Vector Database
A database optimized for storing embeddings and finding nearest neighbors at scale.
Last updated: April 26, 2026
Definition
A vector database stores embeddings and supports approximate-nearest-neighbor (ANN) search across millions to billions of vectors in milliseconds. Common options Apr 2026: Pinecone (managed, simple), Weaviate (open source, hybrid search), Qdrant (open source, fast), Chroma (lightweight, embedded), pgvector (PostgreSQL extension, no extra infrastructure), AWS S3 Vectors (cheapest at scale, integrated with Bedrock). The choice often comes down to existing infrastructure. Pgvector if you already run Postgres, S3 Vectors if you are on AWS Bedrock.
When To Use
Required for production RAG. For prototypes under 10k chunks, an in-memory FAISS index or pgvector works. Above that, pick a managed service.
Related Terms
Building with Vector Database?
I've shipped this pattern in real production systems. If you want a second pair of eyes on your architecture, that's what I do.