[{"data":1,"prerenderedAt":624},["ShallowReactive",2],{"docs-zh-/zh/docs/services/local-llm-services":3},{"id":4,"title":5,"body":6,"description":16,"extension":617,"meta":618,"navigation":619,"path":620,"seo":621,"stem":622,"__hash__":623},"docs_zh/zh/docs/services/local-llm-services.md","本地 LLM 服务",{"type":7,"value":8,"toc":605},"minimark",[9,13,17,22,51,60,63,68,71,100,104,247,251,271,275,278,300,303,392,396,416,419,422,445,448,451,478,481,498,501,504,574,582,585,601],[10,11,5],"h1",{"id":12},"本地-llm-服务",[14,15,16],"p",{},"Doco Translate 支持使用本地 AI 模型在 Mac 上完全运行翻译。这保证了您的文档隐私——数据不会离开您的机器。",[18,19,21],"h2",{"id":20},"为什么使用本地模型","为什么使用本地模型？",[23,24,25,33,39,45],"ul",{},[26,27,28,32],"li",{},[29,30,31],"strong",{},"隐私"," — 您的文档永远不会离开您的 Mac。没有任何文本发送到外部服务器。",[26,34,35,38],{},[29,36,37],{},"无 API 费用"," — 无需支付按请求计费的翻译费用。",[26,40,41,44],{},[29,42,43],{},"离线能力"," — 无需互联网连接即可翻译文档（下载模型后）。",[26,46,47,50],{},[29,48,49],{},"完全控制"," — 精确选择使用哪个模型并配置其行为。",[52,53,54],"blockquote",{},[14,55,56,59],{},[29,57,58],{},"权衡："," 本地模型通常比基于云的 AI 服务产生的翻译质量更低，翻译速度取决于您 Mac 的硬件（CPU、GPU 和内存）。",[18,61,62],{"id":62},"支持的本地服务",[64,65,67],"h3",{"id":66},"ollama","Ollama",[14,69,70],{},"Ollama 是一个流行的开源工具，用于在 macOS、Linux 和 Windows 上本地运行大语言模型。",[23,72,73,83,89],{},[26,74,75,78,79],{},[29,76,77],{},"默认主机："," ",[80,81,82],"code",{},"http://localhost:11434",[26,84,85,88],{},[29,86,87],{},"API 密钥："," 不需要（除非您的 Ollama 实例启用了身份验证）",[26,90,91,78,94],{},[29,92,93],{},"网站：",[95,96,97],"a",{"href":97,"rel":98},"https://ollama.com",[99],"nofollow",[101,102,103],"h4",{"id":103},"设置",[105,106,107,147,192,210],"ol",{},[26,108,109,112],{},[29,110,111],{},"安装 Ollama：",[23,113,114],{},[26,115,116,117,121,122],{},"从 ",[95,118,120],{"href":97,"rel":119},[99],"ollama.com"," 下载或通过 Homebrew 安装：\n",[123,124,129],"pre",{"className":125,"code":126,"language":127,"meta":128,"style":128},"language-bash shiki shiki-themes github-light github-dark","brew install ollama\n","bash","",[80,130,131],{"__ignoreMap":128},[132,133,136,140,144],"span",{"class":134,"line":135},"line",1,[132,137,139],{"class":138},"sScJk","brew",[132,141,143],{"class":142},"sZZnC"," install",[132,145,146],{"class":142}," ollama\n",[26,148,149,152,153,168,171,172],{},[29,150,151],{},"拉取模型：","\n打开终端并运行：",[123,154,156],{"className":125,"code":155,"language":127,"meta":128,"style":128},"ollama pull qwen3.6\n",[80,157,158],{"__ignoreMap":128},[132,159,160,162,165],{"class":134,"line":135},[132,161,66],{"class":138},[132,163,164],{"class":142}," pull",[132,166,167],{"class":142}," qwen3.6\n",[169,170],"br",{},"翻译常用模型：",[23,173,174,180,186],{},[26,175,176,179],{},[80,177,178],{},"qwen3.6"," — 多语言支持强，特别是中文和亚洲语言",[26,181,182,185],{},[80,183,184],{},"llama3.2"," — 通用型，速度和质量平衡良好",[26,187,188,191],{},[80,189,190],{},"gemma4"," — Google 的开源模型，适合欧洲语言",[26,193,194,197,198],{},[29,195,196],{},"启动 Ollama：","\nOllama 安装后自动运行。