[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-openclaw-headless-mode-ollama-zh":3,"article-related-openclaw-headless-mode-ollama-zh":33,"series-tools-3a2cfa8e-b508-4a86-b2fc-23a1d5f72cb1":85},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":25,"views":29,"created_at":30,"published_at":31,"topic_cluster_id":32},"3a2cfa8e-b508-4a86-b2fc-23a1d5f72cb1","openclaw-headless-mode-ollama-zh","Ollama 讓 OpenClaw 可無頭執行","\u003Cp data-speakable=\"summary\">Ollama 現在讓 \u003Ca href=\"\u002Ftag\u002Fopenclaw\">OpenClaw\u003C\u002Fa> 能用無頭模式跑腳本、\u003Ca href=\"\u002Ftag\u002Fdocker\">Docker\u003C\u002Fa> 和 CI 工作。\u003C\u002Fp>\u003Cp>說真的，這改動很實用。\u003Ca href=\"https:\u002F\u002Fdocs.ollama.com\u002Fintegrations\u002Fopenclaw\" target=\"_blank\" rel=\"noopener\">OpenClaw\u003C\u002Fa> 已經進到 \u003Ca href=\"https:\u002F\u002Follama.com\" target=\"_blank\" rel=\"noopener\">Ollama\u003C\u002Fa> 的助理清單。現在只要一條指令，就能少掉一堆互動步驟。\u003C\u002Fp>\u003Cp>官方給的範例很直接：\u003Ccode>ollama launch openclaw --model kimi-k2.5:cloud --yes\u003C\u002Fcode>。它會自動拉模型，也會跳過選單。對自動化來說，這種設計才像樣。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>內容\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>無頭指令\u003C\u002Ftd>\u003Ctd>\u003Ccode>ollama launch openclaw --model kimi-k2.5:cloud --yes\u003C\u002Fcode>\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>必要旗標\u003C\u002Ftd>\u003Ctd>\u003Ccode>--yes\u003C\u002Fcode> 會跳過互動確認\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>模型要求\u003C\u002Ftd>\u003Ctd>無互動模式下必須指定 \u003Ccode>--model\u003C\u002Fcode>\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>建議本機上下文\u003C\u002Ftd>\u003Ctd>至少 64k tokens\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>本機模型例子\u003C\u002Ftd>\u003Ctd>\u003Ccode>gemma4\u003C\u002Fcode> 約需 16 GB VRAM\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>本機模型例子\u003C\u002Ftd>\u003Ctd>\u003Ccode>qwen3.5\u003C\u002Fcode> 約需 11 GB VRAM\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>Ollama 這次到底改了什麼\u003C\u002Fh2>\u003Cp>以前跑 OpenClaw，流程比較像在顧一台剛裝好的機器。你要先啟動，再看提示，再選模型，再確認權限。現在 Ollama 把這些步驟包起來了。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779411978099-6l70.png\" alt=\"Ollama 讓 OpenClaw 可無頭執行\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>它可以自動安裝 OpenClaw。第一次啟動時，也會顯示安全提示。接著會選模型、設 provider、安裝 gateway daemon，還會啟動內建的 web search provider。\u003C\u002Fp>\u003Cp>講白了，就是把「能不能跑」改成「能不能穩定重跑」。這對開發者很重要。因為 CI、Docker、腳本都不愛跟你玩互動式問答。\u003C\u002Fp>\u003Cul>\u003Cli>OpenClaw 不存在時，會自動安裝。\u003C\u002Fli>\u003Cli>第一次啟動會顯示工具存取安全提示。\u003C\u002Fli>\u003Cli>Ollama 會把選定模型設成主模型。\u003C\u002Fli>\u003Cli>gateway 會在背景啟動。\u003C\u002Fli>\u003Cli>OpenClaw 也能走無頭路徑。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>無頭模式為什麼實用\u003C\u002Fh2>\u003Cp>這次真正有價值的，不是名字，而是執行方式。無頭模式代表你可以把 OpenClaw 放進容器、流水線、排程器，還有各種自動化腳本。\u003C\u002Fp>\u003Cp>如果還要人工按選單，那就很難進正式流程。你可能今天能跑，明天因為提示變了就卡住。這種狀況在 CI 裡超煩。\u003C\u002Fp>\u003Cp>Ollama 的文件寫得很清楚。\u003Ccode>--yes\u003C\u002Fcode> 會自動拉模型、跳過選單，而且還要求你指定 \u003Ccode>--model\u003C\u002Fcode>。這樣做的好處很單純，就是可預測。\u003C\u002Fp>\u003Cblockquote>“The \u003Ccode>--yes\u003C\u002Fcode> flag auto-pulls the model, skips selectors, and requires \u003Ccode>--model\u003C\u002Fcode> to be specified.”\u003C\u002Fblockquote>\u003Cp>這句話幾乎就把\u003Ca href=\"\u002Fnews\u002F5-cloudflare-anthropic-deal-zh\">重點\u003C\u002Fa>講完了。它不是在做花俏功能，而是在補自動化最缺的那一塊：穩定啟動。\u003C\u002Fp>\u003Ch2>模型選擇變成重點\u003C\u002Fh2>\u003Cp>Ollama 這次也把模型選擇拉到前台。這很合理。因為 OpenClaw 不是單純聊天工具，而是會碰到工具調用、搜尋、代理流程的 \u003Ca href=\"\u002Ftag\u002Fagent\">agent\u003C\u002Fa> gateway。