<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/">
    <channel>
        <title>智汇AI</title>
        <link>http://easyai.fyi/</link>
        <description>每天看懂AI新闻、知识与工具，给产品经理、运营和AI初学者的人工智能科普站。</description>
        <lastBuildDate>Mon, 29 Jun 2026 18:13:30 GMT</lastBuildDate>
        <docs>https://validator.w3.org/feed/docs/rss2.html</docs>
        <generator>https://github.com/jpmonette/feed</generator>
        <language>zh-CN</language>
        <copyright>All rights reserved 2026, 智汇AI</copyright>
        <item>
            <title><![CDATA[Anthropic AI Agents科普]]></title>
            <link>http://easyai.fyi/article/ai-agents-intro</link>
            <guid>http://easyai.fyi/article/ai-agents-intro</guid>
            <pubDate>Tue, 22 Jul 2025 00:00:00 GMT</pubDate>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-1df5eac3813c80ec9cb5c2aabb4a65b4"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><div class="notion-row"><a target="_blank" rel="noopener noreferrer" class="notion-bookmark notion-block-2385eac3813c80958d63d2806ebd10c8" href="https://www.anthropic.com/engineering/building-effective-agents"><div><div class="notion-bookmark-title">Building Effective AI Agents</div><div class="notion-bookmark-description">Discover how Anthropic approaches the development of reliable AI agents. Learn about our research on agent capabilities, safety considerations, and technical framework for building trustworthy AI.</div><div class="notion-bookmark-link"><div class="notion-bookmark-link-icon"><img src="https://www.anthropic.com/images/icons/favicon-32x32.png?t=2385eac3-813c-8095-8d63-d2806ebd10c8" alt="Building Effective AI Agents" loading="lazy" decoding="async"/></div><div class="notion-bookmark-link-text">https://www.anthropic.com/engineering/building-effective-agents</div></div></div><div class="notion-bookmark-image"><img style="object-fit:cover" src="https://cdn.sanity.io/images/4zrzovbb/website/76b5733c669f0dfb9c7aa7fc512a495867cf12e6-2400x1260.png?t=2385eac3-813c-8095-8d63-d2806ebd10c8" alt="Building Effective AI Agents" loading="lazy" decoding="async"/></div></a></div></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[NotebookLLM]]></title>
            <link>http://easyai.fyi/article/1b85eac3-813c-808d-931f-d347a3ec39a3</link>
            <guid>http://easyai.fyi/article/1b85eac3-813c-808d-931f-d347a3ec39a3</guid>
            <pubDate>Sun, 16 Mar 2025 00:00:00 GMT</pubDate>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-1b85eac3813c808d931fd347a3ec39a3"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><div class="notion-text notion-block-2005eac3813c8077ac06cee312b4a60b">关于NotebookLLM</div><figure class="notion-asset-wrapper notion-asset-wrapper-tweet notion-block-2005eac3813c80318f24e23c2cd31a96"><div style="position:relative;display:flex;justify-content:center;align-self:center;width:100%;max-width:100%;flex-direction:column"><div style="max-width:420px;width:100%;margin-left:auto;margin-right:auto"></div></div></figure><div class="notion-row"><a target="_blank" rel="noopener noreferrer" class="notion-bookmark notion-block-2005eac3813c80ceabaaf7b3ad1eed82" href="https://v2ex.com/t/1098014"><div><div class="notion-bookmark-title">大模型时代，应该用什么知识管理软件？ - V2EX</div><div class="notion-bookmark-description">问与答 - @afterzero - 在笔记时代，我每天在用 flomo 、微信收藏、OneNote 、Markdown 文件、浏览器收藏夹等工具。但是 [我不想把收藏变成一个**把喜欢的东西丢进垃圾堆**的过程。] 大模型可以帮我捡</div><div class="notion-bookmark-link"><div class="notion-bookmark-link-icon"><img src="https://v2ex.com/static/icon-192.png?t=2005eac3-813c-80ce-abaa-f7b3ad1eed82" alt="大模型时代，应该用什么知识管理软件？ - V2EX" loading="lazy" decoding="async"/></div><div class="notion-bookmark-link-text">https://v2ex.com/t/1098014</div></div></div><div class="notion-bookmark-image"><img style="object-fit:cover" src="https://cdn.v2ex.com/gravatar/f5cb03c1357f8c1899ffff051e75603d?s=73&amp;d=retro&amp;t=2005eac3-813c-80ce-abaa-f7b3ad1eed82" alt="大模型时代，应该用什么知识管理软件？ - V2EX" loading="lazy" decoding="async"/></div></a></div></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Livebench-LLM评估和基准测试平台]]></title>
            <link>http://easyai.fyi/article/livebench</link>
            <guid>http://easyai.fyi/article/livebench</guid>
            <pubDate>Sat, 01 Mar 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[大模型评分]]></description>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-1a65eac3813c804cad9ce5c9f596b3fb"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><figure class="notion-asset-wrapper notion-asset-wrapper-embed notion-block-1b95eac3813c80be83d4e3b034a6fbe3"><div style="position:relative;display:flex;justify-content:center;align-self:center;width:100%;max-width:100%;flex-direction:column;height:320px"><iframe class="notion-asset-object-fit" src="https://livebench.ai/#/" title="iframe embed" frameBorder="0" allowfullscreen="" loading="lazy" scrolling="auto"></iframe></div></figure><div class="notion-blank notion-block-1a65eac3813c8048a2f2c5f02bfc5fd7"> </div><div class="notion-blank notion-block-1a65eac3813c80968a5de269facc7ba7"> </div><div class="notion-blank notion-block-1a65eac3813c8070a901cc35ca39bc11"> </div></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[RAG：数据准备]]></title>
            <link>http://easyai.fyi/article/data-prepare</link>
            <guid>http://easyai.fyi/article/data-prepare</guid>
            <pubDate>Mon, 24 Jun 2024 00:00:00 GMT</pubDate>
            <description><![CDATA[数据的组织和结构化影响LLM应用定位数据的性能]]></description>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-5fccdf80d6b346d3b096f815f432e76f"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><figure class="notion-asset-wrapper notion-asset-wrapper-image notion-block-2949338e79a9425cbbf6f4488e7bbc02"><div style="position:relative;display:flex;justify-content:center;align-self:center;width:100%;max-width:100%;flex-direction:column;height:100%"><img style="object-fit:cover" src="https://www.notion.so/image/https%3A%2F%2Fprod-files-secure.s3.us-west-2.amazonaws.com%2F4f82ec79-df5e-4028-9a2c-bd5ba6249e10%2F94159b66-58a8-45c6-8fe6-5ecba99216d3%2FUntitled.png?table=block&amp;id=2949338e-79a9-425c-bbf6-f4488e7bbc02&amp;t=2949338e-79a9-425c-bbf6-f4488e7bbc02&amp;width=708&amp;cache=v2" alt="notion image" loading="lazy" decoding="async"/></div></figure><div class="notion-text notion-block-f8a65c1ac88a4837915f9c689ac7d55d">任何机器学习应用的初始阶段都需要进行数据准备。