--- layout: single title: "大型语言模型驱动的自主代理" date: 2024-05-26 08:00:00 +0800 categories: [编程开发, AI 与大模型] tags: [Agent, LLM] --- ## Application scenarios of AI agents（AI代理的应用场景） AI代理是LLM应用的重要场景，构建代理应用将是2024年的重要技术领域。目前我们主要的智能形式有单AI代理，多AI代理，混合AI代理等三种。 ![](/images/2024/Agent/hybridAgent.png) ### Single AI Agent（单一人工智能代理） 在特定任务场景下完成的工作，比如 GitHub Copilot Chat 下的代理工作区，就是根据用户需求完成特定编程任务的一个例子。基于 LLM 的能力，单个代理可以根据任务执行不同的动作，比如需求分析、项目阅读、代码生成等。它也可以应用于智能家居和自动驾驶。 ### Multi-AI Agents（多人工智能代理） 这就是AI代理之间相互交互的工作。例如上述Semantic Kernel代理实现就是一个例子。脚本生成的AI代理与执行脚本的AI代理进行交互。多代理应用场景在高度协同的工作中非常有帮助，例如软件行业开发、智能生产、企业管理等。 ### Hybrid AI Agent（混合人工智能代理） 这就是人机交互，在同一个环境下做决策。比如智慧医疗、智慧城市等专业领域，可以利用混合智能来完成复杂的专业工作。 - [Building AI Agent Applications Series - Understanding AI Agents](https://techcommunity.microsoft.com/t5/educator-developer-blog/building-ai-agent-applications-series-understanding-ai-agents/ba-p/4046944) ## [Intro of AI agent, & AI agent projects summary](https://medium.com/@henryhengluo/intro-of-ai-agent-ai-agent-projects-summary-52f4a364ab86) | Project | Key Features | |---------------|-------------------------------------------------------------------------------------------------------| | LangChain | - Python and JavaScript libraries for building LLM-backed apps | | | - Modular components for perceiving context, reasoning, chaining, etc. | | | - Reference architectures and templates | | | - Tooling for debugging, testing, deploying chains | | AutoGen | - Orchestrates LLMs and agents for multi-agent conversations | | | - Customizable and conversational agents | | | - Human participation in loops | | | - Toolkit for reasoning, caching, error handling | | PromptAppGPT | - Low-code prompt-based development | | | - Integrations for GPT, DALL-E, and plugins | | | - Online editor, compiler, runner | | | - Auto-generated UI | | | - Built-in agent examples | | AutoGPT | - Toolkit for building custom AI agents | | | - Leverages GPT-3, GPT-4 for agents | | | - Popular open source agent project | | BabyAGI | - Minimalist Python agent | | | - Uses GPT and vector DB | | | - Create, prioritize, execute tasks | | SuperAGI | - Alternative to AutoGPT | | | - Multiple models, vector DBs | | | - GUI and operations console | | | - Performance telemetry | | | - Toolkits and marketplace | | ShortGPT | - Automates video creation workflows | | | - Scripts, prompts, templates | | | - Multilingual voiceover and subtitles | | | - Resource and asset sourcing | | ChatDev | - Multi-agent "virtual software company" | | | - Agents in specialized roles | | | - Workshop model for collaboration | | MetaGPT | - Mimics software company structure | | | - PM, engineer, etc. roles assigned | | | - Agents collaborate on tasks | | Camel | - Early multi-agent framework | | | - Dynamic role assignment | | | - Stages scenarios for collaboration | | JARVIS | - Task planning with ChatGPT | | | - Model selection from Hub | | | - Orchestrates specialist models | | OpenAGI | - Combines expert and LLM models | | | - RLTF for model improvement | | | - Specialized for complex tasks | | XAgent | - Modular dispatcher, planner, actor | | | - Human collaboration abilities | | | - Safety and extensibility | - [Overview of 13 projects (webpage)](https://script.google.com/macros/s/AKfycbwm6MFGFxFKHauvyj-Z34uQTJakbOFIaRTPEp9MOaBvcK5NvlLdvPbZer1lo0LjTVu1tA/exec) ## AI Agent Frameworks - [AutoGPT](https://github.com/Significant-Gravitas/AutoGPT) - [TaskWeaver](https://github.com/microsoft/TaskWeaver) - [CrewAI](https://github.com/joaomdmoura/crewAI) - [BabyAGI](https://github.com/yoheinakajima/babyagi) - [CAMEL](https://github.com/camel-ai/camel) - [MetaGPT](https://github.com/geekan/MetaGPT) - [AutoGen](https://github.com/microsoft/autogen) - [DSPy](https://github.com/stanfordnlp/dspy) - [AutoAgents](https://github.com/Link-AGI/AutoAgents) - [OpenAgents](https://github.com/xlang-ai/OpenAgents) - [Agents](https://github.com/aiwaves-cn/agents) - [AgentVerse](https://github.com/OpenBMB/AgentVerse) - [ChatDev](https://github.com/OpenBMB/ChatDev) - [XAgent](https://github.com/OpenBMB/XAgent) - [Qwen-Agent](https://github.com/QwenLM/Qwen-Agent) - [Lagent](https://github.com/InternLM/lagent) ## Multi-Agent Systems - [CrewAI](https://www.crewai.com/) - [AutoGen](https://microsoft.github.io/autogen/) - [AutoGen Studio](https://github.com/microsoft/autogen/tree/main/samples/apps/autogen-studio) - [AutoGen Studio 2.0](https://autogen-studio.com/autogen-studio-ui) - [Rivet](https://rivet.ironcladapp.com/) - [Agency Swarm](https://github.com/VRSEN/agency-swarm) - [Semantic Kernel](https://learn.microsoft.com/en-us/semantic-kernel/overview/) - [CAMEL](https://github.com/camel-ai/camel) - [AgentVerse](https://github.com/OpenBMB/AgentVerse) - [BotSharp](https://github.com/SciSharp/BotSharp) - AI Agent Team Frameworks - [AI Agent Team Frameworks - Part I](https://www.linkedin.com/pulse/ai-agent-team-frameworks-part-i-p%C3%A9ter-balogh-1xanf) - [AI Agent Team Frameworks - Part II](https://www.linkedin.com/pulse/ai-agent-team-frameworks-part-ii-p%C3%A9ter-balogh-zj1uf) - [AI Agent Team Frameworks - Part III](https://www.linkedin.com/pulse/ai-agent-team-frameworks-part-iii-p%C3%A9ter-balogh-hscuf) - [AI Agent Team Frameworks - Part IV](https://www.linkedin.com/pulse/ai-agent-team-frameworks-part-iv-p%C3%A9ter-balogh-yya8e) - []() ## 参考资料 - [XAgent：用于解决复杂任务的自主代理](https://blog.x-agent.net/blog/xagent/) - [XAgent 为大型模型开发开放模型、数据集、系统和评估工具](https://blog.x-agent.net/projects/) - [Fetch.ai 的 5 大 AI 代理集成](https://fetch.ai/blog/fetch-ai-s-top-5-AI-Agent-integrations) - [使用Qwen-Agent将上下文记忆扩展到百万量级](https://qwenlm.github.io/zh/blog/qwen-agent-2405/)