What is Vector Database?

Full Definition

A vector database is a data storage system designed to efficiently index and query high-dimensional numerical vectors — typically the embeddings produced by AI models — using similarity search rather than exact key lookups. When a query arrives, it is embedded into the same vector space, and the database returns the k most similar stored vectors using algorithms such as HNSW (Hierarchical Navigable Small World) or IVF (Inverted File Index) for approximate nearest-neighbor search. Popular vector databases include Pinecone, Weaviate, Qdrant, Chroma, and pgvector (a PostgreSQL extension). They are the retrieval layer in RAG architectures, powering semantic search, recommendation engines, duplicate detection, and long-term AI agent memory. The key metrics are recall accuracy, query latency, and scalability to billions of vectors.