这包括建立数据输入管道和预处理数据，使其与推理管道兼容。</div><div class="notion-text notion-block-8805737c5a7d4bd98f98095ae0c1a488">在本篇文章中，我们将关注 RAG 的数据准备方面。我们的目标是有效地组织和结构化数据，确保在应用程序中定位答案的最佳性能。</div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-96126f855a854fa3954ba0be176d6f2e" data-id="96126f855a854fa3954ba0be176d6f2e"><span><div id="96126f855a854fa3954ba0be176d6f2e" class="notion-header-anchor"></div><a class="notion-hash-link" href="#96126f855a854fa3954ba0be176d6f2e" title="步骤 1：数据输入Data Ingestion"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title"><span class="notion-yellow_background">步骤 1：数据输入</span><span class="notion-yellow_background"><b>Data Ingestion</b></span></span></span></h4><div class="notion-text notion-block-5ef3baf9bd5f428ea1a45040946cc617">构建消费者友好型聊天机器人始于好的数据选择。</div><ol start="1" class="notion-list notion-list-numbered notion-block-de43121107b14cb18f6b9a6748df0a12"><li>好的选择：确定从用户到 API 的数据源，并建立推送机制，以便对 LLM 应用程序进行持续更新。</li></ol><ol start="2" class="notion-list notion-list-numbered notion-block-e5521f7e04e442d294d1783960db46c7"><li>数据管理很重要：提前实施数据管理政策。对文件来源进行审计和编目，对敏感数据进行编辑，并为上下文培训奠定基础。</li></ol><ol start="3" class="notion-list notion-list-numbered notion-block-5130fbf2662547d8b7c7970c9eacc5fa"><li>质量检查：评估数据的多样性、规模和噪音水平。质量较低的数据集会冲淡响应，因此有必要尽早建立分类机制。</li></ol><ol start="4" class="notion-list notion-list-numbered notion-block-37986043b3534f8c94d0234847f0b57c"><li>保持领先：即使在快节奏的 LLM 开发过程中，要坚持数据治理。这可以降低风险，确保可扩展的、稳健的数据提取。</li></ol><ol start="5" class="notion-list notion-list-numbered notion-block-bd3e74c30f244b8d9add476c75b9da03"><li>实时清理：从 Slack 等平台提取数据？实时过滤噪音、错别字和敏感内容，打造高效的 LLM 应用。</li></ol><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-063c75407d8746248131d3ae469b9ea8" data-id="063c75407d8746248131d3ae469b9ea8"><span><div id="063c75407d8746248131d3ae469b9ea8" class="notion-header-anchor"></div><a class="notion-hash-link" href="#063c75407d8746248131d3ae469b9ea8" title="步骤 2：数据清理Data Cleaning"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title"><span class="notion-yellow_background">步骤 2：数据清理</span><span class="notion-yellow_background"><b>Data Cleaning</b></span></span></span></h4><div class="notion-text notion-block-b92c8b86753248fc854ebff6111be3a9">文件中的每个页面都会转换为<span class="notion-inline-underscore">文档对象</span>，并包含两个基本组件：页面内容和元数据。<b>page_content 和 metadata</b>.</div><div class="notion-text notion-block-bfe795f200eb499eac06009b21aa0c40">”页面内容“是直接从文档页面中提取的文本内容。</div><div class="notion-text notion-block-31e7986c79154d0a8e0196fe1ae47d22">“元数据“是附加详细信息的重要组合，如文档的来源（源文件）、页码、文件类型和其他信息。元数据在生成答案时，会记录它所利用的特定来源。</div><figure class="notion-asset-wrapper notion-asset-wrapper-image notion-block-269e7f998abf4c1a8d2595c735f2acec"><div style="position:relative;display:flex;justify-content:center;align-self:center;width:100%;max-width:100%;flex-direction:column;height:100%"><img style="object-fit:cover" src="https://www.notion.so/image/https%3A%2F%2Fprod-files-secure.s3.us-west-2.amazonaws.com%2F4f82ec79-df5e-4028-9a2c-bd5ba6249e10%2F02d2d370-64a2-45ae-bb8b-d797b6fff6fd%2FUntitled.png?table=block&amp;id=269e7f99-8abf-4c1a-8d25-95c735f2acec&amp;t=269e7f99-8abf-4c1a-8d25-95c735f2acec&amp;width=1400&amp;cache=v2" alt="notion image" loading="lazy" decoding="async"/></div></figure><div class="notion-text notion-block-ad1f96897944403abf91feb7061c3227">为了实现这一目标，可以使用数据加载器等工具，这些工具由 LangChain 和 Llamaindex 等开源库提供。