[{"data":1,"prerenderedAt":2463},["ShallowReactive",2],{"header-counts":3,"footer-counts":6,"alt-main-cherry-studio":8,"alt-list-cherry-studio":540},{"tools":4,"reviews":5},98,27,{"tools":4,"reviews":5,"playbooks":7,"news":5},23,{"id":9,"title":10,"alternatives":11,"api_compatible":16,"body":17,"category":469,"chinese_friendly":470,"cover":471,"description":472,"domestic":473,"extension":474,"faq":475,"free":473,"github":488,"languages":489,"lastVerified":492,"meta":493,"models":16,"navigation":473,"notSuitable":16,"opensource":473,"path":494,"pillar":495,"platforms":496,"priceTable":501,"pricing":510,"published":511,"relatedPlaybooks":512,"relatedReviews":16,"score":515,"self_host":473,"seo":517,"seoTitle":518,"slug":519,"sources":520,"stem":528,"suitable":16,"tagline":529,"tags":530,"updated":523,"verdict":537,"website":538,"__hash__":539},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Fcherry-studio.md","Cherry Studio",[12,13,14,15],"coding\u002Flocal\u002Flobe-chat","coding\u002Flocal\u002Flm-studio","coding\u002Flocal\u002Follama","coding\u002Flocal\u002Fopen-webui",null,{"type":18,"value":19,"toc":454},"minimark",[20,25,29,32,35,82,85,99,105,109,114,131,136,153,156,183,186,337,340,372,376,399,402,428,431],[21,22,24],"h2",{"id":23},"tldr","TL;DR",[26,27,28],"p",{},"Cherry Studio 是一款开源、跨平台（Windows \u002F macOS \u002F Linux \u002F Android）的桌面 AI 客户端，定位『全能 AI 工作台』：把 OpenAI \u002F Anthropic \u002F Google \u002F DeepSeek 等云端模型，以及 Ollama \u002F LM Studio 本地模型，全部聚合到同一个桌面应用里管理。内置 300+ 助手模板、本地 RAG 知识库、Markdown + Mermaid 渲染、MCP 协议支持，所有对话数据本地存储 + WebDAV 备份。AGPL-3.0 开源、GitHub 60k+ stars，企业版可联系商务做私有化部署。",[26,30,31],{},"适合：中文 AI 重度用户、想统一管理多家模型、需要本地知识库 RAG、关注数据本地存储的开发者 \u002F 研究者。不适合：要 Web 端访问 \u002F Docker 自托管 \u002F 团队多人共享 \u002F iOS 端使用。",[21,33,34],{"id":34},"核心能力",[36,37,38,46,52,58,64,70,76],"ul",{},[39,40,41,45],"li",{},[42,43,44],"strong",{},"多模型聚合","：OpenAI \u002F Claude \u002F Gemini \u002F DeepSeek \u002F Qwen \u002F Kimi \u002F Moonshot 等云端 + Ollama \u002F LM Studio 本地",[39,47,48,51],{},[42,49,50],{},"本地 RAG 知识库","：拖拽 PDF \u002F Word \u002F Excel \u002F PPT \u002F 网址 \u002F sitemap → 自动向量化 → 检索增强问答 + 来源追溯",[39,53,54,57],{},[42,55,56],{},"300+ 助手模板","：编程 \u002F 写作 \u002F 翻译 \u002F 学习 \u002F 角色扮演开箱即用，可自定义 System Prompt",[39,59,60,63],{},[42,61,62],{},"MCP 协议","：扩展工具调用 \u002F 联网搜索 \u002F 文件操作",[39,65,66,69],{},[42,67,68],{},"数据本地优先","：对话历史本地存储，WebDAV 同步，不上传第三方",[39,71,72,75],{},[42,73,74],{},"多模态","：图片识别 \u002F PDF 阅读 \u002F Markdown + Mermaid + 代码高亮",[39,77,78,81],{},[42,79,80],{},"AI 绘画 + 翻译","：内置主流 SD \u002F DALL·E \u002F 翻译 API 集成",[21,83,84],{"id":84},"价格",[36,86,87,93],{},[39,88,89,92],{},[42,90,91],{},"开源版","：完全免费，AGPL-3.0",[39,94,95,98],{},[42,96,97],{},"Enterprise","：私有化部署 + 团队协作 + 资源管控，联系销售",[100,101,102],"blockquote",{},[26,103,104],{},"模型 API 费用按你自己绑定的供应商计费；本地 Ollama \u002F LM Studio 零成本。",[21,106,108],{"id":107},"实测mac-m2-中型知识库","实测（Mac M2 + 中型知识库）",[26,110,111],{},[42,112,113],{},"亮点：",[36,115,116,119,122,125,128],{},[39,117,118],{},"中文 UI \u002F 文档 \u002F 社区都顶级，零门槛上手",[39,120,121],{},"本地 RAG 拖入 30+ PDF 后向量化 \u003C 2 分钟（用 bge-m3）",[39,123,124],{},"多模型并排回答：让 Claude \u002F GPT \u002F DeepSeek 同回一个问题做比较",[39,126,127],{},"MCP 接 Brave Search + 自定义工具流畅",[39,129,130],{},"WebDAV 同步坚果云 \u002F 阿里云盘，桌面 + 移动设备数据互通",[26,132,133],{},[42,134,135],{},"踩坑：",[36,137,138,141,144,147,150],{},[39,139,140],{},"没有 Web 端 \u002F Docker 自托管（要这个用 LobeChat）",[39,142,143],{},"iOS 版尚未发布（roadmap 中）",[39,145,146],{},"大型 PDF（>100 MB）向量化偶有失败，要切小",[39,148,149],{},"助手市场质量参差，要自筛",[39,151,152],{},"模型 API 调用全靠你自己付费，新手要先理解 API Key 概念",[21,154,155],{"id":155},"上手",[157,158,159,162,165,174,177,180],"ol",{},[39,160,161],{},"cherry-ai.com 下载客户端（或 GitHub releases）",[39,163,164],{},"设置 → 模型服务 → 填 OpenAI \u002F Claude \u002F DeepSeek API Key",[39,166,167,168],{},"（可选）本地：装 Ollama → Cherry Studio 自动识别 endpoint ",[169,170,171],"a",{"href":171,"rel":172},"http:\u002F\u002Flocalhost:11434",[173],"nofollow",[39,175,176],{},"新建知识库 → 拖文件 \u002F 加网址 → 等向量化",[39,178,179],{},"新对话 → 选模型 → 勾知识库 → 提问",[39,181,182],{},"进阶：自定义助手（System Prompt）+ MCP 扩展工具",[21,184,185],{"id":185},"对比",[187,188,189,210],"table",{},[190,191,192],"thead",{},[193,194,195,199,201,204,207],"tr",{},[196,197,198],"th",{},"维度",[196,200,10],{},[196,202,203],{},"LobeChat",[196,205,206],{},"LM Studio",[196,208,209],{},"Open WebUI",[211,212,213,230,244,260,273,289,304,320],"tbody",{},[193,214,215,219,222,225,227],{},[216,217,218],"td",{},"形态",[216,220,221],{},"桌面",[216,223,224],{},"Web + 桌面",[216,226,221],{},[216,228,229],{},"Docker \u002F 桌面",[193,231,232,234,237,239,242],{},[216,233,44],{},[216,235,236],{},"✅ 云 + 本地",[216,238,236],{},[216,240,241],{},"本地为主",[216,243,236],{},[193,245,246,249,252,254,257],{},[216,247,248],{},"知识库 RAG",[216,250,251],{},"✅ 强",[216,253,251],{},[216,255,256],{},"弱",[216,258,259],{},"✅",[193,261,262,265,267,269,271],{},[216,263,264],{},"MCP",[216,266,259],{},[216,268,259],{},[216,270,256],{},[216,272,259],{},[193,274,275,278,281,284,287],{},[216,276,277],{},"自托管 \u002F Web",[216,279,280],{},"无 Web",[216,282,283],{},"✅ Docker",[216,285,286],{},"无",[216,288,283],{},[193,290,291,294,297,299,302],{},[216,292,293],{},"中文",[216,295,296],{},"5\u002F5",[216,298,296],{},[216,300,301],{},"4\u002F5",[216,303,301],{},[193,305,306,309,312,315,318],{},[216,307,308],{},"开源协议",[216,310,311],{},"AGPL-3.0",[216,313,314],{},"MIT",[216,316,317],{},"闭源（免费）",[216,319,314],{},[193,321,322,325,328,331,334],{},[216,323,324],{},"GitHub Stars",[216,326,327],{},"60k+",[216,329,330],{},"72k+",[216,332,333],{},"–",[216,335,336],{},"126k+",[21,338,339],{"id":339},"避坑",[36,341,342,348,354,360,366],{},[39,343,344,347],{},[42,345,346],{},"API