chore(coderabbit): pause automatic reviews after two reviewed commits

Merged
BO
1d ago
+1-1

CodeRabbit now stops reviewing new pushes on a pull request after it has reviewed two commits, which cuts automatic re-reviews by about 44% on the last measured window (417 to 234 reviews).

.coderabbit.yaml changes one value: reviews.auto_review.auto_pause_after_reviewed_commits goes from 0 (pause disabled, set in #413 to keep reviews active) to 2. The default is 5. CodeRabbit's configuration reference documents the key as an integer, "Set to 0 to disable". It does not say how a paused review is resumed, so the first pull request that pauses shows that; a review can still be requested with a comment.

Why now: 293 of 417 reviews in the measured window re-reviewed a later push. A cap of two could miss findings that only appear on the third and later pushes; 30 of the 36 major findings were on pull requests with more than two reviews, which is an upper bound, not a count of what a cap would remove. The data and the other options considered (cap of three, first review only, quieter profile) are in the review-service volume proposal kept outside this repository.

Validation: the file parses as YAML and validates against the CodeRabbit v2 JSON schema (https://coderabbit.ai/integrations/schema.v2.json; the key is an integer, minimum 0, default 5). The diff is one line. Fork-only: upstream has no .coderabbit.yaml, and none of the open upstream pull requests touch it. No product code, build or test ran for this configuration-only change.

README rows checked: the About this fork feature, command, language and measurement rows are unaffected; no product behavior changes.

Summary by cubic

Pauses CodeRabbit's automatic reviews on a pull request after two commits have been reviewed, cutting automatic re-reviews by about 44% (417 to 234) in the last measured window. Reviews can still be requested by commenting on the PR.

  • Changes auto_pause_after_reviewed_commits from 0 (pause disabled) to 2 in .coderabbit.yaml; CodeRabbit's default is 5.
  • A cap of two could miss findings that only appear on later pushes; 30 of 36 major findings were on PRs with more than two reviews, which is an upper bound, not a count of what this cap removes.