https://blog.mklee.org/tags/architecture/Recent content in Architecture on MK's LabHugo -- 0.146.0zh-twThu, 16 Apr 2026 00:00:00 +0000https://blog.mklee.org/posts/2026-04-16-workspace-template-v3-openai-sdk-borrowings/Thu, 16 Apr 2026 00:00:00 +0000https://blog.mklee.org/posts/2026-04-16-workspace-template-v3-openai-sdk-borrowings/OpenAI 新的 Agents SDK 發布了 Manifest、Capabilities、serialize_session_state 三個概念。我們的 agent workspace template 定位完全不同，但有三個結構性想法值得借過來。https://blog.mklee.org/posts/context-tree-agent-memory/Sat, 28 Mar 2026 00:00:00 +0000https://blog.mklee.org/posts/context-tree-agent-memory/當 AI agent 的記憶系統塞滿了 37 個檔案，65% 是噪音，你需要的不是更大的 context window，而是更好的架構。這是把扁平日誌改造成 journal + knowledge 兩層記憶的實戰記錄。https://blog.mklee.org/posts/ai-agent-memory-three-challenges/Wed, 18 Mar 2026 00:00:00 +0000https://blog.mklee.org/posts/ai-agent-memory-three-challenges/context window 有限、偏好會變、新舊記憶會矛盾——分享我們在 OpenClaw 上實作記憶系統時遇到的三個核心難題與解法。https://blog.mklee.org/posts/agent-self-management/Fri, 06 Mar 2026 00:00:00 +0000https://blog.mklee.org/posts/agent-self-management/一個 AI agent 如何學會「省著點用 AI」——從全 LLM heartbeat 到 Rust binary，從被動維護到主動自省的演化記錄。https://blog.mklee.org/posts/openclaw-memory-architecture/Wed, 18 Feb 2026 00:00:00 +0000https://blog.mklee.org/posts/openclaw-memory-architecture/一個 AI agent 的記憶系統如何從空白 MEMORY.md 演化成帶有 Events Timeline、L0 索引、自動提取規則的結構化架構。