Vector database
A database built to store and search embeddings — the numerical fingerprints AI uses to represent meaning. It finds items by similarity rather than exact match, which is what lets an AI system retrieve the passages most relevant to a question.
An ordinary database is good at exact matches: find the customer with this reference number, or every order placed on this date. But AI often needs something looser — the documents that are about the same topic as a question, even if they share no words with it. That calls for a different kind of store. A vector database holds embeddings, the lists of numbers that capture the meaning of a piece of text or an image, and specialises in finding the ones that sit closest together.
“Closest” is the key idea. Because similar meanings produce similar embeddings, the database can answer a query by returning the items nearest to it in this numerical space, ranked by similarity. Ask about “cancelling a subscription” and it can surface a help article titled “how to end your plan”, despite the different wording.
Vector databases became widely used alongside retrieval-augmented generation, where a system searches a knowledge base for relevant passages and feeds them to a language model. The vector database is the part doing that fast, meaning-based search, and it now underpins a great many AI products built on top of company documents.