开源Multi-agent AI智能体框架aevatar.ai，欢迎大家贡献代码

草根农民 發表於 2025-3-6 14:47:00

开源Multi-agent AI智能体框架aevatar.ai，欢迎大家贡献代码

<div data-page-id="VYFGdrCwGoF2l4x7y9hleVBAgUb" data-lark-html-role="root" data-docx-has-block-data="true">
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe">2025 年是 AI Agent 的元年，我们团队历时 3 个多月，现正式开源Multi-Agent AI 框架，欢迎各位园友前往 GitHub Fork、Star 或提交 PR，共同打造 aevatar.ai 生态。</div>
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe"> </div>
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe">Github地址： </div>
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe"><span class="break-words       tvm-parent-container"><span dir="ltr">aevatar 核心框架:<span class="white-space-pre"> https://github.com/aevatarAI/aevatar-framework <br>aevatar平台:<span class="white-space-pre"> https://github.com/aevatarAI/aevatar-station<br>aevatar 案例:<span class="white-space-pre"> https://github.com/aevatarAI/aevatar-gagents<br></span></span></span></span></span></div>
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe"> </div>
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe">aevatar.ai，一个统一的多智能体平台，旨在解决跨领域和多样化工作负载下开发、部署和管理多种AI智能体的复杂性。通过基于插件的方法和灵活的部署策略——从基于DLL的加载到容器化和分布式执行框架——aevatar.ai允许用户和开发者在一个统一的生态系统中无缝集成专业化的AI解决方案。</div>
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe"> </div>
<div class="ace-line ace-line old-record-id-NsySdaI0ooKrsTx2PStlvybvgUd">关键组件包括：aevatar框架，定义了标准化的智能体接口和生命周期管理；aevatar Station，一个集中化的门户和市场，用于智能体发现、插件处理、请求路由和用户访问控制；以及aevatar Agents，一个包含官方和社区开发的AI模块的仓库，支持多种任务，如自然语言理解、计算机视觉和推荐系统。通过集中化智能体交互和事件流，aevatar.ai减少了集成开销，强制执行一致的安全策略，并提供强大的监控和日志功能以提高可靠性。</div>
<div class="ace-line ace-line old-record-id-NsySdaI0ooKrsTx2PStlvybvgUd"> </div>
<div class="ace-line ace-line old-record-id-C56XdVHeioOWNgxnKqbl0Hk6gMh">通过其开源、模块化的架构，aevatar.ai既适用于小规模使用，也支持大规模企业部署，具备高并发、自动扩展、智能体重用、沙箱化和审计追踪等功能。这些创新促进了一个可持续的AI生态系统，使组织能够快速采用和发展先进的AI能力，同时让开发者专注于创建强大、专业化的智能体，而无需担心基础设施和生命周期管理的复杂性。</div>
<ol class="ace-line ace-line old-record-id-NQiwdIrw0oDfP3xU79xlgAZugaf" start="1">
<li data-list="number">
<h1>引言 (Introduction)

</h1>

</li>

</ol>
<div class="ace-line ace-line old-record-id-E2V6d1j1Lo4jiFxgFYNlJBNWgPf">随着人工智能（AI）技术的快速发展，大型语言模型（LLMs）和智能体的应用已从单一问答交互演变为更高级的能力，如多智能体协作、跨模型合作和复杂业务流程编排。</div>
<div class="ace-line ace-line old-record-id-Uc9bdnUIbox1ISxwBlslfC1Vgmd">然而，目前市场上的AI系统普遍面临平台隔离、模型限制、部署复杂和缺乏可观测性等问题，难以满足企业用户对高效、灵活和安全AI协作的需求。</div>
<div class="image-uploaded gallery old-record-id-Ek1PdLQKno6cPYxIYZ7lk6xrgbf" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"fe3136be-6f30-4e0e-88be-d4b86dd9e759","height":830,"width":1280,"currHeight":830,"currWidth":1280,"natrualHeight":830,"natrualWidth":1280,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FQvXibuYHBoEL4sxtjcml7gMvgRh%2F%3Fpreview_type%3D16","file_token":"QvXibuYHBoEL4sxtjcml7gMvgRh","image_type":"image/png","size":145745,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/introduction-fig-1.7Nbh3Lxs.png" width="936" height="604"></div>
<div class="ace-line ace-line old-record-id-TurYdRkbjo4aFixRNkKljCwUgNu">aevatar.ai作为下一代多AI智能体框架的先行者，旨在构建一个跨平台、跨模型的AI智能体生态系统。通过开放的架构、强大的可视化编排能力和云原生部署方式，它赋能开发者和业务用户在一个系统中统一管理、调度和协调多个智能体，实现“多场景、多模型、多角色”的高效协作。</div>
<div class="ace-line ace-line old-record-id-GhOhdbXAFoCX5zxkm1olJVUbgCh">通过aevatar.ai，我们致力于提供一个灵活、可扩展且符合安全要求的AI解决方案，以推动AI技术的广泛应用和落地。</div>
<div class=" old-record-id-ZvJodzWSqoF0yOx8PNdl5NFigZe" data-type="divider"><hr></div>
<ol class="ace-line ace-line old-record-id-ImTldf3l7o7l39xk1bjl09F2gib" start="2">
<li data-list="number">
<h1 class="heading-1">背景与挑战 (Background & Challenges)</h1>

</li>

</ol><img src="https://whitepaper.aevatar.ai/assets/background-and-challenges-fig-1.CWDIWHSv.jpeg" width="811" height="710">
<h2 class="heading-2 ace-line old-record-id-Zyg3dMhEioC8AvxAJWglfGtkgNh">2.1 AI智能体系统隔离</h2>
<div class="ace-line ace-line old-record-id-D2Nhdssu1o8r1kxZHiAlHR8xg8c">目前，许多AI智能体被隔离在各自的平台中，缺乏统一的通信协议和互操作性。这使得系统之间难以共享数据或协同工作，限制了AI系统的整体效能，尤其是在需要跨平台协作的场景中。</div>
<h2 class="heading-2 ace-line old-record-id-NvizdZSDxoCY6yx7IKXl5404gph">2.2 单一LLM的局限性</h2>
<div class="ace-line ace-line old-record-id-DB4FdwiP9oXXDExVcHal268ggqc">大多数AI智能体依赖于单一的语言模型（如GPT-4或Llama2）。这带来了集中风险；在面对复杂的多步骤任务或多语言场景时，性能可能会受到影响。</div>
<div class="ace-line ace-line old-record-id-Ezk0dk9wGoHUVyxXAVIlQKMpgrb">单一LLM模型的局限性使得系统无法灵活切换或并行使用多个模型，从而限制了其应用范围和性能。</div>
<h2 class="heading-2 ace-line old-record-id-HFsDdVz74oN1rDxCFd7lCchugef">2.3 检索增强生成（RAG）精度不足</h2>
<div class="ace-line ace-line old-record-id-TUBVdxNsro2xnTxfS4SlwsfHgHc">大多数AI智能体使用检索增强生成从专业知识库中检索信息，但由于信息或文档可能不相关、过时或质量低下，难以实现完美的知识库优化和准确性。</div>
<h2 class="heading-2 ace-line old-record-id-GM4BdFCAxonVYHx0RKGlZdiGgyq">2.4 缺乏事件追踪与可观测性</h2>
<div class="ace-line ace-line old-record-id-BQnmdLKbwoIn0cx3h7BlJP0pg6g">现有的AI系统通常缺乏对AI智能体内部状态和历史交互的源头管理。当系统故障或推理异常发生时，难以定位问题并重放事件，增加了维护复杂性和风险。</div>
<h2 class="heading-2 ace-line old-record-id-ZwxcdwZiSouD1MxnrZsloGHHgxf">2.5 高部署与协作成本</h2>
<div class="ace-line ace-line old-record-id-JC3adzJ6EoTcSGxGExKl0qlLgfh">传统的AI系统通常需要复杂的安装、配置和维护过程。缺乏用户友好的工作流编排工具，尤其是在多智能体协作场景中，导致开发和维护成本较高。</div>
<ol class="ace-line ace-line old-record-id-XRw0d23a7oVLrLxr1zhlcRukgyf" start="3">
<li data-list="number">
<h1 class="heading-1">总体目标与愿景 (Vision & Goals)</h1>

</li>

</ol>
<div class="image-uploaded gallery old-record-id-OYR9dimVOo23O0x111BlwmmlgUb" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"3e97edfd-7d1b-4592-87ea-0fd05ad16a09","height":978,"width":1280,"currHeight":978,"currWidth":1280,"natrualHeight":978,"natrualWidth":1280,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FRYMmbyzGIoPLFlx0Nb8lkvA0g3V%2F%3Fpreview_type%3D16","file_token":"RYMmbyzGIoPLFlx0Nb8lkvA0g3V","image_type":"image/png","size":365400,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/vision-and-goals-fig-1.Dhr-kl3k.png" width="1431" height="1094" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJSWU1tYnl6R0lvUExGbHgwTmI4bGt2QTBnM1YiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiT1lSOWRpbVZPbzIzTzB4MTExQmx3bW1sZ1ViIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9SWU1tYnl6R0lvUExGbHgwTmI4bGt2QTBnM1YvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/RYMmbyzGIoPLFlx0Nb8lkvA0g3V/?mount_node_token=OYR9dimVOo23O0x111BlwmmlgUb&mount_point=docx_image" data-width="1280" data-height="978" data-lark-image-uri="drivetoken://RYMmbyzGIoPLFlx0Nb8lkvA0g3V" data-lark-image-width="1280" data-lark-image-height="978"></div>
<h2 class="heading-2 ace-line old-record-id-YbnPdN4W0oIz2HxkwnNlWCuJg0e">3.1 核心：aevatar框架</h2>
<div class="ace-line ace-line old-record-id-H71hdiRYxonIpOxdAf4lD6UYgEc">aevatar框架是aevatar.ai的核心，负责处理基本处理逻辑、智能体交互和核心组件。</div>
<div class="ace-line ace-line old-record-id-Z4PGdIsGyox1WexivlDlUfLKgvb"><strong>Orleans“Grains”作为智能体</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-KEE5dWkmfoBbQRxkaxvlehQUghd" data-list="bullet">Orleans使用“grains”（轻量级、隔离的微对象）来表示参与者或“智能体”。</li>
<li class="ace-line ace-line old-record-id-ARboda9BboPGM7xScPJlSqJigJg" data-list="bullet">一组silos（Orleans中的运行时主机）协调这些grains，使其可以分布在多台服务器上并进行扩展。</li>
<li class="ace-line ace-line old-record-id-NQsmd3yeHoLHqHxyQB6lrdJagOd" data-list="bullet">在此架构中，每个grain实际上是一个智能体（通常称为“GAgent”）。</li>

</ul>
<div class="ace-line ace-line old-record-id-YBOPdc6bpo19NvxYJxlly76HgLe"><strong>多智能体</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Z944dUGdQohb9qxx2TWlvRfagLe" data-list="bullet">图中展示了多层智能体分组。例如，一个“发布GAgent”协调多个“组成员GAgent”实例。</li>
<li class="ace-line ace-line old-record-id-EGchdYmIMo9cOIxRnaLlxoWHgOb" data-list="bullet">事件处理器管理异步触发器或状态变化，使智能体能够实时响应来自其他智能体的数据或更新。</li>

</ul>
<div class="ace-line ace-line old-record-id-Mpj6dVKUPoigf6xUt4Klcewhgqd"><strong>AI集成</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Bjt1d8tBLoRJb9x0uSEl85PBg6c" data-list="bullet">Semantic Kernel提供高级AI编排和提示链功能。</li>

</ul>
<div class="ace-line ace-line old-record-id-CmbcdRZx2oge4OxECFKl6W1JgkI">通过结合以上所有内容，系统可以扩展大量AI智能体，每个智能体执行专门任务，同时通过分组或子组协调它们以实现更复杂的协作目标。</div>
<h2 class="heading-2 ace-line old-record-id-ZOd8d3nfko3r7WxBoWflnlHOgxb">3.2 顶层：aevatar应用</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-YPibdSv9vosdCzxUHhUlWm0SgHd" data-list="bullet"><strong>市场</strong>：一个集中化的平台，用户可以发现、开发、管理和部署各种AI智能体。</li>
<li class="ace-line ace-line old-record-id-SXiAd4LGwobhJZxfuS7l6ZgLgka" data-list="bullet"><strong>智能体</strong>：执行特定任务或功能的独立AI智能体。这些智能体可以独立开发和部署。</li>
<li class="ace-line ace-line old-record-id-EQVndY2R6ouzGYxZI11lRwo6gEw" data-list="bullet"><strong>Webhook</strong>：aevatar可以无缝编排大量外部输入，将现实世界的触发器转换为结构化事件，供G-agent处理，从而实现与各种外部系统的持续、实时交互。</li>

