what if the model in question, say a

lol there is no garuntee 😅 By default the model gets one chance

You could also loop for some max iterations.

In general, it takes a very good LLM to be reliable at this.

thanks for the clarification. what setting changes the "one chance by default."?

Sorry, not a setting, its hardcoded -- youd have to subclass the appropriate component and modify the logic from there

PRs definitely welcome as well