Key 别明文外泄","：客户端配置文件以明文存 Key，机器借出前先清；团队共享用企业版 \u002F 自建中转",[39,349,350,353],{},[42,351,352],{},"知识库别一次塞太多","：单库 1000+ 文档检索质量明显下降，按主题切分多个知识库",[39,355,356,359],{},[42,357,358],{},"嵌入模型选择","：免费 bge-m3 够用；专业用付费 Pro\u002FBAAI\u002Fbge-m3 或 OpenAI text-embedding-3",[39,361,362,365],{},[42,363,364],{},"WebDAV 同步先小范围测","：知识库向量数据较大，先备份对话再开同步",[39,367,368,371],{},[42,369,370],{},"MCP 工具来源要可控","：MCP 是给 AI 真实工具能力，第三方插件审一遍代码",[21,373,375],{"id":374},"适合-不适合","适合 \u002F 不适合",[36,377,378,381,384,387,390,393,396],{},[39,379,380],{},"✅ 中文用户、AI 重度使用 \u002F 多模型管理",[39,382,383],{},"✅ 需要本地 RAG 知识库",[39,385,386],{},"✅ 关注数据隐私 \u002F 本地存储",[39,388,389],{},"✅ 想用 Ollama \u002F LM Studio 本地模型",[39,391,392],{},"❌ 需要 Web 端 \u002F Docker 自托管",[39,394,395],{},"❌ 团队多人共享 \u002F SSO",[39,397,398],{},"❌ iOS 主力用户",[21,400,401],{"id":401},"相关阅读",[36,403,404,410,416,422],{},[39,405,406],{},[169,407,409],{"href":408},"\u002Ftools\u002Fcoding\u002Flocal\u002Flobe-chat","LobeChat 评测",[39,411,412],{},[169,413,415],{"href":414},"\u002Ftools\u002Fcoding\u002Flocal\u002Flm-studio","LM Studio 评测",[39,417,418],{},[169,419,421],{"href":420},"\u002Ftools\u002Fcoding\u002Flocal\u002Follama","Ollama 评测",[39,423,424],{},[169,425,427],{"href":426},"\u002Fplaybook\u002Fonboarding\u002Frag-pipeline-build","RAG Pipeline 搭建 Playbook",[21,429,430],{"id":430},"来源",[157,432,433,440,447],{},[39,434,435,436],{},"Cherry Studio 官网（功能 + 下载）",[169,437,438],{"href":438,"rel":439},"https:\u002F\u002Fwww.cherry-ai.com\u002F",[173],[39,441,442,443],{},"MBLUO Studio — Cherry Studio 评测 2026 ",[169,444,445],{"href":445,"rel":446},"https:\u002F\u002Fmbluostudio.com\u002Ftools\u002Fcherry-studio",[173],[39,448,449,450],{},"Cursor IDE 博客 — Cherry Studio 完全指南（2025-03）",[169,451,452],{"href":452,"rel":453},"https:\u002F\u002Fwww.cursor-ide.com\u002Fblog\u002Fcherry-studio-guide",[173],{"title":455,"searchDepth":456,"depth":456,"links":457},"",3,[458,460,461,462,463,464,465,466,467,468],{"id":23,"depth":459,"text":24},2,{"id":34,"depth":459,"text":34},{"id":84,"depth":459,"text":84},{"id":107,"depth":459,"text":108},{"id":155,"depth":459,"text":155},{"id":185,"depth":459,"text":185},{"id":339,"depth":459,"text":339},{"id":374,"depth":459,"text":375},{"id":401,"depth":459,"text":401},{"id":430,"depth":459,"text":430},"local",5,"\u002Fimg\u002Ftools\u002Fcherry-studio.webp","Cherry Studio 真实评测：开源跨平台桌面 AI 客户端，集成 OpenAI \u002F Claude \u002F Gemini \u002F DeepSeek + Ollama \u002F LM Studio 本地模型，内置 300+ 助手模板 + 本地 RAG 知识库。AGPL-3.0 开源、GitHub 60k+ stars，企业版另询。",true,"md",[476,479,482,485],{"q":477,"a":478},"Cherry Studio 真的免费吗？","是。客户端完全免费、AGPL-3.0 开源，模型调用走你自己的 API Key（OpenAI \u002F Claude \u002F Gemini \u002F DeepSeek 等付费）或本地 Ollama \u002F LM Studio（零成本）。",{"q":480,"a":481},"本地知识库怎么用？","在『知识库』面板新建，拖文件 \u002F 加网址 \u002F 填 sitemap，系统自动向量化（默认 BAAI\u002Fbge-m3 或硅基流动的 Pro 版）；提问时勾选要检索的知识库，AI 会基于检索片段答题并标出来源。",{"q":483,"a":484},"和 LobeChat 怎么选？","都开源、多模型、有 RAG。LobeChat 是 Web + 桌面双形态，可自托管 Docker，72k stars；Cherry Studio 是纯桌面（Win\u002FMac\u002FLinux\u002FAndroid），不支持 Web 部署但桌面体验更精细，60k+ stars。要 Web 访问 \u002F 公司多人共享选 LobeChat；个人重度选 Cherry Studio。",{"q":486,"a":487},"支持 MCP \u002F 插件吗？","支持 MCP（Model Context Protocol）扩展，配合自定义助手（System Prompt）可扩展工具调用、联网搜索等能力。","https:\u002F\u002Fgithub.com\u002FCherryHQ\u002Fcherry-studio",[490,491],"zh","en","2026-08-02",{},"\u002Ftools\u002Fcoding\u002Flocal\u002Fcherry-studio","coding",[497,498,499,500],"windows","macos","linux","android",[502,506],{"plan":91,"price":503,"features":504,"notes":505},"免费","300+ 助手模板 \u002F 云端 + 本地模型 \u002F 知识库 \u002F MCP \u002F WebDAV 备份","AGPL-3.0 开源",{"plan":97,"price":507,"features":508,"notes":509},"联系销售","私有化部署 \u002F 团队协作 \u002F AI 资源管控 \u002F 知识库管理","面向企业团队","开源免费 \u002F 企业版联系销售","2026-06-19",[513,514],"onboarding\u002Frag-pipeline-build","onboarding\u002Fcursor-mcp-deep-integration",{"power":516,"ux":470,"price":470,"cn_support":470,"stability":516},4,{"title":10,"description":472},"Cherry Studio 评测 2026：AI 客户端工具，多模型桌面助手，开源免费","coding\u002Flocal\u002Fcherry-studio",[521,524,526],{"name":522,"url":438,"accessed":523},"Cherry Studio 官网","2026-06-24",{"name":525,"url":445,"accessed":523},"MBLUO Studio — Cherry Studio 评测",{"name":527,"url":452,"accessed":523},"Cursor IDE 博客 — Cherry Studio 指南","tools\u002Fcoding\u002Flocal\u002Fcherry-studio","全能 AI 客户端：多模型聚合 + 本地知识库 + 300+ 助手模板，跨平台桌面应用",[469,531,532,533,534,535,536],"desktop","multi-model","knowledge-base","rag","open-source","china","国产 AI 桌面客户端第一梯队，多模型聚合 + 本地 RAG + 中文体验顶级。需要 Web 部署 \u002F 自托管选 LobeChat；只要桌面体验完整选 Cherry Studio。","https:\u002F\u002Fcherry-ai.com","nhG3iaQ6G7dAWmMarVHUYX3EU47pSjb4yoUqlqZy18k",[541,1049,1529,1996],{"id":542,"title":203,"alternatives":543,"api_compatible":544,"body":560,"category":469,"chinese_friendly":470,"cover":1000,"description":1001,"domestic":473,"extension":474,"faq":1002,"free":473,"github":972,"languages":1015,"lastVerified":492,"meta":1016,"models":16,"navigation":473,"notSuitable":16,"opensource":473,"path":408,"pillar":495,"platforms":1017,"priceTable":1020,"pricing":1028,"published":511,"relatedPlaybooks":1029,"relatedReviews":16,"score":1031,"self_host":473,"seo":1032,"seoTitle":1033,"slug":12,"sources":1034,"stem":1041,"suitable":16,"tagline":1042,"tags":1043,"updated":523,"verdict":1046,"website":1047,"__hash__":1048},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Flobe-chat.md",[519,15,14,13],[545,546,547,548,549,550,551,552,553,554,555,556,557,558,559],"OpenAI","Anthropic","Google","Grok","Mistral","Cohere","阿里通义","百度文心","腾讯混元","Moonshot Kimi","字节豆包","DeepSeek","智谱 GLM","Ollama","Hugging Face",{"type":18,"value":561,"toc":988},[562,564,571,574,576,639,641,654,657,661,665,685,689,706,708,734,736,872,874,912,914,940,942,963,965],[21,563,24],{"id":23},[26,565,566,567,570],{},"LobeChat 是 LobeHub 团队的开源 AI 聊天框架，2023 年发布、GitHub 72k+ stars、MIT 协议。",[42,568,569],{},"Web + 桌面 + Docker 自托管三形态","，把 OpenAI \u002F Claude \u002F Gemini \u002F DeepSeek \u002F Qwen \u002F Kimi \u002F Ollama \u002F LM Studio 等 80+ 模型聚合到一个现代设计的客户端里。