OpenAGI is used as the agent creation package to build agents for AIOS.

From PyPI

pip install pyopenagi

Locally

git clone https://agiresearch/OpenAGI
cd OpenAGI
pip install -e .

To add a new agent, first you need to create a folder under the pyopenagi/agents folder. The folder needs to be the following structure:

- pyopenagi/agents
  - author
    - agent_name
      - agent.py # main code for the agent execution logic
      - config.json # set up configurations for agent
      - meta_requirements.txt # dependencies that the agent needs

If you want to use external tools provided by openagi in your agents, you can follow instructions of setting up tools in How to setup external tools. If you want to add new tools for your developing agent, you need to add a new tool file in the folder.

• Offline tools: locally-deployed models like stable-diffusion model
• Online tools: APIs like rapid api.

If you have developed and tested your agent, and you would like to share your agents, you can use the following to upload your agents

python pyopenagi/agents/interact.py --mode upload --agent <author_name/agent_name>

💡Note that the agent param must exactly match the folder you put your agent locally.

If you want to look at implementations of other agents that people developed, you can use the following command:

python pyopenagi/agents/interact.py --mode download --agent <author_name/agent_name>

For detailed information on how to contribute, see CONTRIBUTE. If you would like to contribute to the codebase, issues or pull requests are always welcome!

Please check out our implementation for our research paper OpenAGI: When LLM Meets Domain Experts.

@article{openagi,
  title={OpenAGI: When LLM Meets Domain Experts},
  author={Ge, Yingqiang and Hua, Wenyue and Mei, Kai and Ji, Jianchao and Tan, Juntao and Xu, Shuyuan and Li, Zelong and Zhang, Yongfeng},
  journal={In Advances in Neural Information Processing Systems (NeurIPS)},
  year={2023}
}