Rabbit is an advanced, modular framework for creating fully autonomous agents capable of controlling a web browser to execute a range of tasks efficiently. This innovative solution harnesses large language models (LLMs) combined with custom tools to facilitate intelligent workflows such as research, information extraction, and automated browser tasks that often involve complex, multi-step processes.

Features

• Agent Loop Execution: Utilize agent_task_loop.py for continual task management.
• Headless Browser Control: Engage with a headless browser through customizable tools for automated operations.
• LLM Integration: Incorporate memory and planning capabilities for enhanced decision-making.
• Extensible SDK: Access a modular SDK located in the rabbit_sdk/ directory, allowing for easy customization and extension.
• Testing Support: Utilize unit and workflow testing to ensure reliability and performance.
• Example Workflows: Explore practical examples that demonstrate real-world browser automation challenges.

Example Workflows

Sentiment Analysis of AI Safety

Run the following command to execute a basic browser task:

cd Rabbit/examples
python3 simple_browser_task.py

This example will perform the following steps:

1. Open various URLs regarding AI and safety.
2. Scrape pertinent content from these pages.
3. Conduct sentiment analysis on the scraped data.
4. Summarize the findings concisely.

Comprehensive Crypto Analysis

To execute a more complex automation workflow, use:

cd Rabbit/examples
python3 complex_workflow.py

This advanced workflow includes:

1. Opening multiple URLs focused on cryptocurrency assets.
2. Scraping relevant data.
3. Running sentiment analyses on the content.
4. Summarizing key findings and generating trading insights.

Development and Testing

To run the testing suite, execute:

python test_agent.py

Future Enhancements

Future plans for Rabbit include adding support for OpenAI and Claude, extending the toolset for data transformation tasks, integrating with a vector database for persistent memory, and developing a web UI for visualizing agent reasoning.