内置 RAG 知识库 + 插件市场 + 助手市场 + 多模型对比 + MCP，是当下综合最强的多模型 AI 客户端之一。",[26,572,573],{},"适合：需要 Web 端访问、Docker 自托管、多模型对比、丰富助手市场的用户；中文重度用户；想给团队 \u002F 家庭部署一个共享 AI 工作台。不适合：只用桌面 + 不需要 Web（Cherry Studio 同样优秀且更精细）、强企业 RBAC + 多租户（Open WebUI 多用户更完善）。",[21,575,34],{"id":34},[36,577,578,583,589,594,600,606,611,616,622,628],{},[39,579,580,582],{},[42,581,44],{},"：OpenAI \u002F Claude \u002F Gemini \u002F DeepSeek \u002F Qwen \u002F Kimi \u002F 豆包 \u002F Groq \u002F Together \u002F OpenRouter \u002F Ollama \u002F LM Studio",[39,584,585,588],{},[42,586,587],{},"多模型对比","：同 prompt 给多模型并排回答",[39,590,591,593],{},[42,592,50],{},"：上传 PDF \u002F Word \u002F 网页 → 向量化 → 检索引用",[39,595,596,599],{},[42,597,598],{},"插件市场","：联网搜索 \u002F 代码执行 \u002F 图像生成 \u002F 翻译等几十款官方插件",[39,601,602,605],{},[42,603,604],{},"助手市场","：几百个预设 AI 角色，一键导入",[39,607,608,610],{},[42,609,62],{},"：扩展任意工具能力",[39,612,613],{},[42,614,615],{},"代码解释器 \u002F 文件上传 \u002F TTS \u002F 多模态",[39,617,618,621],{},[42,619,620],{},"Web + 桌面 + Docker","：三形态，数据可完全本地",[39,623,624,627],{},[42,625,626],{},"LobeHub Cloud","：官方云托管，免部署",[39,629,630,633,634,638],{},[42,631,632],{},"快捷指令 \u002F 工作流","：自定义 prompt 模板，",[635,636,637],"code",{},"\u002Fpodcast-summary"," 类用法",[21,640,84],{"id":84},[36,642,643,649],{},[39,644,645,648],{},[42,646,647],{},"自托管 \u002F 桌面","：完全免费、MIT 开源",[39,650,651,653],{},[42,652,626],{},"：订阅制，云端托管 + 团队协作 + 同步",[26,655,656],{},"模型 API 费用按你自己的供应商付费；本地 Ollama \u002F LM Studio 零成本。",[21,658,660],{"id":659},"实测m2-自托管-docker连-openai-deepseek-本地-ollama","实测（M2 + 自托管 Docker，连 OpenAI + DeepSeek + 本地 Ollama）",[26,662,663],{},[42,664,113],{},[36,666,667,670,673,676,679,682],{},[39,668,669],{},"界面颜值是这一类工具里第一档（深色 \u002F 透明 \u002F 现代感）",[39,671,672],{},"多模型并排对比对选型极其有用：写一道复杂题，Claude \u002F GPT \u002F DeepSeek 直接对比答案",[39,674,675],{},"知识库 RAG 上传 50+ PDF 后检索准确，引用片段可视化",[39,677,678],{},"助手市场拿来即用——「Code Reviewer」「Translation Polish」节省 prompt 编写",[39,680,681],{},"Docker 一键部署，团队 5 人共享流畅",[39,683,684],{},"多平台数据同步（Cloud \u002F WebDAV）",[26,686,687],{},[42,688,135],{},[36,690,691,694,697,700,703],{},[39,692,693],{},"自托管要熟悉 Docker + 反代 + HTTPS",[39,695,696],{},"国内连 OpenAI \u002F Claude 需自带网络方案",[39,698,699],{},"Web 版数据存 LobeHub，隐私敏感场景走桌面 \u002F Docker",[39,701,702],{},"插件市场质量参差，要自筛",[39,704,705],{},"团队多人共享需配 LobeHub Cloud 或自建数据库（Postgres + S3）",[21,707,155],{"id":155},[157,709,710,713,719,722,725,728,731],{},[39,711,712],{},"选形态：Web（chat.lobehub.com 注册即用） \u002F 桌面（GitHub Releases 下载） \u002F Docker",[39,714,715,716],{},"Docker：",[635,717,718],{},"docker run -d -p 3210:3210 -e OPENAI_API_KEY=sk-xxx --name lobe-chat lobehub\u002Flobe-chat",[39,720,721],{},"设置 → AI 服务商 → 添加 OpenAI \u002F Claude \u002F DeepSeek \u002F Ollama",[39,723,724],{},"模型选择器测试对话",[39,726,727],{},"知识库：拖文件 → 等向量化 → 对话引用",[39,729,730],{},"助手市场拉「Code Reviewer」「论文翻译润色」试用",[39,732,733],{},"进阶：插件市场启用联网搜索 \u002F 代码执行；MCP 自定义工具",[21,735,185],{"id":185},[187,737,738,752],{},[190,739,740],{},[193,741,742,744,746,748,750],{},[196,743,198],{},[196,745,203],{},[196,747,10],{},[196,749,209],{},[196,751,206],{},[211,753,754,766,779,792,805,821,834,848,860],{},[193,755,756,758,760,762,764],{},[216,757,218],{},[216,759,620],{},[216,761,221],{},[216,763,229],{},[216,765,221],{},[193,767,768,770,773,775,777],{},[216,769,44],{},[216,771,772],{},"✅ 80+",[216,774,259],{},[216,776,259],{},[216,778,241],{},[193,780,781,783,786,788,790],{},[216,782,587],{},[216,784,785],{},"✅ 一等",[216,787,259],{},[216,789,256],{},[216,791,256],{},[193,793,794,796,798,800,803],{},[216,795,248],{},[216,797,259],{},[216,799,259],{},[216,801,802],{},"✅ + oikb",[216,804,256],{},[193,806,807,810,813,816,819],{},[216,808,809],{},"插件 \u002F 助手市场",[216,811,812],{},"✅ 丰富",[216,814,815],{},"300+ 助手",[216,817,818],{},"Tools",[216,820,256],{},[193,822,823,825,827,829,832],{},[216,824,264],{},[216,826,259],{},[216,828,259],{},[216,830,831],{},"✅ mcpo",[216,833,256],{},[193,835,836,839,842,844,846],{},[216,837,838],{},"多用户",[216,840,841],{},"配 Cloud \u002F 自建",[216,843,286],{},[216,845,785],{},[216,847,286],{},[193,849,850,852,854,856,858],{},[216,851,324],{},[216,853,330],{},[216,855,327],{},[216,857,336],{},[216,859,333],{},[193,861,862,864,866,868,870],{},[216,863,308],{},[216,865,314],{},[216,867,311],{},[216,869,314],{},[216,871,317],{},[21,873,339],{"id":339},[36,875,876,882,888,894,900,906],{},[39,877,878,881],{},[42,879,880],{},"Web 版数据不本地","：隐私敏感选桌面或 Docker 自托管",[39,883,884,887],{},[42,885,886],{},"国内连海外模型走中转","：直连 OpenAI \u002F Claude 不稳，配 OpenRouter \u002F Ofox \u002F 国内中转",[39,889,890,893],{},[42,891,892],{},"Docker 自托管暴露公网","：上反代 + HTTPS + Auth + 备份数据库",[39,895,896,899],{},[42,897,898],{},"嵌入模型中文优化","：默认嵌入对中文一般，配 bge-m3 \u002F 硅基流动 Pro 版",[39,901,902,905],{},[42,903,904],{},"插件市场审一遍","：第三方插件可执行代码，团队部署谨慎启用",[39,907,908,911],{},[42,909,910],{},"同步选 Cloud vs WebDAV","：团队多端走 LobeHub Cloud；个人多设备 WebDAV 即可",[21,913,375],{"id":374},[36,915,916,919,922,925,928,931,934,937],{},[39,917,918],{},"✅ Web + 桌面双形态需求",[39,920,921],{},"✅ Docker 自托管 \u002F 团队共享",[39,923,924],{},"✅ 多模型对比 \u002F 选型",[39,926,927],{},"✅ 中文重度用户",[39,929,930],{},"✅ 助手市场 \u002F 插件生态用户",[39,932,933],{},"❌ 强企业 RBAC + 多租户（Open WebUI 更完善）",[39,935,936],{},"❌ 只要桌面 + 数据完全本地（Cherry Studio 同样优秀）",[39,938,939],{},"❌ 完全不会碰 Docker",[21,941,401],{"id":401},[36,943,944,949,955,959],{},[39,945,946],{},[169,947,948],{"href":494},"Cherry