Vector database
Database optimised for "find the N nearest vectors to this query vector" — the engine behind RAG, semantic search, recommendations, dedup.
Built on ANN (Approximate Nearest Neighbor) rather than exact search. Algorithms: HNSW (de facto, fast + good recall), IVF (Inverted File), DiskANN (Microsoft, disk-based), ScaNN (Google). Distance measures: cosine similarity, dot product, L2 (Euclidean). Dedicated DBs: Pinecone, Qdrant, Weaviate, Chroma, Milvus, LanceDB. Built-into-other-DBs: pgvector (Postgres), Redis Stack, Elasticsearch kNN, MongoDB Atlas Vector.