[{"data":1,"prerenderedAt":1220},["ShallowReactive",2],{"doc-page:\u002Fdocs\u002Fai-local-models":3},{"doc":4,"prev":1187,"next":1195,"resolvedType":8,"readingMinutes":113,"audience":1197,"checklist":1201,"related":1205},{"path":5,"title":6,"description":7,"docType":8,"resourceKind":9,"categoryId":10,"categoryLabel":11,"updatedAt":12,"publishedAt":12,"icon":13,"body":14},"\u002Fdocs\u002Fai-local-models","本地 AI 模型部署","Ollama、LM Studio、vLLM 本地大模型运行与 API 调用","article",null,"ai-tools","AI 工具","2026-02-28","i-carbon-chat-bot",{"type":15,"value":16,"toc":1159},"minimark",[17,21,25,28,32,67,71,74,78,166,169,176,307,311,318,399,403,419,540,544,547,661,665,729,759,763,766,775,778,799,802,805,883,886,943,946,954,965,969,973,1028,1035,1038,1041,1059,1062,1066,1069,1072,1075,1089,1093,1096,1099,1120,1123,1155],[18,19,6],"h1",{"id":20},"本地-ai-模型部署",[22,23,24],"p",{},"在本地运行大语言模型，保护隐私、无需联网、免费使用。",[22,26,27],{},"这页适合作为“本地模型快速起步页”。本地部署真正要先想清楚的不是先下哪个模型，而是你的机器预算、目标任务、交互方式和是否需要 OpenAI 兼容 API。",[29,30,31],"h2",{"id":31},"先按目标选",[33,34,35,43,49,55,61],"ul",{},[36,37,38,42],"li",{},[39,40,41],"strong",{},"想最快跑起来","：优先 Ollama",[36,44,45,48],{},[39,46,47],{},"想用图形界面试模型","：优先 LM Studio",[36,50,51,54],{},[39,52,53],{},"想做高性能 API 服务","：优先 vLLM",[36,56,57,60],{},[39,58,59],{},"想把模型接进本地工具链","：优先带 API 的方案",[36,62,63,66],{},[39,64,65],{},"想离线运行和保护隐私","：优先本地部署路线",[29,68,70],{"id":69},"ollama","Ollama",[22,72,73],{},"最简单的本地模型运行方案。",[75,76,77],"h3",{"id":77},"安装",[79,80,85],"pre",{"className":81,"code":82,"language":83,"meta":84,"style":84},"language-powershell shiki shiki-themes github-light github-dark","# Windows\nwinget install Ollama.Ollama\n\n# macOS\nbrew install ollama\n\n# Linux\ncurl -fsSL https:\u002F\u002Follama.com\u002Finstall.sh | sh\n","powershell","",[86,87,88,97,104,111,117,123,128,134],"code",{"__ignoreMap":84},[89,90,93],"span",{"class":91,"line":92},"line",1,[89,94,96],{"class":95},"sJ8bj","# Windows\n",[89,98,100],{"class":91,"line":99},2,[89,101,103],{"class":102},"sVt8B","winget install Ollama.Ollama\n",[89,105,107],{"class":91,"line":106},3,[89,108,110],{"emptyLinePlaceholder":109},true,"\n",[89,112,114],{"class":91,"line":113},4,[89,115,116],{"class":95},"# macOS\n",[89,118,120],{"class":91,"line":119},5,[89,121,122],{"class":102},"brew install ollama\n",[89,124,126],{"class":91,"line":125},6,[89,127,110],{"emptyLinePlaceholder":109},[89,129,131],{"class":91,"line":130},7,[89,132,133],{"class":95},"# Linux\n",[89,135,137,140,144,147,150,154,157,160,163],{"class":91,"line":136},8,[89,138,139],{"class":102},"curl ",[89,141,143],{"class":142},"szBVR","-",[89,145,146],{"class":102},"fsSL https:",[89,148,149],{"class":142},"\u002F\u002F",[89,151,153],{"class":152},"sj4cs","ollama.com",[89,155,156],{"class":142},"\u002F",[89,158,159],{"class":102},"install.sh ",[89,161,162],{"class":142},"|",[89,164,165],{"class":102}," sh\n",[75,167,168],{"id":168},"基础使用",[22,170,171,172,175],{},"如果你主要想找一个中文能力不错、资源占用相对友好的本地模型，",[86,173,174],{},"qwen2.5:7b"," 往往是比超大模型更容易先跑起来的选择。