Studio 评测",[39,950,951],{},[169,952,954],{"href":953},"\u002Ftools\u002Fcoding\u002Flocal\u002Fopen-webui","Open WebUI 评测",[39,956,957],{},[169,958,421],{"href":420},[39,960,961],{},[169,962,427],{"href":426},[21,964,430],{"id":430},[157,966,967,974,981],{},[39,968,969,970],{},"LobeChat GitHub 仓库（72k+ stars，MIT）",[169,971,972],{"href":972,"rel":973},"https:\u002F\u002Fgithub.com\u002Flobehub\u002Flobe-chat",[173],[39,975,976,977],{},"腾讯云开发者社区 — Lobe Chat 本地化 AI 聊天终极桌面客户端（2026-01）",[169,978,979],{"href":979,"rel":980},"https:\u002F\u002Fcloud.tencent.com\u002Fdeveloper\u002Farticle\u002F2622150",[173],[39,982,983,984],{},"Ofox.ai — LobeChat 完全配置指南 2026（2026-04-17）",[169,985,986],{"href":986,"rel":987},"https:\u002F\u002Fofox.ai\u002Fzh\u002Fblog\u002Flobechat-api-configuration-guide-2026",[173],{"title":455,"searchDepth":456,"depth":456,"links":989},[990,991,992,993,994,995,996,997,998,999],{"id":23,"depth":459,"text":24},{"id":34,"depth":459,"text":34},{"id":84,"depth":459,"text":84},{"id":659,"depth":459,"text":660},{"id":155,"depth":459,"text":155},{"id":185,"depth":459,"text":185},{"id":339,"depth":459,"text":339},{"id":374,"depth":459,"text":375},{"id":401,"depth":459,"text":401},{"id":430,"depth":459,"text":430},"\u002Fimg\u002Ftools\u002Flobe-chat.webp","LobeChat 真实评测：LobeHub 团队开源 AI 聊天框架，GitHub 72k+ stars、MIT 协议。Web + 桌面（Win\u002FMac\u002FLinux\u002FDocker）双形态，支持 OpenAI \u002F Claude \u002F Gemini \u002F DeepSeek \u002F Qwen \u002F Ollama 等 80+ 模型，内置 RAG 知识库 + 插件市场 + 助手市场 + 多模型对比。",[1003,1006,1009,1012],{"q":1004,"a":1005},"Web 版 vs 桌面版 vs Docker 自托管，怎么选？","Web 版（chat.lobehub.com）最快上手但数据存 LobeHub 服务器；桌面版数据本地存、隐私好；Docker 自托管对团队 \u002F 公司部署最优，完全掌控数据。",{"q":1007,"a":1008},"支持哪些模型？","80+ 模型：OpenAI 全系列、Anthropic Claude、Google Gemini、DeepSeek、Qwen、Kimi、Moonshot、字节豆包、Groq、Together、OpenRouter、Ollama \u002F LM Studio 本地模型，以及任何 OpenAI 兼容 API。",{"q":1010,"a":1011},"多模型对比怎么用？","同一对话窗口里把消息广播给多个模型并排回答，选型 \u002F 评估特别有用——直接看 Claude 和 GPT 在同一 prompt 下的回答差异。",{"q":1013,"a":1014},"助手市场是什么？","LobeHub 维护的预设 AI 角色市场（代码审查 \u002F 翻译 \u002F 写作 \u002F 角色扮演等几百个），一键拉到本地用，省去自己写 System Prompt。",[490,491],{},[1018,497,498,499,1019],"web","docker",[1021,1024],{"plan":647,"price":503,"features":1022,"notes":1023},"全功能 \u002F 80+ 模型 \u002F 知识库 \u002F 插件 \u002F 助手市场","MIT 协议",{"plan":626,"price":1025,"features":1026,"notes":1027},"订阅制","云端托管 \u002F 免部署 \u002F 团队协作 \u002F 同步","chat.lobehub.com 注册即用","完全免费（MIT 开源） \u002F LobeHub Cloud 订阅",[513,1030],"onboarding\u002Fclaude-code-getting-started",{"power":470,"ux":470,"price":470,"cn_support":470,"stability":516},{"title":203,"description":1001},"LobeChat - 开源 AI 聊天框架评测与部署 | AIHO",[1035,1037,1039],{"name":1036,"url":972,"accessed":523},"LobeChat GitHub",{"name":1038,"url":979,"accessed":523},"腾讯云开发者社区 — Lobe Chat 终极桌面客户端",{"name":1040,"url":986,"accessed":523},"Ofox.ai — LobeChat 完全配置指南 2026","tools\u002Fcoding\u002Flocal\u002Flobe-chat","现代设计的开源 AI 聊天框架——Web + 桌面双形态、72k+ stars、多模型 + 知识库 + 插件市场",[469,1018,531,532,534,1044,1045,535],"plugin","mcp","颜值与功能双优的多模型 AI 聊天客户端。要 Web + 桌面双形态、自托管 Docker、多模型对比、丰富助手市场——LobeChat 是综合最强；纯桌面体验 Cherry Studio 同样优秀。","https:\u002F\u002Flobehub.com","sAhQ5CHDobJa--z6NbtgysTQDwTBYsoGPXR0zat-5-w",{"id":1050,"title":206,"alternatives":1051,"api_compatible":16,"body":1052,"category":469,"chinese_friendly":456,"cover":1478,"description":1479,"domestic":1480,"extension":474,"faq":1481,"free":473,"github":16,"languages":1494,"lastVerified":492,"meta":1495,"models":16,"navigation":473,"notSuitable":16,"opensource":1480,"path":414,"pillar":495,"platforms":1496,"priceTable":1497,"pricing":1505,"published":511,"relatedPlaybooks":1506,"relatedReviews":16,"score":1507,"self_host":473,"seo":1508,"seoTitle":1509,"slug":13,"sources":1510,"stem":1517,"suitable":16,"tagline":1518,"tags":1519,"updated":523,"verdict":1526,"website":1527,"__hash__":1528},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Flm-studio.md",[14,15,519,12],{"type":18,"value":1053,"toc":1466},[1054,1056,1063,1066,1068,1125,1127,1141,1146,1150,1154,1171,1175,1192,1194,1220,1222,1360,1362,1394,1396,1419,1421,1441,1443],[21,1055,24],{"id":23},[26,1057,1058,1059,1062],{},"LM Studio 是 Windows \u002F macOS \u002F Linux 桌面应用，让你像浏览 App Store 一样发现、下载、运行本地大模型（GGUF \u002F MLX 格式）。底层基于 llama.cpp + MLX，Mac M 系列原生优化。0.3+ 起新增 Headless 模式 + ",[635,1060,1061],{},"lms"," CLI，可在服务器跑 OpenAI 兼容 API（默认 :1234）。个人 \u002F 评估完全免费，商用咨询。",[26,1064,1065],{},"适合：本地 LLM 入门 \u002F 评估、Mac 用户、需要 GUI 调参 \u002F 模型比较、想给 IDE \u002F 应用接本地 OpenAI 兼容 endpoint 的开发者。不适合：多用户并发生产服务（用 vLLM）、嵌入式 \u002F 边缘部署（用 llama.cpp）、纯 CLI 工作流（用 Ollama）。",[21,1067,34],{"id":34},[36,1069,1070,1076,1082,1088,1098,1107,1113,1119],{},[39,1071,1072,1075],{},[42,1073,1074],{},"模型浏览器","：内置 Hugging Face 检索，按 GGUF \u002F MLX \u002F 大小筛选、一键下载",[39,1077,1078,1081],{},[42,1079,1080],{},"聊天界面","：System Prompt \u002F temperature \u002F top-p \u002F context size 可视化调参",[39,1083,1084,1087],{},[42,1085,1086],{},"多模型并存 \u002F 切换","：同时加载多模型在不同会话中比较",[39,1089,1090,1093,1094,1097],{},[42,1091,1092],{},"OpenAI 兼容 Local Server","：",[635,1095,1096],{},"http:\u002F\u002Flocalhost:1234\u002Fv1","，任何 SDK 即接即用",[39,1099,1100,1093,1103,1106],{},[42,1101,1102],{},"Headless \u002F CLI",[635,1104,1105],{},"lms server start --port 1234","，无 GUI 可跑",[39,1108,1109,1112],{},[42,1110,1111],{},"PDF \u002F 文档对话","：内置基础 RAG，丢文件就能聊",[39,1114,1115,1118],{},[42,1116,1117],{},"MLX 原生支持（Mac）","：M1+ 上比 GGUF + Metal 快 30–50%",[39,1120,1121,1124],{},[42,1122,1123],{},"持续批处理","：Codersera 2026 测得 50–90 tok\u002Fs（消费级 GPU + 中等模型）",[21,1126,84],{"id":84},[36,1128,1129,1135],{},[39,1130,1131,1134],{},[42,1132,1133],{},"个人 \u002F 