How efficient is use a function to retrieve RAG data instead of adding files to assistant? Is the Assistant search worse in this case?
Hello there!
Well, retrieval from Assistants is RAG.
By building your own RAG db and function, you essentially have more control over how you could use RAG and how much data you can place in the db. For example, you could use your own RAG function with chat completion, but files added to an Assistant for retrieval can’t be used outside of that Assistant and the assistant API call.
So, it’s less about efficiency, and more about flexibility and ease of use.
Assistants is easier to do
Independent RAG is more robust.
I have a question about how to implement the assistant .
Should the file be placed in the assistant or in the thread?
RAG libraries and functions
Is it a csv file or something? Function is function call?
(I want to be an assistant for system administrators to query, but I encountered the above difficulties.)