<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <author>
    <name>高晓庄</name>
    <email>gaoxiaozhuang@bingosoft.net</email>
  </author>
  <generator uri="https://hexo.io/">Hexo</generator>
  <icon>https://www.gravatar.com/avatar/13ac957b75b5f179b2d3b1674687b479</icon>
  <id>https://xiaozhuanggao.github.io/</id>
  <link href="https://xiaozhuanggao.github.io/" rel="alternate"/>
  <link href="https://xiaozhuanggao.github.io/atom.xml" rel="self"/>
  <rights>All rights reserved 2026, 高晓庄</rights>
  <subtitle>工程师 / 架构师 / 偶尔读点哲学和经济学</subtitle>
  <title>高晓庄 札记</title>
  <updated>2026-11-04T09:19:29.712Z</updated>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="LLM" scheme="https://xiaozhuanggao.github.io/categories/AI/LLM/"/>
    <category term="大模型" scheme="https://xiaozhuanggao.github.io/tags/%E5%A4%A7%E6%A8%A1%E5%9E%8B/"/>
    <category term="混元" scheme="https://xiaozhuanggao.github.io/tags/%E6%B7%B7%E5%85%83/"/>
    <category term="Hy4" scheme="https://xiaozhuanggao.github.io/tags/Hy4/"/>
    <category term="评测方法" scheme="https://xiaozhuanggao.github.io/tags/%E8%AF%84%E6%B5%8B%E6%96%B9%E6%B3%95/"/>
    <id>https://xiaozhuanggao.github.io/2026/11/04/hy4-first-look/</id>
    <link href="https://xiaozhuanggao.github.io/2026/11/04/hy4-first-look/"/>
    <published>2026-11-04T09:00:00.000Z</published>
    <summary>发布当天最值得读的不是那张跑分表，而是官方自己写在最后一节的局限性。本文不含独立实测，只做口径核验与复测方案。</summary>
    <title>腾讯 Hy4 首发速评：770B 跑分该怎么读</title>
    <updated>2026-11-04T09:19:29.712Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="哲学" scheme="https://xiaozhuanggao.github.io/categories/%E5%93%B2%E5%AD%A6/"/>
    <category term="人生哲学" scheme="https://xiaozhuanggao.github.io/tags/%E4%BA%BA%E7%94%9F%E5%93%B2%E5%AD%A6/"/>
    <category term="程乐松" scheme="https://xiaozhuanggao.github.io/tags/%E7%A8%8B%E4%B9%90%E6%9D%BE/"/>
    <category term="善待自己" scheme="https://xiaozhuanggao.github.io/tags/%E5%96%84%E5%BE%85%E8%87%AA%E5%B7%B1/"/>
    <category term="一席" scheme="https://xiaozhuanggao.github.io/tags/%E4%B8%80%E5%B8%AD/"/>
    <category term="随笔" scheme="https://xiaozhuanggao.github.io/tags/%E9%9A%8F%E7%AC%94/"/>
    <id>https://xiaozhuanggao.github.io/2026/09/10/58c38cf9/</id>
    <link href="https://xiaozhuanggao.github.io/2026/09/10/58c38cf9/"/>
    <published>2026-09-10T14:30:00.000Z</published>
    <summary>北大哲学教授程乐松在「一席」的演讲里戳中了一个我们都不愿直面的真相——我们一边复制别人,一边否定当下的自己。</summary>
    <title>善待自己是一剂不保证有效的药方 — 听程乐松讲</title>
    <updated>2026-11-04T09:19:29.712Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="哲学" scheme="https://xiaozhuanggao.github.io/categories/%E5%93%B2%E5%AD%A6/"/>
    <category term="人生哲学" scheme="https://xiaozhuanggao.github.io/tags/%E4%BA%BA%E7%94%9F%E5%93%B2%E5%AD%A6/"/>
    <category term="随笔" scheme="https://xiaozhuanggao.github.io/tags/%E9%9A%8F%E7%AC%94/"/>
    <category term="罗永浩" scheme="https://xiaozhuanggao.github.io/tags/%E7%BD%97%E6%B0%B8%E6%B5%A9/"/>
    <category term="刘震云" scheme="https://xiaozhuanggao.github.io/tags/%E5%88%98%E9%9C%87%E4%BA%91/"/>
