Repolode
Submit
software-mansion

software-mansion/react-native-executorch

A React Native library for running machine learning models directly on phones with no network calls, built on Meta's ExecuTorch runtime.

View on GitHubWebsite
Stars
1,773
Forks
97
Open issues
48
Last commit
3 hours ago

What is it?

React Native ExecuTorch is an on-device AI inference library for React Native apps, built on PyTorch's ExecuTorch runtime. Models run on the user's phone, so inference works offline and user data is not sent to a server.

The library provides ready-to-use hooks for tasks such as LLM chat, computer vision, speech-to-text, text-to-speech and embeddings, with automatic model download, caching and lifecycle management. It also includes a catalog of pre-exported models and supports loading custom .pte models. Lower-level tensor operations and custom pipelines can be written in TypeScript.

Hardware acceleration uses XNNPACK for CPU, Core ML and MLX on Apple Silicon, and Vulkan on Android GPUs. It requires the New Architecture, React Native 0.83+ or Expo SDK 55+ with development builds, iOS 17+ and Android 13+.

Key features

  • 01Runs LLM, vision, speech and embedding models fully on-device with no cloud inference.
  • 02Ready-to-use React hooks for common AI tasks with automatic model caching.
  • 03Hardware acceleration via XNNPACK, Core ML, MLX and Vulkan.
  • 04Pre-exported model catalog, plus support for custom .pte models.
  • 05Custom pipelines can be built in TypeScript with low-level tensor operations.

Use cases

  • Adding an offline chatbot to a React Native app without server costs.
  • Running on-device object detection or OCR in a camera feature.
  • Keeping sensitive speech or text processing on the user's phone.
  • Building a privacy-focused mobile AI product.

Quick start

  1. 1

    Install the library and its peer dependencies with npm

    npm install react-native-executorch react-native-worklets react-native-blob-util
  2. 2

    Use a task hook such as useLLMChatSession with a pre-exported model in your app

Things to know

  • Requires the New Architecture, React Native 0.83+ or Expo SDK 55+ with development builds; Expo Go is not supported.
  • Requires iOS 17.0+ or Android 13+.
  • License is listed as NOASSERTION, so check the license terms before use.
  • Downloads of pre-exported models send an anonymous telemetry event unless disabled.

Topics

#on-device-ai#react-native#mobile-inference#llm-inference#computer-vision#speech-to-text#pytorch#ai#executorch#object-detection

FAQ

What is react-native-executorch?
A React Native library for running machine learning models directly on phones with no network calls, built on Meta's ExecuTorch runtime. React Native ExecuTorch is an on-device AI inference library for React Native apps, built on PyTorch's ExecuTorch runtime. Models run on the user's phone, so inference works offline and user data is not sent to a server.
Is react-native-executorch free to use? What license does it have?
react-native-executorch is open source under the Other license. Check the license terms before using it in commercial products.
Is react-native-executorch still maintained?
Actively maintained. The last commit to react-native-executorch was on Oct 22, 2026.
What can I use react-native-executorch for?
Adding an offline chatbot to a React Native app without server costs. Running on-device object detection or OCR in a camera feature. Keeping sensitive speech or text processing on the user's phone. Building a privacy-focused mobile AI product.
How do I install react-native-executorch?
Install the library and its peer dependencies with npm: npm install react-native-executorch react-native-worklets react-native-blob-util Use a task hook such as useLLMChatSession with a pre-exported model in your app
Where can I get react-native-executorch?
The source code and installation instructions are on GitHub: https://github.com/software-mansion/react-native-executorch

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects