azure_trigger` propagated through windmill-api, windmill-store (resource helper), and added to ee_core Frontend - `triggers/azure/` editor with mode toggle (basic/namespace-push/ namespace-pull) and per-mode config (topic ARM id / namespace + topic name / subscription / filters / push auth / pull options) - Registered in icon map, display names, save functions, badge, wrapper, editor, add-trigger menu OpenAPI - `AzureTrigger`, `AzureTriggerData`, `AzureMode`, `AzureSubscriptionMode`, `AzureDeliveryConfig`, `TestAzureConnection` schemas; `/azure_triggers/*` endpoints; client regenerated Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore: update ee-repo-ref to eaa7c3a9cb37a9ccc93f10a2535d929365acd2d8 This commit updates the EE repository reference after PR #541 was merged in windmill-ee-private. Previous ee-repo-ref: 9689014e8c12c36c1059fd8fa5758d550b8b8bc9 New ee-repo-ref: eaa7c3a9cb37a9ccc93f10a2535d929365acd2d8 Automated by sync-ee-ref workflow. * feat(azure-trigger): secret-auth push, ARM discovery, capture isolation, CLI + parity Frontend: - Split mode selector into Namespace/Basic + Pull/Push - ARM resource dropdowns (namespaces, Basic topics, namespace topics) populated from the service principal; cascade with stale-selection reset on SP / edition change - Remove stale authenticate toggle + audience input (server-managed push_auth_config has replaced them) - Azure listing page: "Create from template" button; "Also delete Azure subscription" toggle in the delete modal; simplified trigger label falling back to path - AzureCapture.svelte: "Test subscription name" with -wm-capture suffix - CompareWorkspaces.svelte: wire Azure for fork/compare - Drop Trigger-deployed/event-loss warning (capture subscription is isolated with -wm-capture) Backend: - Shared-secret push auth (see EE crate for detail) - JSONB push_auth_config column (renamed from delivery_config), #[serde(skip)] so clients/CLI/exports never see it - Drop redundant enabled column; mode supersedes - Azure capture infra: AzureTriggerConfig + set_azure_trigger_config + azure_payload route + TriggerKind::Azure arm; PT15M queue TTL on capture subscriptions so they bound storage after tab close - Granular ACLs, users offboarding, trash, git-sync deployed-object: all include azure_trigger CLI: - Add azure to TRIGGER_TYPES, pushObj dispatch, getTypeStrFromPath, trigger commands (get/update/create/list/template), sync delete switch + regex; e2e test for `trigger new --kind azure` - system_prompts: SCHEMA_MAPPINGS + schema_names include AzureTrigger; auto-generated/* regenerated Skill: - .claude/skills/adding-a-trigger/ checklist covering every file that needs editing when wiring a new trigger type (learned from this PR) ee-repo-ref bumped to b0e490cbf3724b7b64c6a5b010e3bdf24acd873c. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(azure-trigger): ci — ShareModal Kind + regenerated system_prompts - frontend/src/lib/components/ShareModal.svelte: add 'azure_trigger' to the Kind type so the listing page's "Permissions" action compiles (ts2345 — caught by npm_check on CI, missed by fast-check locally). - system_prompts/auto-generated/: regenerate to drop the stale delivery_config / AzureDeliveryConfig fields from the Azure schema (check-freshness on CI). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * refactor(azure-trigger): use workspace constant_time_eq crate Drop hand-rolled constant-time compare in favour of the workspace constant_time_eq crate (same one used by http_trigger_auth). ee-repo-ref bumped to 9659382d47286e7f7f66d01b6f5dd8d4ed34848b. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(azure-trigger): pass placeholder + disabled via inputProps `TextInput`'s `placeholder` and `disabled` go through its `inputProps` prop — CI's `npm run check` caught the stale top-level passing that `npm run check:fast` missed. Align with the DefaultEmailConfigSection pattern. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(azure-trigger): correct LATEST_GIT_SYNC_SCRIPT_PATH version to 28213 The hub deploy of the azure-aware sync-script is version 28213, not 28214. Backend was pinning a non-existent hub script, which broke the git_sync_e2e suite (every deploy's sync step 404'd). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(azure-trigger): add azure_triggers to token scope selector + skill - windmill-api/src/token.rs: `build_trigger_scope_domains` was missing `("azure_triggers", "Azure Event Grid")`, so the CreateToken UI's scope selector didn't surface azure_triggers:read/write. Backend already had `ScopeDomain::AzureTriggers` wired (scopes.rs), this just exposes it. - .claude/skills/adding-a-trigger/SKILL.md: capture both scope-related files under the hardcoded-arrays section so future triggers don't miss the UI surface. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * docs(adding-a-trigger-skill): clarify token.rs scope effect Not a regression — nothing was working before. Skipping TRIGGER_DOMAINS just means the scope works via API/CLI but has no UI checkbox. * docs(adding-a-trigger-skill): trim token.rs bullet * fix(azure-trigger): regen openapi-deref + swap textarea for TextInput - Run build_openapi.sh to regenerate openapi-deref.{yaml,json} with the 12 azure_triggers paths + schemas. These files are served by the runtime (include_str! in windmill-api/src/lib.rs) to external SDK consumers; without this regen the new endpoints wouldn't be advertised. - Replace the raw <textarea> for event type filters with the design-system TextInput in textarea mode (frontend/CLAUDE.md bans raw HTML elements). Addresses cubic + claude PR review items. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>	2026-04-23 16:30:18 +00:00
