The short answer

A vector database stores embeddings — lists of numbers that capture meaning — and finds the ones most similar to a query in milliseconds. It is the retrieval engine behind semantic search, recommendations, and RAG, scaling to billions of items.

Why a normal database is not enough

A relational database is built to match exact values: find the row where id = 42. A vector database is built to match meaning: find the ten items closest to this one in a high-dimensional space. That is a different problem, and it needs a different index.

How similarity search works

Every item is stored as an embedding. At query time you embed the query, then the database returns the nearest vectors by cosine or dot-product distance. Approximate nearest-neighbour (ANN) indexes like HNSW make this fast enough to run over billions of vectors in milliseconds.

Where operators use it

Product discovery, deduplication, recommendations, and grounding LLMs in your own data through RAG. Once your catalogue or documents are embedded, the vector database is the layer every AI feature queries.

The choice that matters

Pinecone, Qdrant, Weaviate, Milvus, and pgvector all do the core job. The real decisions are your embedding model, your chunking strategy, and your re-ranking — those drive retrieval quality far more than which database you pick.

Frequently asked

What is a vector database used for?

Storing embeddings and running similarity search — the retrieval layer behind semantic search, recommendations, deduplication, and Retrieval-Augmented Generation (RAG).

What is the difference between a vector database and a regular database?

A regular database matches exact values and structured queries. A vector database matches by meaning, returning the items whose embeddings are closest to a query vector using nearest-neighbour search.

Do you always need a dedicated vector database?

No. For small workloads, pgvector inside Postgres is often enough. Dedicated vector databases earn their keep at scale, where index performance and filtering over millions of vectors matter.