</ul>
<div class="ace-line ace-line old-record-id-Mh5HdOF4NoELO8xgY0wl3j5ogtb"><strong>示例智能体</strong>：</div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-R0tzdwtozoWARnxt6hSlYabtg2f" data-list="bullet"><strong>Twitter智能体</strong>：监控推文、发布更新或与Twitter互动。</li>
<li class="ace-line ace-line old-record-id-IMEsdSzAroL5vpxFWH1l3lC7gNc" data-list="bullet"><strong>Telegram智能体</strong>：与Telegram进行聊天交互。</li>
<li class="ace-line ace-line old-record-id-Snk7dIrc4oGHoQxTTMBlj8y8gEb" data-list="bullet"><strong>编程智能体</strong>：帮助生成或审查代码。</li>
<li class="ace-line ace-line old-record-id-T6p3denMPobTvKxxSPmlTL24ghc" data-list="bullet"><strong>营销智能体</strong>：执行营销任务，如活动管理。</li>
<li class="ace-line ace-line old-record-id-X7PhdOPO6oJh69x824dlfl5KgAb" data-list="bullet"><strong>运营智能体</strong>：处理运营任务。</li>
<li class="ace-line ace-line old-record-id-MM3SdJr24oxGg0xjRhWlfPIkg3c" data-list="bullet"><strong>aelf智能体</strong>等。</li>

</ul>
<div class="ace-line ace-line old-record-id-YRxodLBIjonCG5xHdIvlamwLgeh">这些是基于底层多智能体框架构建的面向终端用户的“产品”。每个智能体都可以具有专门的逻辑，连接外部API，并利用aevatar核心引擎。</div>
<h2 class="heading-2 ace-line old-record-id-PRzId8u2MoPxlZxIn3kl7C6Wg3c">3.3 环境（Web 2 / Web 3）</h2>
<div class="ace-line ace-line old-record-id-Ui5zdlgtpokm1Uxz6GXl37LNglh">这表示aevatar智能体运行的更广泛环境——既包括传统的Web 2.0环境（如REST API、SaaS服务），也包括Web 3.0环境（如区块链或去中心化服务）。该框架旨在无缝融入这些生态系统。</div>
<h2 class="heading-2 ace-line old-record-id-LfvldjRqcoSmgYxHVHSlBJpHgZd">3.4 LLM集成</h2>
<div class="ace-line ace-line old-record-id-Pnkadx8uzo0G2ZxjFHHl11gGgTc">在图的右侧，您可以看到主要的LLM（大型语言模型）提供商：</div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Ib36dV9rGoIDyDx8VAhlPsljgMf" data-list="bullet">OpenAI / ChatGPT</li>
<li class="ace-line ace-line old-record-id-AOYpdngzOo6zdoxYDfJlEWCBg8e" data-list="bullet">Anthropic</li>
<li class="ace-line ace-line old-record-id-WXUfd69kXoZ401xJgOQlBiALgxg" data-list="bullet">Meta</li>
<li class="ace-line ace-line old-record-id-E1nldPBAyotRtsx5JXMlSOzjgEg" data-list="bullet">Azure OpenAI</li>
<li class="ace-line ace-line old-record-id-NYbod2pylodMD8xxNMMlnNQ9gsb" data-list="bullet">Deepseek</li>
<li class="ace-line ace-line old-record-id-UyzddRmCXosLb1xK1X3lroUagBe" data-list="bullet">以及其他更多…</li>

</ul>
<div class="ace-line ace-line old-record-id-SMFvdMPSlotxY0xBuZGlChPNgsh">这些LLM通过Semantic Kernel连接器集成，使每个智能体都能利用自然语言理解、生成和高级推理功能。</div>
<h2 class="heading-2 ace-line old-record-id-CiYTdeMMaoOjFux8OkKlj3xWgqh">3.5 数据与消息层</h2>
<div class="ace-line ace-line old-record-id-MX99d2SqsojJ2HxfaqVlsKzZgJg">在框架之上，核心数据与消息技术包括：</div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Fa9jdGfU5o8TECxYDZnlJVyBgca" data-list="bullet"><strong>Kafka</strong>：实时消息传递和事件流。</li>
<li class="ace-line ace-line old-record-id-H9sadpKp0opMoexi1iJlB4R4g0e" data-list="bullet"><strong>MongoDB</strong>：基于文档的通用数据存储。</li>
<li class="ace-line ace-line old-record-id-F5y5dlZtroioQuxq62rlXWrxgrf" data-list="bullet"><strong>Elasticsearch</strong>：大规模全文搜索和分析。</li>
<li class="ace-line ace-line old-record-id-Bk11da4AuovzxqxOWy1lNQCggro" data-list="bullet"><strong>Redis</strong>：用于缓存和高速访问的内存数据存储。</li>
<li class="ace-line ace-line old-record-id-HXRbdjc20ofcqyxnSNLlPVKMgwe" data-list="bullet"><strong>Qdrant</strong>：专用向量存储。</li>

</ul>
<div class="ace-line ace-line old-record-id-MjxUdjI4Poj2IsxXkiFlaxlggVd">这些技术支持高吞吐量的数据摄取、搜索、缓存和状态管理，对于大规模智能体交互至关重要。</div>
<h2 class="heading-2 ace-line old-record-id-FSP1d5E6xoIadRxs6BjlJiRvgQh">3.6 部署与DevSecOps</h2>
<div class="ace-line ace-line old-record-id-UcfidOGBKoaEYMxuEM3lxHQfgxh">用于构建、部署和管理aevatar框架的DevSecOps工具包括：</div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-G4EFdcBWMoFgzDx4KlRlhSKugkd" data-list="bullet"><strong>Kubernetes + Docker</strong>：跨集群的容器化和编排。</li>
<li class="ace-line ace-line old-record-id-SAWKds7SfoulsBx5wK5lq9gRgrc" data-list="bullet"><strong>GitHub Actions, GitOps, Argo</strong>：CI/CD管道和“GitOps”风格的部署，用于自动化、版本化发布。</li>
<li class="ace-line ace-line old-record-id-DKuVdXfTjotxajx3DMSldDVugKg" data-list="bullet">“DevSecOps”循环突出了以安全为中心的持续集成/持续部署实践。</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-XJ5EdlKQqoAhmIxcLHYlGvOVgXc">3.7 多云与安全</h2>
<div class="ace-line ace-line old-record-id-BegldL46woL7T2xtuPMlUWmMgpg">最后，支持多云策略，涵盖：</div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-EMzGdDp0yoT4lOxKzbzluAIWgob" data-list="bullet"><strong>GCP、AWS、Azure</strong>：支持的云服务提供商。</li>
<li class="ace-line ace-line old-record-id-ViHRdYAKroB5ygx6ZBKl4JkBgwe" data-list="bullet">额外的安全和可观测性工具，如Grafana（监控仪表板）、Vault（密钥管理）、Elasticsearch/Fluentd/Kibana（EFK日志和分析堆栈）等。</li>

</ul>
<div class="ace-line ace-line old-record-id-QSCudHQKqoDHyBxcT5QlbxVIgOh">这确保了平台可以在不同云基础设施上以安全、容错和成本高效的方式运行。</div>
<h2 class="heading-2 ace-line old-record-id-QrqadGN1soqen1x8QM3lhlMDgIc">3.8 整体架构</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-EILUdecfXoOtaQxRi8flevjJgFg" data-list="bullet">每个aevatar应用（如Twitter智能体或编程智能体）都是一个Orleans“grain”（或一组grains），封装了专门的逻辑。</li>
<li class="ace-line ace-line old-record-id-BTsfd6T1Eo4Ia2xLFloljXi7gNe" data-list="bullet">多智能体或“分组”方法协调大量grains，使它们能够通过Kafka、Redis或直接Orleans消息传递相互传递事件/消息。</li>
<li class="ace-line ace-line old-record-id-DqCndluQxoyhcAxK8nelru2igff" data-list="bullet">Semantic Kernel帮助编排更高级的AI推理、提示链和记忆/知识。</li>
<li class="ace-line ace-line old-record-id-Vcued8mh6oxlNmxXQ6ulK4Leg1g" data-list="bullet">整个设置打包用于云部署（Kubernetes + Docker），并与日志、安全和监控解决方案（Grafana、Vault、EFK）集成。</li>
<li class="ace-line ace-line old-record-id-KQn0dU5teoZncaxWADNlBxMpgrg" data-list="bullet">这种组合提供了一个可扩展、容错且高度可扩展的平台，用于在多个领域（Web 2和Web 3）、多个云上运行AI智能体，同时具备强大的安全性和可观测性。</li>

</ul>
<div class="ace-line ace-line old-record-id-DW3cdngiIolOJOx0jKdlqNaag7f">简而言之，aevatar.ai是一个全栈、云原生的多智能体编排框架，利用Orleans实现基于参与者的扩展，集成Semantic Kernel以提供AI功能，并采用全面的DevSecOps管道和多云部署策略。</div>
<h2 class="heading-2 ace-line old-record-id-Yek6dRzBjo3CfWxeddylxU2Agmb"><strong>3.9 解决上述挑战</strong></h2>
<ol class="ace-line ace-line old-record-id-JqMcd1xqmo93A9xnw9CluM8Kg92" start="1">
<li data-list="number">
<h3 class="heading-3"><strong>多智能体协作</strong></h3>

</li>

</ol>
<div class="image-uploaded gallery old-record-id-NgVldUxZuoqAAmxqs98le6iDgPc" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"afd307f0-678f-4d28-8904-124f2e9208fe","height":1164,"width":1678,"currHeight":1164,"currWidth":1678,"natrualHeight":1164,"natrualWidth":1678,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FXPPqbbAJboN0Qhx8L3Wls8eTgzb%2F%3Fpreview_type%3D16","file_token":"XPPqbbAJboN0Qhx8L3Wls8eTgzb","image_type":"image/jpeg","size":135355,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/vision-and-goals-fig-2.OsxMkGBA.jpeg" width="1428" height="940" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJYUFBxYmJBSmJvTjBRaHg4TDNXbHM4ZVRnemIiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiTmdWbGRVeFp1b3FBQW14cXM5OGxlNmlEZ1BjIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9YUFBxYmJBSmJvTjBRaHg4TDNXbHM4ZVRnemIvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/XPPqbbAJboN0Qhx8L3Wls8eTgzb/?mount_node_token=NgVldUxZuoqAAmxqs98le6iDgPc&mount_point=docx_image" data-width="1678" data-height="1164" data-lark-image-uri="drivetoken://XPPqbbAJboN0Qhx8L3Wls8eTgzb" data-lark-image-width="1678" data-lark-image-height="1164"></div>
<div class="ace-line text-indent ace-line old-record-id-ZBZgd4M1zoFqXZx3cTWlmpyngvb"><em>GAgent: 基于grain的智能体</em></div>
<div class="ace-line ace-line old-record-id-KOzVdjRPHo58BUxfwlFlzqrjgPg">通过分布式参与者模型（基于Orleans）和多智能体管理机制，aevatar.ai实现了多个AI智能体之间的高效互联和复杂事件调度，支持跨平台和跨场景的协作工作流。</div>
<div class="ace-line ace-line old-record-id-LkwqdolXgoIT7ixmGZslZ9ScgUd">aevatar.ai实现了多智能体框架，将AI智能体划分为不同的功能角色。它为多个智能体分配特定职责并将其分组，以在系统中完成用户分配的任务。</div>
<ol class="ace-line ace-line old-record-id-PX6Zd6SdboT6lvxB7q0lr5gfg6d" start="2">
<li data-list="number">
<h3 class="heading-3"><strong>统一的跨模型协作</strong></h3>

