Vector database

A vector database stores embeddings and finds the ones closest in meaning to a query, very quickly.

In practice: When a tool searches thousands of documents by meaning rather than exact keywords, it usually converts the question to an embedding and asks a vector database for the nearest matches. The results are then passed to a language model to write the answer.

Why it matters when picking a tool

This matters mainly if you are building your own AI features. For off-the-shelf tools, what counts is the result: whether search finds the right passage. When choosing developer infrastructure, compare cost at your data size, hosting options and how easily it connects to your existing stack.