What is RAG and why it’s needed?

I have seen Agents going a tad screwy a lot, esp. the more the conversation runs on. The better my prompt at the beginning, the better the results.

What I am interested in is how do we (or is it the LLM?) determine the “end” of a conversation? If results are better at the end, how do we (or can we?!) influence it?

Maybe it’s like talking to a child - “This is your last warning, or you’re off to bed!”

Interested in people’s thoughts!

1 Like