这是indexloc提供的服务,不要输入任何密码
Skip to content

daniel-007/droidrun

 
 

Repository files navigation

GitHub stars Discord Documentation Twitter Follow

DroidRun is a powerful framework for controlling Android devices through LLM agents. It allows you to automate Android device interactions using natural language commands.

✨ Features

  • Control Android devices with natural language commands
  • Supports multiple LLM providers (OpenAI, Anthropic, Gemini, Ollama, DeepSeek)
  • Planning capabilities for complex multi-step tasks
  • LlamaIndex integration for flexible LLM interactions
  • Easy to use CLI with enhanced debugging features
  • Extendable Python API for custom automations
  • Screenshot analysis for visual understanding of the device
  • Execution tracing with Arize Phoenix

📦 Installation

🚀 Option 1: Install from PyPI (Recommended)

pip install droidrun

🔧 Option 2: Install from Source

git clone https://github.com/droidrun/droidrun.git
cd droidrun
pip install -e .

📋 Prerequisites

  1. An Android device connected via USB or ADB over TCP/IP
  2. ADB (Android Debug Bridge) installed and configured
  3. DroidRun Portal app installed on your Android device
  4. API key for at least one of the supported LLM providers:
    • OpenAI
    • Anthropic
    • Google Gemini

🔧 Setting up ADB

ADB (Android Debug Bridge) is required for DroidRun to communicate with your Android device:

  1. Install ADB:

    • Windows: Download Android SDK Platform Tools and extract the ZIP file
    • macOS: brew install android-platform-tools
    • Linux: sudo apt install adb (Ubuntu/Debian) or sudo pacman -S android-tools (Arch)
  2. Add ADB to your PATH:

    • Windows: Add the path to the extracted platform-tools folder to your system's PATH environment variable
    • macOS/Linux: Add the following to your ~/.bashrc or ~/.zshrc:
      export PATH=$PATH:/path/to/platform-tools
  3. Verify ADB installation:

    adb version
  4. Enable USB debugging on your Android device:

    • Go to Settings → About phone
    • Tap Build number 7 times to enable Developer options
    • Go to Settings → System → Developer options (location may vary by device)
    • Enable USB debugging

🛠️ Setup

📱 1. Install DroidRun Portal App

DroidRun requires the DroidRun Portal app to be installed on your Android device:

  1. Download the DroidRun Portal APK from the DroidRun Portal repository
  2. Use DroidRun to install the portal app:
    droidrun setup --path=/path/to/droidrun-portal.apk

Alternatively, you can use ADB to install it manually:

adb install -r /path/to/droidrun-portal.apk

🔑 2. Set up API keys

Create a .env file in your working directory or set environment variables:

# Choose at least one of these based on your preferred provider
export OPENAI_API_KEY="your_openai_api_key_here"
export ANTHROPIC_API_KEY="your_anthropic_api_key_here"
export GEMINI_API_KEY="your_gemini_api_key_here"
export DEEPSEEK_API_KEY="your_deepseek_api_key_here"
# For Ollama, no API key is needed

To load the environment variables from the .env file:

source .env

📱 3. Connect to an Android device

Connect your device via USB or set up wireless ADB:

# List connected devices
droidrun devices

# Connect to a device over Wi-Fi
droidrun connect 192.168.1.100

💻 Using the CLI

DroidRun's CLI is designed to be simple and intuitive. You can use it in two ways:

🚀 Basic Usage

# Format: droidrun "task description" [options]
droidrun "Open the settings app"

🔌 With Provider Options

# Using OpenAI
droidrun "Open the calculator app" --provider OpenAI --model gpt-4o-mini

# Using Anthropic
droidrun "Check the battery level" --provider Anthropic --model claude-3-sonnet-20240229

# Using Gemini
droidrun "Install and open Instagram" --provider Gemini --model models/gemini-2.5-pro-preview-05-06

# Using Ollama (local)
droidrun "Check battery level" --provider Ollama --model llama2

⚙️ Additional Options

# Specify a particular device
droidrun "Open Chrome and search for weather" --device abc123

# Enable vision capabilities
droidrun "Analyze what's on the screen" --vision

# Enable planning for complex tasks
droidrun "Find and download a specific app" --reasoning

# Enable execution tracing (requires Phoenix server running)
droidrun "Debug this complex workflow" --tracing

# Set maximum number of steps
droidrun "Open settings and enable dark mode" --steps 20

📝 Creating a Minimal Test Script

If you want to use DroidRun in your Python code rather than via the CLI, you can create a minimal test script:

