How Does a Vector Database Work?

Most explanations of vector databases turn into a wall of math. Here's the one-minute version instead.

The idea, in one sentence

A vector database doesn't store words — it stores meaning, as a list of numbers, so it can find things that are similar in meaning even when they don't share a single keyword.

What's happening in the animation

1. Every document gets embedded. An embedding model reads "The cat sat on the mat," "Dogs are loyal pets," "Paris is in France," and "I love pizza" — and turns each one into a vector (a point in space).
2. Similar meanings land near each other. The cat and dog sentences — both about pets — end up close together. Paris and pizza end up far away, because they mean something completely different. No shared keywords required.
3. A new query gets embedded the same way. Someone asks "Tell me about puppies." It goes through the exact same embedding model.
4. Similarity search compares the query to every stored vector, then narrows down to the nearest one.
5. The database returns the original content behind that nearest vector — in this case, "Dogs are loyal pets" — even though the query never said "dogs," "loyal," or "pets."

Why this matters

Keyword search asks: does this text contain the same words? Vector search asks: does this text mean the same thing?

That single shift is why vector databases power semantic search, recommendation systems, and RAG (retrieval-augmented generation) pipelines for LLMs — they let you retrieve by meaning, not by string matching.