FALCONN-LIB/FALCONN

Health v2 · maintenance only

Adoption not part of the score

1159 stars · 194 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

FALCONN is a C++ library (with a Python wrapper) implementing locality-sensitive hashing (LSH) data structures for fast nearest neighbor search in high-dimensional spaces, primarily under cosine similarity. It is header-only, depends on Eigen and FFHT, and is optimized for dense and sparse data with multi-probe LSH.

Use cases

• find nearest neighbors by cosine similarity in high-dimensional vectors
• fast similarity search over a million embedding vectors
• maximum inner product search on a memory budget
• nearest neighbor search under Euclidean distance
• build an LSH index for sparse feature vectors
• fast approximate nearest neighbor lookup in milliseconds per query

When to choose

• you need LSH-based approximate nearest neighbor search with a restrictive RAM budget
• you work in C++ or Python with dense or sparse high-dimensional data
• you want a lightweight header-only library with minimal dependencies

When to avoid

• you need graph-based ANN methods like HNSW that often outperform LSH on large RAM budgets
• you need GPU acceleration or distributed indexing
• you need actively developed features beyond the current LSH families

Facets

library · maturity maintenance

search-engine machine-learning machine-learning databases cpp python cross-platform locality-sensitive-hashing nearest-neighbor-search ann cosine-similarity header-only high-dimensional-data algorithms search

Member repositories

For agents

markdown · JSON · MCP: product_card(name="FALCONN-LIB/FALCONN")