[web-shared] Fix trace viewer run span segments for v1 events

[karthikscale3]

Summary

• Fix computeRunSegments to fall back to the run entity's status when no run_completed/run_failed events exist in the event log. V1 runs (specVersion 1) don't emit run lifecycle events, so completed runs were incorrectly rendered as "running" (blue) instead of "succeeded" (green) or "failed" (red).
• Add test script and vitest to @workflow/web-shared so tests run in CI.
• Add unit tests covering the trace builder and segment computation for v1 runs.

Context

PR #1261 refactored the trace viewer to build spans entirely from events instead of fetching Step/Hook entities separately. This works for v2 runs which emit run_created/run_started/run_completed events, but broke v1 runs where these events don't exist.

Changes

• packages/web-shared/src/components/trace-viewer/components/span-segments.ts — read run.status from span.attributes.data as fallback when no terminal event is found
• packages/web-shared/package.json — add "test": "vitest run" and vitest devDependency
• packages/web-shared/test/trace-builder-v1.test.ts — 9 tests covering v1 completed, failed, running, queued states and v2 baseline

[VaguelySerious]

This is more implicit than checking for the absence of a run_created event I think, and also maybe triggers accidentally if we're on a v2 run but haven't loaded the final events for the run yet.

Instead, I think we should, further above, check for the absence of a run_created event, and if so, then look at the run status and construct the failed/completed part, and infer the run_started part too, if available.

[karthikscale3]

yea, good point! will update

[karthikscale3]

updated

[karthikscale3]

Summary

• Fix computeRunSegments to fall back to the run entity's status when no run_completed/run_failed events exist in the event log. V1 runs (specVersion 1) don't emit run lifecycle events, so completed runs were incorrectly rendered as "running" (blue) instead of "succeeded" (green) or "failed" (red).
• Add test script and vitest to @workflow/web-shared so tests run in CI.
• Add unit tests covering the trace builder and segment computation for v1 runs.

Context

PR #1261 refactored the trace viewer to build spans entirely from events instead of fetching Step/Hook entities separately. This works for v2 runs which emit run_created/run_started/run_completed events, but broke v1 runs where these events don't exist.

Changes

• packages/web-shared/src/components/trace-viewer/components/span-segments.ts — read run.status from span.attributes.data as fallback when no terminal event is found
• packages/web-shared/package.json — add "test": "vitest run" and vitest devDependency
• packages/web-shared/test/trace-builder-v1.test.ts — 9 tests covering v1 completed, failed, running, queued states and v2 baseline

[VaguelySerious]

This is more implicit than checking for the absence of a run_created event I think, and also maybe triggers accidentally if we're on a v2 run but haven't loaded the final events for the run yet.

Instead, I think we should, further above, check for the absence of a run_created event, and if so, then look at the run status and construct the failed/completed part, and infer the run_started part too, if available.

[karthikscale3]

yea, good point! will update

[karthikscale3]

updated

[karthikscale3]

Summary

• Fix computeRunSegments to fall back to the run entity's status when no run_completed/run_failed events exist in the event log. V1 runs (specVersion 1) don't emit run lifecycle events, so completed runs were incorrectly rendered as "running" (blue) instead of "succeeded" (green) or "failed" (red).
• Add test script and vitest to @workflow/web-shared so tests run in CI.
• Add unit tests covering the trace builder and segment computation for v1 runs.

Context

PR #1261 refactored the trace viewer to build spans entirely from events instead of fetching Step/Hook entities separately. This works for v2 runs which emit run_created/run_started/run_completed events, but broke v1 runs where these events don't exist.

Changes

• packages/web-shared/src/components/trace-viewer/components/span-segments.ts — read run.status from span.attributes.data as fallback when no terminal event is found
• packages/web-shared/package.json — add "test": "vitest run" and vitest devDependency
• packages/web-shared/test/trace-builder-v1.test.ts — 9 tests covering v1 completed, failed, running, queued states and v2 baseline

[VaguelySerious]

This is more implicit than checking for the absence of a run_created event I think, and also maybe triggers accidentally if we're on a v2 run but haven't loaded the final events for the run yet.

Instead, I think we should, further above, check for the absence of a run_created event, and if so, then look at the run status and construct the failed/completed part, and infer the run_started part too, if available.

[karthikscale3]

yea, good point! will update

[karthikscale3]

updated

[karthikscale3]

Summary

• Fix computeRunSegments to fall back to the run entity's status when no run_completed/run_failed events exist in the event log. V1 runs (specVersion 1) don't emit run lifecycle events, so completed runs were incorrectly rendered as "running" (blue) instead of "succeeded" (green) or "failed" (red).
• Add test script and vitest to @workflow/web-shared so tests run in CI.
• Add unit tests covering the trace builder and segment computation for v1 runs.

Context

PR #1261 refactored the trace viewer to build spans entirely from events instead of fetching Step/Hook entities separately. This works for v2 runs which emit run_created/run_started/run_completed events, but broke v1 runs where these events don't exist.

Changes

• packages/web-shared/src/components/trace-viewer/components/span-segments.ts — read run.status from span.attributes.data as fallback when no terminal event is found
• packages/web-shared/package.json — add "test": "vitest run" and vitest devDependency
• packages/web-shared/test/trace-builder-v1.test.ts — 9 tests covering v1 completed, failed, running, queued states and v2 baseline

[VaguelySerious]

This is more implicit than checking for the absence of a run_created event I think, and also maybe triggers accidentally if we're on a v2 run but haven't loaded the final events for the run yet.

Instead, I think we should, further above, check for the absence of a run_created event, and if so, then look at the run status and construct the failed/completed part, and infer the run_started part too, if available.