它很适合先验证 Ollama、RAG、代码助手或本地问答这类基础链路，再决定是否切到更大的参数规模。",[79,177,181],{"className":178,"code":179,"language":180,"meta":84,"style":84},"language-bash shiki shiki-themes github-light github-dark","# 运行模型（首次会自动下载）\nollama run llama3.1\nollama run qwen2.5:7b\nollama run deepseek-r1:8b\nollama run codellama:13b\n\n# 列出已下载的模型\nollama list\n\n# 下载模型\nollama pull llama3.1:70b\n\n# 删除模型\nollama rm llama3.1\n\n# 查看模型信息\nollama show llama3.1\n","bash",[86,182,183,188,200,209,218,227,231,236,243,248,254,265,270,276,286,291,297],{"__ignoreMap":84},[89,184,185],{"class":91,"line":92},[89,186,187],{"class":95},"# 运行模型（首次会自动下载）\n",[89,189,190,193,197],{"class":91,"line":99},[89,191,69],{"class":192},"sScJk",[89,194,196],{"class":195},"sZZnC"," run",[89,198,199],{"class":195}," llama3.1\n",[89,201,202,204,206],{"class":91,"line":106},[89,203,69],{"class":192},[89,205,196],{"class":195},[89,207,208],{"class":195}," qwen2.5:7b\n",[89,210,211,213,215],{"class":91,"line":113},[89,212,69],{"class":192},[89,214,196],{"class":195},[89,216,217],{"class":195}," deepseek-r1:8b\n",[89,219,220,222,224],{"class":91,"line":119},[89,221,69],{"class":192},[89,223,196],{"class":195},[89,225,226],{"class":195}," codellama:13b\n",[89,228,229],{"class":91,"line":125},[89,230,110],{"emptyLinePlaceholder":109},[89,232,233],{"class":91,"line":130},[89,234,235],{"class":95},"# 列出已下载的模型\n",[89,237,238,240],{"class":91,"line":136},[89,239,69],{"class":192},[89,241,242],{"class":195}," list\n",[89,244,246],{"class":91,"line":245},9,[89,247,110],{"emptyLinePlaceholder":109},[89,249,251],{"class":91,"line":250},10,[89,252,253],{"class":95},"# 下载模型\n",[89,255,257,259,262],{"class":91,"line":256},11,[89,258,69],{"class":192},[89,260,261],{"class":195}," pull",[89,263,264],{"class":195}," llama3.1:70b\n",[89,266,268],{"class":91,"line":267},12,[89,269,110],{"emptyLinePlaceholder":109},[89,271,273],{"class":91,"line":272},13,[89,274,275],{"class":95},"# 删除模型\n",[89,277,279,281,284],{"class":91,"line":278},14,[89,280,69],{"class":192},[89,282,283],{"class":195}," rm",[89,285,199],{"class":195},[89,287,289],{"class":91,"line":288},15,[89,290,110],{"emptyLinePlaceholder":109},[89,292,294],{"class":91,"line":293},16,[89,295,296],{"class":95},"# 查看模型信息\n",[89,298,300,302,305],{"class":91,"line":299},17,[89,301,69],{"class":192},[89,303,304],{"class":195}," show",[89,306,199],{"class":195},[75,308,310],{"id":309},"api-调用","API 调用",[22,312,313,314,317],{},"Ollama 默认在 ",[86,315,316],{},"http:\u002F\u002Flocalhost:11434"," 提供 API。",[79,319,321],{"className":178,"code":320,"language":180,"meta":84,"style":84},"# Chat API\ncurl http:\u002F\u002Flocalhost:11434\u002Fapi\u002Fchat -d '{\n  \"model\": \"llama3.1\",\n  \"messages\": [{\"role\": \"user\", \"content\": \"你好\"}],\n  \"stream\": false\n}'\n\n# Generate API\ncurl http:\u002F\u002Flocalhost:11434\u002Fapi\u002Fgenerate -d '{\n  \"model\": \"llama3.1\",\n  \"prompt\": \"解释什么是 MCP\",\n  \"stream\": false\n}'\n",[86,322,323,328,342,347,352,357,362,366,371,382,386,391,395],{"__ignoreMap":84},[89,324,325],{"class":91,"line":92},[89,326,327],{"class":95},"# Chat API\n",[89,329,330,333,336,339],{"class":91,"line":99},[89,331,332],{"class":192},"curl",[89,334,335],{"class":195}," http:\u002F\u002Flocalhost:11434\u002Fapi\u002Fchat",[89,337,338],{"class":152}," -d",[89,340,341],{"class":195}," '{\n",[89,343,344],{"class":91,"line":106},[89,345,346],{"class":195},"  \"model\": \"llama3.1\",\n",[89,348,349],{"class":91,"line":113},[89,350,351],{"class":195},"  \"messages\": [{\"role\": \"user\", \"content\": \"你好\"}],\n",[89,353,354],{"class":91,"line":119},[89,355,356],{"class":195},"  \"stream\": false\n",[89,358,359],{"class":91,"line":125},[89,360,361],{"class":195},"}'\n",[89,363,364],{"class":91,"line":130},[89,365,110],{"emptyLinePlaceholder":109},[89,367,368],{"class":91,"line":136},[89,369,370],{"class":95},"# Generate API\n",[89,372,373,375,378,380],{"class":91,"line":245},[89,374,332],{"class":192},[89,376,377],{"class":195}," http:\u002F\u002Flocalhost:11434\u002Fapi\u002Fgenerate",[89,379,338],{"class":152},[89,381,341],{"class":195},[89,383,384],{"class":91,"line":250},[89,385,346],{"class":195},[89,387,388],{"class":91,"line":256},[89,389,390],{"class":195},"  \"prompt\": \"解释什么是 MCP\",\n",[89,392,393],{"class":91,"line":267},[89,394,356],{"class":195},[89,396,397],{"class":91,"line":272},[89,398,361],{"class":195},[75,400,402],{"id":401},"nodejs-调用","Node.js 调用",[79,404,406],{"className":178,"code":405,"language":180,"meta":84,"style":84},"pnpm add ollama\n",[86,407,408],{"__ignoreMap":84},[89,409,410,413,416],{"class":91,"line":92},[89,411,412],{"class":192},"pnpm",[89,414,415],{"class":195}," add",[89,417,418],{"class":195}," ollama\n",[79,420,424],{"className":421,"code":422,"language":423,"meta":84,"style":84},"language-typescript shiki shiki-themes github-light github-dark","import { Ollama } from \"ollama\";\n\nconst ollama = new Ollama();\n\nconst response = await ollama.chat({\n  model: \"llama3.1\",\n  messages: [{ role: \"user\", content: \"用 TypeScript 写一个快速排序\" }],\n});\n\nconsole.log(response.message.content);\n","typescript",[86,425,426,443,447,467,471,492,503,520,525,529],{"__ignoreMap":84},[89,427,428,431,434,437,440],{"class":91,"line":92},[89,429,430],{"class":142},"import",[89,432,433],{"class":102}," { Ollama } ",[89,435,436],{"class":142},"from",[89,438,439],{"class":195}," \"ollama\"",[89,441,442],{"class":102},";\n",[89,444,445],{"class":91,"line":99},[89,446,110],{"emptyLinePlaceholder":109},[89,448,449,452,455,458,461,464],{"class":91,"line":106},[89,450,451],{"class":142},"const",[89,453,454],{"class":152}," ollama",[89,456,457],{"class":142}," =",[89,459,460],{"class":142}," new",[89,462,463],{"class":192}," Ollama",[89,465,466],{"class":102},"();\n",[89,468,469],{"class":91,"line":113},[89,470,110],{"emptyLinePlaceholder":109},[89,472,473,475,478,480,483,486,489],{"class":91,"line":119},[89,474,451],{"class":142},[89,476,477],{"class":152}," response",[89,479,457],{"class":142},[89,481,482],{"class":142}," await",[89,484,485],{"class":102}," ollama.",[89,487,488],{"class":192},"chat",[89,490,491],{"class":102},"({\n",[89,493,494,497,500],{"class":91,"line":125},[89,495,496],{"class":102},"  model: ",[89,498,499],{"class":195},"\"llama3.1\"",[89,501,502],{"class":102},",\n",[89,504,505,508,511,514,517],{"class":91,"line":130},[89,506,507],{"class":102},"  messages: [{ role: ",[89,509,510],{"class":195},"\"user\"",[89,512,513],{"class":102},", content: ",[89,515,516],{"class":195},"\"用 TypeScript 写一个快速排序\"",[89,518,519],{"class":102}," }],\n",[89,521,522],{"class":91,"line":136},[89,523,524],{"class":102},"});\n",[89,526,527],{"class":91,"line":245},[89,528,110],{"emptyLinePlaceholder":109},[89,530,531,534,537],{"class":91,"line":250},[89,532,533],{"class":102},"console.",[89,535,536],{"class":192},"log",[89,538,539],{"class":102},"(response.message.content);\n",[75,541,543],{"id":542},"openai-兼容-api","OpenAI 兼容 API",[22,545,546],{},"Ollama 兼容 OpenAI API 格式：",[79,548,550],{"className":421,"code":549,"language":423,"meta":84,"style":84},"import OpenAI from \"openai\";\n\nconst client = new OpenAI({\n  baseURL: \"http:\u002F\u002Flocalhost:11434\u002Fv1\",\n  apiKey: \"ollama\", \u002F\u002F 任意值\n});\n\nconst response = await client.chat.completions.create({\n  model: \"llama3.1\",\n  messages: [{ role: \"user\", content: \"你好\" }],\n});\n",[86,551,552,566,570,586,596,610,614,618,636,644,657],{"__ignoreMap":84},[89,553,554,556,559,561,564],{"class":91,"line":92},[89,555,430],{"class":142},[89,557,558],{"class":102}," OpenAI ",[89,560,436],{"class":142},[89,562,563],{"class":195}," \"openai\"",[89,565,442],{"class":102},[89,567,568],{"class":91,"line":99},[89,569,110],{"emptyLinePlaceholder":109},[89,571,572,574,577,579,581,584],{"class":91,"line":106},[89,573,451],{"class":142},[89,575,576],{"class":152}," client",[89,578,457],{"class":142},[89,580,460],{"class":142},[89,582,583],{"class":192}," OpenAI",[89,585,491],{"class":102},[89,587,588,591,594],{"class":91,"line":113},[89,589,590],{"class":102},"  baseURL: ",[89,592,593],{"class":195},"\"http:\u002F\u002Flocalhost:11434\u002Fv1\"",[89,595,502],{"class":102},[89,597,598,601,604,607],{"class":91,"line":119},[89,599,600],{"class":102},"  apiKey: ",[89,602,603],{"class":195},"\"ollama\"",[89,605,606],{"class":102},", ",[89,608,609],{"class":95},"\u002F\u002F 