<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>LLM Inference on 崔亮 Cui Liang｜AI Agent 工程博客</title>
    <link>https://cuiliang.ai/tags/llm-inference/</link>
    <description>Recent content in LLM Inference on 崔亮 Cui Liang｜AI Agent 工程博客</description>
    <image>
      <title>崔亮 Cui Liang｜AI Agent 工程博客</title>
      <url>https://cuiliang.ai/og-default.png</url>
      <link>https://cuiliang.ai/og-default.png</link>
    </image>
    <generator>Hugo -- 0.147.0</generator>
    <language>zh</language>
    <lastBuildDate>Fri, 27 Feb 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://cuiliang.ai/tags/llm-inference/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>推理服务是怎么影响你的 Agent 的：推理框架与架构决策</title>
      <link>https://cuiliang.ai/posts/inference-frameworks-for-agent-architects/</link>
      <pubDate>Fri, 27 Feb 2026 00:00:00 +0000</pubDate>
      <guid>https://cuiliang.ai/posts/inference-frameworks-for-agent-architects/</guid>
      <description>自顶向下拆解推理框架：延迟构成、上下文成本、模型路由、多模态代价、韧性设计——面向 Agent 架构师的推理层指南。</description>
    </item>
    <item>
      <title>一句话是怎么变成 AI 回复的：LLM 的工作原理</title>
      <link>https://cuiliang.ai/posts/how-a-sentence-becomes-an-ai-response/</link>
      <pubDate>Sun, 15 Feb 2026 00:00:00 +0000</pubDate>
      <guid>https://cuiliang.ai/posts/how-a-sentence-becomes-an-ai-response/</guid>
      <description>从 Tokenization 到自回归生成，用一句「帮我查一下北京明天的天气」走完 LLM 处理全流程。面向 Agent 开发者的灰盒认知指南。</description>
    </item>
    <item>
      <title>Prompt Cache：Agent 成本控制的核心约束</title>
      <link>https://cuiliang.ai/posts/prompt-caching-core-constraints/</link>
      <pubDate>Sat, 17 Jan 2026 00:00:00 +0000</pubDate>
      <guid>https://cuiliang.ai/posts/prompt-caching-core-constraints/</guid>
      <description>从 KV Cache 到 Prompt Cache 的认知跃迁：前缀精确匹配的铁律、Agent 的 I/O 100:1 失衡、以及为什么 Cache-Safe 是 Client 侧工程责任。</description>
    </item>
    <item>
      <title>KV Cache 原理：LLM 推理的底层机制</title>
      <link>https://cuiliang.ai/posts/prompt-caching-kv-cache-fundamentals/</link>
      <pubDate>Sat, 10 Jan 2026 00:00:00 +0000</pubDate>
      <guid>https://cuiliang.ai/posts/prompt-caching-kv-cache-fundamentals/</guid>
      <description>从 Token、Embedding、Attention 到 Prefill/Decode，一次讲透 KV Cache 的底层原理。面向 Agent 工程师的推理基础指南。</description>
    </item>
  </channel>
</rss>