</li>

</ol>
<div class="image-uploaded gallery old-record-id-Javld35PMoMWmIxFlIKlllHGg2d" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"b6ee42b4-cb8f-4328-b1e3-062a8078a8a5","height":224,"width":1154,"currHeight":224,"currWidth":1154,"natrualHeight":224,"natrualWidth":1154,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FNOFEbluvjou8q6xa2yLl8ffEgCh%2F%3Fpreview_type%3D16","file_token":"NOFEbluvjou8q6xa2yLl8ffEgCh","image_type":"image/png","size":13747,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/vision-and-goals-fig-3.a5VxRldC.png" width="1154" height="224" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJOT0ZFYmx1dmpvdThxNnhhMnlMbDhmZkVnQ2giLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiSmF2bGQzNVBNb01XbUl4RmxJS2xsbEhHZzJkIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9OT0ZFYmx1dmpvdThxNnhhMnlMbDhmZkVnQ2gvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/NOFEbluvjou8q6xa2yLl8ffEgCh/?mount_node_token=Javld35PMoMWmIxFlIKlllHGg2d&mount_point=docx_image" data-width="1154" data-height="224" data-lark-image-uri="drivetoken://NOFEbluvjou8q6xa2yLl8ffEgCh" data-lark-image-width="1154" data-lark-image-height="224"></div>
<div class="ace-line ace-line old-record-id-HpUFd4YxHobApEx2mWbljZXGgvd">aevatar.ai 提供了一个多语言模型并行智能体框架。这克服了单一模型的局限性，支持在不同任务中自由切换或并行使用多个模型，从而提升系统灵活性和性能。</div>
<ol class="ace-line ace-line old-record-id-Wl4qd6K0HoxmyMxqqXQlqZkogId" start="3">
<li data-list="number">
<h3 class="heading-3"><strong>多智能体RAG架构</strong></h3>

</li>

</ol>
<div class="image-uploaded gallery old-record-id-Rnp4dQzqcoxOosxS5IhlW7GIgPf" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"1f83fefc-0bf6-4187-9625-cd28f17530ac","height":988,"width":1674,"currHeight":988,"currWidth":1674,"natrualHeight":988,"natrualWidth":1674,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FVXOUbg2Gqo8e3lxa3rflBBJHg6f%2F%3Fpreview_type%3D16","file_token":"VXOUbg2Gqo8e3lxa3rflBBJHg6f","image_type":"image/jpeg","size":136017,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/vision-and-goals-fig-4.C9bIX6ku.jpeg" width="1432" height="835" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJWWE9VYmcyR3FvOGUzbHhhM3JmbEJCSkhnNmYiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiUm5wNGRRenFjb3hPb3N4UzVJaGxXN0dJZ1BmIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9WWE9VYmcyR3FvOGUzbHhhM3JmbEJCSkhnNmYvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/VXOUbg2Gqo8e3lxa3rflBBJHg6f/?mount_node_token=Rnp4dQzqcoxOosxS5IhlW7GIgPf&mount_point=docx_image" data-width="1674" data-height="988" data-lark-image-uri="drivetoken://VXOUbg2Gqo8e3lxa3rflBBJHg6f" data-lark-image-width="1674" data-lark-image-height="988"></div>
<div class="ace-line ace-line old-record-id-BufgdU04boaFgyxNC2XlnWrdged">在多智能体RAG架构下，每个AI智能体代表一个基于特定知识库、检索策略和生成配置的定制化RAG；这在整个系统中最擅长的领域提供答案。通过编排器，用户问题被分配给适当的智能体。或者，也可以并行调用多个智能体，并通过信息整合模块合并答案。这实现了更专业、全面和可扩展的问答或信息生成系统。</div>
<div class="ace-line ace-line old-record-id-OT5qdsNhUovxmGxBydblIkbgg7I">多智能体RAG模型支持：</div>
<ol class="list-number1" start="1">
<li class="ace-line ace-line old-record-id-UUcPdU2s9odSH5xWSG7lFSiBg3b" data-list="number"><strong>灵活扩展</strong>：基于不同业务线或知识领域快速部署新智能体。</li>
<li class="ace-line ace-line old-record-id-VwcFdQB6govxZ7xUSErlAQGMg9d" data-list="number"><strong>降噪</strong>：利用领域特定知识库减少无关信息干扰。</li>
<li class="ace-line ace-line old-record-id-SjRXdsTw0oXkpSx11VTlmORJgtb" data-list="number"><strong>增强可信度</strong>：多个智能体之间的交叉验证。</li>
<li class="ace-line ace-line old-record-id-WWp7dqjb5oEef4xhK3tlYrCggob" data-list="number"><strong>可持续性</strong>：独立维护每个智能体的知识库，便于分而治之。</li>

</ol>
<div class="ace-line ace-line old-record-id-O1mwdl13Mo89oJx5n7vl92Rmg1c">这使得能够构建一个能够持续生成高质量内容的多智能体RAG平台。</div>
<ol class="ace-line ace-line old-record-id-PNzFdX8wgoaQYTxD1tZlg7ODglh" start="4">
<li data-list="number">
<h3 class="heading-3"><strong>可视化与易用性</strong></h3>

</li>

</ol>
<div class="image-uploaded gallery old-record-id-GciQdcFNeoFApFxi1fWlAVxtg7c" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"cf5c3580-4063-4d6b-8dbf-066692ace8cc","height":1138,"width":1886,"currHeight":1138,"currWidth":1886,"natrualHeight":1138,"natrualWidth":1886,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FRyTqbApKQo2NeXxCEBzlrI62gse%2F%3Fpreview_type%3D16","file_token":"RyTqbApKQo2NeXxCEBzlrI62gse","image_type":"image/png","size":319641,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/vision-and-goals-fig-5.CX646vsq.png" width="1546" height="922" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJSeVRxYkFwS1FvMk5lWHhDRUJ6bHJJNjJnc2UiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiR2NpUWRjRk5lb0ZBcEZ4aTFmV2xBVnh0ZzdjIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9SeVRxYkFwS1FvMk5lWHhDRUJ6bHJJNjJnc2UvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/RyTqbApKQo2NeXxCEBzlrI62gse/?mount_node_token=GciQdcFNeoFApFxi1fWlAVxtg7c&mount_point=docx_image" data-width="1886" data-height="1138" data-lark-image-uri="drivetoken://RyTqbApKQo2NeXxCEBzlrI62gse" data-lark-image-width="1886" data-lark-image-height="1138"></div>
<div class="ace-line ace-line old-record-id-VKoZdWOvpo4Pchx3FrZlWb8Kgte">aevatar.ai仪表板提供了低代码/无代码的可视化编排工具，帮助用户轻松设计和监控复杂的工作流。这是降低技术门槛的重要一步，使几乎任何人都能快速上手创建和个性化AI智能体。</div>
<ol class="ace-line ace-line old-record-id-H68tdEfiRoX4NSxTcvBlfjCVg3c" start="5">
<li data-list="number">
<h3 class="heading-3"><strong>安全性与可扩展性</strong></h3>

</li>

</ol>
<div class="image-uploaded gallery old-record-id-O46DddB39ows10xTOg4lcRUQgMe" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"2265db8f-c0b2-4366-899d-a02339b8f0c5","height":308.2010050251256,"width":538,"currHeight":308.2010050251256,"currWidth":538,"natrualHeight":456,"natrualWidth":796,"pluginName":"imageUpload","scale":0.6758793969849246,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FZ41vbVrLioRnCNx3GXOlBilRg1c%2F%3Fpreview_type%3D16","file_token":"Z41vbVrLioRnCNx3GXOlBilRg1c","image_type":"image/jpeg","size":46603,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/vision-and-goals-fig-6.D1omhWmI.jpeg" width="1247" height="720" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJaNDF2YlZyTGlvUm5DTngzR1hPbEJpbFJnMWMiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiTzQ2RGRkQjM5b3dzMTB4VE9nNGxjUlVRZ01lIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9aNDF2YlZyTGlvUm5DTngzR1hPbEJpbFJnMWMvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/Z41vbVrLioRnCNx3GXOlBilRg1c/?mount_node_token=O46DddB39ows10xTOg4lcRUQgMe&mount_point=docx_image" data-width="796" data-height="456" data-lark-image-uri="drivetoken://Z41vbVrLioRnCNx3GXOlBilRg1c" data-lark-image-width="796" data-lark-image-height="456"></div>
<div class="ace-line ace-line old-record-id-Zew8dA30JoSoEuxTRsllTqo2g5e">基于云原生的DevSecOps和微服务架构，aevatar.ai提供了弹性扩展和高并发处理能力，同时确保系统安全性和合规性，满足企业级用户需求。</div>
<ol class="ace-line ace-line old-record-id-Gc1Jdh7DVoPs3yxz4DGlVIUGgge" start="4">
<li data-list="number">
<h1 class="heading-1">架构</h1>

</li>

</ol>
<div class="ace-line ace-line old-record-id-OTANd6E7EondiZxJ66DlvoGugWg"><strong>aevatar.ai</strong> 由三个主要组件组成：aevatar框架、aevatar Station和aevatar Agents。它们协同工作，管理多个AI智能体的整个生命周期——从创建到部署再到持续运营</div>
<div class="image-uploaded gallery old-record-id-PNPsdZGVVorf7GxNaNglq35vgPf" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"3398bed5-9142-4e91-a519-fe41ed001756","height":644,"width":1626,"currHeight":644,"currWidth":1626,"natrualHeight":644,"natrualWidth":1626,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FNgy6b1G2xo7tGtxWaqhl2ltegGf%2F%3Fpreview_type%3D16","file_token":"Ngy6b1G2xo7tGtxWaqhl2ltegGf","image_type":"image/jpeg","size":87739,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/architecture-fig-1.BGaMsfuz.jpeg" width="1430" height="565" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJOZ3k2YjFHMnhvN3RHdHhXYXFobDJsdGVnR2YiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiUE5Qc2RaR1ZWb3JmN0d4TmFOZ2xxMzV2Z1BmIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9OZ3k2YjFHMnhvN3RHdHhXYXFobDJsdGVnR2YvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/Ngy6b1G2xo7tGtxWaqhl2ltegGf/?mount_node_token=PNPsdZGVVorf7GxNaNglq35vgPf&mount_point=docx_image" data-width="1626" data-height="644" data-lark-image-uri="drivetoken://Ngy6b1G2xo7tGtxWaqhl2ltegGf" data-lark-image-width="1626" data-lark-image-height="644"></div>
<ol class="ace-line ace-line old-record-id-YaNgdP8c3oF3noxVRNTl29VZghA" start="1">
<li data-list="number">
<h2 class="heading-2">aevatar框架</h2>

</li>

</ol>
<div class="ace-line ace-line old-record-id-EihadD3ulofUeYxl4c6lb5Omg0c"><strong>aevatar</strong>框架旨在支持AI智能体和事件溯源机制，提供模块化架构以实现可扩展性和可维护性。它利用依赖注入和观察者模式等设计模式来增强灵活性和可扩展性。</div>
<div class="image-uploaded gallery old-record-id-TIevdFE1Go1ic6xBPWalKWAzgxf" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"dd32d252-a146-4895-a458-e8a5430d05d5","height":1792,"width":1614,"currHeight":1792,"currWidth":1614,"natrualHeight":1792,"natrualWidth":1614,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FCkPibpD9WoqwYXxiEF0lVthwgHh%2F%3Fpreview_type%3D16","file_token":"CkPibpD9WoqwYXxiEF0lVthwgHh","image_type":"image/jpeg","size":199180,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/architecture-fig-2.6-a_O059.jpg" width="1412" height="1582" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJDa1BpYnBEOVdvcXdZWHhpRUYwbFZ0aHdnSGgiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiVElldmRGRTFHbzFpYzZ4QlBXYWxLV0F6Z3hmIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9Da1BpYnBEOVdvcXdZWHhpRUYwbFZ0aHdnSGgvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/CkPibpD9WoqwYXxiEF0lVthwgHh/?mount_node_token=TIevdFE1Go1ic6xBPWalKWAzgxf&mount_point=docx_image" data-width="1614" data-height="1792" data-lark-image-uri="drivetoken://CkPibpD9WoqwYXxiEF0lVthwgHh" data-lark-image-width="1614" data-lark-image-height="1792"></div>
<div class="ace-line ace-line old-record-id-GZRddQWHpoXVQNxsmWmljkbTgjb"><strong>设计原则</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-EJ5FdYrUboPhJyxNm0hl8uhLg9c" data-list="bullet"><strong>模块化</strong>：框架设计为模块化，允许开发者根据需要添加或移除组件。</li>
<li class="ace-line ace-line old-record-id-CHcddBeSlojtrRx5K0KlbDRngtf" data-list="bullet"><strong>可扩展性</strong>：新功能可以通过插件添加，而无需修改核心框架。</li>
<li class="ace-line ace-line old-record-id-GoXhdCAI3o3GbSxYl2fl21mkg9g" data-list="bullet"><strong>关注点分离</strong>：每个组件都有特定的职责，提升可维护性和可读性。</li>