任意值\n",[89,611,612],{"class":91,"line":125},[89,613,524],{"class":102},[89,615,616],{"class":91,"line":130},[89,617,110],{"emptyLinePlaceholder":109},[89,619,620,622,624,626,628,631,634],{"class":91,"line":136},[89,621,451],{"class":142},[89,623,477],{"class":152},[89,625,457],{"class":142},[89,627,482],{"class":142},[89,629,630],{"class":102}," client.chat.completions.",[89,632,633],{"class":192},"create",[89,635,491],{"class":102},[89,637,638,640,642],{"class":91,"line":245},[89,639,496],{"class":102},[89,641,499],{"class":195},[89,643,502],{"class":102},[89,645,646,648,650,652,655],{"class":91,"line":250},[89,647,507],{"class":102},[89,649,510],{"class":195},[89,651,513],{"class":102},[89,653,654],{"class":195},"\"你好\"",[89,656,519],{"class":102},[89,658,659],{"class":91,"line":256},[89,660,524],{"class":102},[75,662,664],{"id":663},"自定义模型modelfile","自定义模型（Modelfile）",[79,666,670],{"className":667,"code":668,"language":669,"meta":84,"style":84},"language-dockerfile shiki shiki-themes github-light github-dark","# Modelfile\nFROM llama3.1\n\nPARAMETER temperature 0.7\nPARAMETER top_p 0.9\n\nSYSTEM \"\"\"\n你是一位资深的 TypeScript 开发者。\n回答要简洁直接，优先给代码示例。\n使用中文回答。\n\"\"\"\n","dockerfile",[86,671,672,677,684,688,693,698,702,710,715,720,725],{"__ignoreMap":84},[89,673,674],{"class":91,"line":92},[89,675,676],{"class":95},"# Modelfile\n",[89,678,679,682],{"class":91,"line":99},[89,680,681],{"class":142},"FROM",[89,683,199],{"class":102},[89,685,686],{"class":91,"line":106},[89,687,110],{"emptyLinePlaceholder":109},[89,689,690],{"class":91,"line":113},[89,691,692],{"class":102},"PARAMETER temperature 0.7\n",[89,694,695],{"class":91,"line":119},[89,696,697],{"class":102},"PARAMETER top_p 0.9\n",[89,699,700],{"class":91,"line":125},[89,701,110],{"emptyLinePlaceholder":109},[89,703,704,707],{"class":91,"line":130},[89,705,706],{"class":102},"SYSTEM ",[89,708,709],{"class":195},"\"\"\"\n",[89,711,712],{"class":91,"line":136},[89,713,714],{"class":195},"你是一位资深的 TypeScript 开发者。\n",[89,716,717],{"class":91,"line":245},[89,718,719],{"class":195},"回答要简洁直接，优先给代码示例。\n",[89,721,722],{"class":91,"line":250},[89,723,724],{"class":195},"使用中文回答。\n",[89,726,727],{"class":91,"line":256},[89,728,709],{"class":195},[79,730,732],{"className":178,"code":731,"language":180,"meta":84,"style":84},"ollama create my-coder -f Modelfile\nollama run my-coder\n",[86,733,734,750],{"__ignoreMap":84},[89,735,736,738,741,744,747],{"class":91,"line":92},[89,737,69],{"class":192},[89,739,740],{"class":195}," create",[89,742,743],{"class":195}," my-coder",[89,745,746],{"class":152}," -f",[89,748,749],{"class":195}," Modelfile\n",[89,751,752,754,756],{"class":91,"line":99},[89,753,69],{"class":192},[89,755,196],{"class":195},[89,757,758],{"class":195}," my-coder\n",[29,760,762],{"id":761},"lm-studio","LM