MetaGPT: Modular Multi-Agent System

Project Background

MetaGPT is based on the research paper "MetaGPT: Meta Programming for Multi-Agent Collaborative Framework". The paper demonstrates that software development can be automated and scaled by organizing Large Language Models (LLMs) into structured, role-based multi-agent teams. Inspired by real-world software teams, metaGPT agents each take on a professional role (e.g., Product Manager, Project Manager, Architect, Engineer, QA Engineer) and collaborate in a workflow that mirrors enterprise software engineering.

This implementation adapts the metaGPT research framework to a command-line tool, using the latest Groq LLMs (for example, llama-3.3-70b-versatile). The design is modular—agents, pipeline, LLM interface, and CLI are clearly separated—making it extensible and easy to experiment with orchestration or alternative LLMs.

Repository Structure

metaGPT/
│
├── agents/
│   ├── base_agent.py
│   ├── product_manager.py
│   ├── project_manager.py
│   ├── architect.py
│   ├── engineer.py
│   ├── qa_engineer.py
│   └── __init__.py
│
├── llm/
│   ├── base_llm.py
│   └── groq_llm.py
│
├── core/
│   ├── pipeline.py
│   ├── task.py
│   └── utils.py
│
├── cli/
│   ├── __init__.py
│   └── main.py
│
├── tests/
│   └── test_agents.py
│
├── requirements.txt
├── .env
└── README.md

Component Responsibilities

Folder/File	Purpose
agents/	Agent role classes: Product Manager, Project Manager, Architect, Engineer, QA Engineer (all based on BaseAgent).
llm/	LLM abstraction and Groq API integration.
core/	Pipeline orchestration, task structure, and utilities.
cli/	User-facing command-line entry point.
tests/	Pytest-based tests for key components.
requirements.txt	Python dependencies.
.env	Groq API key and sensitive configuration.
README.md	Project overview and instructions (what you're reading now).

MetaGPT as a CLI Solution: Implementation Journey

The original MetaGPT research explores agent-based workflow, but presents it as a conceptual orchestration (not as a package/repo). This CLI implementation makes the following enhancements:

• Modular Codebase: Each professional role is a class, and all share a unified BaseAgent.
• Groq LLM Integration: Uses up-to-date supported Groq models (e.g., llama-3.3-70b-versatile) with a switchable LLM backend.
• Pipeline Orchestration: Agent outputs are piped in sequence, matching the metaGPT paper’s workflow—requirements → work breakdown → architecture → code → QA.
• Command-Line Interface: The entire process is controlled through a CLI (with Typer), allowing real-time input and output.
• Testing: Includes unit and integration tests using pytest to ensure extensibility and reliability.

Setup and Usage Instructions

1. Clone the Repository

git clone <your-repo-url>  # Replace with your repository URL
cd metaGPT

2. Create and Activate a Virtual Environment (recommended)

python -m venv venv
source venv/bin/activate         # On Windows: venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure the Groq API Key
   1. Register for a Groq LLM account (if you don’t have one) to obtain an API key.
   2. Create a .env file in the project directory and add the API key, like this:

GROQ_API_KEY=your-groq-api-key-here

5. Update the model name (if required) In llm/groq_llm.py, ensure that the model value is set to a currently supported Groq model, e.g.

"model": "llama-3.3-70b-versatile",

Check Groq documentation or dashboard for the latest recommended model names.

6. Run the CLI

python -m cli.main

Tip: Instead of python cli/main.py, using -m resolves imports correctly.

7. Using the CLI

• On launch, you’ll see a welcome prompt and be asked to enter a project idea.
• The CLI will orchestrate the agent team:
  1. Product Manager — requirements, user scenarios, user stories (displayed under “Product Requirements”).
  2. Project Manager — task breakdown, schedule, assignment, progress table (displayed under “Project Plan”).
  3. Architect — system architecture and module decomposition.
  4. Engineer — code or pseudocode implementation of modules.
  5. QA Engineer — test cases, validation strategies, and bug report template.
• Each section will be printed after completion.

8. Example Run Please feel free to look at SampleOutputs for some examples.

9. Run tests using pytest