https://blog.ivanwz.com/tags/%E5%90%91%E9%87%8F%E6%A3%80%E7%B4%A2/Recent content in 向量检索 on 云原生技术栈Hugo -- gohugo.iozh-cnhello@example.com (Ivan)hello@example.com (Ivan)© 2026 IvanFri, 19 Jun 2026 00:00:00 +0000https://blog.ivanwz.com/posts/posts/ai/llm-rag-starter/Wed, 17 Jun 2026 00:00:00 +0000hello@example.com (Ivan)https://blog.ivanwz.com/posts/posts/ai/llm-rag-starter/<p>RAG 是把外部知识接入大模型的常见方式。它并不神秘，本质上是一条从文档到答案的工程链路。</p> <h2 class="relative group">文档切分 <div id="文档切分" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#%e6%96%87%e6%a1%a3%e5%88%87%e5%88%86" aria-label="锚点">#</a> </span> </h2> <p>文档需要被切成适合检索的片段。切分太大，召回不精准；切分太小，上下文容易丢失。常见做法是按标题、段落和固定 token 数组合切分。</p>