What is Agent-E
Agent-E is an intelligent automation system built on the AutoGen agent framework, focused on automation operations within the browser. Agent-E is based on natural language interaction and can perform a variety of complex tasks such as filling out forms, searching and sorting e-commerce products, locating web page content, managing web media playback, conducting deep web searches, automating project management tasks, and providing personal shopping assistance. Agent-E can improve online efficiency, reduce repetitive work, and let users focus more on important matters.
Agent-E's main features
- Form filling: Automatically fill out web forms, including entering personal information.
- E-commerce search and sorting: Search and sort products on e-commerce websites such as Amazon based on criteria such as sales volume or price.
- Content locating: Find specific content on websites, such as sports scores or university contact information.
- Media interaction: Interact with web-based media, including playing YouTube videos and managing playback settings.
- Web search: Perform comprehensive web searches to gather information on various topics.
- Project management automation: Filter issues and automate workflows on project management platforms such as JIRA.
Agent-E's technical principles
- Agent-based architecture: Based on the AutoGen agent framework, it uses agents (such as user agents and browser navigation agents) to perform tasks.
- Skill library: Core functionality is built around a skill library, which contains a series of predefined actions (skills) divided into perception skills and action skills.
- Natural language interaction: Supports allowing users to interact with the browser in natural language, making task execution more intuitive.
- DOM distillation: Based on DOM distillation technology, Agent-E simplifies the HTML DOM into relevant JSON snapshots, focusing on elements relevant to the user's task.
- Change observation: After executing actions, Agent-E monitors state changes and provides them to the LLM in the form of language feedback, guiding more accurate performance.
- Hierarchical planning: Adopts hierarchical planning, decomposing complex tasks into subtasks handled by agents at different levels.
Agent-E's project address
- GitHub repository：https://github.com/EmergenceAI/Agent-E
- arXiv technical paper：https://arxiv.org/pdf/2407.13032
Agent-E's application scenarios
- Online shopping: Automatically search for products, sort results, and add them to the shopping cart, simplifying the shopping process.
- Information retrieval: Quickly gather specific information from the internet, such as news or academic materials, improving research efficiency.
- Form automation: Automatically fill out and submit online forms, reducing manual input and saving time.
- Personal assistant: Provide customized suggestions based on personal preferences, such as restaurant recommendations, improving quality of life.
- Media playback management: Automatically play and control music and video content, enhancing the entertainment experience.