ejaasaari/lorann

Approximate Nearest Neighbor search using reduced-rank regression, with extremely fast queries, tiny memory usage, and rapid indexing on modern vector embeddings.

50
/ 100
Established

Header-only C++17 library with Python bindings that decomposes high-dimensional embeddings using low-rank matrix factorization, enabling SIMD-accelerated distance computations across AVX2, AVX-512, and ARM NEON. Supports multiple data types (float32, float16, bfloat16, uint8, binary) and distance metrics (L2, inner product, cosine, Hamming), with experimental GPU batch query acceleration via PyTorch and index serialization for persistence.

Available on PyPI.

Maintenance 6 / 25
Adoption 12 / 25
Maturity 25 / 25
Community 7 / 25

How are scores calculated?

Stars

51

Forks

3

Language

C++

License

MIT

Last pushed

Dec 11, 2025

Monthly downloads

51

Commits (30d)

0

Dependencies

1

Get this data via API

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

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