    <category term="中年" scheme="https://xiaozhuanggao.github.io/tags/%E4%B8%AD%E5%B9%B4/"/>
    <category term="十字路口" scheme="https://xiaozhuanggao.github.io/tags/%E5%8D%81%E5%AD%97%E8%B7%AF%E5%8F%A3/"/>
    <id>https://xiaozhuanggao.github.io/2026/09/08/1f5a47c0/</id>
    <link href="https://xiaozhuanggao.github.io/2026/09/08/1f5a47c0/"/>
    <published>2026-09-08T14:30:00.000Z</published>
    <summary>听《罗永浩的十字路口》第17期刘震云对谈,从&quot;咸的玩笑&quot;到&quot;心里藏事比较多&quot;,几个让我深夜释怀的瞬间。</summary>
    <title>听《十字路口》刘震云那一期,我突然不焦虑了——一个普通人的几个共鸣</title>
    <updated>2026-11-04T09:19:29.713Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="经济学" scheme="https://xiaozhuanggao.github.io/categories/%E7%BB%8F%E6%B5%8E%E5%AD%A6/"/>
    <category term="反思" scheme="https://xiaozhuanggao.github.io/tags/%E5%8F%8D%E6%80%9D/"/>
    <category term="读书" scheme="https://xiaozhuanggao.github.io/tags/%E8%AF%BB%E4%B9%A6/"/>
    <category term="社会经济学" scheme="https://xiaozhuanggao.github.io/tags/%E7%A4%BE%E4%BC%9A%E7%BB%8F%E6%B5%8E%E5%AD%A6/"/>
    <category term="内卷" scheme="https://xiaozhuanggao.github.io/tags/%E5%86%85%E5%8D%B7/"/>
    <id>https://xiaozhuanggao.github.io/2026/09/05/b7265a56/</id>
    <link href="https://xiaozhuanggao.github.io/2026/09/05/b7265a56/"/>
    <published>2026-09-05T13:00:00.000Z</published>
    <summary>内卷这个词被用烂了。读了点经济学之后我才发现，这其实是经济学的经典命题在互联网时代的具体呈现。</summary>
    <title>关于&quot;内卷&quot;的经济学解释——读《人月神话》和薛兆丰之后</title>
    <updated>2026-11-04T09:19:29.712Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="哲学" scheme="https://xiaozhuanggao.github.io/categories/%E5%93%B2%E5%AD%A6/"/>
    <category term="反思" scheme="https://xiaozhuanggao.github.io/tags/%E5%8F%8D%E6%80%9D/"/>
    <category term="人生哲学" scheme="https://xiaozhuanggao.github.io/tags/%E4%BA%BA%E7%94%9F%E5%93%B2%E5%AD%A6/"/>
    <category term="随笔" scheme="https://xiaozhuanggao.github.io/tags/%E9%9A%8F%E7%AC%94/"/>
    <category term="不确定性" scheme="https://xiaozhuanggao.github.io/tags/%E4%B8%8D%E7%A1%AE%E5%AE%9A%E6%80%A7/"/>
    <category term="人生" scheme="https://xiaozhuanggao.github.io/tags/%E4%BA%BA%E7%94%9F/"/>
    <id>https://xiaozhuanggao.github.io/2026/09/01/332cb110/</id>
    <link href="https://xiaozhuanggao.github.io/2026/09/01/332cb110/"/>
    <published>2026-09-01T14:00:00.000Z</published>
    <summary>二十几岁时我以为人生是可预测系统。这几年我承认：人生最大的事，没有一件是确定的。</summary>
    <title>不确定性是人生唯一确定的事——我不再假装掌控一切</title>
    <updated>2026-11-04T09:19:29.714Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="经济学" scheme="https://xiaozhuanggao.github.io/categories/%E7%BB%8F%E6%B5%8E%E5%AD%A6/"/>
    <category term="反思" scheme="https://xiaozhuanggao.github.io/tags/%E5%8F%8D%E6%80%9D/"/>
    <category term="自由市场" scheme="https://xiaozhuanggao.github.io/tags/%E8%87%AA%E7%94%B1%E5%B8%82%E5%9C%BA/"/>
    <category term="读书" scheme="https://xiaozhuanggao.github.io/tags/%E8%AF%BB%E4%B9%A6/"/>
    <category term="社会经济学" scheme="https://xiaozhuanggao.github.io/tags/%E7%A4%BE%E4%BC%9A%E7%BB%8F%E6%B5%8E%E5%AD%A6/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/30/816e7caf/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/30/816e7caf/"/>