[karthikscale3]

yea, good point! will update

[karthikscale3]

updated

[karthikscale3]

Summary

• Fix computeRunSegments to fall back to the run entity's status when no run_completed/run_failed events exist in the event log. V1 runs (specVersion 1) don't emit run lifecycle events, so completed runs were incorrectly rendered as "running" (blue) instead of "succeeded" (green) or "failed" (red).
• Add test script and vitest to @workflow/web-shared so tests run in CI.
• Add unit tests covering the trace builder and segment computation for v1 runs.

Context

PR #1261 refactored the trace viewer to build spans entirely from events instead of fetching Step/Hook entities separately. This works for v2 runs which emit run_created/run_started/run_completed events, but broke v1 runs where these events don't exist.

Changes

• packages/web-shared/src/components/trace-viewer/components/span-segments.ts — read run.status from span.attributes.data as fallback when no terminal event is found
• packages/web-shared/package.json — add "test": "vitest run" and vitest devDependency
• packages/web-shared/test/trace-builder-v1.test.ts — 9 tests covering v1 completed, failed, running, queued states and v2 baseline

[VaguelySerious]

This is more implicit than checking for the absence of a run_created event I think, and also maybe triggers accidentally if we're on a v2 run but haven't loaded the final events for the run yet.

Instead, I think we should, further above, check for the absence of a run_created event, and if so, then look at the run status and construct the failed/completed part, and infer the run_started part too, if available.

[karthikscale3]

yea, good point! will update

[karthikscale3]

updated

[karthikscale3]

Summary

• Fix computeRunSegments to fall back to the run entity's status when no run_completed/run_failed events exist in the event log. V1 runs (specVersion 1) don't emit run lifecycle events, so completed runs were incorrectly rendered as "running" (blue) instead of "succeeded" (green) or "failed" (red).
• Add test script and vitest to @workflow/web-shared so tests run in CI.
• Add unit tests covering the trace builder and segment computation for v1 runs.

Context

PR #1261 refactored the trace viewer to build spans entirely from events instead of fetching Step/Hook entities separately. This works for v2 runs which emit run_created/run_started/run_completed events, but broke v1 runs where these events don't exist.

Changes

• packages/web-shared/src/components/trace-viewer/components/span-segments.ts — read run.status from span.attributes.data as fallback when no terminal event is found
• packages/web-shared/package.json — add "test": "vitest run" and vitest devDependency
• packages/web-shared/test/trace-builder-v1.test.ts — 9 tests covering v1 completed, failed, running, queued states and v2 baseline

[VaguelySerious]

This is more implicit than checking for the absence of a run_created event I think, and also maybe triggers accidentally if we're on a v2 run but haven't loaded the final events for the run yet.

Instead, I think we should, further above, check for the absence of a run_created event, and if so, then look at the run status and construct the failed/completed part, and infer the run_started part too, if available.

[karthikscale3]

yea, good point! will update

[karthikscale3]

updated

[karthikscale3]

Summary

• Fix computeRunSegments to fall back to the run entity's status when no run_completed/run_failed events exist in the event log. V1 runs (specVersion 1) don't emit run lifecycle events, so completed runs were incorrectly rendered as "running" (blue) instead of "succeeded" (green) or "failed" (red).
• Add test script and vitest to @workflow/web-shared so tests run in CI.
• Add unit tests covering the trace builder and segment computation for v1 runs.

Context

PR #1261 refactored the trace viewer to build spans entirely from events instead of fetching Step/Hook entities separately. This works for v2 runs which emit run_created/run_started/run_completed events, but broke v1 runs where these events don't exist.

Changes

• packages/web-shared/src/components/trace-viewer/components/span-segments.ts — read run.status from span.attributes.data as fallback when no terminal event is found
• packages/web-shared/package.json — add "test": "vitest run" and vitest devDependency
• packages/web-shared/test/trace-builder-v1.test.ts — 9 tests covering v1 completed, failed, running, queued states and v2 baseline

[VaguelySerious]

This is more implicit than checking for the absence of a run_created event I think, and also maybe triggers accidentally if we're on a v2 run but haven't loaded the final events for the run yet.

Instead, I think we should, further above, check for the absence of a run_created event, and if so, then look at the run status and construct the failed/completed part, and infer the run_started part too, if available.

[karthikscale3]

yea, good point! will update

[karthikscale3]

updated

[karthikscale3]

Summary

• Fix computeRunSegments to fall back to the run entity's status when no run_completed/run_failed events exist in the event log. V1 runs (specVersion 1) don't emit run lifecycle events, so completed runs were incorrectly rendered as "running" (blue) instead of "succeeded" (green) or "failed" (red).
• Add test script and vitest to @workflow/web-shared so tests run in CI.
• Add unit tests covering the trace builder and segment computation for v1 runs.

Context

PR #1261 refactored the trace viewer to build spans entirely from events instead of fetching Step/Hook entities separately. This works for v2 runs which emit run_created/run_started/run_completed events, but broke v1 runs where these events don't exist.

Changes

• packages/web-shared/src/components/trace-viewer/components/span-segments.ts — read run.status from span.attributes.data as fallback when no terminal event is found
• packages/web-shared/package.json — add "test": "vitest run" and vitest devDependency
• packages/web-shared/test/trace-builder-v1.test.ts — 9 tests covering v1 completed, failed, running, queued states and v2 baseline

[VaguelySerious]

This is more implicit than checking for the absence of a run_created event I think, and also maybe triggers accidentally if we're on a v2 run but haven't loaded the final events for the run yet.

Instead, I think we should, further above, check for the absence of a run_created event, and if so, then look at the run status and construct the failed/completed part, and infer the run_started part too, if available.

[karthikscale3]

yea, good point! will update

[karthikscale3]

updated