The core idea

A database built to search by distance.

A vector database stores embeddings and answers one question fast: which stored vectors are closest to this query vector?