RunAnywhere Core

Private C++ core library providing unified ML inference across ONNX Runtime, LlamaCPP, and CoreML backends with multi-modality support.

🎯 Overview

RunAnywhere Core is the foundational C/C++ library that powers the RunAnywhere SDK family. It provides a unified, cross-platform API for running machine learning models on-device across multiple inference backends.

Key Features

• 🔄 Multi-Backend Support: ONNX Runtime, LlamaCPP, CoreML, TensorFlow Lite, ExecuTorch
• 🎤 Multi-Modality: Speech-to-Text (ASR), Text-to-Speech (TTS), Text Generation (LLM)
• 📱 Cross-Platform: iOS, Android, macOS, Linux with consistent C API
• ⚡ High Performance: Native C++ implementation with platform-specific optimizations
• 🔌 Clean Integration: Simple C API for easy binding to Swift, Kotlin, and other languages
• 📊 Built-in Telemetry: Performance monitoring and device capability detection

🏗️ Architecture

┌─────────────────────────────────────────────────────┐
│           Platform SDKs (Swift, Kotlin)             │
│  (runanywhere-swift, runanywhere-kotlin, ...)       │
└───────────────────┬─────────────────────────────────┘
                    │ C API
┌───────────────────┴─────────────────────────────────┐
│              RunAnywhere Core (C++)                  │
│  ┌──────────────────────────────────────────────┐   │
│  │        Common Infrastructure                 │   │
│  │  • Telemetry  • Device Info  • Utilities    │   │
│  └──────────────────────────────────────────────┘   │
│  ┌──────────┬──────────┬──────────┬──────────┐      │
│  │  ONNX    │ LlamaCPP │  CoreML  │ TFLite   │      │
│  │ Runtime  │          │          │          │      │
│  │ Backend  │ Backend  │ Backend  │ Backend  │      │
│  └────┬─────┴────┬─────┴────┬─────┴────┬─────┘      │
└───────┼──────────┼──────────┼──────────┼────────────┘
        │          │          │          │
   ┌────┴───┐ ┌───┴────┐ ┌───┴────┐ ┌──┴──────┐
   │ ONNX   │ │ llama  │ │ Core   │ │ TFLite  │
   │Runtime │ │  .cpp  │ │   ML   │ │         │
   └────────┘ └────────┘ └────────┘ └─────────┘

🚀 Quick Start

Prerequisites

• CMake 3.22 or later
• C++17 compatible compiler
• Platform-specific tools:
  • iOS/macOS: Xcode 14+
  • Android: Android NDK r25+
  • Linux: GCC 9+ or Clang 10+

Building

# Download ONNX Runtime dependency (iOS only, first time)
cd /path/to/runanywhere-core
./scripts/download-onnx-ios.sh

# Build for iOS (creates XCFramework)
./scripts/build-ios-onnx.sh

# The built XCFramework will be in dist/RunAnywhereONNX.xcframework

Integration

Copy the built XCFramework to your Swift/Kotlin SDK:

# For Swift SDK
cp -r dist/RunAnywhereONNX.xcframework \
  /path/to/runanywhere-swift/XCFrameworks/

See BUILD.md for detailed build instructions.

📚 Documentation

• Quick Start Guide - Get running in 5 minutes
• ONNX Runtime Setup - Detailed ONNX integration guide
• Modality Support - Multi-modality API documentation
• Implementation Status - Current implementation progress
• Scripts Documentation - Build scripts reference
• Architecture - System design and architecture

🔧 Build Scripts

Script	Purpose
scripts/download-onnx-ios.sh	Download ONNX Runtime iOS dependency
scripts/build-ios-onnx.sh	Build ONNX Runtime XCFramework for iOS

See scripts/README.md for detailed documentation.

📁 Repository Structure

runanywhere-core/
├── cmake/                 # CMake modules for dependency management
├── docs/                  # Documentation
├── include/               # Public C API headers
├── plans/                 # Architecture and planning docs
├── scripts/               # Build automation scripts
├── src/
│   ├── backends/         # Backend implementations
│   │   └── onnx/        # ONNX Runtime backend
│   │       ├── backend/ # Core C++ implementation
│   │       └── bridge/  # Platform bridges (iOS/Android)
│   └── common/          # Shared utilities
├── third_party/          # External dependencies (downloaded)
│   ├── json/            # nlohmann/json (submodule)
│   └── onnxruntime-ios/ # ONNX Runtime (downloaded via script)
├── BUILD.md             # Build instructions
└── README.md            # This file

🎯 Supported Backends

Backend	Status	Platforms	Modalities
ONNX Runtime	✅ Implemented	iOS, Android, macOS, Linux	ASR, TTS, LLM
LlamaCPP	🚧 Planned	iOS, Android, macOS, Linux	LLM
CoreML	🚧 Planned	iOS, macOS	ASR, TTS, LLM
TensorFlow Lite	🚧 Planned	iOS, Android, Linux	ASR, Vision
ExecuTorch	🚧 Planned	iOS, Android	LLM

🎤 Supported Modalities

• Speech-to-Text (ASR): Convert spoken audio to text
• Text-to-Speech (TTS): Generate natural speech from text
• Text Generation (LLM): Large language model inference

🛠️ Development

Prerequisites for Development

# macOS/iOS development
brew install cmake ninja

# Install Xcode Command Line Tools
xcode-select --install

Building for Different Platforms

iOS

./scripts/download-onnx-ios.sh  # First time only
./scripts/build-ios-onnx.sh

Android

# Coming soon

macOS

mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j$(sysctl -n hw.ncpu)

Linux

mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j$(nproc)

🧪 Testing

# Build with tests enabled
cmake .. -DBUILD_TESTS=ON
make
ctest --output-on-failure

📝 License

This is private, proprietary software. All rights reserved.

See LICENSE for details.

🤝 Contributing

This is a private repository. For RunAnywhere team members:

1. Create a feature branch from develop
2. Follow the code style (C++17, Google C++ Style Guide)
3. Write tests for new functionality
4. Submit a Pull Request

See CONTRIBUTING.md for detailed guidelines.

📞 Support

For issues and questions:

• Internal Team: Slack #runanywhere-core channel
• Bug Reports: GitHub Issues (private repo)
• Documentation: See docs/ directory

🔗 Related Projects

• runanywhere-swift - Swift SDK for iOS/macOS
• runanywhere-kotlin - Kotlin SDK for Android
• runanywhere-examples - Example applications

📊 Project Status

Current implementation status:

• ✅ ONNX Runtime backend (iOS)
• ✅ Multi-modality support (ASR, TTS, LLM)
• ✅ iOS XCFramework build pipeline
• ✅ Swift SDK integration
• 🚧 Android support
• 🚧 Additional backends

See IMPLEMENTATION_STATUS.md for detailed progress.

---

Made with ❤️ by the RunAnywhere Team