pgvecto.rs

Why we included this project

Teams already running Postgres can add nearest-neighbor search without spinning up a separate vector store, because pgvecto.rs is a Rust extension built on pgrx that layers embedding similarity directly into the database. That means you query vectors with familiar SQL operators and index types instead of managing another service. Its practical strengths are hybrid filtering, which combines top-k retrieval with relational predicates and joins, plus support for very high-dimensional vectors and compact data types like binary vectors, FP16, and INT8 for memory savings. One thing to weigh before committing: the maintainers now point new deployments toward VectorChord as a more stable successor and provide a migration guide, so the roadmap matters. For evaluating in-database vector search on an existing Postgres workload, it's a well-documented and honest starting point.