[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-llm-wiki-karpathy-knowledge-base-app-zh":3,"tags-llm-wiki-karpathy-knowledge-base-app-zh":27,"related-lang-llm-wiki-karpathy-knowledge-base-app-zh":28,"related-posts-llm-wiki-karpathy-knowledge-base-app-zh":32,"series-tools-c204da96-eaaf-4e34-b1c4-7dc1a28d6085":69},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":9,"image_url":10,"keywords":11,"language":17,"translated_content":9,"views":18,"is_premium":19,"created_at":20,"updated_at":20,"cover_image":10,"published_at":20,"rewrite_status":21,"rewrite_error":9,"rewritten_from_id":9,"slug":22,"category":23,"related_article_id":24,"status":25,"google_indexed_at":26,"x_posted_at":9,"tweet_text":9,"title_rewritten_at":9,"title_original":9,"key_takeaways":9,"topic_cluster_id":9,"embedding":9,"is_canonical_seed":19},"c204da96-eaaf-4e34-b1c4-7dc1a28d6085","Karpathy 的 LLM Wiki 三週變桌面應用","\u003Cp>四月初，Andrej Karpathy 在 GitHub Gist 上發布了一個簡單但深遠的想法：用 AI 將文檔集合漸進地編譯成一份自我維護的知識庫。不同於傳統 RAG 每次查詢都重新構建上下文，這套架構讓知識編譯一次，持續演化。Gist 貼出後迅速累積 5,000 多星。\u003C\u002Fp>\n\n\u003Cp>真正有趣的事發生在三週後。開發者 Nash Su 將這份概念文件轉化為一個名叫 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fnashsu\u002Fllm_wiki\" target=\"_blank\" rel=\"noopener\">LLM Wiki\u003C\u002Fa> 的完整跨平臺桌面應用，現已累積 4,400 星。v0.3.13 剛在四月底發布，搭載 dmg\u002Frpm\u002Fdeb 安裝程式和 Chrome 外掛。速度驚人。\u003C\u002Fp>\n\n\u003Ch2>從想法到執行的架構\u003C\u002Fh2>\n\n\u003Cp>LLM Wiki 實現了 Karpathy 素描的三層架構：原始文檔不可變存放在 raw\u002F 資料夾，LLM 生成的知識庫頁面保存在 wiki\u002F 資料夾，一份 CLAUDE.md 式的綱要檔案定義整個系統的結構和慣例。三個核心操作—— ingest（攝入新來源）、query（查詢並合成）、lint（維護和檢測矛盾）—— 讓知識庫隨時保持一致性和可追溯性。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777388011003-f0pt.png\" alt=\"Karpathy 的 LLM Wiki 三週變桌面應用\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\n\n\u003Cp>但實裝遠不止於此。該應用引入了一個關鍵的兩階段攝入流程：先分析（LLM 提取實體、概念、潛在矛盾，尋找既有維基頁面的關聯），再生成（製作摘要頁、實體頁、概念頁，更新索引，標記需人工檢視的項目）。一份來源文件可以級聯更新 10 到 15 個維基頁面。搭配 SHA256 增量快取和三次重試機制，系統既快又韌性強。\u003C\u002Fp>\n\n\u003Ch2>知識圖譜的新玩法\u003C\u002Fh2>\n\n\u003Cp>最引人注目的創新是知識圖譜設計。系統用四維邊權重量化概念間的關聯：直接維基連結（×3.0 權重）、來源重疊（×4.0，兩頁引用同一文檔）、Adamic-Adar 共同鄰居（×1.5，稀有共享鄰居計重更高）、類型親和力（×1.0）。透過 Sigma.js 和 ForceAtlas2 佈局演算法視覺化，Louvain 社群偵測自動發現話題叢集。\u003C\u002Fp>\n\n\u003Cp>圖譜衍生出兩類洞察：驚奇連結（跨叢集、跨類型的非預期邊）被標記為「認知突破的起點」；知識缺口（孤立頁面、稀疏連結的社群、連接 3+ 叢集的樞紐節點）各配一個「深度研究」按鈕，點擊即啟動 Tavily API 搜尋，合成結果再注入攝入管線。確認對話框防止無意觸發。\u003C\u002Fp>\n\n\u003Cp>檢索層也突破傳統做法。文本分詞（中文用二元切分）、可選的向量 ANN（透過 LanceDB）、從搜尋種子開始的二跳圖遍歷，三者合力達成多階段檢索。使用者可調整上下文視窗（4K 到 1M token），分配比例為 60% 維基 \u002F 20% 對話歷史 \u002F 5% 索引 \u002F 15% 系統提示。文檔報告顯示召回率從 58.2% 提升至 71.4%（啟用向量搜尋時）。\u003C\u002Fp>\n\n\u003Cp>另一個細節：每個專案配一份 purpose.md 意圖檔案。LLM 在每次攝入和查詢時都讀取它，確保知識庫有方向感，不會淪為無向累積。