English | 中文

A Python library for AI-powered UI automation testing. This library provides command-line tools for automated UI testing using computer vision and AI models.

• Android device automation via ADB
• AI-powered UI element recognition and interaction
• Command-line interface for test execution
• Support for multiple AI models (Doubao, Ark, etc.)
• Comprehensive screenshot and interaction capabilities
• Internationalization Support - Auto-detects system language with Chinese/English interface
• Professional Localization - Uses gettext for industry-standard internationalization

Using uv (recommended):

uv add ui-zero

Or using pip:

pip install ui-zero

# Clone the repository
git clone https://github.com/Roland0511/ui-zero.git
cd ui-zero

# Install development dependencies with uv
uv sync --dev

# Or using pip
pip install -e ".[dev]"

This library uses ByteDance's UI-TARS model for AI-powered UI recognition. Follow the deployment guide to deploy the model and obtain an API key.

For Chinese users, see UI-TARS 模型部署教程.

After deployment, set the following environment variables in your system:

# Set the API key in your environment
export OPENAI_API_KEY="your_api_key_here"
export OPENAI_BASE_URL="your_api_base_url_here"

• Python 3.10+
• Android SDK with ADB installed
• Connected Android device with USB debugging enabled

• Android 8.0+ (API level 26+)
• USB debugging enabled
• ADBKeyboard installed for text input support

After installation, you can use the uiz command:

# Run with test case file
uiz --testcase test_case.json

# Run with direct commands
uiz --command "find and click settings icon"

# List available devices
uiz --list-devices

# Use specific device (recommended when multiple devices are connected)
uiz --device DEVICE_ID --command "find and click settings icon"

# Enable debug mode
uiz --debug --command "search for Wi-Fi settings, then click"

# Disable history feature
uiz --no-history --command "open app"

# Show help information
uiz --help

from ui_zero import AndroidAgent, ADBTool

# Initialize ADB and agent
adb_tool = ADBTool()
agent = AndroidAgent(adb_tool)

# Run a test step
result = agent.run("find search bar and type 'hello world'")

# Execute with callbacks
def on_screenshot(img_bytes):
    print(f"Screenshot size: {len(img_bytes)} bytes")

def on_action(prompt, action):
    print(f"Executing action: {action.action}")

result = agent.run(
    "click settings button",
    screenshot_callback=on_screenshot,
    preaction_callback=on_action
)

UI-Zero supports automatic localization based on system language:

The system automatically detects the following environment variables and sets the appropriate language:

• LANG
• LC_ALL
• LC_MESSAGES

from ui_zero.localization import set_language

# Set to Chinese
set_language('zh_CN')

# Set to English
set_language('en_US')

• Chinese (Simplified): zh_CN
• English (US): en_US

# Code formatting
uv run black ui_zero/

# Import sorting
uv run isort ui_zero/

# Type checking
uv run mypy ui_zero/

# Linting
uv run flake8 ui_zero/

# Run all tests
uv run pytest

# Run tests with coverage
uv run pytest --cov=ui_zero

# Run localization tests
uv run pytest tests/test_localization.py -v

Compile translation files:

# Compile all .po files to .mo files
uv run python scripts/compile_translations.py

Manual localization testing:

# Run localization manual tests
uv run python tests/test_localization.py --manual

ui_zero/
├── __init__.py           # Package initialization
├── cli.py               # Command line interface
├── agent.py             # Android automation agent
├── adb.py               # ADB tool class
├── localization.py      # Internationalization manager
├── locale/              # Translation files directory
│   ├── messages.pot     # Translation template
│   ├── zh_CN/LC_MESSAGES/
│   │   ├── ui_zero.po   # Chinese translation source
│   │   └── ui_zero.mo   # Chinese binary file
│   └── en_US/LC_MESSAGES/
│       ├── ui_zero.po   # English translation source
│       └── ui_zero.mo   # English binary file
└── models/              # AI model integrations
    ├── __init__.py
    ├── arkmodel.py      # Ark model base class
    └── doubao_ui_tars.py # Doubao UI-TARS model

Create a test_case.json file:

[
    "find and click settings icon",
    "scroll to bottom",
    "click about phone",
    "go back"
]

Create a test_case.yaml file for more advanced test scenarios:

android:
  # Device ID, optional, defaults to first connected device
  deviceId: <device-id>
tasks:
  - name: <task-name>
    continueOnError: <boolean> # Optional, whether to continue on error, defaults to false
    flow:
      # Execute an interaction, `ai` is shorthand for `aiAction`
      - ai: <prompt>

      # This usage is the same as `ai`
      - aiAction: <prompt>

      # Wait for a condition to be met, with timeout (ms, optional, defaults to 30000)
      - aiWaitFor: <prompt>
        timeout: <ms>

      # Execute an assertion
      - aiAssert: <prompt>
        errorMessage: <error-message> # Optional, error message to print when assertion fails

      # Wait for a certain time (milliseconds)
      - sleep: <ms>

  - name: <another-task-name>
    flow:
      # ...

For complete examples, see test_case.example.yaml in the project root.

1. ADB command not available

   # Ensure Android SDK is installed and added to PATH
   export PATH=$PATH:$ANDROID_HOME/platform-tools

2. Device connection issues

   # Check device connection
   adb devices
   
   # Restart ADB service
   adb kill-server && adb start-server

3. Multiple devices detected

   • When multiple devices are connected, the tool automatically selects the first device
   • Use --device DEVICE_ID to specify a particular device to avoid ambiguity
   • Use --list-devices to see all available devices

4. API key issues

   # Ensure environment variable is set
   echo $ARK_API_KEY

5. Text input issues

   • Ensure ADBKeyboard app is installed on device
   • Enable ADBKeyboard as input method in device settings

Contributions are welcome! Please see the Contributing Guide for details.

MIT License - see the LICENSE file for details.

• Issue Reports: GitHub Issues
• Documentation: Project Documentation
• Repository: GitHub