Vector Database

Overview

A vector database is designed for a different question than a traditional one. Where a conventional database finds records that match exactly, a vector database finds records that are similar in meaning, returning the items whose embeddings sit closest to a query's. This is what makes search by meaning possible: asking for the most relevant passages on a topic, rather than only those containing a precise word.

How it works

Data is first converted into embeddings, vectors capturing meaning, and stored in the database. When a query arrives, it too is embedded, and the database searches for the stored vectors nearest to it. Doing this quickly across millions of vectors requires specialized indexing techniques that find approximate nearest neighbors efficiently, and the search itself can be accelerated on GPUs for low latency at scale.

Why it matters

Vector databases are the retrieval half of retrieval-augmented generation and the backbone of semantic search and recommendation. They are how an AI system reaches into a body of knowledge to ground its answers in relevant, current information. Because that body of knowledge is frequently an organization's own proprietary data, where the vector database lives is directly a matter of data control and sovereignty.