pgvector and PostgreSQL for AI Vector Search Cheat Sheet

pgvector is an open-source PostgreSQL extension that transforms PostgreSQL into a vector database by adding native support for storing, indexing, and querying high-dimensional vector embeddings. Unlike standalone vector databases, pgvector keeps vectors alongside your existing relational data, enabling you to build semantic search, recommendations, and RAG (Retrieval Augmented Generation) systems without introducing a separate database layer. The extension supports approximate nearest neighbor (ANN) search through two index types (IVFFlat and HNSW), three distance metrics (L2, cosine, inner product), and scales from thousands to tens of millions of vectors depending on hardware and tuning. The key mental model: pgvector makes vector similarity search feel like a native PostgreSQL feature—vectors are just another column type, and similarity queries use familiar SQL with specialized operators like <-> for distance.