nnethercott/hannoy

Production-ready KV-backed HNSW implementation in Rust using LMDB

60
/ 100
Established

Supports multiple distance metrics (euclidean, cosine, manhattan, hamming) with quantized variants, and offers Python bindings via PyO3 for integration beyond Rust. The implementation uses compressed bitmaps for graph edges (~200 bytes per vector overhead) and enables dynamic insertions/deletions without full re-indexing, while optional memory-mapping prefetching via `madvise` reduces cold-start latencies by several milliseconds.

71 stars and 46,404 monthly downloads. Available on PyPI.

No Dependents
Maintenance 10 / 25
Adoption 19 / 25
Maturity 18 / 25
Community 13 / 25

How are scores calculated?

Stars

71

Forks

9

Language

Rust

License

MIT

Last pushed

Mar 05, 2026

Monthly downloads

46,404

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/vector-db/nnethercott/hannoy"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.