刪除來源時系統也足夠聰慧：自動清理其摘要頁，移除共享頁面中的該來源引用（但保留共享實體本身），刪除索引項和斷連。\u003C\u002Fp>\n\n\u003Ch2>代理人時代的知識編譯\u003C\u002Fh2>\n\n\u003Cp>這個故事的核心不在單一功能，而在迭代速度和設計哲學的轉變。Karpathy 的想法在 Gist 上流傳——這是開源時代知識循環的新形式。一份建築草圖，共享給社群，由能幹的開發者在數週內具現化成可用產品。RAG 不是錯，但它從預設方案降級為備選方案。真正的前沿是把知識視為編譯後的製品，而非即時查詢目標。mem0 和 MemOS 為 Agent 解決了這個問題；LLM Wiki 為人類使用者解決了它。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777388008222-c6wb.png\" alt=\"Karpathy 的 LLM Wiki 三週變桌面應用\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\n\n\u003Cp>使用者也能直接用 Obsidian 開啟生成的維基目錄——該應用完全相容 Obsidian Vault 格式（含 .obsidian\u002F 配置）。LLM Wiki 作編輯器，Obsidian 作檢視器，兩個工具各盡其能。從來不是 LLM 要取代人類工具，而是代理人時代的工具組合方式從「單一巨人應用」轉向「互操作的專用工具生態」。\u003C\u002Fp>","Karpathy 在四月發布的 LLM Wiki 概念只是一份 Gist 建議，三週後就有開發者推出完整的跨平臺應用。這個速度反映了代理人時代開源迭代的新節奏：架構理念快速循環成可用產品，知識庫從查詢驅動轉向編譯驅動。","oracore-original",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777388011003-f0pt.png",[12,13,14,15,16],"LLM Wiki","Karpathy","Knowledge Graph","Obsidian","RAG Alternative","zh",0,false,"2026-04-28T14:52:19.100239+00:00","done","llm-wiki-karpathy-knowledge-base-app-zh","tools","b7414266-729b-4471-8a0f-6e1f7ea8f04b","published","2026-04-29T09:00:09.752+00:00",[],{"id":24,"slug":29,"title":30,"language":31},"llm-wiki-karpathy-knowledge-base-app-en","From Karpathy Gist to 4,400-Star App: LLM Wiki in Three Weeks","en",[33,39,45,51,57,63],{"id":34,"slug":35,"title":36,"cover_image":37,"image_url":37,"created_at":38,"category":23},"d058a76f-6548-4135-8970-f3a97f255446","why-gemini-api-pricing-is-cheaper-than-it-looks-zh","為什麼 Gemini API 定價其實比看起來更便宜","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778869845081-j4m7.png","2026-05-15T18:30:25.797639+00:00",{"id":40,"slug":41,"title":42,"cover_image":43,"image_url":43,"created_at":44,"category":23},"68e4be16-dc38-4524-a6ea-5ebe22a6c4fb","why-vidhub-huiyuan-hutong-bushi-quan-shebei-tongyong-zh","為什麼 VidHub 會員互通不是「買一次全設備通用」","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778789450987-advz.png","2026-05-14T20:10:24.048988+00:00",{"id":46,"slug":47,"title":48,"cover_image":49,"image_url":49,"created_at":50,"category":23},"7a1e174f-746b-4e82-a0e3-b2475ab39747","why-buns-zig-to-rust-experiment-is-right-zh","為什麼 Bun 的 Zig-to-Rust 實驗是對的","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778767879127-5dna.png","2026-05-14T14:10:26.886397+00:00",{"id":52,"slug":53,"title":54,"cover_image":55,"image_url":55,"created_at":56,"category":23},"e742fc73-5a65-4db3-ad17-88c99262ceb7","why-openai-api-pricing-is-product-strategy-zh","為什麼 OpenAI API 