What is Vector Database

Definition

A vector database is a specialised store designed to index and query high-dimensional vector embeddings efficiently, enabling similarity search that finds items closest in meaning to a query, which underpins semantic search, recommendations, and retrieval for AI applications.

« Back to Glossary Index

Performs fast similarity search over millions of high-dimensional embeddings
Powers semantic retrieval where meaning, not exact keywords, drives results
Scales nearest-neighbour queries that traditional databases handle poorly
Forms the retrieval backbone for RAG and recommendation systems

Glossary: Embeddings

Real World Example

A documentation site stores embeddings of every article in a vector database, so when a user asks a question, the system retrieves the semantically closest passages, even when the wording differs entirely from the query.

FAQs

What does a vector database store?

It stores vector embeddings, numerical representations of data, along with metadata, optimised for similarity search.

How is it different from a relational database?

It is optimised for nearest-neighbour similarity queries over high-dimensional vectors rather than exact matches over structured rows.

What are vector databases used for?

Semantic search, recommendation, retrieval-augmented generation, and any task that matches items by similarity of meaning.