AgentSquare: Automatic LLM Agent Search In Modular Design Space

🌐 Website | 📃 Paper |

AgentSquare

The official implementation for paper AgentSquare: Automatic LLM Agent Search in Modular Design Space with code, prompts and results.

Setup

1. Set up OpenAI API key and store in environment variable OPENAI_API_KEY (see here).

2. Install dependencies

pip install -r requirements.txt

3. Install alfworld environment following instructions here, and webshop environment here

Quick Start

An exemplar script combining different agent modules to solve the task of ALFworld:

cd alfworld
sh run.sh or 
python3 alfworld_run.py \
    --planning deps\
    --reasoning cot\
    --tooluse none\
    --memory dilu\
    --model gpt-3.5-turbo-0125 \

Contribute to AgentSquare

We kindly invite you to contribute to AgentSquare by standardizing your own LLM agents with our proposed I/O interfaces. Let's work together to offer a platform for fully exploiting the potential of successful agent designs and consolidating the collective efforts of LLM agent research community!

Design New Modules

For guidance on standardizing the I/O interfaces of the four types of agent modules, please refer to alfworld/reasoning(planning)_modules.py, which provides a module template and some existing modules, along with a complete interface description available in standard_module_interface.docx. You can submit your standardized modules through this link. The .py file format is preferred, examples can be seen in the module pool folder. We will check your submission timely, once approved we will cite and acknowledge your works in this repository.

How to Add A New Task

You can refer to the workflow.py to integrate it with your encapsulated tasks, just like in alfworld.

Citations

Please considering citing our paper and staring this repo if you use AgentSquare and find it useful, thanks! Feel free to contact [email protected] or open an issue if you have any question.

@misc{shang2024agentsquare,
      title={AgentSquare: Automatic LLM Agent Search in Modular Design Space}, 
      author={Yu Shang and Yu Li and Keyu Zhao and Likai Ma and Jiahe Liu and Fengli Xu and Yong Li},
      year={2024},
      eprint={2410.06153},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
}