评估","：免费，全功能可用",[39,1136,1137,1140],{},[42,1138,1139],{},"商用","：邮件 \u002F 官网联系 LM Studio 团队",[100,1142,1143],{},[26,1144,1145],{},"模型本身免费（开源权重），LM Studio 不抽水任何 token 费用。",[21,1147,1149],{"id":1148},"实测mac-m2-pro-qwen3-coder-7b-gguf-q4_k_m","实测（Mac M2 Pro + Qwen3-Coder-7B GGUF Q4_K_M）",[26,1151,1152],{},[42,1153,113],{},[36,1155,1156,1159,1162,1165,1168],{},[39,1157,1158],{},"模型浏览器极舒服：搜「qwen3-coder」直接列出 GGUF + MLX 各 quant，标硬件兼容度",[39,1160,1161],{},"加载 7B Q4 模型 \u003C 3 秒，生成 ~75 tok\u002Fs",[39,1163,1164],{},"Local Server 开了 Cursor 直接接 baseURL → 本地代码补全零成本",[39,1166,1167],{},"MLX 版同模型 ~110 tok\u002Fs，差距显著",[39,1169,1170],{},"多窗口加载 2 个模型并排测，调 prompt 直观",[26,1172,1173],{},[42,1174,135],{},[36,1176,1177,1180,1183,1186,1189],{},[39,1178,1179],{},"模型库依赖 Hugging Face，国内访问要镜像 \u002F 代理",[39,1181,1182],{},"GPU 显存吃满后会自动 offload 到 CPU，无提示就慢下来",[39,1184,1185],{},"Headless 模式相对 Ollama 偏新，文档稍少",[39,1187,1188],{},"闭源应用（虽免费），不适合企业合规挂钩",[39,1190,1191],{},"中文 UI 可用但部分菜单仍英文",[21,1193,155],{"id":155},[157,1195,1196,1199,1202,1205,1208,1215],{},[39,1197,1198],{},"lmstudio.ai 下载（Mac \u002F Windows \u002F Linux）",[39,1200,1201],{},"打开 → Discover 标签 → 搜模型（如 qwen3-coder、deepseek-v3 GGUF\u002FMLX）→ Download",[39,1203,1204],{},"Chat 标签 → 选模型 → 调参聊天",[39,1206,1207],{},"Local Server 标签 → Start Server → 默认端口 1234",[39,1209,1210,1211,1214],{},"在你的应用里：",[635,1212,1213],{},"baseURL = \"http:\u002F\u002Flocalhost:1234\u002Fv1\"","，API Key 任意",[39,1216,1217,1218],{},"Headless：",[635,1219,1105],{},[21,1221,185],{"id":185},[187,1223,1224,1239],{},[190,1225,1226],{},[193,1227,1228,1230,1232,1234,1236],{},[196,1229,198],{},[196,1231,206],{},[196,1233,558],{},[196,1235,209],{},[196,1237,1238],{},"llama.cpp",[211,1240,1241,1257,1273,1288,1303,1317,1331,1344],{},[193,1242,1243,1245,1248,1251,1254],{},[216,1244,218],{},[216,1246,1247],{},"GUI + CLI",[216,1249,1250],{},"CLI Daemon",[216,1252,1253],{},"Docker UI",[216,1255,1256],{},"二进制",[193,1258,1259,1262,1265,1268,1270],{},[216,1260,1261],{},"模型浏览",[216,1263,1264],{},"✅ 内置",[216,1266,1267],{},"CLI pull",[216,1269,286],{},[216,1271,1272],{},"手动",[193,1274,1275,1278,1280,1283,1286],{},[216,1276,1277],{},"参数调优 GUI",[216,1279,259],{},[216,1281,1282],{},"❌",[216,1284,1285],{},"部分",[216,1287,1282],{},[193,1289,1290,1293,1296,1299,1301],{},[216,1291,1292],{},"OpenAI 兼容 API",[216,1294,1295],{},"✅ :1234",[216,1297,1298],{},"✅ :11434",[216,1300,259],{},[216,1302,259],{},[193,1304,1305,1308,1310,1313,1315],{},[216,1306,1307],{},"MLX (Mac)",[216,1309,259],{},[216,1311,1312],{},"✅ 0.19+",[216,1314,333],{},[216,1316,333],{},[193,1318,1319,1322,1324,1326,1328],{},[216,1320,1321],{},"多用户并发",[216,1323,256],{},[216,1325,256],{},[216,1327,259],{},[216,1329,1330],{},"中",[193,1332,1333,1336,1338,1340,1342],{},[216,1334,1335],{},"开源",[216,1337,317],{},[216,1339,314],{},[216,1341,314],{},[216,1343,314],{},[193,1345,1346,1349,1352,1355,1357],{},[216,1347,1348],{},"上手难度",[216,1350,1351],{},"极低",[216,1353,1354],{},"低",[216,1356,1330],{},[216,1358,1359],{},"高",[21,1361,339],{"id":339},[36,1363,1364,1370,1376,1382,1388],{},[39,1365,1366,1369],{},[42,1367,1368],{},"国内下模型走镜像","：HF 直连慢 \u002F 卡，配 HF_ENDPOINT=hf-mirror.com",[39,1371,1372,1375],{},[42,1373,1374],{},"显存爆 ≠ 报错","：GPU 装不下会无声 offload 到 CPU，关注生成速度，必要时降 quant 或换小模型",[39,1377,1378,1381],{},[42,1379,1380],{},"MLX 优先（Mac M 系列）","：能下 MLX 版就别下 GGUF，速度差距明显",[39,1383,1384,1387],{},[42,1385,1386],{},"Local Server 暴露要谨慎","：默认 0.0.0.0 + 无鉴权，对外开放前加反代 + Bearer",[39,1389,1390,1393],{},[42,1391,1392],{},"闭源合规要核","：企业内部使用前查 license；商用必须联系官方",[21,1395,375],{"id":374},[36,1397,1398,1401,1404,1407,1410,1413,1416],{},[39,1399,1400],{},"✅ 本地 LLM 入门 \u002F 评估",[39,1402,1403],{},"✅ Mac M 系列用户",[39,1405,1406],{},"✅ 想给 Cursor \u002F Cline 接本地 OpenAI 兼容 endpoint",[39,1408,1409],{},"✅ 需要 GUI 调参 \u002F 模型比较",[39,1411,1412],{},"❌ 多用户并发生产服务",[39,1414,1415],{},"❌ 嵌入式 \u002F 边缘设备",[39,1417,1418],{},"❌ 强合规 \u002F 必须开源审计",[21,1420,401],{"id":401},[36,1422,1423,1427,1431,1435],{},[39,1424,1425],{},[169,1426,421],{"href":420},[39,1428,1429],{},[169,1430,954],{"href":953},[39,1432,1433],{},[169,1434,948],{"href":494},[39,1436,1437],{},[169,1438,1440],{"href":1439},"\u002Fplaybook\u002Fonboarding\u002Fclaude-code-getting-started","Claude Code 上手 Playbook",[21,1442,430],{"id":430},[157,1444,1445,1452,1459],{},[39,1446,1447,1448],{},"LM Studio 官网 ",[169,1449,1450],{"href":1450,"rel":1451},"https:\u002F\u002Flmstudio.ai\u002F",[173],[39,1453,1454,1455],{},"Codersera — LM Studio Complete Guide 2026 ",[169,1456,1457],{"href":1457,"rel":1458},"https:\u002F\u002Fcodersera.com\u002Fblog\u002Flm-studio-complete-guide-2026\u002F",[173],[39,1460,1461,1462],{},"Codersera — Ollama vs LM Studio vs vLLM vs llama.cpp vs MLX 2026 ",[169,1463,1464],{"href":1464,"rel":1465},"https:\u002F\u002Fcodersera.com\u002Fblog\u002Follama-vs-lm-studio-vs-vllm-vs-llama-cpp-vs-mlx-2026\u002F",[173],{"title":455,"searchDepth":456,"depth":456,"links":1467},[1468,1469,1470,1471,1472,1473,1474,1475,1476,1477],{"id":23,"depth":459,"text":24},{"id":34,"depth":459,"text":34},{"id":84,"depth":459,"text":84},{"id":1148,"depth":459,"text":1149},{"id":155,"depth":459,"text":155},{"id":185,"depth":459,"text":185},{"id":339,"depth":459,"text":339},{"id":374,"depth":459,"text":375},{"id":401,"depth":459,"text":401},{"id":430,"depth":459,"text":430},"\u002Fimg\u002Ftools\u002Flm-studio.webp","LM Studio 真实评测：跨平台桌面应用，运行本地 GGUF \u002F MLX 大模型。50–90 tok\u002Fs 持续批处理、OpenAI 兼容本地 API（默认端口 1234）、Headless 模式、Mac \u002F Win 双端。对个人开发者免费，企业咨询。",false,[1482,1485,1488,1491],{"q":1483,"a":1484},"和 Ollama 怎么选？","LM Studio 是 GUI 优先（模型浏览器 + 参数面板 + 聊天界面），适合个人 \u002F 评估 \u002F 上手。Ollama 是 CLI \u002F Daemon 优先（后台跑 + REST API），适合应用嵌入 \u002F 脚本调用。