Uh oh!
There was an error while loading. Please reload this page.
Add GitHub Action to suggest PR reviewers based on git history - #4789
Merged
Conversation
Add a reviewer suggestion workflow that posts a PR comment with 1-2 suggested reviewers based on recency-weighted git history of changed files. This is additive only, CODEOWNERS and auto-assign stay unchanged. The action triggers on PR open and ready-for-review events, skips drafts and fork PRs, and updates its comment in-place on re-runs.
Use gh api to resolve commit SHAs to GitHub logins automatically, removing the need to maintain a manual author alias map. Also fix error handling for gh subprocess failures, replace hardcoded CODEOWNERS fallback with a link, add --since to bound git log, and add synchronize trigger so suggestions update on new pushes.
simonfaltum
requested review from
andrewnester, anton-107, denik, pietern and shreyas-goenka
as code ownersMarch 18, 2026 14:55
simonfaltumtemporarily deployed
to
test-trigger-is
March 18, 2026 14:56 — with 
GitHub Actions
Inactive
Switch to deco runner group (IP allow list). Add CODEOWNERS parsing to show eligible reviewers alongside the git-history-based suggestions. Expand from 1-2 to 1-3 suggested reviewers with 1.5x threshold.
Contributor
Eligible reviewersCould not determine reviewers from git history. Based on CODEOWNERS, these people or teams could review: @andrewnester, @anton-107, @denik, @pietern, @shreyas-goenka Suggestions based on git history of 2 changed files (0 scored). See CODEOWNERS for path-specific ownership rules. |
Collaborator
Commit: 81c87b9
20 interesting tests: 9 SKIP, 7 KNOWN, 4 flaky
Top 23 slowest tests (at least 2 minutes):
|
simonfaltumtemporarily deployed
to
test-trigger-is
March 18, 2026 15:06 — with GitHub Actions
Inactive
pietern
approved these changes
Mar 18, 2026
| GITHUB_REPOSITORY: ${{ github.repository }} | ||
| PR_NUMBER: ${{ github.event.pull_request.number }} | ||
| PR_AUTHOR: ${{ github.event.pull_request.user.login }} | ||
| run: python3 tools/suggest_reviewers.py |
Contributor
There was a problem hiding this comment.
Nit: let's use uv here (and setup-uv) to make sure we really run with 3.12.
denik
reviewed
Mar 18, 2026
| if p.name.startswith("out.") or p.name == "output.txt": | ||
| return 0.0 | ||
| if path.startswith(("cmd/workspace/", "cmd/account/")): | ||
| return 0.0 |
Contributor
There was a problem hiding this comment.
FYI, there are non-generated files as well in those directories.
renaudhartert-db
approved these changes
Mar 18, 2026
renaudhartert-db
left a comment
Contributor
There was a problem hiding this comment.
Can we port that to TF and SDKs?
Remove the 0.0 weight override for cmd/workspace/ and cmd/account/ since those directories contain non-generated files too (denik). Use uv + setup-uv to pin Python 3.12 (pietern).
simonfaltumtemporarily deployed
to
test-trigger-is
March 18, 2026 16:48 — with GitHub Actions
Inactive
simonfaltum
enabled auto-merge
March 18, 2026 17:13
Uh oh!
There was an error while loading. Please reload this page.
