🤖 AgentVerse 🪐

A Framework for Multi-LLM Environment Simulation

AgentVerse offers a versatile framework that streamlines the process of creating custom multi-agent environments for large language models (LLMs). Designed to facilitate swift development and customization with minimal effort, our framework empowers researchers to concentrate on their research, rather than being bogged down by implementation details.

---

Contents

• ✨ Features
• 📰 What's New
• 🗓 Coming Soon
• 👾 Simple Demo Video
  • NLP Classroom
  • Prisoner Dilemma
  • Software Design
  • Database Administrator (DBA)
  • Pokemon
• Contents
• 🚀 Getting Started
  • Installation
  • CLI Example
  • Local Website Demo
• 💡 Philosophy
  • Environment
  • Agent
• ✍️ Customize Your Own Environment
  • A Simple Example: Building a Classroom Environment
    • 1. Creating a Task Directory and Configuring the Environment
    • 2. Configuring the Agents
    • 3. Writing an Output Parser
  • Customization Guide for More Complex Environments
• 🔎 Examples

🚀 Getting Started

Installation

pip install -U agentverse

Or you can install the package by manually cloning the latest repository

git clone https://github.com/OpenBMB/AgentVerse.git --depth 1
cd AgentVerse
pip install -r requirements.txt

Some users have reported problems installing the orjson required by gradio. One simple workaround is to install it with Anaconda conda install -c conda-forge orjson.

You also need to export your OpenAI API key as follows

# Export your OpenAI API key
export OPENAI_API_KEY="your_api_key_here"

AgentVerse-ProblemSolving Example

Consulting

Multi-Agent

python3 pipeline.py --task pipeline_projectv --discussion_mode

export OPENAI_API_KEY=YOUR_API_KEY
export LOG_LEVEL=INFO  choose one of them {DEBUG, INFO, WARNING, ERROR, CRITICAL}
python3 pipeline.py --task pipeline_brainstorming --discussion_mode

Single-Agent

python3 pipeline.py --task pipeline_projectv --single_agent

export OPENAI_API_KEY=YOUR_API_KEY
export LOG_LEVEL=INFO  choose one of them {DEBUG, INFO, WARNING, ERROR, CRITICAL}
python3 pipeline.py --task pipeline_brainstorming --single_agent

With the --discussion_mode (or -d in short) parameter, the pipeline will enter the p-p discussion mode, otherwise enter criticizing mode (which is used in the pipeline_pythoncalculator task).

With the --single_agent (or -s in short) parameter, the pipeline will enter single agent CoT mode.

Software Development

multi-Agent

python3 pipeline.py --task pipeline_projectv

export OPENAI_API_KEY=your api key
export LOG_LEVEL=INFO  choose one of them {DEBUG, INFO, WARNING, ERROR, CRITICAL}
python3 pipeline.py --task pipeline_pythoncalculator

Single-Agent

export OPENAI_API_KEY=your api key
export LOG_LEVEL=INFO  choose one of them {DEBUG, INFO, WARNING, ERROR, CRITICAL}
python3 pipeline.py --task pipeline_pythoncalculator --single_agent

Notice: LOG_LEVEL = DEBUG will output all prompts to the terminal and log file. LOG_LEVEL = INFO will only output the parsed response and the pipeline stage info.

Contact

• Weize Chen: [email protected]
• Yusheng Su: [email protected]