</ul>
<div class="ace-line ace-line old-record-id-Vv3AdqueUoMQWKxEqqtldj4GgLg">该框架为开发AI智能体和事件溯源应用提供了灵活的架构，通过其模块化设计实现轻松集成和扩展。通过遵循设计原则和模式，框架确保在添加新功能时仍能保持可扩展性和可维护性。</div>
<ol class="ace-line ace-line old-record-id-EXRidnOp6o6GufxB6mglhIPqgkg" start="2">
<li data-list="number">
<h2 class="heading-2">核心组件概述</h2>

</li>

</ol>
<div class="ace-line ace-line old-record-id-UxYHd21IQo3Z07x9uo1lD1oMgSf"><strong>参与者模型（Actor Model）</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-PvDOdSfGPow4X8xgriHl6UYOgVd" data-list="bullet">负责管理分布式参与者（Grain）的生命周期和通信，为每个智能体提供有状态且可重放的执行环境，确保系统的高并发性和可扩展性。</li>

</ul>
<div class="ace-line ace-line old-record-id-XD1kdxBTaoRwCGxEsiHlhc7Rg1e"><strong>GAgent</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-YkdydO8cjovY7MxLJ4bl4cgDgWd" data-list="bullet">每个子模块（如Telegram、Twitter、MicroAI、SocialAgent等）都是一个独立的GAgent，针对不同平台或场景实现特定的智能体逻辑，支持跨平台扩展。</li>

</ul>
<div class="ace-line ace-line old-record-id-ZQSidMej6o0DRHxQoz2lJVLygWf"><strong>事件溯源（Event Sourcing）</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-XYLedAcpXo39rrxAgk4llGD4gxc" data-list="bullet">提供日志存储、事件重放和快照管理的核心功能。支持多种后端存储选项（如MongoDB和Redis），并确保系统的可追溯性和审计能力。</li>
<li class="ace-line ace-line old-record-id-K6GOdIKhqof4TqxdPJjl1DNvgMf" data-list="bullet">所有关键的智能体事件（如接收的消息、状态更新、模型推理输出）都可以持久化，提供重放和审计能力。</li>

</ul>
<div class="ace-line ace-line old-record-id-UB37dXNSloz5lDxxMvOleD3pgbh"><strong>CQRS（命令查询职责分离）</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Hanqd6ANsoT63exGP1vlMTSjgVe" data-list="bullet">对外提供REST/gRPC接口，并通过读写分离架构支持高效的内部数据查询和索引。结合Elasticsearch等解决方案，实现大规模数据的快速检索。</li>
<li class="ace-line ace-line old-record-id-LmPNd5kdUoksxYxOhHGlRbj5gCf" data-list="bullet">读写分离：系统可以独立处理智能体状态变更的写请求（事件）和外部查询接口。</li>
<li class="ace-line ace-line old-record-id-K1ysdLWCSoArgMxvvZIlRpLvgLd" data-list="bullet">结合Elasticsearch/MongoDB实现快速检索和多维度查询。</li>

</ul>
<div class="ace-line ace-line old-record-id-QVDUddXv2oqdcEx8Ichl3SCFgLc"><strong>aevatar仪表板</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-U44yd4winoUZZ5xbU4ylBhCtghb" data-list="bullet">图形化管理工具，允许用户通过低代码/无代码方式配置多智能体协作流程、监控事件流并编辑业务逻辑。这显著降低了开发门槛，尤其对非技术用户友好。</li>

</ul>
<ol class="ace-line ace-line old-record-id-M3Mpd0eTYohYxZxxrBnltWTSg1T" start="3">
<li data-list="number">
<h2 class="heading-2">GAgent多智能体协作模型</h2>

</li>

</ol>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-GN4Advxo0oytj5xfsiKl4m6Og6g" data-list="bullet">采用<code>GAgentBase<TState, TEvent></code>作为抽象基类，智能体可以继承并实现自己的业务处理方法。</li>

</ul>
<div class="image-uploaded gallery old-record-id-LuEFdgGL5oE5KNxCIgClJLkMgOh" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"2d128b38-fcc1-4cd2-a7b5-33f140ba553d","height":403.72348033373066,"width":582,"currHeight":403.72348033373066,"currWidth":582,"natrualHeight":1164,"natrualWidth":1678,"pluginName":"imageUpload","scale":0.3468414779499404,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FJlSdbF6hcoTOP4xXnl3lpKY1gAg%2F%3Fpreview_type%3D16","file_token":"JlSdbF6hcoTOP4xXnl3lpKY1gAg","image_type":"image/jpeg","size":135355,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/architecture-fig-3.dysNrPJ5.jpeg" width="1429" height="845" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJKbFNkYkY2aGNvVE9QNHhYbmwzbHBLWTFnQWciLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiTHVFRmRnR0w1b0U1S054Q0lnQ2xKTGtNZ09oIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9KbFNkYkY2aGNvVE9QNHhYbmwzbHBLWTFnQWcvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/JlSdbF6hcoTOP4xXnl3lpKY1gAg/?mount_node_token=LuEFdgGL5oE5KNxCIgClJLkMgOh&mount_point=docx_image" data-width="1678" data-height="1164" data-lark-image-uri="drivetoken://JlSdbF6hcoTOP4xXnl3lpKY1gAg" data-lark-image-width="1678" data-lark-image-height="1164"></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-QGzGdZdyTo83Wjx9nurlo3aWgid" data-list="bullet"><strong>GAgent：</strong>管理组内多个智能体的订阅、消息路由和事件协调，支持组内广播、点对点或基于树的事件传输。</li>

</ul>
<ol class="ace-line ace-line old-record-id-Dex0dqr9YoPxYWxzA2OlfgrCgid" start="4">
<li data-list="number">
<h2 class="heading-2">多大型语言模型（LLM）编排</h2>

</li>

</ol>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-QZK2dZyQdoUYMQx0QYSlJw4qgnd" data-list="bullet"><strong>多管齐下</strong>：aevatar.ai通过<code>AIService</code>和<code>AutoGen</code>机制集成对多个LLM（如GPT-4、Claude、Llama2等）的访问。</li>
<li class="ace-line ace-line old-record-id-WxDIdOg0QohqOOxMTOIlxilKgdZ" data-list="bullet"><strong>调度策略</strong>：根据任务类型、资源成本、复杂性等维度动态决定调用哪些模型。</li>
<li class="ace-line ace-line old-record-id-WiNPdlU3qoYSF5xtp1qlqEzwgYc" data-list="bullet"><strong>模型适配层</strong>：在框架层面支持连接更多第三方或私有模型，为企业提供定制化的多语言模型管理。</li>

</ul>
<ol class="ace-line ace-line old-record-id-Zax6dOfj2obrSLxBmVglAWsbgab" start="5">
<li data-list="number">
<h2 class="heading-2">云原生部署与安全合规</h2>

</li>

</ol>
<div class="image-uploaded gallery old-record-id-OQledy3pco5KI9xr9dmlhQYAgIu" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"7cf36efa-c403-4f81-a358-b6ac225053ba","height":1030,"width":2452,"currHeight":1030,"currWidth":2452,"natrualHeight":1030,"natrualWidth":2452,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FWtGkbQpoZo1anJxKXO1l1tEzgSf%2F%3Fpreview_type%3D16","file_token":"WtGkbQpoZo1anJxKXO1l1tEzgSf","image_type":"image/jpeg","size":320357,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/architecture-fig-4.CNQr0aKR.jpeg" width="1431" height="601" data-single-block="true" data-snapshot="eyJ0eXBlIjoiaW1hZ2UiLCJwYXJlbnRfaWQiOiJWWUZHZHJDd0dvRjJsNHg3eTlobGVWQkFnVWIiLCJjb21tZW50cyI6W10sInJldmlzaW9ucyI6W10sImxvY2tlZCI6ZmFsc2UsImhpZGRlbiI6ZmFsc2UsImF1dGhvciI6IjczNzIwMjk1NzQxNjAzNjc2NDciLCJhbGlnbiI6ImNlbnRlciIsImltYWdlIjp7InRva2VuIjoiV3RHa2JRcG9abzFhbkp4S1hPMWwxdEV6Z1NmIiwibWltZVR5cGUiOiJpbWFnZS9qcGVnIiwic2l6ZSI6MzIwMzU3LCJzY2FsZSI6MSwid2lkdGgiOjI0NTIsImhlaWdodCI6MTAzMCwibmFtZSI6ImRldm9wcy1tb2RpZmllZC5qcGciLCJjcm9wIjpbMCwwLDAsMF0sInJvdGF0aW9uIjowLCJjYXB0aW9uIjp7InRleHQiOnsiYXBvb2wiOnsibmV4dE51bSI6MCwibnVtVG9BdHRyaWIiOm51bGx9LCJpbml0aWFsQXR0cmlidXRlZFRleHRzIjp7ImF0dHJpYnMiOm51bGwsInRleHQiOm51bGx9fX19fQ==" data-suite="eyJmaWxlVG9rZW4iOiJXdEdrYlFwb1pvMWFuSnhLWE8xbDF0RXpnU2YiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiT1FsZWR5M3BjbzVLSTl4cjlkbWxoUVlBZ0l1Iiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9XdEdrYlFwb1pvMWFuSnhLWE8xbDF0RXpnU2YvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/WtGkbQpoZo1anJxKXO1l1tEzgSf/?mount_node_token=OQledy3pco5KI9xr9dmlhQYAgIu&mount_point=docx_image" data-width="2452" data-height="1030" data-lark-image-uri="drivetoken://WtGkbQpoZo1anJxKXO1l1tEzgSf" data-lark-image-width="2452" data-lark-image-height="1030"></div>
<div class="ace-line ace-line old-record-id-M4OWdCWlLo0ng3xqg4nlolxqgGg"><strong>Kubernetes部署</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-AcjXdnd1xo5hRDxwghmlIo15gEg" data-list="bullet">Orleans Silo和智能体服务可以容器化，支持自动扩展（HPA）、服务发现和弹性负载均衡。</li>

</ul>
<div class="ace-line ace-line old-record-id-PZDbdDbTYoFeuOxZIwPlzq3kgdh"><strong>DevSecOps & GitOps</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-OZCjdfTT3oEcBExuz6RlteLmgj8" data-list="bullet">提供容器镜像安全扫描、CI/CD集成和基础设施即代码（IaC）部署，确保应用安全性和可追溯性。</li>

</ul>
<div class="ace-line ace-line old-record-id-UqUydQgucowe4ExtEFPlB5KQgge"><strong>安全策略</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-AIZ7dWwrZoDhZTxuHBElUevxgZf" data-list="bullet">通过AuthServer和OAuth/OpenID系统进行身份验证，支持多租户和基于角色的访问控制（RBAC）。</li>

</ul>
<ol class="ace-line ace-line old-record-id-ABe0d1Ny8o6GwixM8jwl0Qsxg2f" start="5">
<li data-list="number">
<h1 class="heading-1">技术细节</h1>

</li>

</ol>
<div class="ace-line ace-line old-record-id-HQwCd0scYoAfmUx9IyHlTlUog3d"><strong>整体流程</strong>：</div>
<ol class="list-number1" start="1">
<li class="ace-line ace-line old-record-id-BxyqdT21voKOL5xmjEMlGfi6gtd" data-list="number"><strong>用户 → aevatar GAgent</strong>：捕获用户的消息或命令。</li>
<li class="ace-line ace-line old-record-id-GTyvdZLtIo9KvPxt2gOlgM2bg4F" data-list="number"><strong>GAgent → aevatar框架</strong>：将结构化事件传递给框架（多智能体协作和AI交互）。</li>
<li class="ace-line ace-line old-record-id-XrvZdltV2orxrjxdjPilvBdXgnf" data-list="number"><strong>aevatar框架 → 核心逻辑（RAG和LLM）</strong>：RAG和LLM解析请求。</li>
<li class="ace-line ace-line old-record-id-Tuwqds5yUo3qAxxqMAxlAYckgGe" data-list="number"><strong>核心逻辑（RAG和LLM）→ 外部服务或知识/记忆库</strong>：检索数据或调用专门操作。</li>
<li class="ace-line ace-line old-record-id-AQwxdlbPcowvahxSt0jlT9haglf" data-list="number"><strong>aevatar框架 → 输出代理</strong>：格式化并准备最终输出。</li>
<li class="ace-line ace-line old-record-id-FTWrdOHUSoXagOxtKCKlU7GxgEe" data-list="number"><strong>输出 → GAgent → 用户</strong>：用户收到响应。</li>

