Hi, I am making a GPT to train sellers, I have a SQL database to return docs, It works of this way:
1.- The user tell that want to take a simulation of a sale.
2.- My GPT use an action. I return a document being information about a simulation with several topics in it, It can be politics about the enterprise, what can do, what can’t do, refunds, prices, etc.
3.- Chatgpt simulate to be a customer that interacts with a seller.
4.- Finally chatgpt must to evaluate the customer to verify that his/her answers are aligned to the information about the simulation gotten.
This is a large query that I use to get docs to put in the prompt:
This is where I add information in the prompt:
This is the prompt that I use to return all the information about documents, customer to simulate and product:
I’ve tried with less docs, but GPTs almost ever give me wrong answers, for example a seller say to chatgpt in a simulation that there is a garantee in 30 days and GPT say that it is correct, but it isn’t. I think that ChatGPT get confused by retrieve a lot of topics in a single action, and put attention in some of them, but I don’t know.
Thank you.