[AGE-3777] fix(frontend): close trace→playground gaps from #4426#4453
Open
ardaerzin wants to merge 2 commits into
Conversation
… links The trace drawer's side panel renders application / variant / environment tags with links built from the trace's references. When a trace is emitted with slug-only refs (the common case for third-party instrumentations), the raw ref objects carry no `id`, so the link templates collapse to empty hrefs and the panel shows non-clickable text. PR #4425 added an imperative resolver (`resolveTraceRefs`) that maps slug-only refs back to concrete `{appId, revisionId}` via the backend's `POST /workflows/revisions/retrieve` endpoint, but only the Playground button consumed it. This change lifts that resolver into a TanStack-Query atom (`resolvedTraceRefsAtomFamily`) so the side panel can subscribe to the same resolved data and produce working links for application, variant, and environment tags on slug-only traces. The atom family is keyed by a stable encoding of the identifying refs (`buildResolvedTraceRefsKey`) so two spans that share the same triple share the same atom and the same query cache entry. A sentinel `EMPTY_TRACE_REFS_KEY` keeps the atom disabled when no identifying ref is present, so consumers can call it unconditionally without a no-op fetch. Scope: application-side refs only. Evaluator slug resolution (different backend endpoint) and ephemeral playground gaps (issue #4426 problems 2 and 3) are out of scope for this branch. Tests cover key encoding, the empty-sentinel path, the disabled-state shape (no projectId, no identifying ref), and the success path that forwards refs to the underlying resolver.
Bundles fixes for the remaining four problems from issue #4426 (the trace→playground flow) on top of #4425's resolver. Problem 1 (slug-only side panel) shipped on the first commit of this branch. Problem 3 — evaluator spans open the graded app (playgroundController.ts): PR #4425's resolver backfills `refs.application_revision.id` for any span carrying app refs. Evaluator spans always carry the graded app's refs alongside the evaluator's, so the application_revision branch fired before the evaluator_revision branch and the user landed on the wrong entity. Extracts the branch-selection logic into a pure helper `selectOpenFromTraceBranch(refs, isEvaluatorSpan)` and gates the application branch on `!isEvaluatorSpan`. The helper is unit-tested in isolation so the regression cannot reappear without a test failure. Problem 2d — environment refs dropped (traceRefResolution.ts, resolvedTraceRefsAtom.ts, retrieveWorkflowRevision): Env refs (environment / environment_variant / environment_revision) are now extracted by `extractReferences`, included in `IDENTIFYING_REF_KEYS` so `hasAppReference` enables the Playground button for env-only traces, forwarded to `POST /workflows/revisions/ retrieve` (the Fern client already accepts these fields), and included in `buildResolvedTraceRefsKey` so the side panel atom resolves them too. End result: env-only traces open the deployed revision instead of disabling the button. Problem 2a/2c — ephemeral has no URI, config panel empty (store.ts createEphemeralWorkflow): Without `data.uri` the playground cannot fetch an OpenAPI schema and falls back to the empty "No configuration needed" state, hiding the parameters we already extract. The constructor now defaults the URI to `agenta:builtin:completion:v0` / `agenta:builtin:chat:v0` based on detected chat mode for non-evaluator ephemerals — same builtin URIs the runnable layer already uses. Evaluator ephemerals continue to take their URI from `deriveBuiltinUriFromSpanName` at the call site. Problem 2b — "Create" spawns a new app instead of forking the source (commit.ts createWorkflowFromEphemeralAtom): Reads `meta.sourceRef.type === "application"` and, when an app id is available, forks the source workflow as a new variant (seed v0 + data v1, mirroring `createWorkflowVariantAtom`) instead of calling `createWorkflow`. Falls back to the new-workflow path when no source id is available (third-party traces with slug-only refs that didn't resolve). This branch lacks dedicated unit tests because the function's dependency surface is heavy; manual QA covers the fork vs new-app decision. Tests - traceRefResolution.test.ts: 13 new cases — `hasAppReference` with env refs, `resolveTraceRefs` forwarding env refs (single / combined / variant / revision), `selectOpenFromTraceBranch` (5 branch scenarios including the issue-#4426 problem-3 regression). - resolvedTraceRefsAtom.test.ts: env refs included in the cache key and produce distinct keys from app-only refs. - retrieveWorkflowRevision.test.ts: forwards env refs in the request body and treats an env ref as identifying. - createEphemeralWorkflow.test.ts (new file): URI derivation for completion / chat / explicit / evaluator-without-uri cases, and parameter pass-through.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
dosubot
Bot
added
the
size:L
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Contributor
Railway Preview Environment
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Closes the trace→playground gaps reported in #4426, stacked on top of #4425 (the resolver). Five problems addressed:
Problem 1 — Slug-only refs render dead tags in the side panel. Lifts #4425's
resolveTraceRefsinto a TanStack-Query atom (resolvedTraceRefsAtomFamily) keyed by a stable encoding of the identifying refs. The trace drawer's side panel subscribes and falls back to resolvedappId/revisionIdwhen the raw refs are slug-only, so Application, Variant, and Environment tags now build proper hrefs.Problem 3 — Evaluator spans open the graded app. #4425's resolver backfills
refs.application_revision.idfor any span carrying app refs. Evaluator spans always carry the graded app's refs alongside their own, so the application_revision branch fired before the evaluator_revision branch. Extracts the gate into a pure helperselectOpenFromTraceBranch(refs, isEvaluatorSpan)and skips the application branch for evaluator spans.Problem 2d — Environment refs dropped.