这些库支持各种格式，从 PDF 和 CSV 到 HTML、Markdown 甚至数据库。</div><div class="notion-text notion-block-24888a217e224df894b64545394ce27f">这种方法的优点是可以通过页码检索文件。</div><h4 class="notion-h notion-h3 notion-h-indent-0 notion-block-d54a2bb0af9641be9b9f322a19be3d03" data-id="d54a2bb0af9641be9b9f322a19be3d03"><span><div id="d54a2bb0af9641be9b9f322a19be3d03" class="notion-header-anchor"></div><a class="notion-hash-link" href="#d54a2bb0af9641be9b9f322a19be3d03" title="步骤 3：分块Chunking"><svg viewBox="0 0 16 16" width="16" height="16"><path fill-rule="evenodd" d="M7.775 3.275a.75.75 0 001.06 1.06l1.25-1.25a2 2 0 112.83 2.83l-2.5 2.5a2 2 0 01-2.83 0 .75.75 0 00-1.06 1.06 3.5 3.5 0 004.95 0l2.5-2.5a3.5 3.5 0 00-4.95-4.95l-1.25 1.25zm-4.69 9.64a2 2 0 010-2.83l2.5-2.5a2 2 0 012.83 0 .75.75 0 001.06-1.06 3.5 3.5 0 00-4.95 0l-2.5 2.5a3.5 3.5 0 004.95 4.95l1.25-1.25a.75.75 0 00-1.06-1.06l-1.25 1.25a2 2 0 01-2.83 0z"></path></svg></a><span class="notion-h-title"><span class="notion-yellow_background">步骤 3：分块</span><span class="notion-yellow_background"><b>Chunking</b></span></span></span></h4><ol start="1" class="notion-list notion-list-numbered notion-block-4507ef94cc49425ca61b1a6ae93695a0"><li><b>为什么要分块？</b></li><ol class="notion-list notion-list-numbered notion-block-4507ef94cc49425ca61b1a6ae93695a0"><div class="notion-text notion-block-32d1fe6abc404c068e836e67b1368ed0">在软件世界里，改变游戏规则的关键在于如何塑造数据--无论是标记符、PDF 还是其他文本文件。</div><div class="notion-text notion-block-3bce64200cd7429b94515d54f9ca7fa1">想象一下：有一份厚厚的 PDF 文件，现在就其内容提出问题。问题出在哪里？传统的方法是将整个文档和您的问题扔给模型，但效果不佳。为什么呢？让我们来谈谈模型“上下文窗口的局限性”。</div><div class="notion-text notion-block-2437dbcfa45040f08889e0612dc704b3">把上下文窗口想象成对文档的窥视，通常仅限于一页或几页。现在，一次共享整个文档？不太现实。不过不用担心！</div><div class="notion-text notion-block-3750be742ba8475dbe80d92c95e7de3a">诀窍在于将<span class="notion-inline-underscore">“数据分块”</span>。将数据分解成易于处理的部分，只将最相关的部分发送给模型。这样，就不会让模型不堪重负，而且还能获得需要的回答。</div><div class="notion-text notion-block-6c86b400c5054d868bfd75e1fae76959">通过将结构化文件分解成易于管理的小块，我们让 LLM 能够以无与伦比的效率处理信息。这种方法不再受页数限制，可确保关键细节不会在混乱中丢失。</div></ol></ol><ol start="2" class="notion-list notion-list-numbered notion-block-649358642a9b456ca23369bd8359cf28"><li><b>分块前的一些点？</b></li><ol class="notion-list notion-list-numbered notion-block-649358642a9b456ca23369bd8359cf28"><div class="notion-text notion-block-cd1529a9813f4cb794577e8659f3371a">文档的结构和长度：</div><li>长文档：书籍、学术文章等</li><li>短文档：社交媒体帖子、客户评论等。</li><li>嵌入模型：分块大小决定了应使用何种嵌入模型。</li><li>预期查询：使用案例是什么？</li></ol></ol><ol start="3" class="notion-list notion-list-numbered notion-block-4c8ec22a21724c228c8f9fbc3ce65ac8"><li><b>数据块大小?</b></li><ol class="notion-list notion-list-numbered notion-block-4c8ec22a21724c228c8f9fbc3ce65ac8"><div class="notion-text notion-block-d42b2b2c71f143408f8ce128f9211bcd"><b>Small chunk size </b>小块：例如：单句 → 生成的上下文信息较少：单句 → 生成的上下文信息较少。