Skip to content

Conversation

@dtsong
Copy link
Owner

@dtsong dtsong commented Jan 24, 2026

Summary

  • Automatic diff chunking when context overflow is detected during LLM review
  • Three strategies: FILE (per-file), LINES (line-count limit), AUTO (best of both)
  • Result merging with fingerprint-based deduplication, token/cost aggregation
  • Integrated into degradation pipeline: full → chunked → reduced → gates-only

Test plan

  • 17 new tests covering file/line/auto chunking and result merging
  • Full suite passes (406 tests, 95.07% coverage)

Closes #40

🤖 Generated with Claude Code

- ChunkStrategy (FILE, LINES, AUTO) splits large diffs into chunks
- merge_review_results combines chunk results with deduplication
- Degradation pipeline triggers chunked review on context overflow
- AUTO strategy: file-based for multi-file, line-based for single large files
- 17 tests covering chunking strategies and result merging

Implements #40

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
@dtsong dtsong merged commit ceeb461 into main Jan 24, 2026
2 checks passed
@dtsong dtsong deleted the feat/40-chunked-review branch January 24, 2026 20:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Context overflow chunked review

2 participants