With function calling in MLX Swift, I experimented with chatting with your health data on-device!

I really learned a lot from the reasoning while vibe coding this feature

Automate up to 70% of your customer support today. Save time and help your users find answers quickly. Try now at

Awesome job, dude! Big fan of your work. This is a killer use case for local models. I built a similar prototype with MLX using Qwen2.5-1.5B-Instruct-4bit running on iPhone 16 Pro.🤙

Amazing! I am fascinated by Garmin's sleep data and I know their algorithm is much better than whatever we have on-device, but we never know what the future holds!

I need this but automatically tell me what workout should I do today based on historical data (workout, sleep, cardiovascular, etc) and goals that I set.

you guys gonna bring us actual apple intelligence before the two decades it will take apple to do [badly]. bravo

Need Tutorial

If you want to learn about tool use, here is a timestamped video by sensei 🙏

ohh, function calling? where do I look for it in mlx' repo?

This PR by @DePasqualeOrg is a good start that shows how to print the tool call method. You will have to parse the call yourself, and provide the output back to the response

@AndyDilanchi