</div><div class="notion-text notion-block-fbec6ba72fc24a4694f23c64b4ffcd12"><b>Large chunk size </b>大块尺寸：例如：整页、多个段落、完整文档：整页、多个段落、完整文档。在这种情况下，语块涵盖的信息更多，可以通过更多的上下文信息提高生成效率。</div></ol></ol><ol start="4" class="notion-list notion-list-numbered notion-block-53b85236185d45b1b2312c664371007f"><li><b>LLM 上下文窗口限制？</b></li><ol class="notion-list notion-list-numbered notion-block-53b85236185d45b1b2312c664371007f"><li><span class="notion-inline-underscore">Top-K Retrieved Chunks：</span>假设 LLM 的上下文窗口大小为 10,000 tkens，我们为给定的用户查询保留了其中的 1000 tokens，再为指令提示和聊天记录保留了其中的 2000 tkens，这样就只剩下 7000 tkens 可用于其他信息。假设我们打算将 K = 10 的前 10 个信息块传入 LLM，这就意味着我们要将剩余的 7000 个信息块除以 10 个信息块，这样每个信息块的最大信息量将为 700 个。</li><li><span class="notion-inline-underscore">分块大小范围</span>：下一步是选择一定范围的潜在块大小进行测试。如前所述，选择时应考虑到内容的性质（如短信息或长文档）、将使用的嵌入模型及其功能（如标记限制）。目的是在保留上下文和保持准确性之间找到平衡。首先要探索各种块的大小，包括捕获更细粒度语义信息的较小块（如 128 或 256 标记）和保留更多上下文的较大块（如 512 或 1024 标记）。</li><div class="notion-text notion-block-c4b43a7b6491476e9e4dd2458dac50f8">评估每种<span class="notion-inline-underscore">分块大小的性能</span>--要测试各种分块大小，可以使用多个索引，或者使用具有多个命名空间的单个索引。使用具有代表性的数据集，为要测试的块大小创建嵌入，并将其保存在索引（或多个索引）中。</div><div class="notion-text notion-block-3b8c074e43a741e2954e3ee6981732c4">然后，可以运行一系列可以评估质量的查询，并比较不同块大小的性能。这很可能是一个迭代的过程，在这个过程中，你会针对不同的查询测试不同的块大小，直到你能根据内容和预期查询确定性能最好的块大小。</div></ol></ol><ol start="5" class="notion-list notion-list-numbered notion-block-04e5081ea16249d5b83c47187da6099a"><li><b>高Context长度的限制?</b></li><ol class="notion-list notion-list-numbered notion-block-04e5081ea16249d5b83c47187da6099a"><div class="notion-text notion-block-f1c9472bf93841dc873f900569e613e7">由于 Transformer 模型的Self Attention 机制，高Context长度会导致时间和内存的二次增长。</div><div class="notion-text notion-block-21a0c5919aaf4f3da89180dbf1d46f9f">在 LlamaIndex 发布的这个示例中，您可以从下表中看到，随着分块大小的增加，平均响应时间略有上升。</div><div class="notion-text notion-block-1096a2684a174b93b799540322c829c8">有趣的是，平均似乎在数据块大小为 1024 时达到顶峰，而平均相关性则随着数据块大小的增大而持续提高，同样在 1024 时达到顶峰。这表明，1024 的数据块大小可以在响应时间和响应质量之间取得最佳平衡。</div><figure class="notion-asset-wrapper notion-asset-wrapper-image notion-block-76b533e77871482fb7839f5066eced44"><div style="position:relative;display:flex;justify-content:center;align-self:center;width:385.984375px;max-width:100%;flex-direction:column"><img style="object-fit:cover" src="https://www.notion.so/image/https%3A%2F%2Fprod-files-secure.s3.us-west-2.amazonaws.com%2F4f82ec79-df5e-4028-9a2c-bd5ba6249e10%2Fb3c2a904-3843-49fc-9c12-52828a0e7790%2FUntitled.png?table=block&amp;id=76b533e7-7871-482f-b783-9f5066eced44&amp;t=76b533e7-7871-482f-b783-9f5066eced44&amp;width=385.984375&amp;cache=v2" alt="notion image" loading="lazy" decoding="async"/></div></figure></ol></ol><ol start="6" class="notion-list notion-list-numbered notion-block-96c8d76e713f45f9bfdf4c10de565eb7"><li><b>分块方法</b></li><ol class="notion-list notion-list-numbered notion-block-96c8d76e713f45f9bfdf4c10de565eb7"><div class="notion-text notion-block-1b4b5a3fe7cc470881e6aea62125c63b">有不同的分块方法，每种方法都可能适用于不同的情况。通过研究每种方法的优缺点，我们的目标是找出适合的应用场景。</div><li><span class="notion-inline-underscore">固定大小的分块</span></li><div class="notion-text notion-block-cb09bd1025fa4c6c9513048115f8a712">我们决定每个分块中的标记数量，同时考虑到可选的重叠。为什么要重叠？为了确保语义上下文的丰富性在各语块之间保持不变。