You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
* Initial plan for Issue #141: 支持Code Review场景下,一次性提交多个review comments的场景
* feat: support batch processing of multiple review comments
Support scenario where reviewers submit multiple pending review comments
and then use "/continue" or "/fix" commands in the final review submission.
The system now handles pull_request_review events with "/continue" or "/fix"
commands by:
- Collecting all review comments associated with the review ID
- Preserving code position information (file paths, line numbers)
- Building complete context for AI to process all issues at once
- Responding via PR comment instead of individual review comments
Key improvements:
- Added GitHub client methods to fetch PR reviews and review comments
- Enhanced webhook handler to process pull_request_review events
- Implemented batch processing in agent with position-aware context
- Supports both multi-comment reviews and review-body-only scenarios
Closes#141
* Looking at the issue information and changes, I'll generate a conventional commit message for this code review batch processing feature:
feat: support batch processing of multiple review comments
Add support for processing multiple review comments at once when reviewer submits "/continue" or "/fix" commands. The system now:
- Creates a processing status comment with @user mention when command is received
- Collects all review comments from the same review ID with position information
- Processes all comments together in a single AI request with complete context
- Updates the initial comment with results instead of creating multiple comments
- Maintains code line position information for accurate context
This enables reviewers to submit multiple pending comments and have the AI agent fix all issues in one batch operation.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: add missing fmt import in webhook handler
Support batch processing of code review comments in pull request reviews.
When a pull_request_review event contains "/continue" or "/fix" commands,
the system now aggregates all related pull_request_review_comment events
by review ID to provide complete context to AI for comprehensive fixes.
Preserves code line position information for accurate comment handling.
Closes#141
* feat: add batch processing support for multiple review comments
Support batch processing of multiple review comments in code review scenarios. When reviewers submit multiple pending review comments and then submit a review with "/continue" or "/fix" command, the system now aggregates all related comments (using review ID) and provides complete context to AI for comprehensive issue resolution.
Key changes:
- Handle pull_request_review events with "/continue" or "/fix" commands
- Aggregate related pull_request_review_comment events by review ID
- Preserve code position information (line numbers) for each comment
- Provide complete context to AI for batch processing of all issues
Fixed gofmt formatting issues in agent.go and handler.go.
Closes#141
* feat: support batch processing of multiple review comments
Implement batch processing for Code Review scenarios where reviewers submit multiple pending review comments and then use "/continue" or "/fix" commands to trigger AI fixes for all comments at once.
Key improvements:
- Remove unnecessary status comment creation to save GitHub API quota
- Process all review comments associated with a review ID as a single batch
- Preserve code position information for each review comment
- Include @user mentions in AI feedback responses
- Clean up unused comment update functions and streamline interaction flow
The system now handles pull_request_review events with "/continue" or "/fix" commands by collecting all associated pull_request_review_comment events (linked by review ID) and providing complete context to AI for comprehensive issue resolution.
Closes#141
* refactor: consolidate review comment processing logic
Remove duplicate processReviewOnlyWithTriggerUser method and unify
the handling of review events with or without comments. The logic
now uses a single flow that adapts the prompt and response message
based on whether review comments exist.
Key changes:
- Removed processReviewOnlyWithTriggerUser method to eliminate duplication
- Unified processing logic in ProcessPRFromReviewWithTriggerUser
- Adjusted prompt format based on comment count (Review vs batch comments)
- Streamlined response messages for different scenarios
- Cleaned up unused methods in github client
Closes#141
* refactor: remove unused ProcessPRFromReview method
Remove the unused ProcessPRFromReview method from agent.go as the codebase now uses ProcessPRFromReviewWithTriggerUser method instead. This cleanup reduces code duplication and maintains consistency in the review comment processing workflow.
Closes#141
* refactor: remove redundant position info text in prompts
Remove "确保保留代码的位置信息" from four prompt statements in batch code review processing. This text was unclear and potentially redundant since position information is already preserved in the review comments structure.
Closes#141
* feat: support batch processing of multiple review comments in Code Review scenarios
Add support for processing multiple review comments submitted together during code review. When a reviewer submits multiple pending review comments followed by a "/continue" or "/fix" command in the final review submission, the system now:
- Detects pull_request_review events with "/continue" or "/fix" prefix
- Aggregates related pull_request_review_comment events using review ID
- Preserves code line position information for each comment
- Provides complete context to AI for comprehensive issue resolution
This enhancement allows AI to address all review feedback in a single iteration rather than processing comments individually.
Closes#141
---------
Co-authored-by: qiniu-ci <qiniu-ci@qiniu.com>
Co-authored-by: Claude <noreply@anthropic.com>
0 commit comments