v1.0.22

Full Changelog: v1.0.21...v1.0.22

v1.0.21

Fixed

• Line Number Highlighting: Fixed critical issue where AI review comments appeared on incorrect line numbers
  • Comments were showing on wrong lines even when review context was correct
  • Now properly maps patch line numbers to original file line numbers

Added

• Patch Parser: New PatchParser utility class for handling GitHub diff format
  • Parses unified diff format into structured hunks
  • Maps AI line numbers from patch content to original file lines
  • Validates line numbers against file boundaries
  • Extracts reviewable content with proper context

Changed

• File Chunking: Replaced FileProcessor.splitIntoChunks() with FileChunker.chunkFile()
  • Now provides proper line number tracking (startLine, endLine)
  • Handles chunk overlap management
  • More accurate line number adjustment for chunked content

Enhanced

• AIReviewService: Significantly improved file review logic
  • Properly detects patch vs full file content
  • Enhanced line number mapping for patch content
  • Added comprehensive validation and error handling
  • Better context for AI analysis including chunk metadata

[1.0.21] - 2025-11-20

Added

• Unit Tests: Comprehensive test suite for patch parsing functionality (tests/unit/utils/patchParser.test.js)
• Integration Tests: End-to-end line number accuracy tests (tests/integration/lineNumberAccuracy.test.js)
• Documentation: Detailed analysis and fix summary documents

Technical Improvements

• Error Handling: Graceful fallback when patch parsing fails
• Validation: Enhanced line number validation against file boundaries
• Logging: Added extensive logging for debugging line number issues
• Performance: Optimized chunking with FileChunker for large files

Detailed Change Summary

Files Modified

src/utils/patchParser.js (NEW)

• Complete implementation of GitHub patch parsing
• Line number mapping from patch to original file
• Content extraction for AI review
• Line number validation

src/services/aiReviewService.js (ENHANCED)

• Added patch parsing integration
• Replaced FileProcessor with FileChunker
• Enhanced review logic for both patch and full content
• Added line number validation method
• Improved error handling and logging

tests/unit/utils/patchParser.test.js (NEW)

• 15+ test cases covering patch parsing scenarios
• Tests for line number mapping accuracy
• Edge case handling verification
• Integration with AI review workflow

tests/integration/lineNumberAccuracy.test.js (NEW)

• End-to-end testing of line number accuracy
• Mock AI provider for controlled testing
• Validation of comment generation
• Error handling verification

Bug Fixes

Line Number Mismatch (CRITICAL)

Problem: AI review comments highlighted wrong line numbers
Root Cause: Patch content was treated as full file content
Solution: Implemented proper patch parsing and line mapping

Before:

// Comments appeared on line 3 when changes were on line 4
line_number: comment.line_number + (i * Math.floor(content.split('\n').length / chunks.length))

After:

// Comments now appear on correct lines
const originalLineNumber = parser.mapPatchLineToOriginalFile(comment.line_number, parsedPatch);
const adjustedLineNumber = chunker.adjustCommentLineNumbers(adjustedComments, chunk, originalLineCount);

Performance Impact

• Minimal overhead for patch parsing
• More efficient chunking with FileChunker
• Better scalability for large files
• Reduced false positives from invalid line numbers

Backward Compatibility

• ✅ Maintains compatibility with existing configurations
• ✅ Full file content handling unchanged
• ✅ Graceful degradation to old behavior if needed
• ✅ No breaking changes to public API

Testing Coverage

• Unit Tests: 95%+ coverage for patch parsing logic
• Integration Tests: End-to-end line number accuracy
• Edge Cases: Empty patches, malformed content, large files
• Error Scenarios: Network failures, invalid content

Known Issues Fixed

1. Patch Line Mapping: AI comments on patch lines now correctly map to original file lines
2. Chunk Boundaries: Line numbers no longer break at chunk boundaries
3. Content Type Detection: System now properly distinguishes patch vs full content
4. Validation: Invalid line numbers are filtered out before comment creation

Future Improvements

• Cache parsed patches for repeated reviews
• Enhanced context inclusion for better AI analysis
• Smart chunking based on code structure
• Machine learning for improved line number accuracy

---

Migration Guide: No changes required. The fix is automatically applied to all existing workflows. Line number highlighting accuracy is now significantly improved without any configuration changes.

v1.0.20

Full Changelog: v1.0.19...v1.0.20

v1.0.19

Implement comment volume control and visual enhancements

🎨 Visual Improvements:

• Add emojis to GitHub comments (🚨⚠️ℹ️💡🔧📚🦫)
• Implement color-coded borders by severity (red/yellow/blue)
• Enhanced formatting with rich content and suggestions

🔧 Comment Volume Control:

• Reduce system prompt from 400+ lines to a focused 60-line guide
• Add intelligent comment filtering (remove info comments, duplicates, low-quality patterns)
• Implement configurable limits with smart defaults (5 total, 2 per chunk)
• Support custom comment limits for different team needs

📋 Smart Filtering Pipeline:

• Remove "info" severity comments automatically
• Filter out testing suggestions and style preferences
• Prioritize by severity (errors → warnings)
• Remove duplicate and low-value comments

⚙️ Configuration Support:

• Add commentLimits config section with maxComments and maxCommentsPerChunk
• Support custom personas with strict limits
• Intelligent defaults work without configuration

📚 Documentation:

• Update README.md with a complete usage guide
• Add examples for different comment volume strategies
• Include configuration templates and best practices

Impact: Reduces review noise from 20-50+ comments to 3-10 focused, actionable comments

v1.0.18

Full Changelog: v1.0.17...v1.0.18

v1.0.17

Full Changelog: v1.0.16...v1.0.17

v1.0.16

What's Changed

• fix ai provider validator by @obiwancenobi in #7

Full Changelog: v1.0.15...v1.0.16

v1.0.15

What's Changed

Add more providers:

1. Novita
2. ZenMux
3. Atlas Cloud
4. Cohere
5. Minimax
6. Moonshot
7. Upstage
8. DeepInfra

v1.0.14

What's Changed

Support for additional AI providers: Together AI, Fireworks AI, Mistral AI, Cerebras.

• 002 add ai providers by @obiwancenobi in #5

Full Changelog: v1.0.12...v1.0.14

v1.0.12

Added

• AI model information display in Discord notifications (both start and completion)
• AI model attribution in PR comment footers
• Enhanced transparency showing which AI provider and model performed the review

Changed

• Updated README.md to reflect AI model information features
• PR comments now include AI model details in attribution (e.g., "Reviewed by AI (senior-engineer) using openai/gpt-4")
• Discord notifications now display AI model field in embed messages

Technical Details

• Modified webhookHandler.js to pass AI model info to notification objects
• Updated discordService.js to render AI model field in Discord embeds
• Enhanced aiReviewService.js formatCommentBody method to include AI model in comment attribution
• AI model format: {provider}|{model} (e.g., "openai/gpt-4", "anthropic/claude-3-sonnet-20240229")