What is Vector Database?

A specialized database engine designed to store, index, and query high-dimensional vector embeddings with ultra-low latency using approximate nearest neighbor (ANN) search algorithms.

Key facts

Utilizes indexing algorithms like HNSW (Hierarchical Navigable Small World) and IVF.
Powers semantic search, recommendation engines, and enterprise RAG architectures.
Measures similarity through cosine distance, dot product, or Euclidean Euclidean metrics.
Examples include Pinecone, Qdrant, Chroma, Milvus, and pgvector for PostgreSQL.

Explanation

Traditional relational databases index scalar values (numbers, strings). Vector databases index multidimensional floating-point arrays generated by embedding models. This enables machines to calculate mathematical similarity between abstract concepts, matching queries with conceptually related documents even if they share zero identical keywords.