Introduction to Function-calling with fine-tuned models ( A "quick-start" with some examples )

Hi everyone, I’m fairly new to ChatGPT API and all its features. I’ve created a few fine-tuned models and had good results with them. Unfortunately, despite the results were good, I need them to be a little different. With that said, I’d love to understand the function-calls in order to be able to try different approaches. It’s a part of the “toolset” that I’m not quite sure how to use it…

So, to summarize, it’s a tool that I really want to try out.

In case some of you are interested in what I’m building right now, I’m simply creating a fine-tuned model to help coworkers to search some information about our company, things that are usually in a guideline or tons of documents. Nothing out of the ordinary, but here is some of the problemas I’m facing:

  • When asking the model about things that were not in the trainning context, it’ll give an answer, even with the temperature very low. Ideally it should say something like “I’m not supposed to answer this”, “I don’t know”, “Something in denial”
  • I’m trying to get some forms back, like contractual forms for example. Here I think the function-call could help. The forms should be exactly the way it was written.

Probably a common chatbot with a flow action would solve this problems… But I’m wondering if i could do them with ChatGPT API.

Thanks a lot for your time!

For those who are also beginning, one of the best articles I’ve read it’s this: Function Calling: Integrate Your GPT Chatbot With Anything - Semaphore

However, considering how fast ChatGPT is changing, by the time you are seeing this, it might be outdated.