</div><div class="notion-text notion-block-6e2e3547622b4ec7801bcc67674d9091">为什么采用固定大小？这是大多数情况下的黄金路径。它不仅计算成本低廉，节省了处理能力，而且使用起来轻而易举。无需复杂的 NLP 库，只需优雅地将固定大小的数据块无缝分解即可。</div><div class="notion-text notion-block-4e03bd9bc7c14b4ea4951363b9ede50a">下面是使用 LangChain 执行固定大小分块的示例：</div><div class="notion-text notion-block-66a2bdee69c54bebb854949239f2fde7"><span class="notion-inline-underscore">b. 专业分块</span></div><div class="notion-text notion-block-81862957fc4e4f86acb4dec4e90e4858">专用分块Markdown 和 LaTeX 是可能会遇到的结构化和格式化内容的两个例子。在这种情况下，可以使用专门的分块方法，在分块过程中保留内容的原始结构。</div><div class="notion-text notion-block-0c0b00608379489db1d9aea9081e3d21">Markdown 是一种轻量级标记语言，常用于格式化文本。通过识别 Markdown 语法（如标题、列表和代码块），可以根据内容的结构和层次对其进行智能划分，从而形成语义更加连贯的分块。例如</div><div class="notion-text notion-block-35ba809f05eb45ddad6da2948734ce34">LaTex 是一种文档编制系统和标记语言，常用于学术论文和技术文档。通过解析 LaTeX 命令和环境，可以创建尊重内容逻辑组织（如章节、小节和方程式）的语块，从而获得更准确和与上下文相关的结果。例如</div><div class="notion-text notion-block-f5bb6d79f26c4903b90629932986e9be"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://medium.com/@vipra_singh/building-llm-applications-data-preparation-part-2-b7306d224245">https://medium.com/@vipra_singh/building-llm-applications-data-preparation-part-2-b7306d224245</a></div></ol></ol></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[ForesightNews的Web3中AI项目全盘点]]></title>
            <link>http://easyai.fyi/article/web3_ai_project</link>
            <guid>http://easyai.fyi/article/web3_ai_project</guid>
            <pubDate>Thu, 07 Mar 2024 00:00:00 GMT</pubDate>
            <description><![CDATA[区块链全景项目盘点]]></description>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-97ac2ffdc2bb4c0da6be0dfe250c6799"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><div class="notion-text notion-block-ccae563740a5467f89114b9cbfbe8a6b">目前行业至少有 140 多个 Web3 + AI 概念项目，已发币的项目共有 85 个，还有部分将于明年发行代币。</div><div class="notion-text notion-block-1aa5eac3813c80989be9ebce29321c8a">这 140 个项目覆盖了基础设施、数据、预测市场、计算与算力、教育、DeFi &amp; 跨链、安全、NFT &amp; 游戏 &amp; 元宇宙、搜索引擎、社交 &amp; 创作者经济、AI 聊天机器人、DID &amp; 消息传递、治理、医疗、交易机器人等诸多方向。</div><div class="notion-text notion-block-1aa5eac3813c80a8bf3fed8184a464ba">其中，基础设施类项目多达 30 个，NFT &amp; 游戏 &amp; 元宇宙类项目有 26 个，数据、计算、AI 聊天机器人类项目也均在 10 个以上。</div><div class="notion-text notion-block-b5c0e4e74f284cadb7dbe0b772241d6f"><a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://docs.google.com/spreadsheets/d/1b1HQEN3KVwRP4zS7wwQdrMzipyyb_gYKqibR4jaMa3o/preview?pli=1#gid=0">https://docs.google.com/spreadsheets/d/1b1HQEN3KVwRP4zS7wwQdrMzipyyb_gYKqibR4jaMa3o/preview?pli=1#gid=0</a></div><div class="notion-blank notion-block-0d3b4b80d9ae41b2975805acc74e2fb0"> </div></main></div>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[当前 MCP的一些实际运用]]></title>
            <link>http://easyai.fyi/article/mcp_project</link>
            <guid>http://easyai.fyi/article/mcp_project</guid>
            <pubDate>Sun, 16 Mar 2025 00:00:00 GMT</pubDate>
