Development

Build and checks

cargo build
cargo test
cargo clippy --all-targets --all-features -- -D warnings

Performance gate

python3 scripts/perf_gate.py

Run the performance gate after search/indexing changes to catch regressions.

Agent token benchmark (PyTorch)

python3 scripts/benchmark_agent_token_efficiency.py --repo /path/to/pytorch

This benchmark measures tokens-to-complete for each scenario: - baseline: grep locate + incremental file-snippet tiers - cgrep: agent locate + incremental agent expand ID tiers

Tier controls:

python3 scripts/benchmark_agent_token_efficiency.py \
  --repo /path/to/pytorch \
  --baseline-file-tiers 2,4,6,8,12 \
  --cgrep-expand-tiers 1,2,4,6,8

This writes: - docs/benchmarks/pytorch-agent-token-efficiency.md - local/benchmarks/pytorch-agent-token-efficiency.json (local-only)