    <published>2026-08-30T14:30:00.000Z</published>
    <summary>20 岁时我是个自由市场原教旨主义者。工作几年后我承认：市场不是万能的，政府也不是。</summary>
    <title>为什么我不再相信&quot;自由市场万能&quot;——读《资本主义与自由》之后</title>
    <updated>2026-11-04T09:19:29.712Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="RAG" scheme="https://xiaozhuanggao.github.io/tags/RAG/"/>
    <category term="评估" scheme="https://xiaozhuanggao.github.io/tags/%E8%AF%84%E4%BC%B0/"/>
    <category term="指标" scheme="https://xiaozhuanggao.github.io/tags/%E6%8C%87%E6%A0%87/"/>
    <category term="工程化" scheme="https://xiaozhuanggao.github.io/tags/%E5%B7%A5%E7%A8%8B%E5%8C%96/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/29/973a4638/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/29/973a4638/"/>
    <published>2026-08-29T02:00:00.000Z</published>
    <summary>没有评估指标,RAG 优化就是盲改。这套评估体系帮我把准确率从 70% 推到 92%。</summary>
    <title>RAG 评估体系——怎么知道你的 RAG 系统好不好</title>
    <updated>2026-11-04T09:19:29.713Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="哲学" scheme="https://xiaozhuanggao.github.io/categories/%E5%93%B2%E5%AD%A6/"/>
    <category term="反思" scheme="https://xiaozhuanggao.github.io/tags/%E5%8F%8D%E6%80%9D/"/>
    <category term="人生哲学" scheme="https://xiaozhuanggao.github.io/tags/%E4%BA%BA%E7%94%9F%E5%93%B2%E5%AD%A6/"/>
    <category term="随笔" scheme="https://xiaozhuanggao.github.io/tags/%E9%9A%8F%E7%AC%94/"/>
    <category term="放下" scheme="https://xiaozhuanggao.github.io/tags/%E6%94%BE%E4%B8%8B/"/>
    <category term="佛学" scheme="https://xiaozhuanggao.github.io/tags/%E4%BD%9B%E5%AD%A6/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/28/276c4dbc/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/28/276c4dbc/"/>
    <published>2026-08-28T15:00:00.000Z</published>
    <summary>放下这两个字被说烂了。我这两年才开始真的理解：它不是放弃，是另一种拿起。</summary>
    <title>这几年我开始理解&quot;放下&quot;——一些不那么好听的真话</title>
    <updated>2026-11-04T09:19:29.713Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="LLM部署" scheme="https://xiaozhuanggao.github.io/tags/LLM%E9%83%A8%E7%BD%B2/"/>
    <category term="vLLM" scheme="https://xiaozhuanggao.github.io/tags/vLLM/"/>
    <category term="TGI" scheme="https://xiaozhuanggao.github.io/tags/TGI/"/>
    <category term="SGLang" scheme="https://xiaozhuanggao.github.io/tags/SGLang/"/>
    <category term="推理优化" scheme="https://xiaozhuanggao.github.io/tags/%E6%8E%A8%E7%90%86%E4%BC%98%E5%8C%96/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/27/c019778a/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/27/c019778a/"/>
    <published>2026-08-27T02:00:00.000Z</published>
    <summary>自部署 LLM 选哪个推理框架?vLLM、TGI、SGLang 性能、吞吐、显存实测对比。</summary>
    <title>LLM 部署优化——vLLM、TGI、SGLang 三大推理框架实测</title>
    <updated>2026-11-04T09:19:29.713Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="工具" scheme="https://xiaozhuanggao.github.io/tags/%E5%B7%A5%E5%85%B7/"/>
    <category term="AI工作流" scheme="https://xiaozhuanggao.github.io/tags/AI%E5%B7%A5%E4%BD%9C%E6%B5%81/"/>