如果没有，请手动启动：",[123,199,201],{"className":125,"code":200,"language":127,"meta":128,"style":128},"ollama serve\n",[80,202,203],{"__ignoreMap":128},[132,204,205,207],{"class":134,"line":135},[132,206,66],{"class":138},[132,208,209],{"class":142}," serve\n",[26,211,212,215],{},[29,213,214],{},"在 Doco Translate 中配置：",[23,216,217,224,230,237,240],{},[26,218,219,220,223],{},"前往",[29,221,222],{},"设置 → 服务 → Ollama","。",[26,225,226,227,229],{},"默认主机（",[80,228,82],{},"）应能开箱即用。",[26,231,232,233,236],{},"使用",[29,234,235],{},"获取模型列表","自动检测您已拉取的模型。",[26,238,239],{},"从下拉菜单中选择模型。",[26,241,242,243,246],{},"点击",[29,244,245],{},"验证服务","测试连接。",[101,248,250],{"id":249},"ollama-使用技巧","Ollama 使用技巧",[23,252,253,259,265],{},[26,254,255,258],{},[29,256,257],{},"模型大小很重要："," 较大的模型（70B+）产生更好的翻译，但需要更多内存且运行更慢。从 7B–8B 模型开始以获得良好平衡。",[26,260,261,264],{},[29,262,263],{},"GPU 加速："," Ollama 在 M 系列 Mac 上自动使用 Apple Silicon GPU 加速。",[26,266,267,270],{},[29,268,269],{},"保持 Ollama 运行："," 在 Doco Translate 中使用之前确保 Ollama 服务正在运行。",[64,272,274],{"id":273},"lm-studio","LM Studio",[14,276,277],{},"LM Studio 是一个桌面应用，用于通过图形界面发现、下载和运行本地 LLM。",[23,279,280,287,292],{},[26,281,282,78,284],{},[29,283,77],{},[80,285,286],{},"http://localhost:1234",[26,288,289,291],{},[29,290,87],{}," 不需要",[26,293,294,78,296],{},[29,295,93],{},[95,297,298],{"href":298,"rel":299},"https://lmstudio.ai",[99],[101,301,103],{"id":302},"设置-1",[105,304,305,319,344,366],{},[26,306,307,310],{},[29,308,309],{},"安装 LM Studio：",[23,311,312],{},[26,313,116,314,318],{},[95,315,317],{"href":298,"rel":316},[99],"lmstudio.ai"," 下载。",[26,320,321,324],{},[29,322,323],{},"下载模型：",[23,325,326,329,338],{},[26,327,328],{},"打开 LM Studio。",[26,330,331,332,334,335,337],{},"使用搜索栏查找模型（例如 ",[80,333,178],{},"、",[80,336,190],{},"）。",[26,339,340,341,223],{},"在您首选的模型变体上点击",[29,342,343],{},"下载",[26,345,346,349],{},[29,347,348],{},"启动本地服务器：",[23,350,351,358,361],{},[26,352,353,354,357],{},"在 LM Studio 中，前往",[29,355,356],{},"本地服务器","选项卡（左侧边栏）。",[26,359,360],{},"选择您下载的模型。",[26,362,242,363,223],{},[29,364,365],{},"启动服务器",[26,367,368,370],{},[29,369,214],{},[23,371,372,377,383,388],{},[26,373,219,374,223],{},[29,375,376],{},"设置 → 服务 → LM Studio",[26,378,379,380,382],{},"如果 LM Studio 的服务器正在运行，默认主机（",[80,381,286],{},"）应能正常工作。",[26,384,232,385,387],{},[29,386,235],{},"检测已加载的模型，或手动输入模型名称。",[26,389,242,390,246],{},[29,391,245],{},[101,393,395],{"id":394},"lm-studio-使用技巧","LM Studio 使用技巧",[23,397,398,404,410],{},[26,399,400,403],{},[29,401,402],{},"一次只能加载一个模型："," LM Studio 一次将一个模型加载到内存中。切换模型需要先卸载当前模型。",[26,405,406,409],{},[29,407,408],{},"量化："," LM Studio 支持各种量化级别（Q4、Q5、Q8）。