[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-memos-plugin-fixes-hermes-agent-memory-zh":3,"tags-memos-plugin-fixes-hermes-agent-memory-zh":27,"related-lang-memos-plugin-fixes-hermes-agent-memory-zh":28,"related-posts-memos-plugin-fixes-hermes-agent-memory-zh":32,"series-ai-agent-d7849e5f-3341-4741-b047-e317e0bd0225":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},"d7849e5f-3341-4741-b047-e317e0bd0225","AI 代理人的記憶危機：為什麼本地模型需要語義記憶引擎","\u003Cp>開源 AI 代理框架 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FNousResearch\u002Fhermes-agent\" target=\"_blank\" rel=\"noopener\">Hermes Agent\u003C\u002Fa> 在開發者社群中獲得了廣泛關注，但其原生記憶層設計暴露了一個被廣泛忽視的問題：當代理超越數百輪對話的規模後，簡單的追加式日誌和關鍵詞文本匹配會迅速崩潰。\u003C\u002Fp>\n\n\u003Ch2>原生代理記憶的三個致命弱點\u003C\u002Fh2>\n\n\u003Cp>Hermes Agent 的記憶機制直接將每個對話輪次寫入 SQLite，並依賴關鍵詞檢索。這種設計在初期看似足夠，但隨著使用時間增長，三個問題逐漸浮現：\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777385820220-wqvv.png\" alt=\"AI 代理人的記憶危機：為什麼本地模型需要語義記憶引擎\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\n\n\u003Cul>\n\u003Cli>\u003Cstrong>重複累積\u003C\u002Fstrong>：相同或相似的事實被反覆記錄，資訊密度逐漸稀釋\u003C\u002Fli>\n\u003Cli>\u003Cstrong>矛盾共存\u003C\u002Fstrong>：「用戶正在節食」和「用戶放棄節食」被同時儲存為獨立記錄，代理無法判斷哪個是當前狀態\u003C\u002Fli>\n\u003Cli>\u003Cstrong>檢索衰減\u003C\u002Fstrong>：隨著記憶條目增加，文本匹配的訊噪比急劇下降，相關記憶常被淹沒在雜訊中\u003C\u002Fli>\n\u003C\u002Ful>\n\n\u003Cp>這不是 Hermes 獨有的問題。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw\" target=\"_blank\" rel=\"noopener\">OpenClaw\u003C\u002Fa> 和其他個人代理框架都存在相同的架構限制。甚至 ChatGPT 的自訂 GPT 記憶功能也面臨類似的可擴展性瓶頸。\u003C\u002Fp>\n\n\u003Ch2>MemOS 的記憶治理模型\u003C\u002Fh2>\n\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FMemTensor\u002FMemOS\" target=\"_blank\" rel=\"noopener\">MemOS\u003C\u002Fa> 是一個開源 AI 記憶作業系統，其 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FMemTensor\u002FMemOS\u002Ftree\u002Fmain\u002Fapps\u002Fmemos-local-plugin\" target=\"_blank\" rel=\"noopener\">Hermes 本地外掛\u003C\u002Fa>提供了不同的方案。寫入管道包含四個步驟：語義分塊 → LLM 摘要化 → 向量嵌入 → 智能去重。關鍵創新在於去重邏輯：不是文本相似度比較，而是讓 LLM 檢視新記憶與既有記憶的語義關係，判定是重複、更新還是新增。在節食例子中，這會合併成單一記錄並保留編輯歷史。\u003C\u002Fp>\n\n\u003Cp>檢索端採用混合搜尋引擎：全文搜尋和向量語義搜尋並行執行，然後透過融合排序、多樣性去重、時間衰減和相關性過濾進行重排。系統在對話開始時會預先取得使用者最新消息相關的記憶，避免檢索延遲。\u003C\u002Fp>\n\n\u003Ch2>分層模型治理的必要性\u003C\u002Fh2>\n\n\u003Cp>Hermes 原生使用單一模型處理記憶相關的所有任務，導致技能品質參差不齊。MemOS 採用三層配置：嵌入層（輕量）→ 摘要層（中等）→ 技能生成層（最強）。更重要的是，它加入了規則和 LLM 評估門檻，只有具備可重複性和實際價值的任務才會轉化為技能。當上層模型不可用時，系統會自動降級到下層模型，保證可靠性。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777385808581-7s0a.png\" alt=\"AI 代理人的記憶危機：為什麼本地模型需要語義記憶引擎\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\n\n\u003Cp>這個設計模式直指一個更廣泛的趨勢：代理框架正從「執行每個請求」演進到「治理執行的後果」。記憶不再是被動的日誌，而是需要主動管理的知識資產。\u003C\u002Fp>\n\n\u003Ch2>多代理協作與隱私邊界\u003C\u002Fh2>\n\n\u003Cp>MemOS 支援兩層協作架構。第一層允許同機多個 Hermes 實例共享通用記憶池和技能，但保持隔離；第二層跨機器提供 Hub-Client 架構，私有資料始終保留在本地，只有顯式共享的內容對團隊可見。這對需要隱私控制的企業或多人工作流至關重要。\u003C\u002Fp>","Hermes Agent 雖然擁有 12 萬星標，但其原生記憶系統採用純文本追加式儲存，無法有效去重和檢索。MemOS 本地外掛透過語義分塊、LLM 驗證和混合檢索，解決了自主 AI 代理在長期使用中的記憶衰退問題。這反映了一個關鍵趨勢：代理執行環境正從玩具級原型進入需要真實記憶治理的生產系統。","oracore-original",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777385820220-wqvv.png",[12,13,14,15,16],"Hermes Agent","MemOS","AI Memory","MemTensor","Agent Framework","zh",2,false,"2026-04-28T14:15:27.48825+00:00","done","memos-plugin-fixes-hermes-agent-memory-zh","ai-agent","90d31b81-cb67-46e8-82b7-ea3410e6acac","published","2026-04-29T09:00:09.819+00:00",[],{"id":24,"slug":29,"title":30,"language":31},"memos-plugin-fixes-hermes-agent-memory-en","Hermes Agent Hits a Memory Wall. MemOS Shows What Comes Next.","en",[33,39,45,51,57,63],{"id":34,"slug":35,"title":36,"cover_image":37,"image_url":37,"created_at":38,"category":23},"38406a12-f833-4c69-ae22-99c31f03dd52","switch-ai-outputs-markdown-to-html-zh","怎麼把 