Smart Home Automation Agent

A LangGraph-powered intelligent home automation system that uses OpenAI's language models and the BigTool framework to control various smart home devices through natural language interactions.

🏠 What This Project Does

This project demonstrates an AI agent that can understand natural language commands and execute them on a comprehensive smart home system. The agent can control:

• Smart Lights: Turn on/off, dim, change colors, control individual rooms or all lights
• Thermostat: Set temperature, change modes (heating/cooling/auto), schedule temperature changes
• Security System: Lock/unlock doors, check door status, manage security cameras, motion detection
• Plant Care: Water plants, check soil moisture, set watering schedules
• Additional Smart Devices: Window blinds, smart vacuum, music system, alarms, air quality monitoring, smart fans

🔧 What is BigTool and Why Use It?

LangGraph BigTool is a framework that enables AI agents to work with large numbers of tools efficiently. Instead of loading all tools into the agent's context at once (which would exceed token limits), BigTool:

1. Creates a searchable tool registry using vector embeddings
2. Dynamically retrieves relevant tools based on the user's query
3. Scales to hundreds or thousands of tools without context overflow
4. Improves response accuracy by providing only the most relevant tools for each task

In our smart home system, we have 25+ different tools. BigTool allows the agent to:

• Find the right tool for "turn off kitchen lights" (retrieves turn_off_light)
• Handle complex queries like "I'm cold" (retrieves thermostat and heating tools)
• Scale to even more devices without performance degradation

🚀 Setup Instructions

Prerequisites

1. Install uv (Python package manager):

   curl -LsSf https://astral.sh/uv/install.sh | sh

2. Set up environment variables: Create a .env file in the project root:

   # Required for OpenAI integration
   OPENAI_API_KEY=your_openai_api_key_here

Installation

1. Clone the repository:

   git clone <repository-url>
   cd agent-big-tool

2. Install dependencies using uv:

   uv sync

🏃‍♂️ Running the Project

Execute the smart home automation demo:

uv run python graph.py

This will run through several example queries:

• "Turn off the lights in the kitchen"
• "Set the thermostat to 70 degrees"
• "Lock the front door"
• "What's happening in the living room?"
• "It's a little dirty, can you vacuum the living room?"
• "I'm cold, can you turn on the heat?"

🏗️ Agent Architecture

The system uses LangGraph to create a sophisticated agent workflow:

graph TD
    A[User Query] --> B[System Message]
    B --> C[LLM Processing]
    C --> D[Tool Retrieval]
    D --> E{Tools Found?}
    E -->|Yes| F[Execute Relevant Tools]
    E -->|No| G[Request Clarification]
    F --> H[Tool Results]
    H --> I[Generate Response]
    I --> J[Final Answer]
    G --> J

    subgraph "BigTool Framework"
    D --> K[Vector Search]
    K --> L[Tool Registry]
    L --> M[Embedding Match]
    M --> N[Return Best Tools]
    N --> F
    end

    subgraph "Smart Home Tools"
    O[Lighting Control]
    P[Thermostat Control]
    Q[Security System]
    R[Plant Care]
    S[Additional Devices]
    end

    F --> O
    F --> P
    F --> Q
    F --> R
    F --> S

Loading

🛠️ Key Components

1. Agent Configuration (graph.py)

• LLM: Uses OpenAI's GPT-4o for natural language understanding
• Tool Registry: Maps unique IDs to tool instances
• Vector Store: InMemoryStore with OpenAI embeddings for tool retrieval
• Graph Compilation: Creates the executable agent workflow

2. Tool Collection (tools.py)

• 25+ Smart Home Tools: Comprehensive coverage of home automation
• Langchain Tool Format: Properly decorated with descriptions for LLM understanding
• Realistic Responses: Simulated but realistic tool responses

3. BigTool Integration

• Semantic Tool Search: Finds relevant tools based on query similarity
• Efficient Tool Loading: Only loads tools needed for specific queries
• Scalable Architecture: Can handle hundreds of tools without performance issues

📦 Dependencies

• langchain: Core framework for LLM applications
• langchain-openai: OpenAI integration for chat and embeddings
• langgraph: Workflow orchestration framework
• langgraph-bigtool: BigTool framework for large-scale tool management
• python-dotenv: Environment variable management

🎯 Example Interactions

The agent can handle various types of natural language commands:

• Direct Commands: "Turn off the kitchen lights"
• Contextual Requests: "I'm cold" → Adjusts thermostat/heating
• Complex Queries: "Secure the house for the night" → Locks doors, checks cameras
• Maintenance Tasks: "The living room needs cleaning" → Starts vacuum

🔮 Extending the System

To add new smart home devices:

1. Create new tools in tools.py following the @tool decorator pattern
2. Add tools to HOME_AUTOMATION_TOOLS list
3. The BigTool framework will automatically make them searchable
4. Test with natural language queries that would use the new tools

📄 License

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