eng-dev-ecosystem-bot
commented
Mar 18, 2026
Collaborator
Commit: a2e290d
46 interesting tests: 15 flaky, 14 RECOVERED, 9 KNOWN, 7 FAIL, 1 SKIP
Top 50 slowest tests (at least 2 minutes):
|
denik pushed a commit
that referenced
this pull request
May 20, 2026
## Why The CLI's CODEOWNERS catch-all assigns 6 people to every PR outside a few narrow paths. This creates review noise and diffuses responsibility. We want targeted reviewer suggestions based on who actually worked on the changed code recently. ## Changes Before: Every PR touching core code auto-assigns all 6 CODEOWNERS. No signal about who is best suited to review. Now: A new GitHub Action analyzes git history of the changed files and posts a PR comment with two sections: - **Suggested reviewers** (1-3 people best suited based on recency-weighted git history) - **Eligible reviewers** (everyone from CODEOWNERS who could review, minus the suggested ones) This is additive only. CODEOWNERS and auto-assign stay unchanged. How it works: - Triggers on PR open, synchronize, and ready-for-review (skips drafts and fork PRs) - Classifies changed files by type (source=1.0, tests=0.3, acceptance=0.2, generated=0.0) - Scores contributors using recency-weighted commit history (half-life 150 days) - Resolves git author names to GitHub logins via the GitHub API (no hardcoded alias table to maintain) - Parses `.github/CODEOWNERS` to find eligible reviewers for the changed paths - Updates the comment in-place on re-runs (no notification churn) Implementation: a single Python script (`tools/suggest_reviewers.py`, 281 lines) and a minimal workflow YAML. ## Test plan - [x] Action ran on this PR itself and posted a comment successfully - [x] Verified script parses cleanly, passed `make checks`, passed `ruff format` - [x] Dry-run tested against 4 recent merged PRs with different characteristics: **PR #4784** (66 files, by pietern, big DABs rename): ``` ## Suggested reviewers - @denik -- recent work in `./`, `bundle/`, `cmd/bundle/generate/` Confidence: high ## Eligible reviewers @andrewnester, @anton-107, @lennartkats-db, @shreyas-goenka, @simonfaltum ``` Correctly identifies Denis as the clear top reviewer (2x second place score). All 6 CODEOWNERS shown as eligible. **PR #4782** (13 files, by denik, bundle engine priority): ``` ## Suggested reviewers - @andrewnester -- recent work in `bundle/schema/`, `bundle/internal/schema/`, `cmd/bundle/generate/` - @pietern -- recent work in `bundle/schema/`, `cmd/bundle/generate/`, `bundle/internal/schema/` - @shreyas-goenka -- recent work in `bundle/schema/`, `bundle/internal/schema/`, `cmd/bundle/` Confidence: medium ## Eligible reviewers @anton-107, @simonfaltum ``` Suggests 3 reviewers when scores are close. Remaining CODEOWNERS shown as eligible. **PR #4785** (2 files, by MarioCadenas, apps bug fix): ``` ## Suggested reviewers - @arsenyinfo -- recent work in `cmd/apps/` - @pietern -- recent work in `cmd/apps/` - @pkosiec -- recent work in `cmd/apps/` Confidence: low ## Eligible reviewers @databricks/eng-apps-devex ``` Correctly suggests apps-area contributors (not the catch-all CODEOWNERS). Shows the apps team as eligible. Low confidence since only 2 files. **PR #4774** (28 files, by denik, direct engine grants): ``` ## Suggested reviewers - @andrewnester -- recent work in `bundle/direct/dresources/`, `acceptance/bundle/invariant/` - @shreyas-goenka -- recent work in `bundle/direct/dresources/` Confidence: high ## Eligible reviewers @anton-107, @pietern, @simonfaltum ``` Correctly identifies the two main bundle/direct contributors. High confidence with clear score separation.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
The CLI's CODEOWNERS catch-all assigns 6 people to every PR outside a few narrow paths. This creates review noise and diffuses responsibility. We want targeted reviewer suggestions based on who actually worked on the changed code recently.
Changes
Before: Every PR touching core code auto-assigns all 6 CODEOWNERS. No signal about who is best suited to review.
Now: A new GitHub Action analyzes git history of the changed files and posts a PR comment with two sections:
This is additive only. CODEOWNERS and auto-assign stay unchanged.
How it works:
.github/CODEOWNERSto find eligible reviewers for the changed pathsImplementation: a single Python script (
tools/suggest_reviewers.py, 281 lines) and a minimal workflow YAML.Test plan
make checks, passedruff formatPR #4784 (66 files, by pietern, big DABs rename):
Correctly identifies Denis as the clear top reviewer (2x second place score). All 6 CODEOWNERS shown as eligible.
PR #4782 (13 files, by denik, bundle engine priority):
Suggests 3 reviewers when scores are close. Remaining CODEOWNERS shown as eligible.
PR #4785 (2 files, by MarioCadenas, apps bug fix):
Correctly suggests apps-area contributors (not the catch-all CODEOWNERS). Shows the apps team as eligible. Low confidence since only 2 files.
PR #4774 (28 files, by denik, direct engine grants):
Correctly identifies the two main bundle/direct contributors. High confidence with clear score separation.