Could you share a prompt that makes large language models role-play to generate multi-turn tool-calling conversations?

[TOMATODA]

Our team currently needs to synthesize training data based on some internal APIs.Your paper's approach of using multiple model role-playing to generate high-quality datasets and filter out low-quality ones is very informative.Since we lack experience in data synthesis, would you be willing to share the prompts you used for having large language models generate data and filter data?Really appreciate that.