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779411952121-jhs5.png\" alt=\"Ollama 讓 OpenClaw 可無頭執行\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>官方把模型分成雲端和本機兩種。雲端選項像 \u003Ca href=\"https:\u002F\u002Follama.com\u002Fsearch?q=kimi-k2.5\" target=\"_blank\" rel=\"noopener\">kimi-k2.5:cloud\u003C\u002Fa>，主打 multimodal reasoning 和 subagents。\u003Ca href=\"https:\u002F\u002Follama.com\u002Fsearch?q=qwen3.5\" target=\"_blank\" rel=\"noopener\">qwen3.5:cloud\u003C\u002Fa> 則偏 reasoning、coding 和 vision。\u003C\u002Fp>\u003Cp>本機部分也很務實。\u003Ca href=\"https:\u002F\u002Follama.com\u002Fsearch?q=gemma4\" target=\"_blank\" rel=\"noopener\">gemma4\u003C\u002Fa> 大約要 16 GB VRAM。\u003Ca href=\"https:\u002F\u002Follama.com\u002Fsearch?q=qwen3.5\" target=\"_blank\" rel=\"noopener\">qwen3.5\u003C\u002Fa> 約 11 GB VRAM。對很多筆電或工作站來說，這\u003Ca href=\"\u002Fnews\u002Fwei-shen-me-2026-indiana-fever-jiu-shi-yao-xian-zai-ying-zh-zh\">就是要\u003C\u002Fa>先算帳的地方。\u003C\u002Fp>\u003Cul>\u003Cli>\u003Ccode>kimi-k2.5:cloud\u003C\u002Fcode> 適合多模態代理工作。\u003C\u002Fli>\u003Cli>\u003Ccode>qwen3.5:cloud\u003C\u002Fcode> 偏推理、寫碼、視覺。\u003C\u002Fli>\u003Cli>\u003Ccode>glm-5.1:cloud\u003C\u002Fcode> 主打推理與 code generation。\u003C\u002Fli>\u003Cli>\u003Ccode>minimax-m2.7:cloud\u003C\u002Fcode> 偏快、偏省，適合生產力任務。\u003C\u002Fli>\u003Cli>本機跑法至少要顧 64k tokens 與記憶體空間。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>這和舊的 Clawdbot 路線怎麼接\u003C\u002Fh2>\u003Cp>OpenClaw 以前叫 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenclaw\" target=\"_blank\" rel=\"noopener\">Clawdbot\u003C\u002Fa>，Ollama 也還保留這個別名。這點不錯。因為很多人早就把舊指令寫進筆記或腳本。\u003C\u002Fp>\u003Cp>所以 \u003Ccode>ollama launch clawdbot\u003C\u002Fcode> 還能用。這種相容性很重要。你不想因為改名，就把一整批自動化流程弄壞。\u003C\u002Fp>\u003Cp>另外還有手動設定路線。你可以用 \u003Ccode>ollama launch openclaw --config\u003C\u002Fcode> 改模型。也能用 \u003Ccode>openclaw configure --section web\u003C\u002Fcode> 調整搜尋。若要串訊息\u003Ca href=\"\u002Fnews\u002Fkpmg-anthropic-digital-gateway-claude-zh\">平台\u003C\u002Fa>，\u003Ccode>openclaw configure --section channels\u003C\u002Fcode> 可以連 WhatsApp、Telegram、Slack、Discord、iMessage。\u003C\u002Fp>\u003Cul>\u003Cli>舊指令 \u003Ccode>ollama launch clawdbot\u003C\u002Fcode> 仍可用。\u003C\u002Fli>\u003Cli>模型可透過 \u003Ccode>--config\u003C\u002Fcode> 調整。\u003C\u002Fli>\u003Cli>web 搜尋可獨立設定。\u003C\u002Fli>\u003Cli>channels 區塊可連接多個訊息平台。\u003C\u002Fli>\u003Cli>launcher 和設定分開，對維運比較友善。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>開發者現在該怎麼看\u003C\u002Fh2>\u003Cp>如果你想把 OpenClaw 放進自動化，建議先用 headless 指令實測。先確認模型能順利載入，再看 gateway 是否正常啟動。不要直接塞進 production，這種東西最怕環境差一點就翻車。\u003C\u002Fp>\u003Cp>如果你要跑本機模型，先看 VRAM 和 context window。16 GB、11 GB、64k tokens，這些數字都不是裝飾品。它們直接決定你能不能穩定跑。\u003C\u002Fp>\u003Cp>我覺得這次更新的價值很明確。它把 OpenClaw 從「可以手動玩」拉到「可以進流程」。如果你的工作流需要的是 agent gateway，不是單純聊天視窗，那這個無頭模式就很有用。\u003C\u002Fp>\u003Cp>接下來最該做的事很簡單：拿你的實際模型跑一次，記錄啟動時間、記憶體占用、失敗率。那才知道它適不適合放進 CI 或 Docker。","Ollama 為 OpenClaw 加上無頭執行模式，適合 Docker、CI\u002FCD 和腳本。