extractReferencesnow readsenvironment/environment_variant/environment_revision, the resolver forwards them toPOST /workflows/revisions/retrieve(the Fern client already accepts these fields), andhasAppReferenceenables the Playground button for env-only traces. The side panel atom key includes env refs so env-only traces also resolve their environment-tag link.Problem 2a / 2c — Ephemeral has no URI, config panel empty even when parameters exist.
createEphemeralWorkflownow defaults the URI toagenta:builtin:completion:v0oragenta:builtin:chat:v0(chat detection mirrors the existing logic) for non-evaluator ephemerals, so the playground can fetch the schema and the parameters we already extract render against a real form. Evaluator ephemerals continue to take their URI fromderiveBuiltinUriFromSpanName.Problem 2b — "Create" spawns a disconnected new app.
createWorkflowFromEphemeralAtomnow readsmeta.sourceRefand, when the ephemeral was opened from an existing application's trace, forks the source workflow as a new variant (seed v0 + data v1) instead of callingcreateWorkflow. Falls back to the new-workflow path when no source id is available (e.g. third-party traces with slug-only refs that didn't resolve).Testing
Verified locally
pnpm test --runinweb/packages/agenta-playground→ 60/60 unit tests pass (35 existing + 25 new).pnpm exec vitest run tests/unitinweb/packages/agenta-entities→ 344/344 unit tests pass (327 existing + 17 new).pnpm lint-fixacross theweb/workspace → clean.Added or updated tests
web/packages/agenta-playground/tests/unit/traceRefResolution.test.ts— 13 new cases.hasAppReferencewith env refs;resolveTraceRefsforwarding env refs (single / variant / revision / combined with app);selectOpenFromTraceBranchcovering app-revision, evaluator-revision, evaluator-with-also-app-revision (regression guard for Problem 3), and empty-string id handling.web/packages/agenta-playground/tests/unit/resolvedTraceRefsAtom.test.ts— env-ref cache-key encoding (env-only produces a non-empty key, different env slugs produce different keys, app+env distinct from app-only).web/packages/agenta-entities/tests/unit/retrieveWorkflowRevision.test.ts— env refs forwarded in the request body and treated as identifying refs.web/packages/agenta-entities/tests/unit/createEphemeralWorkflow.test.ts(new file) — URI derivation for completion / chat / explicit / evaluator-without-uri cases plus parameter pass-through.Problem 2b's fork-as-variant branch in
createWorkflowFromEphemeralAtomhas no dedicated unit test because the function's dependency surface (5+ atoms, 3+ API functions, callback registry) makes a focused mock setup prohibitive. The branching itself is straightforward — manual QA covers it.QA follow-up
ag.refs.application.idis missing butslugis present). Confirm the Applications / Variants / Environments tags in the right side panel render with working links.ag.refs.environment.*reference (no application-side refs). Confirm the Playground button is enabled and clicking it resolves to the deployed revision.Demo
N/A for the data-layer changes (Problems 2a/2c/2d/3) — purely behavioral. Problem 1 and 2b have visible UI effects but no new components were added, so screenshots aren't substantially different from before/after the data wiring.
Checklist
Contributor Resources