两者都基于 llama.cpp，在 Mac M 系列上都已用 MLX。",{"q":1486,"a":1487},"支持 MLX 吗？","支持。Mac M1+ 上可加载 MLX 格式模型，速度比 GGUF + Metal 快 30–50%。模型搜索时筛选 MLX 即可。",{"q":1489,"a":1490},"OpenAI 兼容 API 怎么用？","开 Local Server → 默认端口 1234 → `http:\u002F\u002Flocalhost:1234\u002Fv1`。任何 OpenAI SDK 把 baseURL 改这个就能跑本地模型，零代码改动。",{"q":1492,"a":1493},"Headless 模式？","0.3+ 起支持 `lms server start` CLI 启动后台服务，无 GUI 即可跑 OpenAI 兼容 API，适合服务器 \u002F SSH 场景。",[491,490],{},[497,498,499],[1498,1501],{"plan":1133,"price":503,"features":1499,"notes":1500},"全功能 GUI + Headless API + GGUF\u002FMLX","供个人 \u002F 评估使用",{"plan":1139,"price":1502,"features":1503,"notes":1504},"联系咨询","团队部署 \u002F 商用 license","邮件 \u002F 官网联系","免费（个人 \u002F 评估） \u002F 企业 \u002F 商用咨询",[513,1030],{"power":516,"ux":470,"price":470,"cn_support":456,"stability":516},{"title":206,"description":1479},"LM Studio 评测 2026：本地运行开源大模型，图形化界面，AI 模型管理",[1511,1513,1515],{"name":1512,"url":1450,"accessed":523},"LM Studio 官网",{"name":1514,"url":1457,"accessed":523},"Codersera — LM Studio Complete Guide 2026",{"name":1516,"url":1464,"accessed":523},"Codersera — Ollama vs LM Studio vs vLLM 2026","tools\u002Fcoding\u002Flocal\u002Flm-studio","本地 LLM 的 GUI 首选——模型浏览器 + GGUF\u002FMLX 推理 + OpenAI 兼容 API + Mac 原生优化",[469,1520,1521,1522,1523,1524,1525],"gui","gguf","mlx","llama-cpp","mac","openai-compatible","Mac \u002F Windows 桌面本地 LLM 的 GUI 首选——上手最快、模型浏览最舒服、自带 OpenAI 兼容 API。批量服务 \u002F 多用户场景用 vLLM；纯 CLI \u002F 嵌入应用走 Ollama。","https:\u002F\u002Flmstudio.ai","Pj3jNb1Z4S55e91UkW1ZMgI86ps_Wm7yCc7zeXPF05U",{"id":1530,"title":558,"alternatives":1531,"api_compatible":1532,"body":1533,"category":469,"chinese_friendly":456,"cover":1952,"description":1953,"domestic":1480,"extension":474,"faq":1954,"free":473,"github":1967,"languages":1968,"lastVerified":492,"meta":1969,"models":16,"navigation":473,"notSuitable":16,"opensource":473,"path":420,"pillar":495,"platforms":1970,"priceTable":1971,"pricing":1975,"published":511,"relatedPlaybooks":1976,"relatedReviews":16,"score":1977,"self_host":473,"seo":1978,"seoTitle":1979,"slug":14,"sources":1980,"stem":1986,"suitable":16,"tagline":1987,"tags":1988,"updated":523,"verdict":1993,"website":1994,"__hash__":1995},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Follama.md",[13,15,519,12],[545,546,547,548,549,550,551,552,553,554,555,556,557,558,559],{"type":18,"value":1534,"toc":1940},[1535,1537,1544,1547,1549,1617,1619,1622,1626,1630,1650,1654,1685,1687,1721,1723,1838,1840,1872,1874,1897,1899,1917,1919],[21,1536,24],{"id":23},[26,1538,1539,1540,1543],{},"Ollama 是本地 LLM 的 Daemon 事实标准——后台跑、暴露 REST API（11434）+ CLI、Modelfile 配置、GGUF 一站式。MIT 开源，跨 Win \u002F Mac \u002F Linux。0.19+ 起 Mac M 系列底层切 MLX 推理。模型库覆盖 Llama \u002F Qwen \u002F DeepSeek \u002F Gemma \u002F Mistral 等主流开源模型，",[635,1541,1542],{},"ollama pull"," 一键拉。",[26,1545,1546],{},"适合：给 Cursor \u002F Cline \u002F Continue \u002F Open WebUI 接本地 OpenAI 兼容 endpoint、个人 \u002F 评估 \u002F 原型、嵌入应用、自动化脚本。不适合：GUI 偏好用户（用 LM Studio）、多用户并发生产服务（用 vLLM）、模型浏览 \u002F 调参界面（用 LM Studio）。",[21,1548,34],{"id":34},[36,1550,1551,1557,1565,1571,1578,1593,1599,1605,1611],{},[39,1552,1553,1556],{},[42,1554,1555],{},"后台 Daemon","：开机自启，应用调用零延迟",[39,1558,1559,1093,1562],{},[42,1560,1561],{},"CLI",[635,1563,1564],{},"ollama pull \u002F run \u002F list \u002F show \u002F create \u002F serve",[39,1566,1567,1570],{},[42,1568,1569],{},"Modelfile","：类 Dockerfile 注册任意 GGUF，配 SYSTEM \u002F PARAMETER \u002F TEMPLATE",[39,1572,1573,1093,1575],{},[42,1574,1292],{},[635,1576,1577],{},"http:\u002F\u002Flocalhost:11434\u002Fv1\u002Fchat\u002Fcompletions",[39,1579,1580,1093,1583,1586,1587,1586,1590],{},[42,1581,1582],{},"原生 API",[635,1584,1585],{},"\u002Fapi\u002Fchat","、",[635,1588,1589],{},"\u002Fapi\u002Fgenerate",[635,1591,1592],{},"\u002Fapi\u002Fembeddings",[39,1594,1595,1598],{},[42,1596,1597],{},"模型库","：官方注册表内置 Llama \u002F Qwen \u002F DeepSeek \u002F Gemma \u002F Mistral \u002F GPT-OSS 等",[39,1600,1601,1604],{},[42,1602,1603],{},"MLX 加速（Mac）","：0.19+ 起 M 系列自动用 MLX",[39,1606,1607,1610],{},[42,1608,1609],{},"量化","：默认 Q4_K_M、支持 Q5 \u002F Q8 \u002F FP16",[39,1612,1613,1616],{},[42,1614,1615],{},"跨平台","：Win \u002F Mac \u002F Linux 安装包，Docker 官方镜像",[21,1618,84],{"id":84},[26,1620,1621],{},"完全免费、MIT 开源、商用免费。",[21,1623,1625],{"id":1624},"实测m2-pro-qwen3-coder-7b-q4","实测（M2 Pro + Qwen3-Coder-7B Q4）",[26,1627,1628],{},[42,1629,113],{},[36,1631,1632,1638,1641,1644,1647],{},[39,1633,1634,1637],{},[635,1635,1636],{},"ollama run qwen3-coder:7b"," 一行起飞，3 秒进交互",[39,1639,1640],{},"REST API 配 Cursor \u002F Cline \u002F Continue 几乎全工具开箱即用",[39,1642,1643],{},"Modelfile 写自定义编码助手（low temperature + system prompt + 16K context）几分钟搞定",[39,1645,1646],{},"多模型并存，按需切换，内存占用合理",[39,1648,1649],{},"Mac M 系列 MLX 后比旧 GGUF 模式快显著",[26,1651,1652],{},[42,1653,135],{},[36,1655,1656,1666,1672,1679,1682],{},[39,1657,1658,1659,1662,1663],{},"默认 ",[635,1660,1661],{},"num_ctx"," 偏小（2048），跑长上下文要在 Modelfile 加 ",[635,1664,1665],{},"PARAMETER num_ctx 16384",[39,1667,1668,1669],{},"模型默认走 0.0.0.0:11434 ↔ Docker 容器互访要 ",[635,1670,1671],{},"--add-host=host.docker.internal:host-gateway",[39,1673,1674,1675,1678],{},"国内 ",[635,1676,1677],{},"ollama.com\u002Flibrary"," 下载偶有慢，可手动 HF 下 GGUF + Modelfile 自建",[39,1680,1681],{},"多用户并发吞吐显著低于 vLLM",[39,1683,1684],{},"没有 GUI，模型浏览 \u002F 参数面板要走 LM Studio \u002F Open WebUI 配合",[21,1686,155],{"id":155},[157,1688,1689,1695,1701,1706,1712,1718],{},[39,1690,1691,1694],{},[635,1692,1693],{},"curl -fsSL https:\u002F\u002Follama.ai\u002Finstall.sh | sh","（Mac \u002F Linux）；Windows winget",[39,1696,1697,1700],{},[635,1698,1699],{},"ollama pull qwen3-coder:7b","（按需换模型）",[39,1702,1703,1705],{},[635,1704,1636],{}," 