fix(llm): make JSON extraction string-aware so braces in text cannot drop findings

[Liohtml]

Summary

parse_llm_json counted braces without tracking string context, so any } inside a JSON string value — entirely plausible in an LLM's free-text overall_impression/findings — truncated the object mid-string. The resulting decode error was swallowed by parse_llm_response, which fell back to treating the whole response as plain text: all structured findings silently discarded.

Fixes #136

Changes

• src/medcheck/llm/base.py: parse_llm_json now uses json.JSONDecoder().raw_decode (string/escape-aware), trying candidate { positions in order until one decodes to an object — this also recovers when a stray { appears in prose before the real JSON, which the old scanner could not
• tests/unit/test_llm/test_base.py: 3 new tests — braces inside string values, stray { before the object, markdown-fenced JSON

Testing

• Existing tests pass (uv run pytest) — 175 passed (all existing parser tests unchanged and green)
• New tests added for new functionality — 3 new parser tests
• Coverage does not decrease (uv run pytest --cov-fail-under=85) — 87.23%
• Linting passes (uv run ruff check .)
• Type checking passes (uv run mypy src/medcheck --strict)
• Pre-commit hooks pass — ruff hooks pass locally; full pre-commit not run

Additional Notes

Error behavior is unchanged for genuinely JSON-free responses: parse_llm_json still raises ValueError and parse_llm_response still falls back to raw text.

🤖 Generated with Claude Code

https://claude.ai/code/session_01W4svt5QTs4WUMSy4HwiVt9

---

Generated by Claude Code

[coderabbitai]

Warning

Review limit reached

@Liohtml, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 39 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 5807a6d6-ea3c-4cf8-ae16-0acd45f753b5

📥 Commits

Reviewing files that changed from the base of the PR and between 63a57ac and a76c6af.

📒 Files selected for processing (2)

• src/medcheck/llm/base.py
• tests/unit/test_llm/test_base.py

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/dependabot-prs-q6u94z

---

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.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!