Definition
A vector database stores each document chunk, image or record as a vector produced by an embedding model, alongside the original content and metadata. Given a query vector, it returns the nearest neighbours, which are the items closest in meaning even if they share no words.
Specialised products exist, and many general databases (PostgreSQL with pgvector, search engines, cloud data platforms) now offer vector search too. For most companies the choice matters less than the quality of the content, the embeddings and the access controls applied to what comes back.