Skip to main content

Release Notes v1.4.0

Release Date: October 16, 2025
Version: 1.4.0
Codename: "Real ExecuTorch + QNN Integration"

🚀 Major Features​

Real ExecuTorch + QNN Integration​

  • ✅ Actual Model Inference: Real Llama3.2-1B running on device
  • ✅ Qualcomm QNN Backend: Hardware acceleration via AI Engine Direct
  • ✅ v79 Context Binaries: Optimized for SoC Model-69
  • ✅ ARM64-v8a Support: Native Android architecture

Professional Documentation​

  • ✅ EdgeAI Documentation Platform: Modern, responsive documentation
  • ✅ Technical Guides: Comprehensive setup and architecture docs
  • ✅ GitHub Pages: Automatic deployment and hosting
  • ✅ Mobile Optimized: Perfect mobile documentation experience

🔧 Technical Improvements​

Native C++ Engine​

  • Real Model Loading: Actual PyTorch model weights
  • Tokenization: Real SentencePiece tokenizer
  • Inference Loop: Complete generation pipeline
  • Decoding: Human-readable text output

Android Integration​

  • JNI Interface: Seamless Kotlin ↔ C++ communication
  • Asset Management: Efficient model file handling
  • Error Handling: Robust error reporting and recovery
  • Performance: Optimized for mobile devices

📚 Documentation​

New EdgeAI Documentation Platform​

  • Getting Started: Quick setup guide
  • Technical Docs: Deep dive into architecture
  • Setup Guides: Step-by-step configuration
  • Release Notes: Version history and changelog

Content Structure​

  • Project Structure: Complete codebase overview
  • Qualcomm AI HUB Setup: Detailed configuration guide
  • Architecture: Technical implementation details
  • Troubleshooting: Common issues and solutions

🎯 Performance​

Optimizations​

  • Hardware Acceleration: Qualcomm QNN backend
  • Memory Management: Efficient model loading
  • Response Quality: Improved text generation
  • Speed: Faster inference times

Compatibility​

  • Android 7.0+: Wide device support
  • ARM64-v8a: Modern architecture targeting
  • Qualcomm Chipsets: Optimized for Snapdragon processors

🐛 Bug Fixes​

  • Fixed native library loading issues
  • Resolved tokenization problems
  • Improved error handling and reporting
  • Fixed CSS styling issues in documentation

🔄 Migration from Previous Versions​

From v1.3.0​

  • Documentation moved to EdgeAI Documentation Platform
  • Improved native inference engine
  • Better error handling and logging

Breaking Changes​

  • None - fully backward compatible

📦 Installation​

APK Download​

  • Release APK: Available on GitHub Releases
  • Debug APK: For development and testing
  • Size: Optimized for mobile distribution

Requirements​

  • Android 7.0 (API level 24) or higher
  • ARM64-v8a architecture
  • Qualcomm Snapdragon processor (recommended)

🎉 What's Next​

Upcoming Features​

  • Model Fine-tuning: Custom model training
  • Multi-language Support: Internationalization
  • Performance Monitoring: Real-time metrics
  • Plugin System: Extensible architecture

Community​

  • Open Source: Full source code available
  • Contributions: Welcome community contributions
  • Support: Active community support

📞 Support​

  • GitHub Issues: Report bugs and request features
  • Discussions: Community support and discussions
  • Documentation: Comprehensive guides and tutorials

Built with ❤️ for the AI community