What is a Vector database?
A vector database stores embeddings and quickly finds the nearest matches to a query vector, which is how semantic search and RAG systems retrieve relevant content.
In more detail
Comparing a query against millions of vectors one by one is slow, so these databases use approximate nearest-neighbor indexes to return close matches in milliseconds. Options include purpose-built products such as Pinecone, Qdrant and Weaviate, and extensions to databases you may already run, such as pgvector for PostgreSQL. For many business workloads under a few million chunks, the extension is enough and saves you another system to secure. Whatever you choose, store metadata with each vector (source file, owner, date, access group) so you can filter by permission before the model sees anything. Retrieval that ignores permissions is the most common leak in knowledge assistants.
Example
A 200-person consultancy keeps its handbook, past reports and policies in PostgreSQL with pgvector. Each chunk carries the team that owns it, so an associate's search never returns partner-only documents.