            <content:encoded><![CDATA[<div id="notion-article" class="mx-auto overflow-hidden "><main class="notion light-mode notion-page notion-block-f47837c11c0e4d69badd16885156225b"><div class="notion-viewport"></div><div class="notion-collection-page-properties"></div><div class="notion-text notion-block-1b85eac3813c80ca943de3f5c4ce9117">MCP（模型上下文协议）是一个由Anthropic于2024年11月提出的开放标准，旨在解决AI模型与外部数据源和工具连接的标准化问题。</div><div class="notion-text notion-block-1b85eac3813c80caa908c7ea507b4bdd">MCP被描述为AI助手连接到内容库、商业工具和开发环境的一种通用接口，帮助AI模型生成更相关、更准确的响应。</div><div class="notion-text notion-block-1b85eac3813c805f94b6d3c5f16ff8dc">根据Anthropic的官方介绍 (<a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://www.anthropic.com/news/model-context-protocol">Anthropic: Introducing the Model Context Protocol</a>)，MCP的目标是打破数据孤岛，解决AI模型因信息孤立而受限的问题。</div><div class="notion-text notion-block-1b85eac3813c808c89f0fb57dd5e89cb">传统上，每个新数据源都需要自定义实现，这使得系统难以扩展。MCP通过提供一个统一的协议，取代了碎片化的集成方式，使AI系统能够更轻松地访问所需上下文。</div><div class="notion-text notion-block-1b85eac3813c8014bb18df68344a9e23">Medium上的文章 (<a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://medium.com/@kenzic/getting-started-model-context-protocol-e0a80dddff80">Medium: Getting Started with Model Context Protocol</a>)进一步解释，MCP被比喻为AI的USB-C端口，类似于USB-C如何标准化设备连接，MCP为AI模型与各种数据源和工具的连接提供了一种标准化方法。这简化了集成，打破了数据孤岛，并释放了AI提供高质量输出的潜力。</div><div class="notion-text notion-block-1b85eac3813c80c1bafafec553c301ac">MCP采用客户端-服务器架构，具体来说，AI应用（如Claude桌面应用）作为MCP主机，连接到MCP服务器，这些服务器暴露特定的功能。例如，MCP服务器可以提供文件系统操作、GitHub API集成、Google Drive访问或PostgreSQL数据库查询等能力 (<a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://modelcontextprotocol.io/docs/concepts/architecture">Model Context Protocol Documentation: Core architecture</a>)。</div><div class="notion-text notion-block-1b85eac3813c80c4a31cd5e3f6b8f56d">MCP的应用范围非常广泛，包括但不限于Blender的3D场景生成、Perplexity的实时网络搜索、QGIS的地图绘制、PubMed的学术数据库访问、Supabase数据库连接、Gradio客户端的工具集成、通知声音播放、Weaviate的向量搜索能力以及Figma设计的代码生成 (<a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://x.com/minchoi/status/1900931746448756879">X post by Min Choi</a>)。</div><div class="notion-text notion-block-1b85eac3813c8057b60ce79f9cfd1ff0">MCP在学术领域的应用，例如通过PubMed数据库连接Claude，这允许AI直接访问学术文章，极大地提升了研究效率 (<a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://x.com/MushtaqBilalPhD/status/1899495850977419754">X post by Mushtaq Bilal, PhD</a>)。安装过程包括下载Claude桌面应用、安装Node.js、获取Brave API密钥，并配置JSON文件以启用MCP服务器，这可以在15分钟内完成。</div><div class="notion-text notion-block-1b85eac3813c80d7a5f0e4ce2d7cb58e">MCP的另一个优势是灵活性，它允许开发者在不同LLM提供商和供应商之间切换，同时遵循最佳实践来保护数据安全 (<a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://modelcontextprotocol.io/introduction">Model Context Protocol Documentation: Introduction</a>)。GitHub上的MCP项目 (<a target="_blank" rel="noopener noreferrer" class="notion-link" href="https://github.com/modelcontextprotocol">Model Context Protocol GitHub</a>)欢迎社区贡献，提供了详细的文档和教程，鼓励开发者参与改进。</div><table class="notion-simple-table notion-block-1b85eac3813c8077ac33e729b32a4712"><tbody><tr class="notion-simple-table-row notion-block-1b85eac3813c8036a0e2ce6f70076e8f"><td class="" style="width:120px"><div class="notion-simple-table-cell">序号</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">应用示例</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">描述</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c801d8452dde6af6919a1"><td class="" style="width:120px"><div