Studio",[22,764,765],{},"图形界面的本地模型管理工具。",[79,767,769],{"className":81,"code":768,"language":83,"meta":84,"style":84},"winget install LMStudio.LMStudio\n",[86,770,771],{"__ignoreMap":84},[89,772,773],{"class":91,"line":92},[89,774,768],{"class":102},[22,776,777],{},"特点：",[33,779,780,783,786,793,796],{},[36,781,782],{},"可视化模型搜索和下载（Hugging Face）",[36,784,785],{},"内置聊天界面",[36,787,788,789,792],{},"兼容 OpenAI API（",[86,790,791],{},"http:\u002F\u002Flocalhost:1234\u002Fv1","）",[36,794,795],{},"支持 GGUF 格式模型",[36,797,798],{},"硬件检测和推荐配置",[29,800,801],{"id":801},"推荐模型",[75,803,804],{"id":804},"通用对话",[806,807,808,824],"table",{},[809,810,811],"thead",{},[812,813,814,818,821],"tr",{},[815,816,817],"th",{},"模型",[815,819,820],{},"大小",[815,822,823],{},"说明",[825,826,827,839,850,861,872],"tbody",{},[812,828,829,833,836],{},[830,831,832],"td",{},"Llama 3.1 8B",[830,834,835],{},"~4.7 GB",[830,837,838],{},"Meta，综合能力强",[812,840,841,844,847],{},[830,842,843],{},"Qwen 2.5 7B",[830,845,846],{},"~4.4 GB",[830,848,849],{},"阿里，中文优秀",[812,851,852,855,858],{},[830,853,854],{},"DeepSeek R1 8B",[830,856,857],{},"~4.9 GB",[830,859,860],{},"推理能力强",[812,862,863,866,869],{},[830,864,865],{},"Gemma 2 9B",[830,867,868],{},"~5.4 GB",[830,870,871],{},"Google，轻量高效",[812,873,874,877,880],{},[830,875,876],{},"Phi-3.5 Mini",[830,878,879],{},"~2.2 GB",[830,881,882],{},"微软，超小但聪明",[75,884,885],{"id":885},"代码生成",[806,887,888,898],{},[809,889,890],{},[812,891,892,894,896],{},[815,893,817],{},[815,895,820],{},[815,897,823],{},[825,899,900,911,922,932],{},[812,901,902,905,908],{},[830,903,904],{},"CodeLlama 13B",[830,906,907],{},"~7.4 GB",[830,909,910],{},"Meta，代码专用",[812,912,913,916,919],{},[830,914,915],{},"DeepSeek Coder V2",[830,917,918],{},"~8.9 GB",[830,920,921],{},"代码能力顶级",[812,923,924,927,929],{},[830,925,926],{},"Qwen 2.5 Coder 7B",[830,928,846],{},[830,930,931],{},"代码补全优秀",[812,933,934,937,940],{},[830,935,936],{},"StarCoder2 7B",[830,938,939],{},"~4.0 GB",[830,941,942],{},"BigCode，多语言",[29,944,945],{"id":945},"硬件要求",[79,947,952],{"className":948,"code":950,"language":951},[949],"language-text","7B 模型：8GB+ 内存\u002F显存\n13B 模型：16GB+ 内存\u002F显存\n70B 模型：48GB+ 内存\u002F显存（或量化后 32GB+）\n","text",[86,953,950],{"__ignoreMap":84},[33,955,956,959,962],{},[36,957,958],{},"GPU 推理远快于 CPU",[36,960,961],{},"Apple Silicon 的统一内存对大模型友好",[36,963,964],{},"NVIDIA GPU 需要 CUDA 支持",[29,966,968],{"id":967},"web-ui","Web UI",[75,970,972],{"id":971},"open-webui","Open WebUI",[79,974,976],{"className":178,"code":975,"language":180,"meta":84,"style":84},"docker run -d -p 3000:8080 \\\n  --add-host=host.docker.internal:host-gateway \\\n  -v open-webui:\u002Fapp\u002Fbackend\u002Fdata \\\n  --name open-webui \\\n  