Introducing ToolsGen 🛠️: A modular Python library for synthesizing tool-calling datasets

[atasoglu]

I built a tool to solve a problem I kept running into: creating quality datasets for training LLMs to use tools.

ToolsGen takes your JSON tool definitions and automatically generates realistic user requests, corresponding tool calls, and evaluates them using an LLM-as-a-judge pipeline. It outputs datasets ready to use with Hugging Face.

What makes it useful:

• Generates realistic user requests + tool calls from JSON definitions
• LLM-as-a-judge quality scoring with multi-dimensional rubrics
• Multiple sampling strategies (random, parameter-aware, semantic)
• OpenAI-compatible API support
• Outputs JSONL with train/val splits

Still early days (API isn't stable yet), but it's already helping me generate tool-calling datasets much faster.

Check it out: https://github.com/atasoglu/toolsgen

Here is an example dataset generated with this library: https://huggingface.co/datasets/atasoglu/nano-tool-calling-v1

Happy to hear feedback or ideas!

[julien-c]

This is quite cool! I like it