This project provides a framework for generating customized software using natural language prompts, leveraging a multi-agent system of LLM-powered agents simulating a software company. It's designed for researchers and developers interested in exploring LLM-driven collaboration and automated software development, offering a highly customizable and extendable platform.

How It Works

ChatDev simulates a software company with specialized agents (CEO, CPO, CTO, Programmer, Tester, etc.) collaborating through functional seminars. This multi-agent approach, organized in a chain-like topology, facilitates task decomposition and execution for software creation, aiming to mimic human team dynamics for efficient development. Recent advancements like MacNet introduce DAG-based collaboration for greater scalability and versatility beyond software development.

Quick Start & Requirements

• Install: Clone the repository, create a Python 3.9+ environment (e.g., conda create -n ChatDev_conda_env python=3.9), activate it, and run pip3 install -r requirements.txt.
• Prerequisites: OpenAI API key.
• Run: python3 run.py --task "[description_of_your_idea]" --name "[project_name]".
• Docs: Wiki, Web Access

Highlighted Details

• Supports multi-agent collaboration with customizable roles and workflows (ChatChain, Phase, Role).
• Features like Git mode, Human-Agent-Interaction, and Art mode enhance development flexibility.
• Offers incremental development capabilities to build upon existing codebases.
• Research papers on Experiential Co-Learning and MacNet are available, with MacNet enhancing agent collaboration scalability.

Maintenance & Community

The project is actively developed, with recent updates introducing MacNet and research on iterative refinement. Community contributions are encouraged, with links to a Discord server for support and discussion.

Licensing & Compatibility

Source code is licensed under Apache 2.0. Data used is licensed under CC BY-NC 4.0, restricting commercial use of trained models.

Limitations & Caveats

The CC BY-NC 4.0 data license imposes non-commercial restrictions on models trained with project data, limiting commercial applications. The core functionality relies on OpenAI API keys, incurring associated costs.