Skip to content

fix(senpi-codemode): format eval elapsed duration - #743

Merged
code-yeongyu merged 2 commits into
mainfrom
fix/tool-call-elapsed-rendering
Aug 6, 2026
Merged

fix(senpi-codemode): format eval elapsed duration#743
code-yeongyu merged 2 commits into
mainfrom
fix/tool-call-elapsed-rendering

Conversation

@code-yeongyu

@code-yeongyu code-yeongyu commented Aug 6, 2026

Copy link
Copy Markdown
Owner

Summary

  • format completed eval durations with the existing compact elapsed formatter
  • keep the simple-result path consistent with cell headers, agent progress, and nested tool-call widgets
  • add focused boundary and adjacent-frame regression coverage

Changes

  • resultMetadata() now renders values such as 1_250ms, 61_000ms, and 3_720_000ms as 1s, 1m 1s, and 1h 2m
  • very short completed/error results consistently render <1s
  • fork behavior is recorded in packages/senpi-codemode/changes.md

QA & Evidence

  • RED: focused boundary test failed 4/4 on raw took <N>ms
  • RED: integrated frame differed only at took 1250ms versus took 1s
  • GREEN: focused elapsed suite 5/5
  • GREEN: full @code-yeongyu/senpi-codemode suite 515 passed, 6 pre-existing skipped
  • npm run check passed
  • package build passed
  • xterm.js parsed-cell assertions passed 2/2
  • agent-browser screenshot confirmed phase summarizing | took 1h 2m
  • two independent visual QA reviewers returned PASS with no blockers

Local evidence: local-ignore/qa-evidence/20260806-eval-elapsed-rendering/

Risks & Residuals

  • intentionally does not change live footer, working-status, thinking-duration, or timeout-error formatting
  • no schema, persistence, API, or concurrency behavior changes

Summary by cubic

Fix inconsistent eval duration labels by using the compact formatter for final results, so times show as “took 1s”, “1m 1s”, etc., instead of raw milliseconds. This matches cell headers and nested tool-call widgets in senpi-codemode, and is documented in the changelog.

  • Bug Fixes
    • Route durationMs through the compact formatter in resultMetadata() to render “took 1s”, “1m 1s”, “1h 2m”.
    • Show “took <1s” for very short completed/error results.
    • Add test/eval-result-duration.test.ts and update existing tests; record the change in packages/senpi-codemode/CHANGELOG.md and changes.md.
    • Scope: no changes to live footer, working-status, thinking-duration, or timeout-error formatting.

Written for commit 250f8f9. Summary will update on new commits.

Review in cubic

@code-yeongyu
code-yeongyu merged commit 36b0a9b into main Aug 6, 2026
15 checks passed
@code-yeongyu
code-yeongyu deleted the fix/tool-call-elapsed-rendering branch August 6, 2026 05:03
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant