CodeGraph CLI MCP Server

# CodeGraph Benchmarking Suite This project includes a comprehensive Criterion-based benchmarking suite with baseline comparison and automated report generation. ## Quick Start - Run all benches and save a baseline (default: `baseline`): - `make bench` - Environment: - `BASELINE_NAME` (default `baseline`) to name the saved baseline - `ENABLE_FAISS_BENCH=1` to include FAISS-dependent benches (requires FAISS available) - Compare current results to a baseline: - `make bench-compare` (env: `BASELINE_NAME`, `THRESHOLD` default `0.10`) - Generate a Markdown report: - `make bench-report` (writes to `benchmarks/reports/benchmark_report_latest.md`) Artifacts are copied under `benchmarks/artifacts/<timestamp>/criterion` and reports under `benchmarks/reports/`. ## Notes - Criterion baselines are saved via `--save-baseline <name>` and stored in `target/criterion/<bench>/base/<name>/`. - The comparison script parses `estimates.json` and flags regressions above the threshold (default 10%). - Vector and API benches may require enabling `codegraph-vector` feature `faiss`. In CI or environments without FAISS, leave `ENABLE_FAISS_BENCH=0` to skip them. ## Adding New Benches - Prefer per-crate benches under `crates/<crate>/benches/your_bench.rs`. - Use `criterion_group!` and `criterion_main!` and keep measurement windows modest for CI. - Use `Throughput` and parameterized inputs for meaningful, comparable metrics.