MetaGPT: The Multi-Agent Framework

Assign different roles to GPTs to form a collaborative software entity for complex tasks.

1. MetaGPT takes a one line requirement as input and outputs user stories / competitive analysis / requirements / data structures / APIs / documents, etc.
2. Internally, MetaGPT includes product managers / architects / project managers / engineers. It provides the entire process of a software company along with carefully orchestrated SOPs.
   1. Code = SOP(Team) is the core philosophy. We materialize SOP and apply it to teams composed of LLMs.

Software Company Multi-Role Schematic (Gradually Implementing)

Examples (fully generated by GPT-4)

For example, if you type python startup.py "Design a RecSys like Toutiao", you would get many outputs, one of them is data & api design

It requires around $0.2 (GPT-4 api's costs) to generate one example with analysis and design, around $2.0 to a full project.

Installation

Traditional Installation

# Step 1: Ensure that NPM is installed on your system. Then install mermaid-js.
npm --version
sudo npm install -g @mermaid-js/mermaid-cli

# Step 2: Ensure that Python 3.9+ is installed on your system. You can check this by using:
python --version

# Step 3: Clone the repository to your local machine, and install it.
git clone https://github.com/geekan/metagpt
cd metagpt
python setup.py install

Note:

• If already have Chrome, Chromium, or MS Edge installed, you can skip downloading Chromium by setting the environment variable PUPPETEER_SKIP_CHROMIUM_DOWNLOAD to true.

• Some people are having issues installing this tool globally. Installing it locally is an alternative solution,

  npm install @mermaid-js/mermaid-cli

• don't forget to the configuration for mmdc in config.yml

  PUPPETEER_CONFIG: "./config/puppeteer-config.json"
  MMDC: "./node_modules/.bin/mmdc"

Installation by Docker

# Step 1: Download metagpt official image and prepare config.yaml
docker pull metagpt/metagpt:v0.3.1
mkdir -p /opt/metagpt/{config,workspace}
docker run --rm metagpt/metagpt:v0.3.1 cat /app/metagpt/config/config.yaml > /opt/metagpt/config/key.yaml
vim /opt/metagpt/config/key.yaml # Change the config

# Step 2: Run metagpt demo with container
docker run --rm \
    --privileged \
    -v /opt/metagpt/config/key.yaml:/app/metagpt/config/key.yaml \
    -v /opt/metagpt/workspace:/app/metagpt/workspace \
    metagpt/metagpt:v0.3.1 \
    python startup.py "Write a cli snake game"

# You can also start a container and execute commands in it
docker run --name metagpt -d \
    --privileged \
    -v /opt/metagpt/config/key.yaml:/app/metagpt/config/key.yaml \
    -v /opt/metagpt/workspace:/app/metagpt/workspace \
    metagpt/metagpt:v0.3.1

docker exec -it metagpt /bin/bash
$ python startup.py "Write a cli snake game"

The command docker run ... do the following things:

• Run in privileged mode to have permission to run the browser
• Map host directory /opt/metagpt/config to container directory /app/metagpt/config
• Map host directory /opt/metagpt/workspace to container directory /app/metagpt/workspace
• Execute the demo command python startup.py "Write a cli snake game"

Build image by yourself

# You can also build metagpt image by yourself.
git clone https://github.com/geekan/MetaGPT.git
cd MetaGPT && docker build -t metagpt:custom .

Configuration

• Configure your OPENAI_API_KEY in any of config/key.yaml / config/config.yaml / env
• Priority order: config/key.yaml > config/config.yaml > env

# Copy the configuration file and make the necessary modifications.
cp config/config.yaml config/key.yaml

Variable Name	config/key.yaml	env
OPENAI_API_KEY # Replace with your own key	OPENAI_API_KEY: "sk-..."	export OPENAI_API_KEY="sk-..."
OPENAI_API_BASE # Optional	OPENAI_API_BASE: "https://<YOUR_SITE>/v1"	export OPENAI_API_BASE="https://<YOUR_SITE>/v1"

Tutorial: Initiating a startup

python startup.py "Write a cli snake game"
# Use code review will cost more money, but will opt for better code quality.
python startup.py "Write a cli snake game" --code_review True 

After running the script, you can find your new project in the workspace/ directory.

Preference of Platform or Tool

You can tell which platform or tool you want to use when stating your requirements.

python startup.py "Write a cli snake game based on pygame"

Usage

NAME
    startup.py - We are a software startup comprised of AI. By investing in us, you are empowering a future filled with limitless possibilities.

SYNOPSIS
    startup.py IDEA <flags>

DESCRIPTION
    We are a software startup comprised of AI. By investing in us, you are empowering a future filled with limitless possibilities.

POSITIONAL ARGUMENTS
    IDEA
        Type: str
        Your innovative idea, such as "Creating a snake game."

FLAGS
    --investment=INVESTMENT
        Type: float
        Default: 3.0
        As an investor, you have the opportunity to contribute a certain dollar amount to this AI company.
    --n_round=N_ROUND
        Type: int
        Default: 5

NOTES
    You can also use flags syntax for POSITIONAL ARGUMENTS

Code walkthrough

from metagpt.software_company import SoftwareCompany
from metagpt.roles import ProjectManager, ProductManager, Architect, Engineer

async def startup(idea: str, investment: float = 3.0, n_round: int = 5):
    """Run a startup. Be a boss."""
    company = SoftwareCompany()
    company.hire([ProductManager(), Architect(), ProjectManager(), Engineer()])
    company.invest(investment)
    company.start_project(idea)
    await company.run(n_round=n_round)

You can check examples for more details on single role (with knowledge base) and LLM only examples.

QuickStart

It is difficult to install and configure the local environment for some users. The following tutorials will allow you to quickly experience the charm of MetaGPT.

• MetaGPT quickstart

Citation

For now, cite the Arxiv paper:

@misc{hong2023metagpt,
      title={MetaGPT: Meta Programming for Multi-Agent Collaborative Framework}, 
      author={Sirui Hong and Xiawu Zheng and Jonathan Chen and Yuheng Cheng and Jinlin Wang and Ceyao Zhang and Zili Wang and Steven Ka Shing Yau and Zijuan Lin and Liyang Zhou and Chenyu Ran and Lingfeng Xiao and Chenglin Wu},
      year={2023},
      eprint={2308.00352},
      archivePrefix={arXiv},
      primaryClass={cs.AI}
}

Contact Information

If you have any questions or feedback about this project, please feel free to contact us. We highly appreciate your suggestions!

• Email: [email protected]
• GitHub Issues: For more technical inquiries, you can also create a new issue in our GitHub repository.

We will respond to all questions within 2-3 business days.

Demo

demo-cli-blackjack-compress.mp4