pgvector

Open-source PostgreSQL extension for vector similarity search.

Updated April 2026

Overview

Website: github.com
Segment: Vector Databases
Posture: Vector DB Extension

Product overview

pgvector provides an extension to PostgreSQL enabling storage, indexing (HNSW, IVFFlat), and similarity search (cosine, L2, etc.) on high-dimensional vectors for AI applications like semantic search, recommendations, and RAG. It allows keeping vectors alongside relational data with full ACID compliance, JOINs, and Postgres features, used by developers on platforms like AWS Aurora, Heroku Postgres, Supabase, and CockroachDB. Distinct from dedicated vector DBs by avoiding infrastructure sprawl, leveraging existing Postgres expertise, and being completely free/open-source.,

Revenue model

Free open-source; no direct revenue or pricing—costs via PostgreSQL hosting (e.g., AWS RDS compute/storage, Supabase subscriptions $25+/mo).

Moat

Pgvector's key competitive moat is its seamless, native integration as an open-source PostgreSQL extension, enabling efficient vector similarity search alongside traditional relational data without requiring a separate database, which creates high switching costs for Postgres users and leverages the ecosystem's massive scale, ACID compliance, and familiarity. This is bolstered by strong performance in high-throughput scenarios (e.g., 11.4x higher query throughput than Qdrant at 99% recall) and ongoing optimizations like pgvector 0.8.0's 9x faster queries, making it defensible for hybrid workloads despite scalability limits in distributed setups.

Headwinds

Dependence on PostgreSQL ecosystem and potential performance limitations compared to purpose-built vector databases.