Vector Database

A vector database stores huge numbers of embeddings (meaning-coordinates) and quickly finds the ones that are "closest in meaning" to a given input. If a regular database is a filing cabinet built for finding exact matches, a vector database is a bookshelf where similar things are already grouped together.

It works behind the scenes in RAG: internal documents are converted into embeddings and stored here, and when a question comes in, the relevant passages are pulled out and handed to the AI. Specialized startups drew a lot of attention early in the AI boom, and now the trend is for established databases to build vector search in as a built-in feature.