<?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>Agent Architecture on Yuchen Du</title><link>https://dudusoar.github.io/tags/agent-architecture/</link><description>Recent content in Agent Architecture on Yuchen Du</description><generator>Hugo -- 0.164.0</generator><language>en-us</language><lastBuildDate>Fri, 17 Jul 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://dudusoar.github.io/tags/agent-architecture/index.xml" rel="self" type="application/rss+xml"/><item><title>Separate Schemas, Prompts, Tools, Workflows, and Memory in an LLM System</title><link>https://dudusoar.github.io/posts/llm-system-architecture/01-five-layer-llm-system/</link><pubDate>Fri, 17 Jul 2026 00:00:00 +0000</pubDate><guid>https://dudusoar.github.io/posts/llm-system-architecture/01-five-layer-llm-system/</guid><description>Why an LLM decision system needs five separable layers rather than one increasingly large prompt.</description></item><item><title>Treat LLM Evidence Tools as Read-Only Interfaces</title><link>https://dudusoar.github.io/posts/llm-system-architecture/06-tools-as-read-only-evidence-interfaces/</link><pubDate>Fri, 17 Jul 2026 00:00:00 +0000</pubDate><guid>https://dudusoar.github.io/posts/llm-system-architecture/06-tools-as-read-only-evidence-interfaces/</guid><description>Evidence tools expose structured facts and deterministic calculations without making decisions on the model&amp;#39;s behalf.</description></item></channel></rss>