较低量化（Q4）使用更少内存但可能降低质量。",[26,411,412,415],{},[29,413,414],{},"服务器必须运行："," LM Studio 本地服务器必须处于活动状态，Doco Translate 才能连接。",[18,417,418],{"id":418},"配置本地服务",[14,420,421],{},"Doco Translate 中的本地服务设置与云 AI 服务类似，但有一些区别：",[23,423,424,430,436],{},[26,425,426,429],{},[29,427,428],{},"无需 API 密钥"," — 本地服务默认不需要身份验证。如果您在本地服务上配置了身份验证，可以输入凭据。",[26,431,432,435],{},[29,433,434],{},"自定义主机"," — 如果您的本地服务运行在不同的端口或机器上，可以更改主机。",[26,437,438,441,442,444],{},[29,439,440],{},"模型选择"," — 使用",[29,443,235],{},"自动检测可用模型，或手动添加模型。",[18,446,447],{"id":447},"自定义本地服务",[14,449,450],{},"如果您运行的是与 OpenAI API 格式兼容的其他本地 LLM 服务器：",[105,452,453,462,465,472,475],{},[26,454,219,455,458,459,223],{},[29,456,457],{},"设置 → 服务","并点击",[29,460,461],{},"自定义服务",[26,463,464],{},"为您的服务输入名称。",[26,466,467,468,471],{},"选择 ",[29,469,470],{},"OpenAI"," 协议。",[26,473,474],{},"输入本地服务器的主机地址。",[26,476,477],{},"根据需要配置模型名称和其他设置。",[14,479,480],{},"这适用于任何 OpenAI 兼容服务器，包括：",[23,482,483,486,489,492,495],{},[26,484,485],{},"vLLM",[26,487,488],{},"text-generation-webui",[26,490,491],{},"LocalAI",[26,493,494],{},"llama.cpp server",[26,496,497],{},"任何自定义 API 服务器",[18,499,500],{"id":500},"性能考量",[14,502,503],{},"使用本地模型的翻译速度取决于几个因素：",[505,506,507,520],"table",{},[508,509,510],"thead",{},[511,512,513,517],"tr",{},[514,515,516],"th",{},"因素",[514,518,519],{},"影响",[521,522,523,534,544,554,564],"tbody",{},[511,524,525,531],{},[526,527,528],"td",{},[29,529,530],{},"模型大小",[526,532,533],{},"较小的模型（7B）更快；较大的模型（70B+）更慢但更准确",[511,535,536,541],{},[526,537,538],{},[29,539,540],{},"量化",[526,542,543],{},"较低量化 = 更快但不太准确",[511,545,546,551],{},[526,547,548],{},[29,549,550],{},"硬件",[526,552,553],{},"Apple Silicon M 系列芯片提供最佳性能",[511,555,556,561],{},[526,557,558],{},[29,559,560],{},"内存",[526,562,563],{},"较大的模型需要更多内存（8B ≈ 5GB，70B ≈ 40GB）",[511,565,566,571],{},[526,567,568],{},[29,569,570],{},"并发",[526,572,573],{},"较低的并发设置（1–2）更适合本地模型以避免过载",[52,575,576],{},[14,577,578,581],{},[29,579,580],{},"建议："," 从 7B–8B 模型开始，仅在质量不足时增加模型大小。将本地服务的最大并发页面数设置为 1 或 2，以避免 Mac 过载。",[583,584],"hr",{},[14,586,587,78,590,594,595,78,598],{},[29,588,589],{},"上一步：",[95,591,593],{"href":592},"./ai-services","AI 服务"," · ",[29,596,597],{},"下一步：",[95,599,461],{"href":600},"./custom-services",[602,603,604],"style",{},"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}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);}",{"title":128,"searchDepth":606,"depth":606,"links":607},2,[608,609,614,615,616],{"id":20,"depth":606,"text":21},{"id":62,"depth":606,"text":62,"children":610},[611,613],{"id":66,"depth":612,"text":67},3,{"id":273,"depth":612,"text":274},{"id":418,"depth":606,"text":418},{"id":447,"depth":606,"text":447},{"id":500,"depth":606,"text":500},"md",{},true,"/zh/docs/services/local-llm-services",{"title":5,"description":16},"zh/docs/services/local-llm-services","jRYLAwKdH2_tDj7OZuL1rYUdCt061cLxwL-faCfVVdY",1780845388943]