Free Open Source iPhone Chatbot

Описание: 🔗 Onyx repository: https://github.com/kiraa-ai/project-onyx

Run AI on your iPhone, iPad, or Mac for FREE, completely offline.

Introducing Project Onyx.

Token costs aren't going down, they're going up. If you're running AI in the cloud, that meter keeps ticking. But it doesn't have to be that way.

At WWDC this week, the headline went to Siri, but the announcement that actually matters was tucked away in the developer sessions: Core AI, Apple's new on-device framework that replaces Core ML and runs across the whole chip: CPU, GPU, and Neural Engine. Apple just put its weight behind keeping intelligence local: no token costs, no snooping, no data retention.

Core AI is still in beta, so in the meantime I built Project Onyx — a small, open-source app that runs a large language model entirely on your own device using Apple's MLX framework. No cloud. No API key. No account. Turn on airplane mode and it still works.

The name? Onyx = "On Your MLX" — and it's a stone you can't see into, which is the whole point: total privacy.

Onyx is a skeleton — a clean starting point with model loading, on-device inference, and a basic chat interface. Fork it, rebuild it, swap the model, bolt it onto your own product. It's yours.
What you need to get started:

A Mac + Xcode (free)
A free Apple ID

Deploy straight to your own device and hit Run. Step-by-step links are below.
The era where private, on-device AI was only for big companies? That's ending — and you can be one of the people building what comes next.


① Install Xcode & deploy to your device
Install Xcode (free, Mac App Store): https://apps.apple.com/us/app/xcode/i...
Xcode downloads & support (Apple): https://developer.apple.com/support/x...
Running your app on a device — official Apple guide: https://developer.apple.com/documenta...

Enabling Developer Mode on iPhone/iPad: https://developer.apple.com/documenta...

Building and running an app (Apple): https://developer.apple.com/documenta...

Create a free Apple ID (no paid account needed to test): https://account.apple.com

② The engine — Apple MLX (what makes it run on-device)
MLX (core framework): https://github.com/ml-explore/mlx
MLX Swift: https://github.com/ml-explore/mlx-swift
MLX Swift Examples — LLMEval + a full example chat app to learn from: https://github.com/ml-explore/mlx-swi...
MLX documentation: https://ml-explore.github.io/mlx/
Apple Machine Learning Research — running LLMs with MLX: https://machinelearning.apple.com/res...

③ Where to get models (free, open, ready to run)
MLX Community on Hugging Face (4-bit & 8-bit ready-to-run models): https://huggingface.co/mlx-community
Browse all MLX-compatible models: https://huggingface.co/models?library...

Using MLX at Hugging Face (how-to docs):
https://huggingface.co/docs/hub/en/mlx
Hugging Face (home of open models): https://huggingface.co

④ Learn Swift & build it out
Swift (free, open source): https://www.swift.org
Apple Developer — Swift: https://developer.apple.com/swift/
Apple's SwiftUI tutorials: https://developer.apple.com/tutorials...

⑤ Other ways to run local LLMs on a Mac (no Xcode required)
Ollama: https://ollama.com
LM Studio: https://lmstudio.ai

⚠️ Note: Onyx is a skeleton starter app provided as-is, with no support.
Free to fork and modify under Apache 2.0