ghcr.io\u002Fopen-webui\u002Fopen-webui:main\n",[86,977,978,996,1003,1013,1023],{"__ignoreMap":84},[89,979,980,983,985,987,990,993],{"class":91,"line":92},[89,981,982],{"class":192},"docker",[89,984,196],{"class":195},[89,986,338],{"class":152},[89,988,989],{"class":152}," -p",[89,991,992],{"class":195}," 3000:8080",[89,994,995],{"class":152}," \\\n",[89,997,998,1001],{"class":91,"line":99},[89,999,1000],{"class":152},"  --add-host=host.docker.internal:host-gateway",[89,1002,995],{"class":152},[89,1004,1005,1008,1011],{"class":91,"line":106},[89,1006,1007],{"class":152},"  -v",[89,1009,1010],{"class":195}," open-webui:\u002Fapp\u002Fbackend\u002Fdata",[89,1012,995],{"class":152},[89,1014,1015,1018,1021],{"class":91,"line":113},[89,1016,1017],{"class":152},"  --name",[89,1019,1020],{"class":195}," open-webui",[89,1022,995],{"class":152},[89,1024,1025],{"class":91,"line":119},[89,1026,1027],{"class":195},"  ghcr.io\u002Fopen-webui\u002Fopen-webui:main\n",[22,1029,1030,1031,1034],{},"访问 ",[86,1032,1033],{},"http:\u002F\u002Flocalhost:3000","，自动连接本地 Ollama。",[29,1036,1037],{"id":1037},"推荐落地顺序",[22,1039,1040],{},"建议按这个顺序开始：",[1042,1043,1044,1047,1050,1053,1056],"ol",{},[36,1045,1046],{},"先选一个入门运行器",[36,1048,1049],{},"先跑一个小模型验证速度和内存占用",[36,1051,1052],{},"再补更适合自己任务的模型",[36,1054,1055],{},"再接 Web UI、API 或代理工具",[36,1057,1058],{},"最后再考虑量化、批量调用和长期驻留服务",[29,1060,1061],{"id":1061},"常见问题",[75,1063,1065],{"id":1064},"模型能跑但效果不理想","模型能跑，但效果不理想",[22,1067,1068],{},"通常说明模型大小、量化级别或任务类型不匹配。先换模型，再考虑调提示词。",[75,1070,1071],{"id":1071},"本地很慢",[22,1073,1074],{},"优先检查：",[33,1076,1077,1080,1083,1086],{},[36,1078,1079],{},"是否走了 CPU 推理",[36,1081,1082],{},"内存是否不足",[36,1084,1085],{},"模型是否过大",[36,1087,1088],{},"是否同时开了太多本地服务",[75,1090,1092],{"id":1091},"应该先学本地模型还是直接用云-api","应该先学本地模型还是直接用云 API",[22,1094,1095],{},"如果你更看重隐私、离线、零调用成本，本地模型值得先学；如果你更看重效果和省维护，云 API 往往更直接。",[29,1097,1098],{"id":1098},"延伸阅读",[33,1100,1101,1108,1114],{},[36,1102,1103],{},[1104,1105,1107],"a",{"href":1106},"\u002Fdocs\u002Flocal-llm-deployment","本地 LLM 部署指南",[36,1109,1110],{},[1104,1111,1113],{"href":1112},"\u002Fdocs\u002Fai-tools","AI 工具合集",[36,1115,1116],{},[1104,1117,1119],{"href":1118},"\u002Fdocs\u002Fai-api-usage","AI API 接入指南",[29,1121,1122],{"id":1122},"参考链接",[33,1124,1125,1133,1140,1147],{},[36,1126,1127,1132],{},[1104,1128,70],{"href":1129,"rel":1130},"https:\u002F\u002Follama.com\u002F",[1131],"nofollow"," — 官网与模型库",[36,1134,1135,1139],{},[1104,1136,762],{"href":1137,"rel":1138},"https:\u002F\u002Flmstudio.ai\u002F",[1131]," — 图形界面工具",[36,1141,1142,1146],{},[1104,1143,972],{"href":1144,"rel":1145},"https:\u002F\u002Fgithub.com\u002Fopen-webui\u002Fopen-webui",[1131]," — Web 聊天界面",[36,1148,1149,1154],{},[1104,1150,1153],{"href":1151,"rel":1152},"https:\u002F\u002Fhuggingface.co\u002F",[1131],"Hugging Face"," — 模型仓库",[1156,1157,1158],"style",{},"html pre.shiki code .sJ8bj, html code.shiki .sJ8bj{--shiki-default:#6A737D;--shiki-dark:#6A737D}html pre.shiki code .sVt8B, html code.shiki .sVt8B{--shiki-default:#24292E;--shiki-dark:#E1E4E8}html pre.shiki code .szBVR, html code.shiki .szBVR{--shiki-default:#D73A49;--shiki-dark:#F97583}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}",{"title":84,"searchDepth":99,"depth":99,"links":1160},[1161,1162,1170,1171,1175,1176,1179,1180,1185,1186],{"id":31,"depth":99,"text":31},{"id":69,"depth":99,"text":70,"children":1163},[1164,1165,1166,1167,1168,1169],{"id":77,"depth":106,"text":77},{"id":168,"depth":106,"text":168},{"id":309,"depth":106,"text":310},{"id":401,"depth":106,"text":402},{"id":542,"depth":106,"text":543},{"id":663,"depth":106,"text":664},{"id":761,"depth":99,"text":762},{"id":801,"depth":99,"text":801,"children":1172},[1173,1174],{"id":804,"depth":106,"text":804},{"id":885,"depth":106,"text":885},{"id":945,"depth":99,"text":945},{"id":967,"depth":99,"text":968,"children":1177},[1178],{"id":971,"depth":106,"text":972},{"id":1037,"depth":99,"text":1037},{"id":1061,"depth":99,"text":1061,"children":1181},[1182,1183,1184],{"id":1064,"depth":106,"text":1065},{"id":1071,"depth":106,"text":1071},{"id":1091,"depth":106,"text":1092},{"id":1098,"depth":99,"text":1098},{"id":1122,"depth":99,"text":1122},{"path":1188,"title":1189,"description":1190,"docType":8,"resourceKind":9,"categoryId":1191,"categoryLabel":1192,"updatedAt":1193,"publishedAt":1193,"icon":1194},"\u002Fdocs\u002Fnavigation","站内导航与索引","DomiVault 内容分类索引，按系统、工具、开发、网络和媒体快速跳转。","navigation","导航","2026-03-01","i-carbon-compass",{"path":1106,"title":1107,"description":1196,"docType":8,"resourceKind":9,"categoryId":10,"categoryLabel":11,"updatedAt":12,"publishedAt":12,"icon":13},"使用 Ollama、vLLM、LM Studio 在本地运行大语言模型",[1198,1199,1200],"希望把零散经验整理成长期可复用工作流的人","正在使用 AI 工具、Agent 或自动化工作流的人","希望阅读时顺手建立自己的操作清单或收藏体系的人",[1202,1203,1204],"先浏览标题、摘要和目录，带着问题阅读会更高效","确认模型供应商、API Key、CLI 工具链与本地资源是否已准备好","如果页面里提到相关文档，尽量一起打开对照，效果通常更完整",[1206,1211,1212,1216],{"path":1207,"title":1208,"description":1209,"docType":8,"resourceKind":9,"categoryId":10,"categoryLabel":11,"updatedAt":1210,"publishedAt":1210,"icon":13},"\u002Fdocs\u002Fskills-guide","AI Agent Skills 指南","理解 skills 的作用、目录结构、编写方式，以及它与 MCP 的关系","2026-03-08",{"path":1106,"title":1107,"description":1196,"docType":8,"resourceKind":9,"categoryId":10,"categoryLabel":11,"updatedAt":12,"publishedAt":12,"icon":13},{"path":1213,"title":1214,"description":1215,"docType":8,"resourceKind":9,"categoryId":10,"categoryLabel":11,"updatedAt":12,"publishedAt":12,"icon":13},"\u002Fdocs\u002Fai-coding-rules","AI 编程助手规则配置","Cursor Rules、Claude Projects、Kiro Steering 等 AI 编程助手的规则与上下文配置",{"path":1217,"title":1218,"description":1219,"docType":8,"resourceKind":9,"categoryId":10,"categoryLabel":11,"updatedAt":12,"publishedAt":12,"icon":13},"\u002Fdocs\u002Fai-agent-security","AI 代理安全配置指南","OpenClaw、nanobot 等 AI 代理的安全配置最佳实践",1776215711221]