featured · github
pgvector: Vector Search Without a New Database
Add semantic search to Postgres instead of running a separate vector store. Same database, fewer moving parts, less operational overhead.
pgvector/pgvector ↗You're already using Postgres for your app's core data. pgvector lets you store embeddings and run similarity searches inside that same database instead of spinning up Pinecone, Weaviate, or another specialized tool.
Why it matters: Every extra database is another thing to keep in sync, monitor, and pay for. With pgvector, your user data, transactions, and embeddings all live in one place. You index vectors the same way you'd index regular columns. Query them with standard SQL. If you're building search, recommendations, or anything that needs "find similar things," pgvector handles it without the infrastructure tax.
The tradeoff is modest: it won't outperform a dedicated vector database at extreme scale, but for most products shipping today, that's a non-issue. The maturity is real. It's been battle-tested across production deployments.
Share kit
pgvector: Stop paying for a separate vector database
Vector search is moving into traditional databases. pgvector runs semantic search inside Postgres, no new infrastructure required. Skip the sync layer, keep one source of truth for your data, cut operational overhead. Open source, production-ready, worth evaluating if you're already on Postgres.
Vector search doesn't need its own database. pgvector runs semantic search inside Postgres, same SQL interface, same data store for OLTP and embeddings. If you're already on Postgres, this kills the case for a separate vector DB.
The vector DB wave arrived. Now it's consolidating back into the database you already use. pgvector brings semantic search into Postgres, eliminating sync complexity and cutting infrastructure costs for founders and teams shipping AI-assisted features. Open source. Production-ready.
Just realized we've been paying for separate vector databases when Postgres could handle it all along. pgvector crossed 12k stars by solving a real infrastructure problem: vector similarity search lives inside your existing database now, not in another silo. The shift is quiet but material. You get embeddings + OLTP in one system. No sync headaches. No new managed service bill. Postgres does both. If you're already on Postgres and shipping RAG or semantic search, this eliminates a whole operational layer. Worth a 20-minute spike if you haven't looked yet. https://github.com/pgvector/pgvector
pgvector: vector similarity search inside Postgres. 12k stars. No separate vector DB. No sync drift. Just embeddings + your data in one place. If you're already on Postgres, this is a no-brainer spike. https://github.com/pgvector/pgvector