Repository navigation
Fail unsuccessful reviews and preserve safe execution diagnostics - #11
Merged
Merged
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Failed inference previously returned success after recording its safe fallback, making an unsuccessful AI review a green Actions check. Always propagate the primary failure after reporting it; keep the separately protected publisher eligible to publish only a non-approving result. Progress now explicitly distinguishes fallback publication from completed AI analysis while retaining the actual GitHub workflow conclusion.
Add bounded, allowlisted Docker/Codex/RPC/proxy diagnostics. Discard arbitrary stderr, provider text, credentials, prompts and source. Only the proxy's actual HTTP response supplies upstream status; a local AppError default is not provider HTTP evidence. Preserve Responses endpoint/policy, OIDC, credential isolation and merge restrictions. The historical generic log cannot establish the original inference failure's specific cause.
Validation: Node 24
npm run check(140 tests, typecheck, lint, build), clean-source check, real workerd/D1, Actionlint, Docker build and real pinned Codex with successful/failing synthetic Responses, npm audit, strict image scan and Wrangler dry-run. No real inference, deployment, credential changes or merge performed.Rollout requires matching immutable source/Worker trust/caller pins after this PR is approved and merged. Pin PRs remain Draft until both use the actual merged source SHA. No database migration, new permissions or OAuth reconnect required. See docs/review-rollout.md.