</ol>
<div class="image-uploaded gallery old-record-id-SCEidHdi6oM1h2xUsqjl1tO6g3b" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"9c566b73-2ed0-455d-a617-28542e0f989f","height":720,"width":1640,"currHeight":720,"currWidth":1640,"natrualHeight":720,"natrualWidth":1640,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FHdBcbYd3xoajE0xz4oRlpkdTg3c%2F%3Fpreview_type%3D16","file_token":"HdBcbYd3xoajE0xz4oRlpkdTg3c","image_type":"image/jpeg","size":86314,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/technical-fig-1.Czoewzuf.jpeg" width="1429" height="589" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJIZEJjYllkM3hvYWpFMHh6NG9SbHBrZFRnM2MiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiU0NFaWRIZGk2b00xaDJ4VXNxamwxdE82ZzNiIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9IZEJjYllkM3hvYWpFMHh6NG9SbHBrZFRnM2MvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/HdBcbYd3xoajE0xz4oRlpkdTg3c/?mount_node_token=SCEidHdi6oM1h2xUsqjl1tO6g3b&mount_point=docx_image" data-width="1640" data-height="720" data-lark-image-uri="drivetoken://HdBcbYd3xoajE0xz4oRlpkdTg3c" data-lark-image-width="1640" data-lark-image-height="720"></div>
<div class="ace-line ace-line old-record-id-S9xtdzqhZobZwDxxClnlLaqrgug"><strong>详细流程</strong>：</div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-AzrUdL1XMooqP4xnjTKlA7tcgkh" data-list="bullet"><strong>智能体创建与初始化</strong>：客户端请求GAgentFactory创建智能体，智能体使用StateLogEventStorage初始化状态，并通过StreamProvider设置订阅。</li>
<li class="ace-line ace-line old-record-id-Akexdct7toaWVjxTKS9lPKLEgfe" data-list="bullet"><strong>事件发布与处理</strong>：客户端（或其他系统）向智能体发布事件，智能体将事件追加到事件存储中，更新其内存状态，并在必要时发布到外部流。</li>
<li class="ace-line ace-line old-record-id-RDLQdg49Wok5nfxzRBKlXrLcgqf" data-list="bullet"><strong>状态恢复</strong>：在需要时，智能体从存储中检索快照，并应用所有后续事件，最终获得最新状态。</li>

</ul>
<div class="ace-line ace-line old-record-id-IwhUdJqQyo4TgKxvF7jlJ8qPgfc"><strong>显著特点与优势</strong>：</div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-NBtCdL5E5oqW9Jxk1pDlU8QXgxb" data-list="bullet"><strong>多智能体协作</strong>：系统可以将复杂任务拆分为更小的专门子任务，每个任务由适当的智能体处理。</li>
<li class="ace-line ace-line old-record-id-MQPGd1Nvjo93R0xLun2larA5gyh" data-list="bullet"><strong>动态流程</strong>：智能体按需激活和调用（虚拟参与者模型），允许在任务可拆分的场景中实现并发或并行调用。</li>
<li class="ace-line ace-line old-record-id-TB26dtWawoEspixt9iJlXr42gCb" data-list="bullet"><strong>与外部服务集成</strong>：知识模块可以无缝整合实时数据、领域文档或高级处理能力。</li>
<li class="ace-line ace-line old-record-id-JXjNdmwnfobkLrxafF1lD5BKgGg" data-list="bullet"><strong>检索增强生成（RAG）</strong>：智能体可以查询向量数据库或记忆存储，利用最新的上下文数据增强LLM或其他逻辑。</li>
<li class="ace-line ace-line old-record-id-V33YdKyQXoTjo8xkqsKlYFJGg2d" data-list="bullet"><strong>可扩展性与扩展性</strong>：每个组件都可以水平扩展，新的智能体或工具可以在不改变主要架构的情况下引入。</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-AqFSd2dAaoDwLXxKWBXlB2gkgSd">5.1 Orleans参与者模型与可扩展性</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-YDXldDP5po4EsCxo5dHl7kOZgsf" data-list="bullet"><strong>分布式参与者</strong>：每个智能体作为一个Grain，存储自己的状态和事件历史。Orleans处理调度和消息传递，消除了手动管理并发锁和网络通信的需求。</li>
<li class="ace-line ace-line old-record-id-RoLSdi1GYoA3LvxQVBwlHyCcglb" data-list="bullet"><strong>水平扩展</strong>：当系统需要处理更多对话或更高并发时，添加Silo节点可以扩展智能体实例并自动平衡负载。</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-War4dpsUHoLliuxTT5jl1XDBgWd">5.2 GAgentBase设计与事件驱动架构</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Eswhddbv5oDU5uxUXJsle7mYghh" data-list="bullet">
<div><strong>GAgentBase<TState, TStateLogEvent></strong>：</div>
<ul class="list-bullet2">
<li class="ace-line ace-line old-record-id-XqNfdSfCtoGnlNxsGFQlXLswg5e" data-list="bullet">继承<code>JournaledGrain<TState, StateLogEventBase<TStateLogEvent>></code>，天然具备事件溯源能力。</li>
<li class="ace-line ace-line old-record-id-ZvnLdQtPfojr97xBuhwleXGfggc" data-list="bullet">通过<code>PublishToAsync/SubscribeToAsync</code>等方法，允许智能体自由组合和交互，形成多对多或多级事件流拓扑。</li>

</ul>

</li>
<li class="ace-line ace-line old-record-id-PprBdM0AqojmMqxNz7Kljgvsg6e" data-list="bullet">
<div><strong>EventWrapper</strong>：</div>
<ul class="list-bullet2">
<li class="ace-line ace-line old-record-id-EAVAdAZSyoPeQexv26ElZCBjgGg" data-list="bullet">为所有事件添加ID、时间戳和上下文等元数据，便于审计和调试，避免传统“黑箱AI”问题。</li>

</ul>

</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-MBQxd9hJ0oQTrrx7PR3lSOETgRh">5.3 低代码/无代码编排与可视化</h2>
<div class="image-uploaded gallery old-record-id-H2GCdU54aofUK1xyL2jlRXwKgZf" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"29a2b265-54ac-4048-8bad-d63a875ed8a2","height":1664,"width":2560,"currHeight":1664,"currWidth":2560,"natrualHeight":1664,"natrualWidth":2560,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FXGnkbXG8Nov9OKxf2nplRZY9gLg%2F%3Fpreview_type%3D16","file_token":"XGnkbXG8Nov9OKxf2nplRZY9gLg","image_type":"image/jpeg","size":344115,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/technical-fig-2.Cb7FqvU2.jpeg" width="1429" height="929" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJYR25rYlhHOE5vdjlPS3hmMm5wbFJaWTlnTGciLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiSDJHQ2RVNTRhb2ZVSzF4eUwyamxSWHdLZ1pmIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9YR25rYlhHOE5vdjlPS3hmMm5wbFJaWTlnTGcvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/XGnkbXG8Nov9OKxf2nplRZY9gLg/?mount_node_token=H2GCdU54aofUK1xyL2jlRXwKgZf&mount_point=docx_image" data-width="2560" data-height="1664" data-lark-image-uri="drivetoken://XGnkbXG8Nov9OKxf2nplRZY9gLg" data-lark-image-width="2560" data-lark-image-height="1664"></div>
<div class="ace-line ace-line old-record-id-VCbxdAEi4ow7iPxq3TUlyDMpghc"> </div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-FJvodDbbooHMipxkEvrltjeRgrr" data-list="bullet"><strong>拖放流程设计</strong>：用户可以在仪表板上拖放智能体节点、配置事件路由并设置模型策略，而无需编写复杂的后端代码。</li>
<li class="ace-line ace-line old-record-id-Uq0Zdga6QobxNOxWEl4la4bkgyb" data-list="bullet"><strong>实时监控与日志重放</strong>：集成事件溯源日志，允许通过aevatar仪表板查看任何时刻的事件序列或智能体状态，帮助业务优化和维护故障排查。</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-ILgTd2IPRo0NXWxXB7zliWfbgQe">5.4 可观测性与监控</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Ue3xdUdsNoyjNuxkjUFlHCfAgHf" data-list="bullet"><strong>分布式追踪</strong>：集成OpenTelemetry、Jaeger或Zipkin，可视化跟踪跨智能体/Grain的调用链。</li>
<li class="ace-line ace-line old-record-id-P6xjdP2S4omJEwxey47lcRqXgWk" data-list="bullet"><strong>指标与警报</strong>：收集系统指标（如QPS、延迟、错误率等），并基于Prometheus/Grafana实现实时警报。</li>
<li class="ace-line ace-line old-record-id-PzqVdZOeHo0Q48x5z4Ql8XYugXg" data-list="bullet"><strong>Orleans仪表板</strong>：可选的内置Orleans仪表板，显示运行时数据（如Grain激活数量、消息处理速率等）。</li>

</ul>
<ol class="ace-line ace-line old-record-id-EUAPdEdT5oE672xDNA2lS3H3g3c" start="6">
<li data-list="number">
<h1 class="heading-1">关键特性</h1>

</li>

</ol>
<h2 class="heading-2 ace-line old-record-id-VZBmd2j9Qo1KHZxiKISle323gwb">6.1 多模型并行处理/动态切换</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-EbahdWXqsoCm0xxicEGlwX3VgZd" data-list="bullet">根据业务需求自动（或手动）在不同LLM之间切换。</li>
<li class="ace-line ace-line old-record-id-Dp1sdTJ5SoyzhNxoD7Jl205PgAd" data-list="bullet">能够分配多个模型同时处理子任务并合并结果。</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-PatQdoifVoG7sHxAc9nltFJKgjb">6.2 高级任务编排与协作</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-CzkKdcB6Do0drOxcF8clJIyZgCU" data-list="bullet">GAgent提供基于事件的协作机制，允许多个智能体并行处理复杂业务流程。</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-CyZId4BJzoBUbwx8i6Tltsl6gBf">6.3 RAG集成</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Smnbdta7xoYXs1xnt6tlsqCLgmX" data-list="bullet">连接向量数据库/文档搜索引擎，使智能体能够从大规模知识库中检索并生成答案。</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-DELFdWM2toSErIxAXp8lgJGvgrd">6.4 跨平台扩展</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Dg0QddTMaoYjRsxKFtLlGpJfgRg" data-list="bullet">通过插件与Telegram、X、Slack等平台集成，快速构建多渠道聊天/通信场景。</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-Neq0dU5moos1IPxMkJzl20KAg9e">6.5 开发者与非开发者友好</h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Lgfid7xc1o1AG1xlgjZly2dugVe" data-list="bullet"><strong>面向开发者</strong>：提供可编程的插件框架。</li>
<li class="ace-line ace-line old-record-id-NU3vdb8TmoGJIgxOfZHl0Cq0gag" data-list="bullet"><strong>面向非技术人员</strong>：通过仪表板的低代码/无代码管理，快速上手。</li>

</ul>
<ol class="ace-line ace-line old-record-id-NKh6dLYRVo5HTLxxA0MlITAUgkd" start="7">
<li data-list="number">
<h1 class="heading-1">用例</h1>