直接聊",[39,1707,1708,1709],{},"应用接入：baseURL = ",[635,1710,1711],{},"http:\u002F\u002Flocalhost:11434\u002Fv1",[39,1713,1714,1715],{},"自定义：写 Modelfile → ",[635,1716,1717],{},"ollama create my-coder -f Modelfile",[39,1719,1720],{},"进阶：装 Open WebUI 做前端 \u002F 多人共享",[21,1722,185],{"id":185},[187,1724,1725,1740],{},[190,1726,1727],{},[193,1728,1729,1731,1733,1735,1738],{},[196,1730,198],{},[196,1732,558],{},[196,1734,206],{},[196,1736,1737],{},"vLLM",[196,1739,1238],{},[211,1741,1742,1758,1770,1783,1796,1812,1824],{},[193,1743,1744,1746,1749,1752,1755],{},[216,1745,218],{},[216,1747,1748],{},"CLI + Daemon",[216,1750,1751],{},"GUI + Headless",[216,1753,1754],{},"Python Server",[216,1756,1757],{},"C++ 二进制",[193,1759,1760,1762,1764,1766,1768],{},[216,1761,155],{},[216,1763,1351],{},[216,1765,1351],{},[216,1767,1330],{},[216,1769,1359],{},[193,1771,1772,1774,1776,1779,1781],{},[216,1773,1261],{},[216,1775,1561],{},[216,1777,1778],{},"✅ GUI",[216,1780,286],{},[216,1782,286],{},[193,1784,1785,1788,1790,1792,1794],{},[216,1786,1787],{},"OpenAI 兼容",[216,1789,1298],{},[216,1791,1295],{},[216,1793,259],{},[216,1795,259],{},[193,1797,1798,1801,1804,1807,1810],{},[216,1799,1800],{},"多用户吞吐",[216,1802,1803],{},"弱（~40 tok\u002Fs）",[216,1805,1806],{},"中（50–90）",[216,1808,1809],{},"强（800–12500）",[216,1811,1330],{},[193,1813,1814,1816,1818,1820,1822],{},[216,1815,1307],{},[216,1817,1312],{},[216,1819,259],{},[216,1821,1285],{},[216,1823,333],{},[193,1825,1826,1828,1830,1833,1836],{},[216,1827,1335],{},[216,1829,314],{},[216,1831,1832],{},"闭源",[216,1834,1835],{},"Apache 2.0",[216,1837,314],{},[21,1839,339],{"id":339},[36,1841,1842,1848,1854,1860,1866],{},[39,1843,1844,1847],{},[42,1845,1846],{},"num_ctx 一定要设","：默认 2K 太小，跑代码 \u002F 长文档要 16K+",[39,1849,1850,1853],{},[42,1851,1852],{},"Modelfile 模板别漏 TEMPLATE","：错的 chat template 会让模型输出乱码 \u002F 不停",[39,1855,1856,1859],{},[42,1857,1858],{},"KV cache 爆表 = 速度悬崖","：32B 模型 32K 上下文，KV cache 可能 12+ GB，超显存自动 offload 慢 10×",[39,1861,1862,1865],{},[42,1863,1864],{},"不要 0.0.0.0 直接对公网","：默认无鉴权，对外暴露走反代 + Bearer \u002F mTLS",[39,1867,1868,1871],{},[42,1869,1870],{},"Mac 让它自动用 MLX","：升 0.19+；不要手动强制 GGUF + Metal",[21,1873,375],{"id":374},[36,1875,1876,1879,1882,1885,1888,1891,1894],{},[39,1877,1878],{},"✅ 应用 \u002F IDE 接本地模型（Cursor \u002F Cline \u002F Continue）",[39,1880,1881],{},"✅ 个人 \u002F 评估 \u002F 脚本自动化",[39,1883,1884],{},"✅ Modelfile 自定义系统 prompt + 参数",[39,1886,1887],{},"✅ Mac M 系列 MLX 用户",[39,1889,1890],{},"❌ 多用户并发生产服务（用 vLLM）",[39,1892,1893],{},"❌ GUI 调参 \u002F 模型浏览（配 LM Studio \u002F Open WebUI）",[39,1895,1896],{},"❌ 极致单卡吞吐研究（直接 llama.cpp \u002F vLLM）",[21,1898,401],{"id":401},[36,1900,1901,1905,1909,1913],{},[39,1902,1903],{},[169,1904,415],{"href":414},[39,1906,1907],{},[169,1908,954],{"href":953},[39,1910,1911],{},[169,1912,948],{"href":494},[39,1914,1915],{},[169,1916,427],{"href":426},[21,1918,430],{"id":430},[157,1920,1921,1928,1935],{},[39,1922,1923,1924],{},"Markaicode — Import GGUF Models into Ollama 2026（2026-05-15）",[169,1925,1926],{"href":1926,"rel":1927},"https:\u002F\u002Fmarkaicode.com\u002Fimport-gguf-models-ollama-guide",[173],[39,1929,1930,1931],{},"ComputingForGeeks — Ollama Models Cheat Sheet 2026 ",[169,1932,1933],{"href":1933,"rel":1934},"https:\u002F\u002Fcomputingforgeeks.com\u002Follama-models-cheat-sheet",[173],[39,1936,1461,1937],{},[169,1938,1464],{"href":1464,"rel":1939},[173],{"title":455,"searchDepth":456,"depth":456,"links":1941},[1942,1943,1944,1945,1946,1947,1948,1949,1950,1951],{"id":23,"depth":459,"text":24},{"id":34,"depth":459,"text":34},{"id":84,"depth":459,"text":84},{"id":1624,"depth":459,"text":1625},{"id":155,"depth":459,"text":155},{"id":185,"depth":459,"text":185},{"id":339,"depth":459,"text":339},{"id":374,"depth":459,"text":375},{"id":401,"depth":459,"text":401},{"id":430,"depth":459,"text":430},"\u002Fimg\u002Ftools\u002Follama.webp","Ollama 真实评测：本地 LLM 的事实标准 Daemon，CLI + REST API，模型库 + Modelfile + GGUF 一站式。0.19+ 在 Mac M 系列用 MLX 加速；OpenAI 兼容端点 11434；MIT 开源 + 跨平台。",[1955,1958,1961,1964],{"q":1956,"a":1957},"和 LM Studio 怎么选？","Ollama = Daemon + CLI，开机自启在 11434 端口跑，应用 \u002F IDE 调它最方便。LM Studio = GUI，模型浏览 \u002F 调参 \u002F 聊天体验更好。两者底层都基于 llama.cpp，Mac M 系列上都已切 MLX。",{"q":1959,"a":1960},"Modelfile 是什么？","类 Dockerfile 的模型配置：`FROM .\u002Fxxx.gguf` + PARAMETER \u002F TEMPLATE \u002F SYSTEM。把任意 GGUF 注册成本地模型供调用。`ollama create my-model -f Modelfile`。",{"q":1962,"a":1963},"OpenAI 兼容端点？","`http:\u002F\u002Flocalhost:11434\u002Fv1`。任何 OpenAI SDK 改 baseURL 即用。也可走原生 `\u002Fapi\u002Fchat`、`\u002Fapi\u002Fgenerate`。",{"q":1965,"a":1966},"并发能力？","单用户原型场景顺滑（~40 tok\u002Fs peak），多用户并发明显不如 vLLM（vLLM 的 PagedAttention + 连续批处理高 16–20×）。生产并发选 vLLM。","https:\u002F\u002Fgithub.com\u002Follama\u002Follama",[491],{},[497,498,499,1019],[1972],{"plan":91,"price":503,"features":1973,"notes":1974},"完整 CLI + REST API + Modelfile + 模型库 + MIT 协议","全平台、商用免费","完全免费 + 开源（MIT）",[513,1030],{"power":516,"ux":516,"price":470,"cn_support":456,"stability":470},{"title":558,"description":1953},"Ollama 评测 2026：本地运行大模型，开源 AI 模型管理工具，私有化部署指南",[1981,1983,1985],{"name":1982,"url":1926,"accessed":523},"Markaicode — Import GGUF 2026",{"name":1984,"url":1933,"accessed":523},"ComputingForGeeks — Ollama Cheat Sheet 2026",{"name":1516,"url":1464,"accessed":523},"tools\u002Fcoding\u002Flocal\u002Follama","本地 LLM 的 Daemon——CLI + REST API 后台跑，给 Cursor \u002F Cline \u002F Open WebUI 接本地模型最低门槛",[469,1989,1990,1991,1992,1521,1522,1525,535],"daemon","cli","rest-api","modelfile","本地 LLM 的 Daemon 事实标准，CLI \u002F Modelfile \u002F REST API 三件套配合最广泛。