定價是產品策略，不是註腳","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778749859485-chvz.png","2026-05-14T09:10:26.003818+00:00",{"id":58,"slug":59,"title":60,"cover_image":61,"image_url":61,"created_at":62,"category":23},"c757c5d8-eda9-45dc-9020-4b002f4d6237","why-claude-code-prompt-design-beats-ide-copilots-zh","為什麼 Claude Code 的提示設計贏過 IDE Copilot","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778742645084-dao9.png","2026-05-14T07:10:29.371901+00:00",{"id":64,"slug":65,"title":66,"cover_image":67,"image_url":67,"created_at":68,"category":23},"4adef3ab-9f07-4970-91cf-77b8b581b348","why-databricks-model-serving-is-right-default-zh","為什麼 Databricks Model Serving 是生產推論的正確預設","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778692245329-a2wt.png","2026-05-13T17:10:30.659153+00:00",[70,75,80,85,90,95,100,105,110,115],{"id":71,"slug":72,"title":73,"created_at":74},"de769291-4574-4c46-a76d-772bd99e6ec9","googles-biggest-gemini-launches-in-2026-zh","Google 2026 最大 Gemini 盤點","2026-03-26T07:26:39.21072+00:00",{"id":76,"slug":77,"title":78,"created_at":79},"855cd52f-6fab-46cc-a7c1-42195e8a0de4","surepath-real-time-mcp-policy-controls-zh","SurePath 推出即時 MCP 政策控管","2026-03-26T07:57:40.77233+00:00",{"id":81,"slug":82,"title":83,"created_at":84},"9b19ab54-edef-4dbd-9ce4-a51e4bae4ebb","mcp-in-2026-the-ai-tool-layer-teams-use-zh","2026 年 MCP：團隊真的在用的 AI 工具層","2026-03-26T08:01:46.589694+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"af9c46c3-7a28-410b-9f04-32b3de30a68c","prompting-in-2026-what-actually-works-zh","2026 提示工程，真正有用的是什麼","2026-03-26T08:08:12.453028+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"05553086-6ed0-4758-81fd-6cab24b575e0","garry-tan-open-sources-claude-code-toolkit-zh","Garry Tan 開源 Claude Code 工具包","2026-03-26T08:26:20.068737+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"042a73a2-18a2-433d-9e8f-9802b9559aac","github-ai-projects-to-watch-in-2026-zh","2026 必看 20 個 GitHub AI 專案","2026-03-26T08:28:09.619964+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"a5f94120-ac0d-4483-9a8b-63590071ac6a","claude-code-vs-cursor-2026-zh","Claude Code 與 Cursor 深度對比：202…","2026-03-26T13:27:14.279193+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"0975afa1-e0c7-4130-a20d-d890eaed995e","practical-github-guide-learning-ml-2026-zh","2026 機器學習入門 GitHub 實用指南","2026-03-27T01:16:49.712576+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"bfdb467a-290f-4a80-b3a9-6f081afb6dff","aiml-2026-student-ai-ml-lab-repo-review-zh","AIML-2026：像課綱的學生實驗 Repo","2026-03-27T01:21:51.467798+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"80cabc3e-09fc-4ff5-8f07-b8d68f5ae545","ai-trending-github-repos-and-research-feeds-zh","AI Trending：把 AI 資源收成一張表","2026-03-27T01:31:35.262183+00:00"]