</li>

</ol>
<div class="image-uploaded gallery old-record-id-G4HKdv0FeoBBJnxhdSylIkhfg5e" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"7b188ecf-6374-4542-97dd-bef51a7050a5","height":427.35373317013466,"width":764,"currHeight":427.35373317013466,"currWidth":764,"natrualHeight":914,"natrualWidth":1634,"pluginName":"imageUpload","scale":0.4675642594859241,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FORGgb01zKoJD9Bxl5thlcCl0gIO%2F%3Fpreview_type%3D16","file_token":"ORGgb01zKoJD9Bxl5thlcCl0gIO","image_type":"image/jpeg","size":114990,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/use-cases-fig-1.BOS_xRKt.jpeg" width="1431" height="754" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJPUkdnYjAxektvSkQ5QnhsNXRobGNDbDBnSU8iLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiRzRIS2R2MEZlb0JCSm54aGRTeWxJa2hmZzVlIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9PUkdnYjAxektvSkQ5QnhsNXRobGNDbDBnSU8vP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/ORGgb01zKoJD9Bxl5thlcCl0gIO/?mount_node_token=G4HKdv0FeoBBJnxhdSylIkhfg5e&mount_point=docx_image" data-width="1634" data-height="914" data-lark-image-uri="drivetoken://ORGgb01zKoJD9Bxl5thlcCl0gIO" data-lark-image-width="1634" data-lark-image-height="914"></div>
<div class="ace-line ace-line old-record-id-LdF0dJnqAouvHWxSLTxlwwLIgeh"> </div>
<h2 class="heading-2 ace-line old-record-id-ST2hdkNQboXCrXxkeAblYYXtgMg"><strong>7.1 多智能体协作/自动化</strong></h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Pe2ddSTgaog6eJxVEAQle0z9gJd" data-list="bullet">
<div><strong>跨部门智能体</strong>:</div>
<ul class="list-bullet2">
<li class="ace-line ace-line old-record-id-GaLodWOiIowbMqxNvOtlukLRgzh" data-list="bullet">Finance Agent（财务智能体）自动处理报销工作流程</li>
<li class="ace-line ace-line old-record-id-KQL8dlysboDx5XxYFEulGXBAgze" data-list="bullet">HR Agent（人力资源智能体）负责简历/履历筛选</li>
<li class="ace-line ace-line old-record-id-ChMod559JoWXJrxBALalNj6xgae" data-list="bullet">IT Agent（IT智能体）管理工单</li>
<li class="ace-line ace-line old-record-id-X1v9d6XUFok1Ojxezpaly0Fvgpf" data-list="bullet">各智能体可相互通知事件或汇总审批结果</li>

</ul>

</li>
<li class="ace-line ace-line old-record-id-B7uXdZEhyoH8iwxGunKlRHOkgNd" data-list="bullet">
<div><strong>低代码管理</strong>:</div>
<ul class="list-bullet2">
<ul class="list-bullet2">
<li class="ace-line ace-line old-record-id-TphSdasgTokveNxsxbblrRK9g7g" data-list="bullet">用户可在 aevatar Dashboard 中配置流程并设置触发条件</li>
<li class="ace-line ace-line old-record-id-XjB5dexEioPfRhx3u02lVwltgUg" data-list="bullet">智能体在接到指令后，会根据事件流程自动执行</li>

</ul>

</ul>
<div class="image-uploaded gallery old-record-id-T9J2dxleUokxCdxgnnXllvfHgSb" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"b65d49b6-f1d8-4c30-bd96-ba66f97bb532","height":197.970802919708,"width":382,"currHeight":197.970802919708,"currWidth":382,"natrualHeight":710,"natrualWidth":1370,"pluginName":"imageUpload","scale":0.27883211678832115,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FDhg3bMCnnoXLFuxdZqAlr3sCgFe%2F%3Fpreview_type%3D16","file_token":"Dhg3bMCnnoXLFuxdZqAlr3sCgFe","image_type":"image/jpeg","size":70771,"comments":[]}]}"><img src="https://whitepaper.aevatar.ai/assets/use-cases-fig-2.BYIsk07w.jpeg" width="1194" height="621" data-single-block="true" data-snapshot="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" data-suite="eyJmaWxlVG9rZW4iOiJEaGczYk1Dbm5vWExGdXhkWnFBbHIzc0NnRmUiLCJvYmpUeXBlIjoiZG9jeCIsIm9ialRva2VuIjoiVDlKMmR4bGVVb2t4Q2R4Z25uWGxsdmZIZ1NiIiwib3JpZ2luU3JjIjoiaHR0cHM6Ly9pbnRlcm5hbC1hcGktZHJpdmUtc3RyZWFtLXNnLmxhcmtzdWl0ZS5jb20vc3BhY2UvYXBpL2JveC9zdHJlYW0vZG93bmxvYWQvcHJldmlldy9EaGczYk1Dbm5vWExGdXhkWnFBbHIzc0NnRmUvP3ByZXZpZXdfdHlwZT0xNiJ9" data-src="https://internal-api-drive-stream-sg.larksuite.com/space/api/box/stream/download/all/Dhg3bMCnnoXLFuxdZqAlr3sCgFe/?mount_node_token=T9J2dxleUokxCdxgnnXllvfHgSb&mount_point=docx_image" data-width="1370" data-height="710" data-lark-image-uri="drivetoken://Dhg3bMCnnoXLFuxdZqAlr3sCgFe" data-lark-image-width="1370" data-lark-image-height="710"></div>
<div class="ace-line ace-line old-record-id-TmJodfpRPoezMPxxgg2lZw6Egze"> </div>

</li>

</ul>
<div class="ace-line ace-line old-record-id-BVaidBng3oDcuXxoiJwlQMcMgVb">以上基于任务的编排，使 G‐agents（G-智能体）既能独立运行又可无缝协同，充分利用各自的专业能力，更高效地完成复杂目标。</div>
<h2 class="heading-2 ace-line old-record-id-E5MXdwRIcoRHBgxNQFTlog9Fguc"><strong>7.2 多语言客服/社交媒体智能体</strong></h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-CZOidnGFZo0TWax3KwwlW45VgIN" data-list="bullet">
<div><strong>Telegram/X 适配</strong>:</div>
<ul class="list-bullet2">
<li class="ace-line ace-line old-record-id-U28Xd0IS6oIQi3xfSg9ltMq9ghd" data-list="bullet">可部署多个智能体进行交互</li>
<li class="ace-line ace-line old-record-id-JRgbdTHBEoeOyCxQGrklCPHjgZc" data-list="bullet">支持多种通信渠道与多语言服务</li>

</ul>

</li>

</ul>
<h2 class="heading-2 ace-line old-record-id-T9iWdSfc7oeS7DxiR1zlx2vUg6g"><strong>7.3 区块链/金融/制造业中的行业应用</strong></h2>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-F0YldrhiVoSx8zxH9FNlgnZdgu3" data-list="bullet">
<div><strong>智能合约分析</strong>:</div>
<ul class="list-bullet2">
<li class="ace-line ace-line old-record-id-RMM2dzf7eovzBpxKTOYl0l3qgK9" data-list="bullet">智能体从区块链中获取智能合约文本</li>
<li class="ace-line ace-line old-record-id-TFZNd7CbZoYUgWxGIJvlWy5lgUg" data-list="bullet">执行风险检测和语言解析</li>
<li class="ace-line ace-line old-record-id-D77Tdc1Vgo5hTtxqQUQlqow9gHe" data-list="bullet">发现异常时，向运营智能体推送警示</li>

</ul>

</li>
<li class="ace-line ace-line old-record-id-Nj7cdJOw7oQeDWxU6P2lS4V2g6e" data-list="bullet">
<div><strong>数据驱动的决策</strong>:</div>
<ul class="list-bullet2">
<li class="ace-line ace-line old-record-id-F9DpduwffoL3fhxCkRXlqiDQgTb" data-list="bullet">在制造行业，智能体可实时分析 IoT 传感器数据</li>
<li class="ace-line ace-line old-record-id-Eb2Xdy5o8ojtn1xFAVOlZKfjgyh" data-list="bullet">并结合 LLM（大型语言模型）提供故障诊断或生产建议</li>

</ul>

</li>

</ul>
<ol class="ace-line ace-line old-record-id-UPTZdVdTwoCANRxVZJjlYkKzgqc" start="8">
<li data-list="number">
<h1 class="heading-1">当前的 AI 框架格局</h1>

</li>

</ol>
<h2 class="heading-2 ace-line old-record-id-ENY9dYfIto0Xn8xZNZWlKj54gYf">8.1 对比表：aevatar intelligence x ElizaOS x G.A.M.E</h2>
<div>
<table class="ace-table" data-ace-table-col-widths="183;242;190;238"><colgroup><col width="183"><col width="242"><col width="190"><col width="238"></colgroup>
<tbody>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-E9xndNBAGoC7F7xVIEJl4lwgg4f"><strong>对比方向</strong></div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-EJoCdHJx9odAeuxI9QdlfU2BgTh"><strong>aevatar intelligence</strong></div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-LIMgd8KjXo2XjgxhblelxCfngNf"><strong>ElizaOS</strong></div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-OgE2dRicwom20oxJVXbleCdBgDg"><strong>G.A.M.E</strong></div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-E5kWdWdMzo9ZQixC0djliyptgnE"><strong>主要优势(Key Strength)</strong></div>
<div class="ace-line ace-line old-record-id-E9xJdmfCZoKgpMxM95Qlz61dgOd"> </div>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Mk5Mdc4Teo3V6jx7AeolMpTUgKe" data-list="bullet">用户无需编写代码或少量代码即可使用</li>
<li class="ace-line ace-line old-record-id-QG6ddrFdDoCMplxfo96lHoOCgfd" data-list="bullet">各个Agent能按照需求、复杂度各自以不同LLM驱动并协作</li>
<li class="ace-line ace-line old-record-id-YsWhdZ7CAoB8jAxwx0sllLMXgwb" data-list="bullet">设定Agent协作的逻辑、协作的流程</li>
<li class="ace-line ace-line old-record-id-C09sdpt07ofijBxDx7ZlrdJCg7C" data-list="bullet">能回放事件，以分析Agent的工作流程</li>

</ul>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-UAzPdivQ6oDY1AxlbCVlIcgbg5e" data-list="bullet">功能集与插件集不断增长</li>
<li class="ace-line ace-line old-record-id-VPyRdumesoKgYyx4VKdlflWsgde" data-list="bullet">完全可定制并可控</li>

</ul>
<div class="ace-line ace-line old-record-id-RBLFd2dLvoKdpAxLeQrloVxPgCc"> </div>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-TeD3dOFaXo6eqrxzHivlsFa4g6b" data-list="bullet">低代码、低复杂度的上线方式</li>

</ul>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-BIgUdaSoSokugKxx8oBlBaz9gNg"><strong>能力(Capabilities)</strong></div>
<div class="ace-line ace-line old-record-id-B3HzdCyFnokiCKxnjuDl4hLWgvd"> </div>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-DkkPdTVe8o6mD5xF7MjlFU0Cggg" data-list="bullet">不同语言模型驱动的Agent们可以同时协作</li>
<li class="ace-line ace-line old-record-id-SzO7dbX3Xog1wWxD2VMle5MagRd" data-list="bullet">Agent和工作流可以轻易创建、复制、调整、扩展</li>

</ul>
<div class="ace-line ace-line old-record-id-HbKFdfck2omEAwxel0tlBywygIc"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-BMm1dWQZeo60CexfViqlfnD6gFe">各个Agent只能同时使用同一种语言模型协作，之间目前无法交互、协作、集体做决策</div>
<div class="ace-line ace-line old-record-id-DIPAd0zCtoF4sixD3eXlv3vrgah"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Ro2gdw8pXon0M0xJkLtl1VBngNc">各个Agent只能同时使用同一种语言模型协作，之间目前无法交互、协作、集体做决策</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-VBrCd6x55otpTmxg5eBlrcXlg2g"><strong>多语言模型(LLM)编排</strong></div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-DlLPddKNBoYI0exPx4clpIRJgoh">通过 AutoGen 进行多LLM编排，适用于在任何类型的应用中进行复杂推理和决策</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-NEIHdU9sCoP1etxWjNtlfr2ogLg">仅限于单一模型的 API 集成，没有多LLM自动化，不具备跨应用的灵活性</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-MSdMdLjOOo95gHx45JIlzWZZgBf">针对虚拟世界中的自然语言交互进行了优化，不适用于通用型应用</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-ALyKdaimlotLkhxYauYlg0iPg2f"><strong>框架设计(Design)</strong></div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-VRctdkvfnoXwJfxMgz2l6PJAgUd">模块化＋延展性插件＋动态集群管理系统</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Nakjdw6FEo6spxxsQGtlNb7Dg6b">模块化＋延展性插件系统</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-FR1PdFvfQouC1VxyeFClZIVqgRg">模块化＋环境无关性 (environment agnostic)</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-UhpedoEl3oF4AgxHFSglHRi0gWh"><strong>目标用户(Target Audience)</strong></div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-UOy2dfVNBoSaLSxxnhQlNgP0gyf">终端用户与技术型开发者均适用</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-FhBYdQG15oySt4xNAkYlPsr0gWg">技术型开发者</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-KPcgdbrw9omTdNxCbOJln0dUglb">非技术型用戶</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-EXm5d4LYsorfH0xRm5VlVo9Igbc"><strong>编程语言(Coding Language)</strong></div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-US4Id1wGkoOMulx6lU4lpCHUgpl">无代码或低代码（no-code or low-code）</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-NeYvdHGNHoYSscxHmcMlv9hlgDd">TypeScript/JavaScript</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-ZIV9dlxhIoQYsMx6vuvljIU5gXb">低代码 (Low-code)</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-LbUNd3lK2o2bAIxYA06lDItzgwf"><strong>可扩展性 (Scalability)</strong></div>
<div class="ace-line ace-line old-record-id-I3sqdLzaQo52mUxMKnylVFQygje"> </div>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Fks4dno47o23VoxlimJlPWpWggf" data-list="bullet">使用 Orleans，一个结合了微服务和 Actor 模型的分布式框架，可针对大规模代理网络实现可扩展性和高可用性。</li>
<li class="ace-line ace-line old-record-id-WvM8d8GFRorxYdxZXcBlM2OdgSg" data-list="bullet">基于容器化部署，使用 Kubernetes 实现跨云能力、自动伸缩、高可用性和高并发。</li>

