pg_embedding

PostgreSQL extension for high-speed vector similarity search using the Hierarchical Navigable Small World (HNSW) indexing algorithm.

pg_embedding is a PostgreSQL extension engineered for efficient, high-dimensional vector similarity search. It leverages the Hierarchical Navigable Small World (HNSW) indexing algorithm to optimize Approximate Nearest Neighbor (ANN) queries. Benchmarks have shown it can deliver up to 20x faster search speeds than the IVFFlat index used by pgvector, maintaining 99% accuracy for graph-based searches. The extension supports critical distance metrics: Euclidean (L2), Cosine, and Manhattan, providing a robust solution for integrating vector embeddings directly into your Postgres database.

https://github.com/neondatabase/pg_embedding