v2.20.0

Added

  • /ai-pr-review full is an alias for /ai-pr-review review-full (#995). Both start the same full-mode run and get the same reaction and reply. The command is not case-sensitive. A token that is not an exact alias or command name still gets the “did not recognize” reply. The help reply and the docs list the alias.

Fixed

  • A Critical finding that an agent and an analyzer both reported could get no inline comment or annotation (#1000). The inline cap (25 by default) fills from the front of the findings list, and the step that collapses repeated analyzer findings moved these findings behind every agent-only finding. On a review with many findings, the Critical finding fell past the cap. On Bitbucket it got no Code Insights annotation. The list now keeps its severity order on GitHub, GitLab, and Bitbucket, so the most severe findings get the inline slots first. More than 25 findings of the same severity can still push one past the cap.

Changed

  • Internal cleanup. The command table BASH_ONLY_COMMANDS in ai_pr_review/slash/parser.py is now WORKFLOW_COMMANDS, because those commands are routed by the slash commands workflow. Stale references to the removed bash engine are gone from comments, docstrings, and three docs lines. No behavior changes.
  • Dependency update. The trufflesecurity/trufflehog image tag is 3.98.1 (#1001).

See CHANGELOG.md for full technical detail.