AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation (2308.08155v2)

Abstract: AutoGen is an open-source framework that allows developers to build LLM applications via multiple agents that can converse with each other to accomplish tasks. AutoGen agents are customizable, conversable, and can operate in various modes that employ combinations of LLMs, human inputs, and tools. Using AutoGen, developers can also flexibly define agent interaction behaviors. Both natural language and computer code can be used to program flexible conversation patterns for different applications. AutoGen serves as a generic infrastructure to build diverse applications of various complexities and LLM capacities. Empirical studies demonstrate the effectiveness of the framework in many example applications, with domains ranging from mathematics, coding, question answering, operations research, online decision-making, entertainment, etc.

• The paper presents AutoGen’s core contribution: a flexible, open-source framework enabling customizable multi-agent conversation programming for LLM applications.
• It introduces a novel paradigm that integrates LLMs, human inputs, and tools to manage task-oriented dialogues through dynamic inter-agent interactions.
• Empirical studies demonstrate that AutoGen enhances performance in complex problem solving and mathematical tasks while reducing development effort.

Introduction

AutoGen is an open-source framework introduced to facilitate the construction of applications powered by LLMs through the innovative use of multi-agent conversations. The framework provides a generic platform accommodating various modes of operation, integrating LLMs, human inputs, and tools. Developers can craft customizable, conversable agents with the ability to engage in conversation to perform tasks.

Agent Customization and Conversation Programming

A central feature of AutoGen is its support for creating conversable agents – entities designed to interact in a task-oriented dialogue, capable of sending and receiving messages. These agents can manifest a variety of capabilities, drawing on LLMs, human inputs, and tool functionalities. Agents can be customized to suit the specific needs of an application, with reusable and extendable capabilities across different application needs.

AutoGen also introduces conversation programming, where inter-agent dialogs drive the computation and control flow of tasks. This design pattern hinges on agent responses that are conversation-centric, ensuring actions taken are relevant to the dialog context. The system enforces a natural progression of agent responses through an auto-reply mechanism. Programming and natural language control are blended to manage conversation patterns effectively.

Empirical Studies and Applications

Empirical studies affirm AutoGen's marked efficacy in numerous domains, showcasing applications extending from mathematics to complex problem solving. Agents in AutoGen have exhibited outstanding task performance, with empirical proofs underscoring the flexibility and strength of conversable agents and the conversation programming paradigm. Multi-agent conversations have enabled dynamic, flexible chat patterns that can evolve based on task requirements, offering innovative solutions with lesser developmental efforts.

Conclusion

AutoGen encapsulates a paradigm shift in conceptualizing and implementing multi-agent systems for varied LLM applications. With its emphasis on customizable agent design and a unified interface for conversable agents, AutoGen manages to converge the robust capabilities of LLMs into pragmatic, flexible applications that cater to diverse complexities and task requirements. The ability to program agent interactions both with other agents and with humans stands at the heart of AutoGen's innovative approach, paving the pathway for future advancements and research opportunities in the field of AI and LLMs.