Vector database
Also known as: vector DB · vector store
A vector database (vector store) is a database optimized for storing embeddings and answering “what’s closest to this vector?” fast.
It’s the retrieval engine of RAG:
- Store embeddings of your documents/chunks.
- Query with the embedding of your question.
- Retrieve the nearest neighbors (most relevant chunks).
Examples: Pinecone, Weaviate, Qdrant, Chroma, pgvector.
You don't need a big one: for small-to-medium corpora, an in-memory library or a Postgres extension (pgvector) is plenty. Reach for a dedicated vector DB at scale.