RAG is not the only reason vector search matters

RAG made vector search visible. That is not the same as saying it is the only use case that matters.

A lot of developers first encounter vector search through RAG.

That makes sense. It is one of the most visible AI application patterns today: embed content, retrieve the closest matches, pass them to the model.

But I think that first impression can be misleading.

RAG is not the only reason vector search matters.

It is just one of the first use cases that made the value obvious.

What actually makes vector search useful is something more general: similarity.

Once similarity becomes the thing you care about, vector search starts to show up in more places than people often expect. Retrieval is one example. But so are semantic search, related-content matching, duplicate detection, and anomaly detection.

RAG may be the most familiar example, but it should not define how we think about vector search.

That feels like the better mental model to me.

Not: vector search is mainly for chat.

But: vector search is a building block for similarity-driven applications.

RAG just happens to be one very visible way it shows up.

Related: my Qdrant + .NET examples repo

https://github.com/ovnecron/qdrant-dotnet-examples

P.S. In case you missed it: Microsoft announced the public preview of Fireworks AI on Microsoft Foundry - worth a look if you care about open models and production AI on Azure.