What is the difference between RAG and a vector database?

A vector database is a tool, and RAG is a technique — a vector database is one piece you might use inside a RAG system, not the whole system itself.

RAG is the overall pattern

RAG stands for Retrieval-Augmented Generation. Instead of relying only on what a model memorized during training, RAG first retrieves relevant documents or facts, then hands them to the AI as context so it can generate a grounded, current answer. That's the whole recipe: retrieve, then generate.

A vector database is one way to do the "retrieve" part

A vector database stores your documents as embeddings — numerical representations that capture meaning — so you can search by similarity instead of just matching keywords. That's genuinely useful for RAG, since it can find relevant text even when the wording doesn't match exactly.

Why they're not interchangeable

You can build RAG without a vector database — plenty of systems retrieve using plain keyword search, a SQL query, or an API call, and it still counts as RAG. And a vector database has plenty of jobs outside RAG too, like product recommendations, image similarity search, or deduplication.

Think of it this way: RAG is the recipe, and a vector database is one popular ingredient. You can swap out the ingredient and still cook the same dish, and you can use that ingredient in a completely different recipe.