| Field | Value |
|---|---|
| Name | code-review-graph |
| Full Name | tirth8205/code-review-graph |
| URL | https://github.com/tirth8205/code-review-graph |
| Description | Local knowledge graph for Claude Code. Builds a persistent map of your codebase so Claude reads only what matters -- 6.8x fewer tokens on reviews and up to 49x on daily coding tasks. |
| Language | Python |
| License | MIT |
| Stars | 3,731 |
| Forks | 331 |
| Created | 2026-02-26 |
| Age | 29 days |
| Signal | Score |
|---|---|
| Newcomer Score | 9.7 |
| Momentum Score | 9.0 |
| Trend Score | 5.6 |
| Signal Type | newcomer |
| Stars/Day Avg | 128.7 |
| Recent Commits (30d) | 127 |
- Stars 1K-50K: 3.7K stars -- within range
- Core logic clear: Tree-sitter parsing -> graph construction -> context retrieval
- 1-day reimplement: Core algorithms (AST parsing, graph building, query engine) are well-scoped
- Something to learn: Incremental code graph construction, tree-sitter in Go, smart context selection algorithms
- Tree-sitter parsing: Parse source files into AST, extract functions/classes/imports/calls
- Knowledge graph construction: Build directed graph of code relationships (calls, imports, inheritance, references)
- Incremental updates: Only re-parse changed files, update graph edges efficiently
- Context retrieval: Given a diff/query, walk the graph to find the minimal set of relevant code
- Token optimization: Score and rank code sections by relevance, fit within token budgets
- Performance: Go's concurrency for parallel file parsing across large codebases
- Single binary: No Python environment setup, no pip dependencies
- Memory efficiency: Go's lower memory footprint for large graphs
- Native tree-sitter: go-tree-sitter bindings are mature and fast
- Faster incremental updates: goroutine-per-file parallel parsing
- Smaller binary size: vs Python + tree-sitter + networkx + all dependencies
- Tree-sitter grammar integration in Go
- Graph data structures for code analysis (adjacency lists, topological sort)
- Diff-aware context selection algorithms
- Token budget optimization (knapsack-style relevance packing)
| Project | Stars | Why Excluded |
|---|---|---|
| Scrapling | 33.5K | Browser-dependent (Playwright), web scraping is UI-heavy |
| opik | 18.5K | Dashboard/UI-heavy, too complex for 1 day |
| prompt-optimizer | 25.6K | Web app UI, core logic is LLM calls (not algorithmic) |
| edgequake | 1.5K | Too similar to lightrag-go (already done) |
| TrendRadar | 49.8K | Similar to our git-trend-sync |
| agent-orchestrator | 5.5K | Tightly coupled to specific tools (tmux, worktrees) |
| helicone | 5.4K | UI-heavy observability dashboard |
- parser/: Tree-sitter integration, multi-language support (Go, Python, TypeScript, Rust)
- graph/: In-memory directed graph with persistence (JSON/binary serialization)
- index/: File watcher + incremental update engine
- query/: Context retrieval given diff hunks or symbol names
- cli/: Command-line interface for build, query, serve