What is Vector Database?
A database optimized for storing embeddings and finding nearest neighbors by semantic similarity at scale.
Definition
A vector database indexes high-dimensional embedding vectors and answers approximate nearest-neighbor queries—"find the 10 items most similar to this one"—in milliseconds across millions of entries, using ANN algorithms like HNSW or IVF rather than exact scans. Similarity is measured by cosine or dot-product distance, typically with metadata filtering combined in the same query. Dedicated engines include Pinecone, Weaviate, Qdrant, and Milvus, while pgvector brings vector search into Postgres and many teams prefer that operational simplicity. Vector stores are the retrieval backbone of RAG systems, semantic search, recommendations, and deduplication pipelines.