只要加上 --yes，還要指定模型，就能跳過互動流程。","docs.ollama.com","https:\u002F\u002Fdocs.ollama.com\u002Fintegrations\u002Fopenclaw",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779411978099-6l70.png","tools","zh","3c939d79-8282-4818-aa82-af78d041658c",[17,18,19,20,21,22,23,24],"Ollama","OpenClaw","無頭模式","CI\u002FCD","Docker","LLM","agent gateway","AI coding agent",[26,27,28],"Ollama 為 OpenClaw 加上無頭執行，適合腳本、Docker 和 CI。","無互動模式下必須加 --yes，還要指定 --model。","模型與硬體配置變成重點，尤其是 VRAM 與 64k tokens。",4,"2026-05-22T01:05:28.381201+00:00","2026-05-22T01:05:28.228+00:00","c3c88dd2-a940-438a-b359-0e5a24562273",{"tags":34,"relatedLang":44,"relatedPosts":48},[35,36,38,40,42],{"name":19,"slug":19},{"name":17,"slug":37},"ollama",{"name":18,"slug":39},"openclaw",{"name":21,"slug":41},"docker",{"name":20,"slug":43},"cicd",{"id":15,"slug":45,"title":46,"language":47},"openclaw-headless-mode-ollama-en","OpenClaw can now run headless in Ollama","en",[49,55,61,67,73,79],{"id":50,"slug":51,"title":52,"cover_image":53,"image_url":53,"created_at":54,"category":13},"5656a6ab-9e07-41be-9cea-3440fb8846e2","nvidia-lg-ai-collaboration-playbook-zh","Nvidia 和 LG 把 AI 合作變成模板","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781056994999-8eng.png","2026-06-10T02:02:46.590133+00:00",{"id":56,"slug":57,"title":58,"cover_image":59,"image_url":59,"created_at":60,"category":13},"e48be66d-d7de-419e-b5fd-805f0784ef15","ollama-best-free-ai-path-2026-zh","Ollama 是 2026 年真正適合工作的免費 AI 路徑","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781056077878-11pc.png","2026-06-10T01:47:24.632993+00:00",{"id":62,"slug":63,"title":64,"cover_image":65,"image_url":65,"created_at":66,"category":13},"9b53427c-8c2a-4960-a773-f14d4528caae","awesome-production-ml-turns-chaos-into-stack-zh","這份 MLOps 清單把混亂拆成堆疊","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781055220958-dmar.png","2026-06-10T01:33:14.850634+00:00",{"id":68,"slug":69,"title":70,"cover_image":71,"image_url":71,"created_at":72,"category":13},"d5af1522-28aa-4cfb-8779-1ecf168bc0b5","bentoml-turns-model-serving-into-python-apis-zh","BentoML 把模型服務變成 Python API","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781054310299-c1gm.png","2026-06-10T01:17:56.193093+00:00",{"id":74,"slug":75,"title":76,"cover_image":77,"image_url":77,"created_at":78,"category":13},"63d8b456-ad6b-475e-86e9-d4677ca226aa","magenta-realtime-2-score-inside-daw-zh","Magenta RealTime 2 讓你在 DAW 裡即時改曲","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781046204038-8tox.png","2026-06-09T23:02:55.9651+00:00",{"id":80,"slug":81,"title":82,"cover_image":83,"image_url":83,"created_at":84,"category":13},"f60261ff-a42e-4cfb-9f90-97785e633289","open-source-ai-tools-beat-claude-paid-tiers-zh","開源 AI 工具在價值上已經贏過 Claude 付費方案","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781045266035-on7t.png","2026-06-09T22:47:20.195939+00:00",[86,91,96,101,106,111,116,121,126,131],{"id":87,"slug":88,"title":89,"created_at":90},"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":92,"slug":93,"title":94,"created_at":95},"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":97,"slug":98,"title":99,"created_at":100},"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":102,"slug":103,"title":104,"created_at":105},"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":107,"slug":108,"title":109,"created_at":110},"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":112,"slug":113,"title":114,"created_at":115},"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":117,"slug":118,"title":119,"created_at":120},"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":122,"slug":123,"title":124,"created_at":125},"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":127,"slug":128,"title":129,"created_at":130},"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",{"id":132,"slug":133,"title":134,"created_at":135},"3ce6e6e2-bac5-463e-9f8d-45caabcc61f7","awesome-ai-for-science-research-tools-map-zh","AI 科研工具清單，開始像地圖了","2026-03-27T01:46:50.521945+00:00"]