In my experience, solving the problem of determining when a chatbot should stop asking questions is made much easier by using OpenAI Assistant's structured response capabilities in JSON. Here's how it works:

In the JSON, you define the parameters you want to be returned, and by using the 'required' parameter, you can specify which parameters must always be included. For example, what parameters you could include:

"current_question" - Tracks the question currently being asked.

"stop_questions" - A boolean set to false by default and updated to true once all questions are asked, ensuring the assistant knows when to stop.

When I faced a similar challenge, this setup worked really well. By prompting GPT to produce structured outputs, it strictly followed the specified format and stopped when required.

For more information, check OpenAI's documentation: https://platform.openai.com/docs/guides/structured-outputs