zai-org/Open-AutoGLM
Open AutoGLM
Open-AutoGLM is an autonomous agent framework designed to perform complex user workflows on mobile devices. By translating natural language instructions into precise sequences of taps, scrolls, and text inputs, the system enables the automation of mobile application interactions and testing.
The platform distinguishes itself through a combination of vision-language processing and reinforcement learning. It converts graphical user interfaces into structured data, allowing agents to parse screen elements and map natural language commands to coordinate-based actions. To ensure reliability, the system employs heuristic-based error recovery to navigate around interface interruptions such as pop-ups, advertisements, and network delays.
The framework provides a secure, containerized environment for executing these tasks, which isolates agent processes to protect sensitive data and maintain audit trails. Additionally, it functions as a training platform where agents refine their decision-making policies through repeated reinforcement learning cycles within virtualized mobile environments.
Features
- UI Automation - Translates natural language commands into precise mobile UI interactions for automated testing and workflow execution.
- Large Language Models - Leverages large language models to enable autonomous navigation and multi-step workflow execution.
- Reinforcement Learning Environments - Provides a training platform for refining agent behaviors through reinforcement learning in virtualized mobile interfaces.
- Mobile Testing Frameworks - Automates complex mobile user workflows by translating natural language into precise interface interactions.
- Autonomous Agents - Develops intelligent agents capable of navigating diverse mobile interfaces through reinforcement learning.
- Vision-Language Grounding Models - Maps natural language instructions to spatial coordinates on mobile interfaces using vision-language grounding models.
- Cloud Sandbox Provisioning - Provides a secure, containerized cloud environment for executing and auditing automated mobile tasks.
- Action Sequence Composers - Chains complex user input sequences into executable workflows for mobile device automation.
- Agent Task Execution - Executes automated software tasks within secure, isolated containers with audit logging.
- Model Feedback Loops - Implements feedback loops to iteratively refine agent decision-making policies based on task completion performance.
- Multimodal Models - Agentic multimodal model for automated device interaction.
- Agent Security Runtimes - Enforces security policies and isolation for agent execution environments within cloud containers.
- Interaction Navigators - Navigates through pop-ups and network latency to maintain reliable operation chains in mobile applications.
- Browser Automation Interfaces - Provides interfaces for programmatic interaction with mobile environments to bypass obstacles during automated testing.
- Heuristic Selectors - Uses heuristic-based selectors to navigate dynamic mobile interfaces and recover from unexpected interruptions like pop-ups.
- Interface Representations - Converts raw graphical user interfaces into structured data to maintain context during automated navigation.
- Sandboxed Execution Environments - Provides isolated runtime environments for executing automated agent tasks securely within virtualized containers.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
