ChatDev is an open-source conversational platform developed by OpenBMB. It emulates an entire software development team, with agents forming a multi-agent organizational structure. These agents collaborate by participating in specialized functional seminars, including tasks such as designing, coding, testing, and documenting. The main goal of ChatDev is to offer an easy-to-use, highly customizable, and extendable framework based on large language models (LLMs). It includes features like Git integration for version control, Human-Agent-Interaction mode for interactive participation in the design stage, and Art mode for generating images used in the software.
AutoGen, developed by Microsoft, facilitates multi-agent conversations for task completion. This framework allows developers to define multiple agents with specific roles in what is essentially an AI group chat. These agents can converse with one another to provide a solution and can even integrate human feedback. Developers can use AutoGen to build a full spectrum of conversational patterns regarding topics like conversation autonomy, the number of agents involved, and the topology of agent dialogues. AutoGen provides an enhanced inference API that can replace openai.Completion or openai.ChatCompletion.
There are other open-source autonomous agent frameworks worth mentioning. For instance, AGENTS is an open-source library designed to support important features including planning, memory, tool usage, multi-agent communication, and fine-grained symbolic control. It is user-friendly and research-friendly, enabling non-specialists to build, customize, test, tune, and deploy state-of-the-art autonomous language agents without much coding.
SuperAGI is another open-source infrastructure for building components, tools, frameworks, and models to enable open-source AGI. It offers multiple open-source models via Python, Node.js, and HTTP APIs to power autonomous agents.
Adala is an open-source framework for autonomous data labeling agents. Adala's agents are designed to learn and improve at data tasks like classification and labeling when provided with ground truth datasets.
MetaGPT - MetaGPT takes a one line requirement as input and outputs user stories / competitive analysis / requirements / data structures / APIs / documents, etc.