    <category term="n8n" scheme="https://xiaozhuanggao.github.io/tags/n8n/"/>
    <category term="自动化" scheme="https://xiaozhuanggao.github.io/tags/%E8%87%AA%E5%8A%A8%E5%8C%96/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/26/ab2c739d/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/26/ab2c739d/"/>
    <published>2026-08-26T02:00:00.000Z</published>
    <summary>把 AI 嵌入工作流,自动化重复任务。n8n + LLM 是 2025 年最强组合。</summary>
    <title>AI 工作流自动化——n8n + LLM 实战</title>
    <updated>2026-11-04T09:19:29.713Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="哲学" scheme="https://xiaozhuanggao.github.io/categories/%E5%93%B2%E5%AD%A6/"/>
    <category term="随笔" scheme="https://xiaozhuanggao.github.io/tags/%E9%9A%8F%E7%AC%94/"/>
    <category term="道德经" scheme="https://xiaozhuanggao.github.io/tags/%E9%81%93%E5%BE%B7%E7%BB%8F/"/>
    <category term="老子" scheme="https://xiaozhuanggao.github.io/tags/%E8%80%81%E5%AD%90/"/>
    <category term="无为" scheme="https://xiaozhuanggao.github.io/tags/%E6%97%A0%E4%B8%BA/"/>
    <category term="道家" scheme="https://xiaozhuanggao.github.io/tags/%E9%81%93%E5%AE%B6/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/25/3a7667a8/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/25/3a7667a8/"/>
    <published>2026-08-25T15:15:00.000Z</published>
    <summary>上善若水是道德经最有名的一句话，但真正让我半夜睡不着的，是另一段不那么出名的话。</summary>
    <title>道德经里那段让我半夜睡不着的话</title>
    <updated>2026-11-04T09:19:29.712Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="Agent" scheme="https://xiaozhuanggao.github.io/tags/Agent/"/>
    <category term="MCP" scheme="https://xiaozhuanggao.github.io/tags/MCP/"/>
    <category term="工具集成" scheme="https://xiaozhuanggao.github.io/tags/%E5%B7%A5%E5%85%B7%E9%9B%86%E6%88%90/"/>
    <category term="协议" scheme="https://xiaozhuanggao.github.io/tags/%E5%8D%8F%E8%AE%AE/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/23/51d6a071/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/23/51d6a071/"/>
    <published>2026-08-23T02:00:00.000Z</published>
    <summary>MCP 是 Anthropic 推出的&quot;AI 时代的 USB-C&quot;。一次接入,所有模型都能用。</summary>
    <title>MCP(Model Context Protocol)实战——AI 工具集成的新标准</title>
    <updated>2026-11-04T09:19:29.713Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="哲学" scheme="https://xiaozhuanggao.github.io/categories/%E5%93%B2%E5%AD%A6/"/>
    <category term="人生哲学" scheme="https://xiaozhuanggao.github.io/tags/%E4%BA%BA%E7%94%9F%E5%93%B2%E5%AD%A6/"/>
    <category term="随笔" scheme="https://xiaozhuanggao.github.io/tags/%E9%9A%8F%E7%AC%94/"/>
    <category term="王阳明" scheme="https://xiaozhuanggao.github.io/tags/%E7%8E%8B%E9%98%B3%E6%98%8E/"/>
    <category term="知行合一" scheme="https://xiaozhuanggao.github.io/tags/%E7%9F%A5%E8%A1%8C%E5%90%88%E4%B8%80/"/>
    <category term="心得" scheme="https://xiaozhuanggao.github.io/tags/%E5%BF%83%E5%BE%97/"/>
    <category term="儒家" scheme="https://xiaozhuanggao.github.io/tags/%E5%84%92%E5%AE%B6/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/20/609d5f1f/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/20/609d5f1f/"/>
