Skip to content
The Internet Compass

AI

Vector Database

A vector database stores embeddings and retrieves the nearest matches to a query vector using approximate nearest-neighbour indexes such as HNSW or IVF, trading exactness for speed at scale.

Dedicated vector databases compete with vector extensions to existing databases; for most workloads under a few million vectors, the extension is sufficient and operationally simpler.

Production retrieval systems typically combine vector search with keyword search, because pure semantic matching handles exact identifiers and rare terms poorly.