coreai-models  by apple

On-device AI development toolkit for Apple platforms

Created 1 month ago
1,361 stars

Top 28.8% on SourcePulse

GitHubView on GitHub
1 Expert Loves This Project
Project Summary

This repository provides model export recipes, Python primitives, and Swift runtime utilities for building on-device AI applications using Apple's Core AI framework. It targets developers aiming to deploy machine learning models efficiently on macOS and iOS devices, offering a streamlined path from popular open-source models to native integration. The primary benefit is simplifying the complex process of model conversion and runtime deployment for Apple's ecosystem.

How It Works

The project is structured around four key components: model export recipes for converting models (e.g., from Hugging Face) to Core AI format, reusable Python primitives for authoring custom models in PyTorch, Swift runtime utilities (as a Swift package) for integrating exported models into apps, and specialized "skills" plugins designed to enhance coding agents' interaction with Core AI. This approach facilitates a cohesive workflow, enabling developers to leverage existing ML models and build performant on-device AI features.

Quick Start & Requirements

Installation requires uv (install via brew install uv or curl -LsSf https://astral.sh/uv/install.sh | sh). Running and app integration necessitate macOS and iOS 27.0+ and Xcode 27.0+. Export recipes are detailed within the models/ directory, often requiring reference to individual model READMEs. Models are exported as .aimodel files, potentially including resource folders for complex models like language or diffusion models. Command-line tools are available for direct model execution on macOS. Supported models can be listed using uv run coreai.model.registry --list-models after cloning the repository.

Highlighted Details

  • Agent Skills: Includes plugins like working-with-coreai, model-authoring, and model-compression-exploration to empower coding agents with Core AI deployment and optimization capabilities.
  • Model Catalog: Features a curated collection of export recipes for popular open-source models, simplifying their adaptation for Core AI.
  • Swift Runtime: Provides a Swift package (coreai-models) offering utilities for seamless integration of Core AI models into native macOS and iOS applications.

Maintenance & Community

The project is currently not accepting code contributions (pull requests will be closed) to maintain focus on a curated, well-tested model gallery and core utilities. Feedback, bug reports, and model requests are actively welcomed via GitHub Issues.

Licensing & Compatibility

This project is licensed under the BSD 3-Clause License, which is generally permissive for commercial use and integration into closed-source applications.

Limitations & Caveats

Code contributions are explicitly not accepted at this time, with the project prioritizing maintenance of its curated model gallery and core Swift runtime utilities. User feedback through GitHub Issues is the primary channel for suggesting improvements or new models.

Health Check
Last Commit

1 day ago

Responsiveness

Inactive

Pull Requests (30d)
42
Issues (30d)
28
Star History
635 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.