    <published>2026-08-20T13:30:00.000Z</published>
    <summary>第一次读知行合一觉得是鸡汤，工作几年后重读，发现这是我活过这些年的真实写照。</summary>
    <title>重读&quot;知行合一&quot;——王阳明对我这些年的影响</title>
    <updated>2026-11-04T09:19:29.714Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="AI编程" scheme="https://xiaozhuanggao.github.io/tags/AI%E7%BC%96%E7%A8%8B/"/>
    <category term="Cursor" scheme="https://xiaozhuanggao.github.io/tags/Cursor/"/>
    <category term="Claude Code" scheme="https://xiaozhuanggao.github.io/tags/Claude-Code/"/>
    <category term="Copilot" scheme="https://xiaozhuanggao.github.io/tags/Copilot/"/>
    <category term="工具" scheme="https://xiaozhuanggao.github.io/tags/%E5%B7%A5%E5%85%B7/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/19/a0f8504b/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/19/a0f8504b/"/>
    <published>2026-08-19T02:00:00.000Z</published>
    <summary>AI 编程助手横评:Cursor、Claude Code、GitHub Copilot 谁更强?我从真实项目出发对比。</summary>
    <title>Cursor vs Claude Code vs Copilot——AI 编程助手横评</title>
    <updated>2026-11-04T09:19:29.712Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="向量数据库" scheme="https://xiaozhuanggao.github.io/tags/%E5%90%91%E9%87%8F%E6%95%B0%E6%8D%AE%E5%BA%93/"/>
    <category term="Chroma" scheme="https://xiaozhuanggao.github.io/tags/Chroma/"/>
    <category term="Milvus" scheme="https://xiaozhuanggao.github.io/tags/Milvus/"/>
    <category term="Qdrant" scheme="https://xiaozhuanggao.github.io/tags/Qdrant/"/>
    <category term="Weaviate" scheme="https://xiaozhuanggao.github.io/tags/Weaviate/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/18/7ca1eca5/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/18/7ca1eca5/"/>
    <published>2026-08-18T02:00:00.000Z</published>
    <summary>向量数据库是 RAG 的核心组件。4 个主流选哪个?我的实测对比告诉你答案。</summary>
    <title>向量数据库横评——Chroma、Milvus、Qdrant、Weaviate 怎么选</title>
    <updated>2026-11-04T09:19:29.714Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="知识图谱" scheme="https://xiaozhuanggao.github.io/tags/%E7%9F%A5%E8%AF%86%E5%9B%BE%E8%B0%B1/"/>
    <category term="RAG" scheme="https://xiaozhuanggao.github.io/tags/RAG/"/>
    <category term="GraphRAG" scheme="https://xiaozhuanggao.github.io/tags/GraphRAG/"/>
    <category term="Neo4j" scheme="https://xiaozhuanggao.github.io/tags/Neo4j/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/16/1db12911/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/16/1db12911/"/>
    <published>2026-08-16T02:00:00.000Z</published>
    <summary>基础 RAG 是&quot;找相似文档&quot;,GraphRAG 是&quot;理解实体关系&quot;。跨文档推理场景下效果提升 40%。</summary>
    <title>GraphRAG 实战——用知识图谱增强 RAG 检索</title>
    <updated>2026-11-04T09:19:29.712Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="RAG" scheme="https://xiaozhuanggao.github.io/tags/RAG/"/>
    <category term="HyDE" scheme="https://xiaozhuanggao.github.io/tags/HyDE/"/>
    <category term="Self-RAG" scheme="https://xiaozhuanggao.github.io/tags/Self-RAG/"/>
    <category term="CRAG" scheme="https://xiaozhuanggao.github.io/tags/CRAG/"/>
