- [2024.01.15] We release Easytool for easier tool usage.
- The code and datasets are available at easytool.
- The paper is available at EASYTOOL: Enhancing LLM-based Agents with Concise Tool Instruction.
LLM-based agents usually employ tool documentation to grasp the selection and usage of tools from different sources, but these documentations could be inconsistent in formats, redundant with excessive length, and lacking demonstrations for instructions.
EasyTool is an easy but effective method to create clear, structured, and unified instructions from tool documentations for improving LLM-based agents in using tools.
- Prepare requirements:
pip install -r requirements.txt - Data Construction:
python3 data_process.py
Before running any of the commands, ensure that you have set the necessary API keys. Replace "" with your actual keys.
export OPENAI_API_KEY="your_openai_api_key_here"
export RAPIDAPI_KEY="your_rapidapi_key_here"You need first get the tool execution code (./data/toolenv/tools.) from the following link: Google Drive or Tsinghua Cloud and then save them to ./toolenv/tools To inference with LLMs, run the following commands:
unzip data_toolbench/tool_instruction/API_description_embeddings.zip -d data_toolbench/tool_instruction/
export OPENAI_API_KEY=""
export RAPIDAPI_KEY=""
python3 main.py \
--model_name gpt-3.5-turbo \
--task toolbench \
--data_type G2 \
--tool_root_dir ./toolenv/tools
python3 main.py \
--model_name gpt-3.5-turbo \
--task toolbench \
--data_type G3 \
--tool_root_dir ./toolenv/tools
python3 main.py \
--model_name gpt-3.5-turbo \
--task toolbench_retrieve \
--data_type G2 \
--tool_root_dir ./toolenv/tools
python3 main.py \
--model_name gpt-3.5-turbo \
--task toolbench_retrieve \
--data_type G3 \
--tool_root_dir ./toolenv/toolsTo inference with LLMs, run the following commands:
export OPENAI_API_KEY=""
python3 main.py \
--model_name gpt-3.5-turbo \
--task funcqa \
--data_type funcqa_mh
python3 main.py \
--model_name gpt-3.5-turbo \
--task funcqa \
--data_type funcqa_ohTo inference with LLMs, run the following commands:
export OPENAI_API_KEY=""
python3 main.py \
--model_name gpt-3.5-turbo \
--task restbench If you find this work useful in your method, you can cite the paper as below:
@article{yuan2024easytool,
title = {EASYTOOL: Enhancing LLM-based Agents with Concise Tool Instruction},
author = {Siyu Yuan and Kaitao Song and Jiangjie Chen and Xu Tan and Yongliang Shen and Ren Kan and Dongsheng Li and Deqing Yang},
journal = {arXiv preprint arXiv:2401.06201},
year = {2024}
}