Vector database

A database built to search embeddings quickly — the memory behind most RAG systems.

Builder path · 2 of 6

What it is

A vector database stores embeddings and finds the ones closest to a query. Options range from extensions of familiar databases, such as pgvector for PostgreSQL, to dedicated search engines.

Example

Thousands of product sheets are split into passages, embedded and stored; a question returns the most relevant passages almost instantly.

What can go wrong

One shared index for all users or clients can leak data between them.

How to secure it

Separate indexes or filter strictly by tenant and permissions, encrypt data at rest, and remove vectors when the source document is deleted.

Related service: Applied AI