fix(github_rest_api): include title type details in auto-merge eligibility error message

[dclong]

Summary

• fix: include title type details in auto-merge eligibility error message

Changed files

Modified

• github_rest_api/github.py (+10/-3)

Commits

• 49c4a98 fix: include title type details in auto-merge eligibility error message

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

[dclong]

Claude finished @dclong's task in 1m 40s —— View job

---

Code Review: PR #118

• Gather context (diff, changed files)
• Run /code-review:code-review on the PR
• Post review findings
• Decide on auto-merge eligibility

Note: The sandbox denied running python -m pytest / python3 via Bash in this environment, so I reviewed the diff manually instead of executing the test suite. Please re-run CI to confirm tests/test_github.py still passes (the existing test_field_gate_failure_rejects/test_field_gate_failure_passes_eligible_pr tests only assert is None/is not None, so they don't need updates for this change).

Summary

This PR enriches the auto-merge eligibility failure message in _field_gate_failure (github_rest_api/github.py:180-188) to include the PR's detected (or missing) Conventional-Commits title type alongside the list of eligible types, which makes failures much easier to diagnose from logs.

Findings

No correctness bugs found. Two minor, non-blocking observations:

1. Scope creep in the same commit (github_rest_api/github.py:36): DEFAULT_AUTO_MERGE_TYPES was expanded from ("chore", "docs", "deps") to ("build", "chore", "ci", "docs", "deps"). That's an unrelated behavioral change bundled into a commit whose title/PR description only describes an error-message fix. Not incorrect, but worth splitting into its own commit/PR for a clean history — and it's worth confirming the added build/ci types are an intentional policy change.
2. Minor duplicate work (github_rest_api/github.py:181-182): _conventional_type(title) is now computed twice — once inside _title_type_allowed(title, allowed_types) and again on the next line to build the message. Harmless (cheap regex match) but could be avoided by having _title_type_allowed return the parsed type, or by computing type_ once and deriving the boolean from it.

Both are style/cleanup items, not blocking issues.

AUTO_MERGE_APPROVED