AI 輸出改成 HTML","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778743243861-8901.png","2026-05-14T07:20:21.545364+00:00",{"id":40,"slug":41,"title":42,"cover_image":43,"image_url":43,"created_at":44,"category":23},"c7c69fe4-97e3-4edf-a9d6-a79d0c4495b4","anthropic-cat-wu-proactive-ai-assistants-zh","Cat Wu 談 Claude 的主動式 AI","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778735455993-gnw7.png","2026-05-14T05:10:30.453046+00:00",{"id":46,"slug":47,"title":48,"cover_image":49,"image_url":49,"created_at":50,"category":23},"e1d6acda-fa49-4514-aa75-709504be9f93","how-to-run-hermes-agent-on-discord-zh","如何在 Discord 執行 Hermes Agent","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778724655796-cjul.png","2026-05-14T02:10:34.362605+00:00",{"id":52,"slug":53,"title":54,"cover_image":55,"image_url":55,"created_at":56,"category":23},"4104fa5f-d95f-45c5-9032-99416cf0365c","why-ragflow-is-the-right-open-source-rag-engine-to-self-host-zh","為什麼 RAGFlow 是最適合自架的開源 RAG 引擎","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778674262278-1630.png","2026-05-13T12:10:23.762632+00:00",{"id":58,"slug":59,"title":60,"cover_image":61,"image_url":61,"created_at":62,"category":23},"7095f05c-34f5-469f-a044-2525d2010ce9","how-to-add-temporal-rag-in-production-zh","如何在正式環境加入 Temporal RAG","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778667053844-osvs.png","2026-05-13T10:10:30.930982+00:00",{"id":64,"slug":65,"title":66,"cover_image":67,"image_url":67,"created_at":68,"category":23},"10479c95-53c6-4723-9aaa-2fde5fb19ee7","github-agentic-workflows-ai-github-actions-zh","GitHub 把 AI 代理放進 Actions","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778551884342-8io7.png","2026-05-12T02:11:02.069769+00:00",[70,75,80,85,90,95,100,105,110,115],{"id":71,"slug":72,"title":73,"created_at":74},"4ae1e197-1d3d-4233-8733-eafe9cb6438b","claude-now-uses-your-pc-to-finish-tasks-zh","Claude 開始幫你操作電腦","2026-03-26T07:20:48.457387+00:00",{"id":76,"slug":77,"title":78,"created_at":79},"5bede67f-e21c-413d-9ab8-54a3c3d26227","googles-2026-ai-agent-report-decoded-zh","Google 2026 AI Agent 報告解讀","2026-03-26T11:15:22.651956+00:00",{"id":81,"slug":82,"title":83,"created_at":84},"2987d097-563f-46c7-b76f-b558d8ef7c2b","kimi-k25-review-stronger-still-not-legend-zh","Kimi K2.5 評測：更強，但還不是神作","2026-03-27T07:15:55.277513+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"95c9053b-e3f4-4cb5-aace-5c54f4c9e044","claude-code-controls-mac-desktop-zh","Claude Code 也能操控 Mac 了","2026-03-28T03:01:58.58121+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"dc58e153-e3a8-4c06-9b96-1aa64eabbf5f","cloudflare-100x-faster-ai-agent-sandbox-zh","Cloudflare 的 AI 沙箱跑超快","2026-03-28T03:09:44.142236+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"1c8afc56-253f-47a2-979f-1065ff072f2a","openai-backs-isara-agent-swarm-bet-zh","OpenAI 挺 Isara 的 agent swarm …","2026-03-28T03:15:27.513155+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"7379b422-576e-45df-ad5a-d57a0d9dd467","openai-plan-automated-ai-researcher-zh","OpenAI 想做自動化 AI 研究員","2026-03-28T03:17:42.090548+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"48c9889e-86df-450b-a356-e4a4b7c83c5b","harness-engineering-ai-agent-reliability-2026-zh","駕馭工程：從「馬具」到「作業系統」，AI Agent 可靠性的終極密碼","2026-03-31T06:42:53.556721+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"e41546b8-ba9e-455f-9159-88d4614ad711","openai-codex-plugin-claude-code-zh","OpenAI 把 Codex 放進 Claude Code","2026-04-01T09:21:54.687617+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"96d8e8c8-1edd-475d-9145-b1e7a1b02b65","mcp-explained-from-prompts-to-production-zh","MCP 怎麼把提示詞變工作流","2026-04-01T09:24:39.321274+00:00"]