Better Code Review Graph

Parses your codebase with Tree-sitter, builds a structural graph of functions, classes, and imports, then lets Claude query only the code that matters. Ships with fixed multi-word search (AND logic instead of literal substring), configurable embeddings via Qwen or OpenAI, and qualified call resolution that actually returns results. The v2.0 release adds temporal columns for snapshot queries at any commit SHA, a diff action that returns added/removed/modified buckets without re-parsing, and an optional security scanner (regex-based tier 1 by default, ~120 Semgrep rules at tier 2). Also generates LLM-powered docstring summaries per function to boost semantic search recall and exports graphs as GraphML, JSON-LD, Dot, or Cypher. Fork of tirth8205/code-review-graph with production CI, type safety, and actual test coverage.