Term
Vector database
A database optimised for storing embeddings and running approximate nearest-neighbour search at scale. Common options: pgvector (Postgres extension), Pinecone, Weaviate, Qdrant, Chroma, Milvus.
Background
Vector databases provide indexing structures (HNSW, IVF-PQ) that make k-NN search over millions of vectors fast and approximate. They are the storage layer for RAG. In 2026, pgvector remains the most popular choice for embedded use cases because it adds vector support to an existing Postgres instance, eliminating a separate service. Pinecone leads managed offerings; Qdrant and Weaviate cover self-hosted needs. Vector DB choice is rarely the bottleneck — embedding quality and chunking strategy matter more.