← Back to glossary

RAG and search

Vector database

A data system optimized to store vectors and retrieve items whose numerical representations are most similar to a query.

Business example

Document chunks are stored with embeddings, source identifiers, and permission metadata for fast retrieval.

How it works

1.Stores vectors alongside metadata.
2.Indexes them for approximate nearest-neighbor search.
3.Filters results by tenant, permissions, source, or freshness.

Common misconceptions

A vector database is the complete RAG system.

It is one retrieval component; ingestion, authorization, ranking, generation, and evaluation remain necessary.

Related concepts

Embeddings Semantic search Retrieval-Augmented Generation (RAG)

Related reading

RAG is not uploading PDFs: permissions, freshness, traceability, and reliable sources

Sources

https://github.com/pgvector/pgvector

Reviewed: 2026-08-04

Reviewed by: Javier Chulvi Bernad · LLM Engineer · Madrid