    <category term="检索增强" scheme="https://xiaozhuanggao.github.io/tags/%E6%A3%80%E7%B4%A2%E5%A2%9E%E5%BC%BA/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/14/770527e3/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/14/770527e3/"/>
    <published>2026-08-14T02:00:00.000Z</published>
    <summary>基础 RAG 准确率 70% 出头。HyDE / Self-RAG / CRAG 能把它推到 90%+。怎么做?</summary>
    <title>RAG 进阶——HyDE、Self-RAG、CRAG 三种增强检索实战</title>
    <updated>2026-11-04T09:19:29.713Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="AI框架" scheme="https://xiaozhuanggao.github.io/tags/AI%E6%A1%86%E6%9E%B6/"/>
    <category term="Dify" scheme="https://xiaozhuanggao.github.io/tags/Dify/"/>
    <category term="Coze" scheme="https://xiaozhuanggao.github.io/tags/Coze/"/>
    <category term="FastGPT" scheme="https://xiaozhuanggao.github.io/tags/FastGPT/"/>
    <category term="RAG" scheme="https://xiaozhuanggao.github.io/tags/RAG/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/11/f27d38d2/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/11/f27d38d2/"/>
    <published>2026-08-11T02:00:00.000Z</published>
    <summary>三大 AI 应用编排平台实测对比:功能、上手难度、扩展性、生产能力。</summary>
    <title>Dify vs Coze vs FastGPT——AI 应用编排平台横评</title>
    <updated>2026-11-04T09:19:29.712Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="RAG" scheme="https://xiaozhuanggao.github.io/tags/RAG/"/>
    <category term="LangChain" scheme="https://xiaozhuanggao.github.io/tags/LangChain/"/>
    <category term="框架" scheme="https://xiaozhuanggao.github.io/tags/%E6%A1%86%E6%9E%B6/"/>
    <category term="LlamaIndex" scheme="https://xiaozhuanggao.github.io/tags/LlamaIndex/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/08/caeb3bda/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/08/caeb3bda/"/>
    <published>2026-08-08T02:00:00.000Z</published>
    <summary>LangChain 和 LlamaIndex 是 AI 应用最常用的两个框架。但它们的定位完全不同。</summary>
    <title>LlamaIndex vs LangChain——两个框架的真正区别</title>
    <updated>2026-11-04T09:19:29.713Z</updated>
  </entry>
  <entry>
    <author>
      <name>高晓庄</name>
      <email>gaoxiaozhuang@bingosoft.net</email>
    </author>
    <category term="AI" scheme="https://xiaozhuanggao.github.io/categories/AI/"/>
    <category term="Agent" scheme="https://xiaozhuanggao.github.io/tags/Agent/"/>
    <category term="记忆" scheme="https://xiaozhuanggao.github.io/tags/%E8%AE%B0%E5%BF%86/"/>
    <category term="向量数据库" scheme="https://xiaozhuanggao.github.io/tags/%E5%90%91%E9%87%8F%E6%95%B0%E6%8D%AE%E5%BA%93/"/>
    <category term="知识图谱" scheme="https://xiaozhuanggao.github.io/tags/%E7%9F%A5%E8%AF%86%E5%9B%BE%E8%B0%B1/"/>
    <id>https://xiaozhuanggao.github.io/2026/08/05/987aaa5f/</id>
    <link href="https://xiaozhuanggao.github.io/2026/08/05/987aaa5f/"/>
    <published>2026-08-05T02:00:00.000Z</published>
    <summary>Agent 没有记忆就是个&quot;金鱼&quot;。短期、长期、向量、图四种记忆,各适合什么场景?</summary>
    <title>Agent 记忆系统——短期、长期、向量、图四种记忆怎么选</title>
    <updated>2026-11-04T09:19:29.712Z</updated>
  </entry>
</feed>