centdixandClaude Opus 4.5	aea74445a3	fix: add flow conversation token scope (#8903) ... * fix: add flow conversation token scope Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: make flow conversations scope plural Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: update flow chat service import Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>	2026-04-22 13:42:22 +00:00
hugocasaandClaude Opus 4.6	d938625785	feat: add download all logs button for flow jobs (#8748) ... * feat: add download all logs button for flow jobs Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: use recursive CTE to include all nested flow jobs in log download Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: start iteration index at 1 and interleave children with parents Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: distinguish branch vs loop iteration in log section headers Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: include flownode and singlestepflow kinds in branch/iteration labels Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: improve branch labels (branchone: default/1/2, branchall: 1/2) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: resolve module types from flow_node table for nested structures Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: use full path in iteration/branch labels and show step kind name Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: show iteration index for simple module forloop optimized jobs Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: handle aiagent jobs as intermediate flow jobs with tool call children Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * refactor: reuse existing get_logs_from_store/disk instead of duplicating Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * sqlx * sqlx --------- Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	2026-04-07 17:59:16 +00:00
Ruben FiszelandClaude Opus 4.6	f5fc9f8485	fix: require mcp: scope for MCP endpoints instead of blanket bypass (#8597) ... Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	2026-03-29 12:39:48 +00:00
Ruben FiszelandClaude Opus 4.6	9ff8a85af6	refactor: extract windmill-api into subcrates for parallel compilation (#7845) ... * refactor: extract windmill-api into 4 subcrates (api-auth, store, api-sse, api-jobs) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor: eliminate refresh_token OnceLock bridge in windmill-store Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: eliminate FromRequestParts OnceLock bridge in windmill-api-auth Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor: wire subcrates into workspace and clean up unused re-exports Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: resolve cargo check --all-features errors in subcrate wiring Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * sqlx * all * chore: update ee-repo-ref for warning fixes Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor: extract windmill-trigger crate and expand windmill-api-jobs Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: extract windmill-trigger-kafka crate from windmill-api Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: extract windmill-trigger-postgres crate from windmill-api Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: extract windmill-trigger-websocket and windmill-trigger-mqtt crates Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: extract windmill-trigger-nats, sqs, gcp, and email crates Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: extract windmill-trigger-http crate from windmill-api Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: move token creation and permission helpers to windmill-api-auth Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: extract windmill-native-triggers crate from windmill-api Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * sqlx * all * refactor: extract windmill-api-embeddings crate and fix CI warnings Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: resolve type mismatch in oauth2_oss and remaining warnings Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: use correct HTTP_CLIENT config in embeddings crate (30s timeout, cert override) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * all * fix: gate oauth_refresh_ee on oauth2 feature to fix warnings Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * all --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>	2026-02-07 22:12:55 +00:00