OMG-Agent: A Natural Language-Based Android Automation Assistant
OMG-Agent is an open-source automation tool designed specifically for the Android platform. It combines natural language commands with a GUI model, allowing users to directly control their phones with simple text descriptions (such as "Open WeChat and send a message to Zhang San"), greatly lowering the barrier to mobile automation.
Core Functions and Technical Highlights
- Natural Language DrivenNo need to write complex scripts; you can drive your phone to complete specific tasks through conversational commands.
- A robust model ecosystemIt has built-in support for mainstream mobile GUI large models such as AutoGLM and GELab-Zero, and is compatible with the OpenAI interface.
- Real-time device interactionIt enables efficient screenshot acquisition and operation execution based on the ADB interface, supporting both real devices and emulators.
- Flexible deployment and interfaceSupports cross-platform deployment and provides a bilingual (Chinese and English) interface with light and dark theme switching.
- Open source and scalableThe project is completely open source, and developers can perform secondary development according to specific business scenarios.
Quick Start Guide
1. Environment Preparation (Installing ADB)
Choose the appropriate command to install the ADB environment based on your operating system:
- Windows:
scoop install adb - macOS:
brew install android-platform-tools - Linux:
apt install adb
2. Project Deployment
# Clone the repository: git clone https://github.com/safphere/OMG-Agent.git cd OMG-Agent # Install dependencies and start the application: pip install -r requirements.txt python run.py3. Equipment Configuration
- MobileEnable "Developer options" and enable "USB debugging".
- Input method:Install ADBKeyboard To ensure that the text input is correct.
- connectConnect your phone via USB cable and complete device authorization.
4. Operating Procedures
After starting the program, execute the following commands in sequence:Refresh device $rightarrow$ Start casting $rightarrow$ Input natural language commands $rightarrow$ Click to execute。
Comparison of built-in GUI models
| Model Name | source | Core features |
|---|---|---|
| AutoGLM-Phone-9B | Zhipu AI | Deeply optimized for mobile GUI operation, ensuring precise command execution. |
| GELab-Zero-4B-preview | Leaping Stars | Lightweight design, suitable for general mobile agent tasks. |
Applicable Scenarios
- Geek Player: Try using AI to remotely take over the phone and achieve personalized automated processes.
- Technology developersResearch UI automation testing or explore Agent technology.
- R&D team: To conduct rapid prototyping and functional testing of AI Agent products.
- Efficiency ExpertAutomation workers who need multiple devices to work together to handle repetitive tasks.
Resource Acquisition
GitHub repository: safphere/OMG-Agent
Backup download: Quark Cloud Drive Download
End of text
