🤖 MML (Motion Mind Local) - v1.1.0 (Stable)

MML (Motion Mind Local) is a fully offline and private local AI chat application. It runs .gguf format Large Language Models (LLMs) directly on your device's processor and memory. No internet or cloud server is required.

---

✨ What's New (v1.1.0 Updates)

• 💾 Persistent Chat History: Chat history is now automatically saved. Your previous conversations remain even after closing and reopening the app.
• 🔄 Auto-Loading Engine: The previously loaded model automatically loads when the app opens.
• 🧹 Advanced Memory Management: Improved logic to perfectly clear old memory (RAM) and cache files before loading a new model.
• 🎨 New Branding: High-resolution new logo and optimized user interface.
• 📖 Interactive Guide: Detailed in-app guide for model downloading and settings.

---

🚀 Core Features

• Fully Offline: Chat with AI without any internet connection.
• Privacy First: Your data and messages never leave your device.
• Custom Model Support: Use any GGUF model (Llama, Phi, Gemma, Qwen) based on your device's RAM.
• Real-time Streaming: Token-by-token responses for a live typing experience.
• Smart Context: AI remembers previous messages and responds intelligently (Sliding Window Context).

---

🛠 Technical Architecture

• Framework: Expo SDK 54 (New Architecture Enabled)
• Engine: llama.rn (Native binding for llama.cpp)
• Storage: expo-file-system (Model caching & chat history management)
• UI: React Native with Custom Hooks (useLlamaManager, useChatLayout)

---

📖 Model Setup Guide

Follow these steps for the best performance:

1. Download: Download a .gguf model from Hugging Face.
2. Format: Always try to use Q4_K_M quantization (best balance between speed and intelligence).
3. Recommendations:
   • 4GB RAM: Llama-1B or Qwen-1.5B
   • 6GB+ RAM: Gemma-2B or Phi-3-Mini
   • 12GB+ RAM: Llama-3-8B

---

🏗 Logic Flow

• Model Lifecycle: On app launch, it checks the models/ folder and auto-loads the model using initLlama.
• Memory Safety: Uses context.release() to manage device RAM and prevent crashes with large models.
• Persistence: Last 20 messages are saved in chat_history.json file.

---

👤 Developer

MD. Saiful Alom Siam
MML - Bringing AI to your pocket, offline and private.

---

📄 License

This project is created for personal and educational purposes.