tensorchord/pgvecto.rs

Implements the VBASE indexing method for combined vector search with relational filtering and joins, supports extended vector types (binary, FP16, INT8) and up to 65535 dimensions with runtime SIMD dispatch, and manages indexes separately from PostgreSQL's native storage for optimized performance. Built as a Rust-based PostgreSQL extension using pgrx, it provides three distance metrics (Euclidean, dot product, cosine) through SQL operators for seamless integration with standard queries.

2,158 stars. No commits in the last 6 months.