</ul>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-JHKpdTcF6ojiAZxI1MXlnW3kgOf" data-list="bullet">使用 Node.js，多进程架构，但缺少分布式编程模型。</li>

</ul>
<div class="ace-line ace-line old-record-id-SdDXdO3EwoRakmx7CmklNQUmgTh"> </div>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-UNF0dytqroM0E9xoM1LlaaQMgFd" data-list="bullet">依赖于 Photon 或 SpatialOS 之类的游戏专用后端来实现实时性能。</li>

</ul>
<div class="ace-line ace-line old-record-id-BlAZdTA6ToI7e4xS3cyl2BAzg7d"> </div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-EfbcdanwFoR3uLxTRSulTq0OgXc"><strong>用例(Use Cases)</strong></div>
<div class="ace-line ace-line old-record-id-WMfldjLDxo4RDwxtPOdlcxfigmb"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-NNiPdcXQHoHc4vx03AKlOh7Bgpb">为区块链和金融等行业中的通用、可扩展、多领域逻辑而构建</div>
<div class="ace-line ace-line old-record-id-KDaEdwBm2oFMWdxnfs2lbaeQgnd"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-JjFWd5Bu2oqcq6xSrmYlMspxgug">为较小的网页项目和社区驱动的原型开发而构建</div>
<div class="ace-line ace-line old-record-id-TZlGd4PYjovVgDxYIjalCbVDgPb"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Kwnkd8Fk4oFhwGxZWVSltbhtgah">为游戏和元宇宙场景（包含通证经济集成）而构建</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-LZhfdEfofoK58LxiC5YloSsQg9f"><strong>云原生 & DevOps (Cloud Native & DevOps)</strong></div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-LuaWdh3K7ogKUixTMDxliooEgRh">先进的云原生 Kubernetes 部署，通过 DevSecOps & GitOps 提供强大的安全性</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Y59VdayXfoRzYqxVtf4lE0Jrgmf">专注于速度，但缺乏广泛的自动化和合规机制</div>
<div class="ace-line ace-line old-record-id-IfkxdgwkyoNPbhxw5wUlaEHKgKg"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-QiRFdE372oxM9gxn62Ul8gH6gtg">专注于性能，但未提供完整的云原生工具</div>
<div class="ace-line ace-line old-record-id-Yc1cdEemxoJEBUxi7bUlCzbJg6f"> </div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-UleCdEFgco7pGfxtdQQlTri1gAh"><strong>维护 (Maintainability)</strong></div>
<div class="ace-line ace-line old-record-id-UPusdjNh7o8wMFxGByDla8Wrgti"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-DiysdmeDXoAZ4fxAMnulN3FRgje">Agent-as-a-Service 通过Plugin进行功能迭代，无须用户自己部署服务。优化了对Agent开发和部署的操作。</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-KsHfdBuOZomqfFxXhbuliU2mgEb">Supabase 借助其后端即服务（Backend-as-a-Service）平台，为开发运维提供易用的部署方式</div>
<div class="ace-line ace-line old-record-id-AqsWd2Yb8o8vvLx46bJlxTubgrc"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-ZCXPdRhawoG4xFxTXm4lRcm6gEe">未确定 —— 闭源</div>
<div class="ace-line ace-line old-record-id-Qip6dHS3aox3lrxwqyvlLd9qg7m"> </div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-NnStdJ0cqomLUExxnrOliLcZgBh"><strong>代码获取 (Code Access)</strong></div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-IYZvd0dzooy1Cax7kxJlSN6Rgwd">开源</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Du3WdHRG6oLvfzxDlM3lK8lPglf">开源</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-WN9ZdAr9eoyg7bxgqmblPUpVgHg">闭源（黑盒）</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-LFMadudFJob6z8x5TAvlerVJgQf"><strong>平台集成</strong> <strong>(Platform Integrations)</strong></div>
<div class="ace-line ace-line old-record-id-KHGydJNR3oQwWtxBocYlhiryg4c"> </div>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-G68nde7aLo15c4xHat6l8jZcgvb" data-list="bullet">Twitter</li>
<li class="ace-line ace-line old-record-id-CO5Ld7LAOoSyruxyXanlbYxGgmg" data-list="bullet">Telegram</li>

</ul>
<div class="ace-line ace-line old-record-id-PHjEdp9vrokQ9DxYigilAijqgTe"> </div>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-W3FCdmd85oyG46xlCx8lF1F7gHg" data-list="bullet">Discrod</li>
<li class="ace-line ace-line old-record-id-PnWsd2ZHNogeUExOKG5lAvvsgKh" data-list="bullet">Twitter</li>
<li class="ace-line ace-line old-record-id-QQZLdcvaooipxExnNdilP995gWg" data-list="bullet">Telegram</li>

</ul>

</td>
<td rowspan="1" colspan="1">
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-S6addOePdoBNfMxcPmFlIMJrgQg" data-list="bullet">Discrod</li>
<li class="ace-line ace-line old-record-id-TX4IdMzXEo7VV0xpSgZl1zZIgKP" data-list="bullet">Twitter</li>
<li class="ace-line ace-line old-record-id-JicOdIYx8oe3uUxI7fnlBE9Wgeh" data-list="bullet">Telegram</li>

</ul>

</td>

</tr>

</tbody>

</table>

</div>
<h2 class="heading-2 ace-line old-record-id-ZkojdkJOyom3edxazgkl4jE4gNe">8.2 技术与商业价值</h2>
<div class="ace-line ace-line old-record-id-Zs8ed5R2go8CxrxkUUWlsXU8g9e"><strong>强大的多语言模型协作</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-O1VWdsuTgo27qSxC33BlCXmpgCh" data-list="bullet">允许在同一个业务流程中调用多个大型语言模型</li>
<li class="ace-line ace-line old-record-id-ZYcFddgaxoQP8RxG0bUlZ1YCgKh" data-list="bullet">优化成本与性能</li>

</ul>
<div class="ace-line ace-line old-record-id-W4BmdcwGKoQNqJxs4m8lO9I5g7f"><strong>易用性</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-HEted2fQuoMUdwxBUDZlXauAgXd" data-list="bullet">aevatar Marketplace 提供低/无代码编辑器</li>
<li class="ace-line ace-line old-record-id-KQ3zdVWrIo9GvDx3yPOltkxzgbd" data-list="bullet">大幅缩短智能体开发与运营周期</li>

</ul>
<div class="ace-line ace-line old-record-id-NjGhdSrWWoYUJHxuayjlpB8vgth"><strong>高并发与可追溯性</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-JRFVdwlVfoztwaxHHdJlqkACgye" data-list="bullet">基于 Actor + Event Sourcing</li>
<li class="ace-line ace-line old-record-id-W6Ind4K3Lo7EPsxAa42lXeR6gzC" data-list="bullet">系统可轻松处理数万级并发操作</li>
<li class="ace-line ace-line old-record-id-PpEjdFj8Qo9TioxTvEqljwjfg0c" data-list="bullet">所有交互历史可回放并进行审计</li>

</ul>
<div class="ace-line ace-line old-record-id-KsxkdmKgeofwRoxeO3WloqcBg9f"><strong>合规与安全</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-CCsgdFV1io7LrJxpAGxlbPzKgSf" data-list="bullet">结合云原生 & DevSecOps & GitOps</li>
<li class="ace-line ace-line old-record-id-UpoidNMYhog92Zx0Tt1lPEaTgGb" data-list="bullet">在 Kubernetes 中实现自动化部署</li>
<li class="ace-line ace-line old-record-id-N44kdg7pBoDGQ3xoAyZlKfPSgYg" data-list="bullet">确保安全与合规要求</li>

</ul>
<ol class="ace-line ace-line old-record-id-GB1Jdi9jhoRMxQxWsI2lgdSQgHg" start="9">
<li data-list="number">
<h1 class="heading-1">路线图</h1>

</li>

</ol>
<h2 class="heading-2 ace-line old-record-id-ThckdsUGboFpfjxXNWElAmTugYb">9.1 短期规划</h2>
<div class="image-uploaded gallery old-record-id-BIvFdVcBSo77KYxbJIElALMIg3f" data-type="image" data-ace-gallery-json="{"items":[{"uuid":"f78f1e08-484d-4cd4-bc57-33615b6f7d71","height":1600,"width":1954,"currHeight":1600,"currWidth":1954,"natrualHeight":1600,"natrualWidth":1954,"pluginName":"imageUpload","scale":1,"src":"https%3A%2F%2Finternal-api-drive-stream-sg.larksuite.com%2Fspace%2Fapi%2Fbox%2Fstream%2Fdownload%2Fpreview%2FMyolbqptoog3ZrxuiymlY8dKg7g%2F%3Fpreview_type%3D16","file_token":"Myolbqptoog3ZrxuiymlY8dKg7g","image_type":"image/png","size":466136,"comments":[]}]}">
<div data-page-id="Oyb0dwIoUo1mvoxg0PGlLFK6g7d" data-lark-html-role="root" data-docx-has-block-data="true">
<table class="ace-table" data-ace-table-col-widths="177;235;255;182"><colgroup><col width="177"><col width="235"><col width="255"><col width="182"></colgroup>
<tbody>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-I6WTd73mZoMBNHxywRRlcyJ6gXe-Z8uldLL9YoQH8sxKO47lHhfEg8d">Teams</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-Zsxid1gttofWnbx9E7jlWM9xgZb-QOsBdjFvsovHGcxu3XclT49kgVh">Done</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-CslldrSgJoCBIdxTYVwlzjkVgKc-YTXXdazuFoSD28xCSPJlphwRgye">Phase 1</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-Uanodcb4aoNXlCxwfQSlWuEdgcd-JJgSdWiOWojd6Hx10NHlZcYNgi3">Phase 2</div>

