Vector stores and indexes

Vector databases store embeddings and run approximate nearest-neighbour (ANN) search with metadata filters.

Evolving: the basics hold, details shift. Reviewed 17 Oct 2026.

Key ideas

HNSW is the default ANN index: fast and accurate, memory-hungry. IVF and disk-based indexes trade recall for scale.
Options: pgvector (Postgres), OpenSearch, Amazon S3 Vectors, Qdrant, Weaviate, Pinecone, Milvus, LanceDB.
If you already run Postgres and have under ~10M vectors, pgvector is usually enough.
Check filtered-search performance; pre-filtering by tenant or permissions can collapse recall in some engines.
Plan for re-indexing, backups and multi-tenancy (namespace, partition or row-level filter).

Adding a dedicated vector DB before proving the existing database cannot do the job.