Agentic Smart Home

Agentic Smart Home

An intelligent smart home controller built with Flutter, powered by an on-device Gemma 2B LLM. This application demonstrates the power of agentic AI in mobile apps, allowing users to control their smart home environment through natural language voice commands fully offline, ensuring privacy and speed without needing an internet connection.

📱 Demo

RPReplay_Final1763722328.mov

✨ Features

• On-Device Intelligence: Uses Google's Gemma 2B model running locally on the device via flutter_gemma for privacy and offline capability.
• Agentic UI: Built with flutter_ui_agent to demonstrate how to easily integrate agentic capabilities into Flutter apps.
• Natural Language Control: Speak naturally to control your home (e.g., "Turn on the living room lights and set the color to blue").
• Multi-Step Actions: The agent understands and executes complex, multi-part commands in a single request.
• Smart Navigation: Automatically navigates to the relevant room page when a command is issued for a device in another location.
• Voice Feedback: Visual feedback for speech recognition and agent processing states.
• Room Management: Controls for lights, ambient color, thermostats, TVs, and garage gates across multiple rooms (Living Room, Bedroom, Kitchen, Garage).

🛠️ Tech Stack

• Framework: Flutter
• LLM: flutter_gemma
  • Enables offline, on-device inference using Google's Gemma 2B model. No internet connection required for intelligence.
• UI Components: flutter_ui_agent
  • Provides the core agentic UI elements, including the chat interface and action widgets.
• Speech Recognition: speech_to_text
  • Handles offline voice-to-text conversion, allowing for fast and private command input.
• Navigation: synced_page_views
  • Manages the synchronized room navigation, ensuring the UI updates smoothly when the agent switches rooms.
• State Management: ListenableBuilder & ChangeNotifier (Provider-free architecture)

🚀 Getting Started

Prerequisites

• Flutter SDK installed
• iOS device or Android device with GPU support.
• Gemma Model: You need the gemma-2b-it-gpu-int4.bin model file.

Installation

1. Clone the repository:

   git clone https://github.com/MAUstaoglu/agentic_smart_home.git
   cd agentic_smart_home

2. Download the Model:

   • Download the gemma-2b-it-gpu-int4.bin model
   • Place the file in assets/models/gemma-2b-it-gpu-int4.bin.

3. Install Dependencies:

   flutter pub get

4. Run the App:

   flutter run --profile # Recommended for better performance with LLMs

💡 Usage

1. Grant Permissions: Allow microphone access when prompted.
2. Speak: Tap the microphone icon and say a command.
   • Example: "Turn on the kitchen light."
   • Example: "Set the living room temperature to 24 degrees."
   • Example: "Go to the garage and open the gate."
   • Example: "Turn off all lights in the bedroom."
3. Watch the Magic: The agent will process your request, navigate if necessary, and execute the actions.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.