#!/usr/bin/env python3
import asyncio
from droidrun.agent.droid import DroidAgent
from droidrun.agent.utils.llm_picker import load_llm
from droidrun.tools import load_tools

async def main():
    # Load tools
    tool_list, tools_instance = await load_tools()
    
    # Load LLM
    llm = load_llm(
        provider_name="Gemini",  # Case sensitive: OpenAI, Ollama, Anthropic, Gemini, DeepSeek
        model="models/gemini-2.5-pro-preview-05-06",
        temperature=0.2
    )
    
    # Create and run the agent
    agent = DroidAgent(
        goal="Open the Settings app and check the Android version",
        llm=llm,
        tools_instance=tools_instance,
        tool_list=tool_list,
        vision=True,      # Enable vision for screen analysis
        reasoning=True    # Enable planning for complex tasks
    )
    
    # Run the agent
    result = await agent.run()
    print(f"Success: {result['success']}")
    if result.get('reason'):
        print(f"Reason: {result['reason']}")

if __name__ == "__main__":
    asyncio.run(main())

You can also use LlamaIndex directly:

import asyncio
from llama_index.llms.gemini import Gemini
from droidrun.agent.droid import DroidAgent
from droidrun.tools import load_tools

async def main():
    # Load tools
    tool_list, tools_instance = await load_tools()
    
    # Create LlamaIndex LLM directly
    llm = Gemini(
        model="models/gemini-2.5-pro-preview-05-06",
        temperature=0.2
    )
    
    # Create and run the agent
    agent = DroidAgent(
        goal="Open the Settings app and check the Android version",
        llm=llm,
        tools_instance=tools_instance,
        tool_list=tool_list
    )
    
    # Run the agent
    result = await agent.run()
    print(f"Success: {result['success']}")

if __name__ == "__main__":
    asyncio.run(main())

❓ Troubleshooting

🔑 API Key Issues

If you encounter errors about missing API keys, ensure:

  1. You've set the correct environment variable for your chosen provider
  2. The API key is valid and has appropriate permissions
  3. You've correctly sourced your .env file or exported the variables manually

📱 Device Connection Issues

If you have trouble connecting to your device:

  1. Ensure USB debugging is enabled on your Android device
  2. Check that your device is recognized by ADB: adb devices
  3. For wireless connections, make sure your device and computer are on the same network

🤖 LLM Provider Selection

If DroidRun is using the wrong LLM provider:

  1. Explicitly specify the provider with --provider (in CLI) or llm_provider= (in code)
  2. When using Gemini, ensure you have set GEMINI_API_KEY and specified --provider gemini

📊 Tracing Issues

If you're using the tracing feature:

  1. Make sure to install Arize Phoenix: pip install "arize-phoenix[llama-index]"
  2. Start the Phoenix server before running your command: phoenix serve
  3. Access the tracing UI at http://localhost:6006 after execution

🎬 Demo Videos

  1. Shopping Assistant: Watch how DroidRun searches Amazon for headphones and sends the top 3 products to a colleague on WhatsApp.

    Prompt: "Go to Amazon, search for headphones and write the top 3 products to my colleague on WhatsApp."

    Shopping Assistant Demo

  2. Social Media Automation: See DroidRun open X (Twitter) and post "Hello World".

    Prompt: "Open up X and post Hello World."

    Social Media Automation Demo

💡 Example Use Cases

  • Automated UI testing of Android applications
  • Creating guided workflows for non-technical users
  • Automating repetitive tasks on Android devices
  • Remote assistance for less technical users
  • Exploring Android UI with natural language commands

🗺️ Roadmap

🤖 Agent:

  • Improve memory: Enhance context retention for complex multi-step tasks
  • Expand planning capabilities: Add support for more complex reasoning strategies
  • Add Integrations: Support more LLM providers and agent frameworks (LangChain, Agno etc.)

⚙️ Automations:

  • Create Automation Scripts: Generate reusable scripts from agent actions that can be scheduled or shared

☁️ Cloud:

  • Hosted version: Remote device control via web interface without local setup
  • Add-Ons: Marketplace for extensions serving specific use cases
  • Proxy Hours: Cloud compute time with tiered pricing for running automations
  • Droidrun AppStore: Simple installation of Apps on your hosted devices

👥 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

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

Security Checks

To ensure the security of the codebase, we have integrated security checks using bandit and safety. These tools help identify potential security issues in the code and dependencies.

Running Security Checks

Before submitting any code, please run the following security checks:

  1. Bandit: A tool to find common security issues in Python code.

    bandit -r droidrun
  2. Safety: A tool to check your installed dependencies for known security vulnerabilities.

    safety check

You can also run both checks using the provided script:

python -m droidrun.tools.security_check

About

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 98.7%
  • Shell 1.3%