fix: surface real causes behind sandbox result errors (PRDENG-58)#68
Open
nojibe wants to merge 1 commit into
Open
Conversation
Three related fixes so a failed sandbox run shows operators why it failed instead of a wall of generic errors. - KeyPointCoverageTable: replace the generic "Error loading evaluation data" empty state with a diagnosis that reports the actual coverageResult.error (generation failure vs. evaluation failure vs. missing/empty coverage). - SharedEvaluationComponents: add a dedicated "Not Evaluated" bucket for criteria whose coverageExtent is undefined/NaN. N/A criteria no longer render under the green "Passed Criteria" header. - llm-coverage-evaluator: when all judges fail, categorize each judge's failure reason (rate-limited, auth, parse, upstream 5xx, etc.) and include the summary in the assessment error, plus attach the full judgeLog so the existing per-point Log expander in the UI can show operators the trace. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
Claude Code Review
This repository is configured for manual code reviews. Comment @claude review for a one-time review, or @claude review always to subscribe this PR to a review on every future push.
Tip: disable this comment in your organization's Code Review settings.
Contributor
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Addresses PRDENG-58 — a sandbox run showed three different generic errors and no way to tell what actually failed. This PR replaces each with a diagnostic that surfaces the real cause.
KeyPointCoverageTable— the empty-state card ("Error loading evaluation data for this model.") now diagnoses why there's no data: reports the actualcoverageResult.errorstring(s) when present, and distinguishes generation failure vs. evaluation failure vs. missing coverage vs. empty assessments.SharedEvaluationComponents— adds a dedicated Not Evaluated bucket for criteria whosecoverageExtentisundefined/NaN. N/A criteria no longer render under the green Passed Criteria header (which was misleading when judges failed and every criterion was unscored).llm-coverage-evaluator— when all judges fail in consensus mode, categorizes each judge's failure reason (rate-limited, auth, parse error, timeout, upstream gateway, etc.) and appends a summary to the assessment error. Also attaches the per-pointjudgeLogso operators can expand the existing Log section in the UI and see the full trace. Backup-judge failures are included in the summary.Test plan
pnpm test:cli— 45 files / 521 tests passpnpm test:web— 49 files / 596 tests passllm-coverage-evaluator.test.tsupdated from equality-against-old-literal to regex match against the new richer prefixGeneration failed: .../All N judge(s) failed... Reasons: ...messages🤖 Generated with Claude Code