fix(qa): persist direct-API qa_reviewer verdict via a qa_status.json artifact

[OBenner]

Summary

The direct-API (generic_edit) qa_reviewer had the same latent status-persistence bug that #342 fixed for the qa_fixer. build_qa_reviewer_prompt — especially its previous_error self-correction branch — tells the model to record its verdict by editing implementation_plan.json's qa_signoff with Claude's Write/Edit tools and absolute paths. On a non-Claude provider running through the generic_edit runtime, those tools don't exist (the real tools are write_file/replace_text) and resolve_workspace_path rejects absolute paths — so the reviewer could never persist qa_signoff, and run_qa_reviewer_via_runtime returned "error" instead of "approved"/"rejected". QA on e.g. OpenAI never recorded a verdict.

This applies the qa_fixer pattern, adapted for the reviewer's richer verdict (approved/rejected + issues_found, vs the fixer's single fixes_applied flag).

Changes (apps/backend/qa/reviewer.py)

• build_qa_reviewer_prompt gains runtime_status_relspec: str | None = None (default None → the Claude path is byte-for-byte unchanged). When set, an addendum appended last OVERRIDES the Write/Edit/absolute-path guidance: edit via write_file with workspace-relative paths, and record the verdict by writing <relspec>/qa_status.json with status ("approved"/"rejected") plus issues_found when rejected, then call finish.
• run_qa_reviewer_via_runtime folds <spec_dir>/qa_status.json into qa_signoff via the new merge_runtime_qa_reviewer_status_artifact when no usable verdict landed in the plan, then re-reads. The existing update_qa_signoff_with_coverage and Graphiti save_session_memory steps are preserved.
• run_qa_reviewer_runtime_session async-closes the provider session (OpenAICompatibleSession.aclose()) in a finally, mirroring _aclose_agent_session in qa/fixer.py, so the AsyncOpenAI connection pool is released in-loop rather than at interpreter exit.

Tests

tests/test_qa_reviewer_runtime.py (mirrors the qa_fixer suite): scripted runtime session writes qa_status.json → approved/rejected merged into qa_signoff; no-artifact stays "error"; prompt-addendum content + its precedence over the self-correction block; per-session async close.

apps/backend/.venv/bin/python -m pytest tests/test_qa_reviewer_runtime.py -p no:cacheprovider -q
# 16 passed

(The Claude-path tests/test_qa_reviewer.py fails to collect in this environment — a pre-existing agents.runtime.qa_phase_routing import error unrelated to this change, confirmed by reproducing it with these edits stashed out.)

Stacking

Stacked on #342 (the qa_fixer fix), which adds OpenAICompatibleSession.aclose(). The base is set to that branch so the diff shows only the reviewer changes; GitHub will retarget this PR to develop automatically once #342 merges.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

🗂️ Base branches to auto review (3)

• develop
• release/*
• hotfix/*

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 7657813b-de15-4320-af19-15437767c661

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/naughty-wozniak-899d70

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[sonarqubecloud]

Quality Gate passed

Issues
7 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[OBenner]

Closing as a duplicate. This exact fix is already merged to develop via #340 ("persist direct-API QA reviewer verdict and close its async client"), which is more complete — it also clears the status artifact (_clear_runtime_qa_signoff_artifact) and adds a dedicated test_openai_compat_session_lifecycle.py. Nothing to merge here.