pgvecto.rs

pgvecto.rs is an open-source Postgres extension written in Rust with 2,100+ stars on GitHub that adds vector similarity search to PostgreSQL, built for the embedding workloads behind RAG and AI search applications.

Overview

If your application data already lives in Postgres, adding a separate vector database means syncing two systems forever. pgvecto.rs, from the TensorChord team, takes the extension route instead: install it into Postgres and your existing tables gain vector similarity search, written in Rust on the pgrx framework. Its headline difference from pgvector is filtered search done properly — its VBASE method is designed for single-vector TopK queries combined with relational filters and joins, so a request for the ten nearest vectors within one tenant or category returns ten real results instead of a filtered-down handful. It supports high-dimensional vectors and the indexing needed for low-latency nearest-neighbor search at scale. The team now also develops VectorChord, a newer implementation with better stability and performance, and publishes a migration guide — pgvecto.rs remains available and documented, with Docker images published for quick deployment.

Key Features

• Native Postgres extension – vector search inside your existing database, no extra system.
• Filtered search that works – VBASE method for TopK plus filter plus join queries.
• Written in Rust – built on pgrx for performance and memory safety.
• High-dimensional vectors – designed for modern embedding models and RAG workloads.
• Docker images – published images for fast local and production deployment.
• Migration path – documented route to the newer VectorChord implementation.

💳 Pricing

pgvecto.rs is free and open source under the Apache License 2.0 for self-hosting in your own Postgres. The VectorChord project from the same team also offers cloud services with their own pricing — check the official site for current plans.

Who Is It For

Backend and platform engineers building RAG, semantic search or recommendation features on Postgres who want vector search where their data already lives, with filtering that behaves correctly.