</td>

</tr>
<tr>
<td rowspan="3" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-YmGxdScbwokbaKxhjMTlVEOogTb-ULkid5XTJoByVJxYJe8lLhmqgdd">aevatar-framework</div>
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-YmGxdScbwokbaKxhjMTlVEOogTb-NsFNd3Vd5olPjtxYSXMlcJgtgZg"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-ALEPdyyBvo2Un2xzSXrlgCIngrh-HWLIddK97oRiEaxX2FRl5d7igSc">🏆基础搭建 1.Multi-agent基础框架</div>
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-ALEPdyyBvo2Un2xzSXrlgCIngrh-E1mfd4G9coYwu9xJlomlji11gNd"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-HQHzd2mz9oGVnaxzgovlfN0ZgVf-DzHcdV9m1oEPNLxTndXlvwGTgMc">🏆AI Gagent升级 1.更多LLM支持 2.Memory升级 3.Knowledge base 4.RAG升级</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-ZTubdU3mPod3DxxUUO4l0v2MgJd-D8F2dGsuoo4OiMxbGj3l1R0lgVg">🏆AI Gagent升级 1.多模态 2.Knowledge base共享 3.自我反馈</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-Ip0HdwaHyorIzxxndQGldvpogmb-OUWJdd1V3o8XT8xW3tnlJEzKggi">🏆Marketplace 1.Gagent Marketplace标准 2.Ai组件 Marketplace标准</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-CiuSd81rxojeWixEjwelecpSgLg-PrildWAhwo5xrDxyBfplbSSFg8e">🏆编排 1.自然语言生成 2.可视化工作流面板</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-RJPedHcOAoqrdBxGpy5l97ZYgKe-ARHEdkI2ToDF6UxxsbQldttfgUm">🏆编排升级 1.反馈与评估模块</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-TsyudnCpEoDPXlxP5jelL0C8gwb-DsYudvyiLoZARoxb7H3lr4l3gfg"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-On00d7UlToVuaIxnvjaloNpegrd-AUZEdeRiro0ufuxjQ0blvyiPgkf">4.任务调度</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-TO3UdYROio8etTxbhhClxdI3g0e-QE2hddlQNo7MoFxpC2yli5Jcgbf"> </div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-BfJzdjtyEoDpoixyzmDlxrs7gV9-HoKWdUWSTolnIexVQ3alsgSfg6e"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-ACxZdNBrXoAWJNxqqiolOxstgIf-TlzldPyDioeiS8xzDWWl4hGSgRh"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-PqiXdD6ZMoFY1Axmeh2lDwkkghf-FQq7dji68o343PxTMA8lrLwrgig">🏆Aevatar workflow sdk 1.权限体系 2.Gagent构建 3.Workflow构建</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-T58vdHU73o6lgqxIBoxlVTURgRf-JqIWdRE6Koxf6Mxy45LlUhaTgaf"> </div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-HQxPdyNRcodUrfxIGDBlOyaYgKh-Keu3dINVvoLioKxVyeZl0HL1gTd"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-QUBHdyvUMoUuc4x7xAblvCqJg5h-CCIadr81no7o9Dxg1bTl5BssgRd"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-U2Dsdk0kzoZzH5xRgDelD9qigSf-SwvMdrsAfogHwTxWpi0ltocdgYc">🏆Gagent类型丰富 1.系统Gagent 2.更丰富的社交媒体组件</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-MtATdlQjGo2fMuxfSLYlnt9Agfg-Xw6qd13h7oX3E5xCMthlQUv9gUg">🏆Gagent类型丰富 1.Aelf chain组件</div>

</td>

</tr>
<tr>
<td rowspan="2" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-KGFrdFN6Yo4ZnGx2mESlUMP2gXd-HPhudrTvuoZC3Ax7X7UlsTKMg7g">aevatar-applications</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-IZM5d4d9BohxmMxt13LlL2fKgDd-RMPvdkN4QoAIpDxm3sKlc78ug2g">🏆Pumpfun</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-QUzIdSpfSoGIEjxEqz7lCRFmgtf-TGixdpk5KopkBaxNbCxlyvzlg7g">🏆Aevatar workflow sdk 1.权限体系 2.Gagent构建 3.Workflow构建</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-P4rad7IRxo4rAgxzMIPlJvaMgJh-AqrHduI6vo0aauxmkHslbLP1gvf">🏆Station1.0 1.Agents即服务 2.Marketplace 3.Gagent&Swarms构建 4.Dashboard</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-KUOIdqVtMowUgWxyOl3le5RGgPe-AXgSdcoqioHI1KxBCojlkwelgFb"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-FgqFd43C3oLFmixuY4klgCzZgQh-DTeLdUFyFoWuLoxfl02l7Z1MgOc">🏆神秘Ai游戏</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-BDDIdja2AoY0InxGhD7lkq6rgae-CPLCdxTmVoNVanxqYkwlKRJngOc">🏆AI大赛</div>

</td>

</tr>
<tr>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-K5oIdisKborKiCx8VLulUuLDgXf-X9uDdaKcPoRSfUxDH57lOA79gMg"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-CwxOd4GRJoq2PcxePTslYnxdgjF-Gp8od2Dk1oxWx9x9Gf0lbLJkgTh"> </div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-HsICd3JaXoqmHbxdJB3lneqUgAg-YiohdhWtUoDxsOxkY5blMyp7gHd">AI大赛</div>

</td>
<td rowspan="1" colspan="1">
<div class="ace-line ace-line old-record-id-Oyb0dwIoUo1mvoxg0PGlLFK6g7d-JZ7EdIHoDoQsVtxT69flqmcXg9d-XB4wdRGkfozJtmxHMArlNPvSg6d-E0bxdE1i4oeYO7xVfY3l3YGzgnd"> </div>

</td>

</tr>

</tbody>

</table>

</div>

</div>
<h2 class="heading-2 ace-line old-record-id-FMRkdXTolospXwx8d4QlEhf5gXc">9.2 长期规划</h2>
<div class="ace-line ace-line old-record-id-KCHFd56yjoufTnx77FyltYmJgxe"><strong>增强的向量检索 (RAG) 能力</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-XUKgdXWG4ogxGQxrieZlL6CvgLb" data-list="bullet">原生支持向量数据库</li>
<li class="ace-line ace-line old-record-id-T2fPdDA5KoAviWxtXF4lO3Q1ghc" data-list="bullet">优化海量文档分块检索与 AI 答案生成</li>

</ul>
<div class="ace-line ace-line old-record-id-BdKzddROGoOfKqxPL74leZnugYc"><strong>增强的智能体插件市场</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-BEK5dcZCaoSW1lxnPwnlvukmgGe" data-list="bullet">推出多行业插件生态系统</li>
<li class="ace-line ace-line old-record-id-UajZdWfKyojxh9xyyjWlICEKgUg" data-list="bullet">
<div>提供可即插即用的智能体模块，覆盖如：</div>
<ul class="list-bullet2">
<li class="ace-line ace-line old-record-id-BhXEdiBbQoQDVvxeq6TlUl20gbf" data-list="bullet">金融风控</li>
<li class="ace-line ace-line old-record-id-RTx1dy2smoh5vlxYa1Dlhwgsg2z" data-list="bullet">供应链管理</li>
<li class="ace-line ace-line old-record-id-NHXJdQMAbo2KAoxHIwTlGIKHgUk" data-list="bullet">医疗健康</li>

</ul>

</li>

</ul>
<div class="ace-line ace-line old-record-id-EJa7dqtRto92nUxN9mTlhM2Qg5r"><strong>服务网格与零信任安全</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-RZXxdGTtrokxPnxT3wXlx5S8ghc" data-list="bullet">进一步深化服务网格集成</li>
<li class="ace-line ace-line old-record-id-EojUd2711ol2cnxgzOfluAVig0I" data-list="bullet">加强数据加密、流量管控与访问策略</li>

</ul>
<div class="ace-line ace-line old-record-id-BTIXdLEAXoQyrmxJ4A2lOyN9g0g"><strong>RLHF（基于人类反馈的强化学习）机制</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-QjA2d7sWVoSE6zxKYMalze4Sgvd" data-list="bullet">支持对智能体进行实时人类反馈训练</li>
<li class="ace-line ace-line old-record-id-SHIBd9QD0oT5KoxqzVnlBWVdgCs" data-list="bullet">持续优化对话质量、逻辑推理及行为决策</li>

</ul>
<div class="ace-line ace-line old-record-id-S6EgdfrYmo9hjvxJnBcl1VYmgud"><strong>无限的智能体协作</strong></div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-XdUUdTnCBoVLsgxhIQclbaragxc" data-list="bullet">探索与第三方 AI 系统及边缘计算设备的互联</li>
<li class="ace-line ace-line old-record-id-E5FYdwhinosr6CxqSEllDpmVgFb" data-list="bullet">将多智能体协作从云原生拓展至物联网或其他 AI 智能体平台</li>

</ul>
<ol class="ace-line ace-line old-record-id-SQkjdCJC2ocYNIxjjFKlFJM3g77" start="10">
<li data-list="number">
<h1 class="heading-1">总结</h1>

</li>

</ol>
<div class="ace-line ace-line old-record-id-EbyMd5MBJoA7tBxdR3klj1BQgbb">随着多模型、多智能体协作逐渐成为主流，aevatar.ai 作为下一代 AI 智能体经济的先行者，提供了跨平台、跨语言模型、低门槛且高度可扩展的解决方案。</div>
<div class="ace-line ace-line old-record-id-Tg3cdL2gCo2faexDketlTYSbg3g">通过充分利用 Orleans Actor 模型、事件溯源以及云原生架构，aevatar.ai 实现了以下核心价值：</div>
<ul class="list-bullet1">
<li class="ace-line ace-line old-record-id-Y7QKdwH2ZoWQLFxYWROlOEvVgJe" data-list="bullet"><strong>全面的多智能体协作</strong>：突破单一模型和封闭生态的限制，让不同 AI 智能体之间实现信息共享与高效通信。</li>
<li class="ace-line ace-line old-record-id-MuljdEsZGo3WMWx9Ob9lzrpZgod" data-list="bullet"><strong>可视化与低代码</strong>：大幅降低开发和维护门槛，帮助不同层级的用户快速落地 AI 智能体解决方案。</li>
<li class="ace-line ace-line old-record-id-CPrXd38DGorqhXxCpFsljHKTgVb" data-list="bullet"><strong>高并发与可追溯性</strong>：分布式 Actor 与事件溯源，确保在大规模场景中的稳定性与可审计性。</li>
<li class="ace-line ace-line old-record-id-JwpEdpagcoQVhLxztWOlVAedgGh" data-list="bullet"><strong>安全与可扩展性</strong>：云原生 DevSecOps 方案，既满足行业定制化需求，又确保合规。</li>

</ul>
<div class="ace-line ace-line old-record-id-XYS9dMhfGoF9GvxxvQhlhU1bgCd">展望未来，aevatar.ai 将持续迭代升级，致力于打造功能全面、稳健的 Agent-as-a-Service 平台，为更多行业和个人用户带来便捷而强大的 AI 协作体验。</div>
<div class="ace-line ace-line old-record-id-MVtAdd8kgobk4Oxs6mjlPbV2gHC"> </div>
<div class="ace-line ace-line old-record-id-MVtAdd8kgobk4Oxs6mjlPbV2gHC">我们诚邀社区共同参与生态建设，携手推动 AI 智能体系统的开放与成功。</div>
<div class="ace-line ace-line old-record-id-MVtAdd8kgobk4Oxs6mjlPbV2gHC">
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe"> </div>
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe">Github地址： </div>
<div class="ace-line ace-line old-record-id-PA4Id9ViBo5YkVxNfkvlHT3Pgqe"><span class="break-words       tvm-parent-container"><span dir="ltr"><span class="white-space-pre">aevatar 核心框架:<span class="white-space-pre"> https://github.com/aevatarAI/aevatar-framework </span><br>aevatar平台:<span class="white-space-pre"> https://github.com/aevatarAI/aevatar-station<br>aevatar 案例:<span class="white-space-pre"> https://github.com/aevatarAI/aevatar-gagents</span></span></span></span></span></div>

</div>
<div class="ace-line ace-line old-record-id-FC8odDHlnoSyZhxrBpmll4rsgud"> </div>

</div>

</div>
<div id="MySignature" role="contentinfo">
<div id="AllanboltSignature">
         <p id="PSignature" style="border-top: #e0e0e0 1px dashed; border-right: #e0e0e0 1px dashed;
            border-bottom: #e0e0e0 1px dashed; border-left: #e0e0e0 1px dashed; padding-top: 10px;
            padding-right: 10px; padding-bottom: 10px; padding-left: 60px; background: url(https://images.cnblogs.com/cnblogs_com/lloydsheng/239039/o_copyright.gif) #e5f1f4 no-repeat 1% 50%;
            font-family: 微软雅黑; font-size: 12px;">
            <br />
            作者：<b><span style="font-size: 12px; color: red;">圣殿骑士</span></b>
            <br />
            出处：http://www.cnblogs.com/KnightsWarrior/
            <br />
                  关于作者：专注于微软平台项目架构、管理和企业解决方案。自认在面向对象，面向服务以及微服务领域有一定的造诣，熟悉设计模式、TDD、极限编程、领域驱动、架构设计、敏捷开发和项目管理。现主要从事.NET/.NET Core, Go, JavaScript/TypeScript, Azure/AWS等云计算方面的项目开发、架构、管理和企业培训工作。如有问题或建议，请多多赐教！
            <br />
            本文版权归作者和博客园共有，欢迎转载，但未经作者同意必须保留此段声明，且在文章页面明显位置给出原文连接。如有问题，可以邮件：<strong>KnightsWarrior(at)msn(dot)com
</strong> 微博：<strong>圣殿骑士微博</strong> 联系我，非常感谢。
            <br />

         </p>
   </div>
<div id="weibowidget"></div><br><br>
来源：https://www.cnblogs.com/KnightsWarrior/p/18755136/aevatarAI

頁: [1]

琼殿技术论坛's Archiver

开源Multi-agent AI智能体框架aevatar.ai，欢迎大家贡献代码