class="notion-simple-table-cell">1</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">Blender MCP</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">Claude可通过提示直接生成3D场景</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c8079a0c2c803e783b612"><td class="" style="width:120px"><div class="notion-simple-table-cell">2</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">Perplexity MCP</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">AI助手可进行实时网络搜索</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c8007a47ff9e8a502a84a"><td class="" style="width:120px"><div class="notion-simple-table-cell">3</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">MCP QGIS</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">使用QGIS进行地图绘制</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c802bacecd06d810d5507"><td class="" style="width:120px"><div class="notion-simple-table-cell">4</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">Firecrawl MCP</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">通过提示克隆任何网站</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c80b3a49bd38410b971cf"><td class="" style="width:120px"><div class="notion-simple-table-cell">5</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">PubMed学术数据库</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">连接Claude以访问学术文章</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c80d697c8c6d65af93dc6"><td class="" style="width:120px"><div class="notion-simple-table-cell">6</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">Supabase数据库</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">连接数据库进行数据查询</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c80de87a0f4e52c43853c"><td class="" style="width:120px"><div class="notion-simple-table-cell">7</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">MCP Gradio客户端</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">使用Gradio接口与MCP服务器交互</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c80b4ab26ed20e77a90ec"><td class="" style="width:120px"><div class="notion-simple-table-cell">8</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">通知MCP</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">任务完成后播放声音通知</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c8094af52e4edd49210d6"><td class="" style="width:120px"><div class="notion-simple-table-cell">9</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">Weaviate MCP</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">连接Weaviate的向量搜索能力</div></td></tr><tr class="notion-simple-table-row notion-block-1b85eac3813c804c948ee4a97c763d14"><td class="" style="width:120px"><div class="notion-simple-table-cell">10</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">Figma MCP</div></td><td class="" style="width:120px"><div class="notion-simple-table-cell">从Figma设计轻松生成代码</div></td></tr></tbody></table><div class="notion-sync-block notion-block-1b85eac3813c80e3a1b8f28767976d6d"><div class="notion-text notion-block-1b85eac3813c80f6b686f9191c7fceba">‣</div></div><div class="notion-blank notion-block-1b85eac3813c807184ded0b8506de77a"> </div></main></div>]]></content:encoded>
        </item>
    </channel>
</rss>