Definition
A vector database is a database designed to store embeddings, the numerical representations of meaning, and to find the items closest to a given query very quickly. A traditional database answers questions such as "which orders were placed yesterday". A vector database answers "which documents are most similar to this question". It is the storage and search layer that makes semantic search and retrieval-augmented generation work at scale.
In a company, the vector database sits between your documents and your AI assistant. Manuals, tickets, emails and policies are chunked, embedded and indexed. When an employee or customer asks a question, the system embeds the question, retrieves the nearest chunks and hands them to the language model. Options range from dedicated products to vector extensions in databases you already run, such as PostgreSQL, which is often the pragmatic choice.
Vector search has become a feature of most mainstream databases rather than a separate category, and many teams no longer need a specialised product unless they index hundreds of millions of items. The misconception is that the vector database is the intelligence. It is plumbing. Chunking strategy, metadata filters, hybrid search with keywords and re-ranking determine whether the right document is found, and the language model does the reasoning.
In practice
A law firm indexes 200,000 past contract clauses in a vector database. Lawyers ask for "non-compete clauses enforceable in Germany" and get the closest precedents ranked in under a second.
Why it matters
If you plan an AI assistant over your company knowledge, you will need vector search. Knowing it can often live in your existing database saves money and one more vendor relationship.
Frequently asked questions
- Do I need a dedicated vector database for RAG?
- Not necessarily. For most companies with thousands to a few million documents, vector extensions in existing databases such as PostgreSQL work well and simplify operations. Dedicated vector databases make sense at very large scale or when you need specialised features and latency guarantees.
- What is the difference between a vector database and a regular database?
- A regular database finds exact matches and ranges on structured fields. A vector database finds the nearest neighbours in meaning by comparing embeddings. Many modern databases now do both, letting you filter by customer or date and rank by semantic similarity in one query.