I am using the GPT real-time API to make function calls, and I want the model’s response to pause as soon as a function call is detected. The model should not generate any further response until the function’s execution is complete and its response is returned.
This is already what it’s doing. (If you’re talking about the OpenAI Realtime API)
In the function you can describe “This function is for xyz, when executing it, do not say anything to the user and wait until it is finished”.
Just don’t post a response.create until your function is done processing.
On the contrary, if you want the opposite, so the AI saying “Hold on one second while I do that for you” you would pass that in the functions description instead.
Cheers. 