Vector databases store and search embeddings.

An embedding is a list of numbers produced by a machine learning model. The useful property is that similar items tend to have similar vectors. Text with similar meaning, images with similar content, or users with similar behavior can be placed near each other in vector space.

This enables queries that ordinary keyword search does not handle well:

A vector database does not understand meaning by itself. The embedding model creates the representation. The vector database stores those vectors, indexes them, filters them, and retrieves the nearest matches efficiently.