contextd is a local-first semantic context daemon for AI agents. It watches files, indexes them via ONNX embeddings, and exposes search through REST API, CLI, and MCP.
https://github.com/sandy-sachin7/contextd
cargo build --release
cargo test # unit tests (43 + 4 + 5 + 3 = 55)
cargo test --test load_test # integration tests
cargo test --test watcher_test
cargo test --test zero_config_test # auto-download tests
python3 scripts/test_mcp_local.py # MCP E2E tests
python3 scripts/memory_stress_test.py # memory stress test
bash scripts/playground.sh # one-line playground.github/workflows/ci.yml— fmt check, build, test, clippy.github/workflows/release.yml— cross-compile for 5 targets on tag.github/workflows/docker.yml— Docker build + push.github/workflows/audit.yml— daily cargo-audit.github/workflows/publish.yml— crates.io publish
- Rust edition 2021, stable toolchain
- Conventional commits:
feat:,fix:,docs:,chore:,style:,refactor: - All new code needs tests
- Document public API
- Use anyhow for error handling