GUI 偏好用户走 LM Studio；多用户并发生产用 vLLM；其他场景几乎默认 Ollama。","https:\u002F\u002Follama.com","yL3ZqN3rlWsImgvBFSYlFraTqj9ki9gSJlMbXVmruyg",{"id":1997,"title":209,"alternatives":1998,"api_compatible":1999,"body":2000,"category":469,"chinese_friendly":516,"cover":2415,"description":2416,"domestic":1480,"extension":474,"faq":2417,"free":473,"github":2430,"languages":2431,"lastVerified":492,"meta":2432,"models":16,"navigation":473,"notSuitable":16,"opensource":473,"path":953,"pillar":495,"platforms":2433,"priceTable":2435,"pricing":2442,"published":511,"relatedPlaybooks":2443,"relatedReviews":16,"score":2444,"self_host":473,"seo":2445,"seoTitle":2446,"slug":15,"sources":2447,"stem":2454,"suitable":16,"tagline":2455,"tags":2456,"updated":523,"verdict":2460,"website":2461,"__hash__":2462},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Fopen-webui.md",[12,519,14,13],[545,546,547,548,549,550,551,552,553,554,555,556,557,558,559],{"type":18,"value":2001,"toc":2403},[2002,2004,2007,2010,2012,2074,2076,2079,2083,2087,2115,2119,2143,2145,2179,2181,2288,2290,2333,2335,2358,2360,2378,2380],[21,2003,24],{"id":23},[26,2005,2006],{},"Open WebUI（原 Ollama WebUI）是 MIT 开源、自托管 AI 平台，最常见用法是 Docker 跑起来给 Ollama 套一个 ChatGPT 风格前端。GitHub 126k+ stars、282M+ Docker pulls，事实上的本地 AI 前端首选。支持任意 OpenAI 兼容后端 + RAG 知识库 + 多用户账号 + 工具调用 + MCP-OpenAPI 代理 + 联网搜索 + 语音 + 图像生成。",[26,2008,2009],{},"适合：团队 \u002F 家庭 \u002F 公司部署一份共享、要 Web 端访问、多用户分账号、SearXNG 联网搜索、Confluence \u002F S3 \u002F GitHub 数据源同步。不适合：单人桌面体验（用 Cherry Studio）、零运维 \u002F 不愿碰 Docker。",[21,2011,34],{"id":34},[36,2013,2014,2020,2026,2032,2038,2044,2050,2056,2062,2068],{},[39,2015,2016,2019],{},[42,2017,2018],{},"多模型后端","：Ollama \u002F OpenAI \u002F vLLM \u002F Anthropic \u002F Groq \u002F LocalAI \u002F 任意 OpenAI 兼容",[39,2021,2022,2025],{},[42,2023,2024],{},"多用户 + RBAC","：注册 \u002F 邀请 \u002F 角色权限 \u002F 工作区隔离",[39,2027,2028,2031],{},[42,2029,2030],{},"RAG 知识库","：上传文档 \u002F 网址 \u002F SearXNG 联网搜索 → 向量化 → 对话引用",[39,2033,2034,2037],{},[42,2035,2036],{},"Tools \u002F Functions","：Python 写函数即扩展（联网 \u002F 计算器 \u002F 自定义 API）",[39,2039,2040,2043],{},[42,2041,2042],{},"mcpo","：MCP-to-OpenAPI 代理，任意 MCP 服务器接进来",[39,2045,2046,2049],{},[42,2047,2048],{},"oikb","：知识库同步本地文件夹 \u002F GitHub \u002F S3 \u002F Confluence 等 40+ 源",[39,2051,2052,2055],{},[42,2053,2054],{},"open-terminal \u002F cptr","：给 AI 真实终端 + 文件 + 沙箱执行",[39,2057,2058,2061],{},[42,2059,2060],{},"图像生成","：Stable Diffusion \u002F DALL·E \u002F 自托管接入",[39,2063,2064,2067],{},[42,2065,2066],{},"语音输入 \u002F TTS","：内置",[39,2069,2070,2073],{},[42,2071,2072],{},"企业 LTS","：custom branding + SLA + 长期支持版本（联系销售）",[21,2075,84],{"id":84},[26,2077,2078],{},"完全免费、MIT 开源、商用免费。Enterprise 提供品牌定制 + SLA + LTS。",[21,2080,2082],{"id":2081},"实测ubuntu-2404-ollama-后端-5-人小团队","实测（Ubuntu 24.04 + Ollama 后端 + 5 人小团队）",[26,2084,2085],{},[42,2086,113],{},[36,2088,2089,2096,2099,2106,2109,2112],{},[39,2090,2091,2092,2095],{},"单条 ",[635,2093,2094],{},"docker run"," 五分钟上线",[39,2097,2098],{},"自带的多用户 + 角色权限省去重新搭 Auth",[39,2100,2101,2102,2105],{},"RAG 直传 30 个 PDF 后向量化顺利，对话中 ",[635,2103,2104],{},"#知识库"," 引用准确",[39,2107,2108],{},"mcpo 把 GitHub MCP 服务器接进来，团队对话里直接 issue \u002F PR 操作",[39,2110,2111],{},"模型切换流畅，OpenAI + Ollama 并存",[39,2113,2114],{},"SearXNG 联网搜索给模型实时信息，过时知识截止问题缓解",[26,2116,2117],{},[42,2118,135],{},[36,2120,2121,2124,2130,2137,2140],{},[39,2122,2123],{},"Docker 镜像 ~1.5GB，首次拉取偏慢",[39,2125,1658,2126,2129],{},[635,2127,2128],{},"0.0.0.0"," 公网暴露要加 HTTPS + 反代",[39,2131,2132,2133,2136],{},"嵌入模型 ",[635,2134,2135],{},"sentence-transformers"," 中文效果一般，建议换 bge-m3",[39,2138,2139],{},"多用户共享 Ollama 时并发吞吐瓶颈在 Ollama，不在 Open WebUI（生产用 vLLM 后端）",[39,2141,2142],{},"版本升级要看 changelog，部分 minor 含 breaking 改动",[21,2144,155],{"id":155},[157,2146,2147,2153,2160,2163,2166,2169,2172],{},[39,2148,2149,2150],{},"装 Docker → ",[635,2151,2152],{},"docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:\u002Fapp\u002Fbackend\u002Fdata --name open-webui --restart always ghcr.io\u002Fopen-webui\u002Fopen-webui:main",[39,2154,2155,2156,2159],{},"浏览器开 ",[635,2157,2158],{},"http:\u002F\u002Flocalhost:3000"," → 注册第一个账号（管理员）",[39,2161,2162],{},"设置 → Connections → 连接 Ollama \u002F 加 OpenAI Key",[39,2164,2165],{},"Models → Pull \u002F Discover 模型",[39,2167,2168],{},"Workspaces → 建知识库 → 上传文档",[39,2170,2171],{},"Tools → 启用 \u002F 写自定义函数",[39,2173,2174,2175,2178],{},"生产部署：Nginx 反代 + Let's Encrypt + 备份 ",[635,2176,2177],{},"\u002Fapp\u002Fbackend\u002Fdata"," volume",[21,2180,185],{"id":185},[187,2182,2183,2197],{},[190,2184,2185],{},[193,2186,2187,2189,2191,2193,2195],{},[196,2188,198],{},[196,2190,209],{},[196,2192,203],{},[196,2194,10],{},[196,2196,206],{},[211,2198,2199,2211,2223,2237,2250,2264,2276],{},[193,2200,2201,2203,2205,2207,2209],{},[216,2202,218],{},[216,2204,229],{},[216,2206,224],{},[216,2208,221],{},[216,2210,221],{},[193,2212,2213,2215,2217,2219,2221],{},[216,2214,838],{},[216,2216,785],{},[216,2218,259],{},[216,2220,286],{},[216,2222,286],{},[193,2224,2225,2228,2231,2233,2235],{},[216,2226,2227],{},"RAG",[216,2229,2230],{},"✅ 强 + oikb",[216,2232,259],{},[216,2234,259],{},[216,2236,256],{},[193,2238,2239,2242,2244,2246,2248],{},[216,2240,2241],{},"工具 \u002F MCP",[216,2243,831],{},[216,2245,259],{},[216,2247,259],{},[216,2249,256],{},[193,2251,2252,2255,2258,2260,2262],{},[216,2253,2254],{},"自托管",[216,2256,2257],{},"✅ Docker \u002F K8s",[216,2259,283],{},[216,2261,286],{},[216,2263,286],{},[193,2265,2266,2268,2270,2272,2274],{},[216,2267,324],{},[216,2269,336],{},[216,2271,330],{},[216,2273,327],{},[216,2275,333],{},[193,2277,2278,2280,2282,2284,2286],{},[216,2279,308],{},[216,2281,314],{},[216,2283,314],{},[216,2285,311],{},[216,2287,317],{},[21,2289,339],{"id":339},[36,2291,2292,2298,2306,2315,2321,2327],{},[39,2293,2294,2297],{},[42,2295,2296],{},"不要裸 0.0.0.0 + HTTP 暴露公网","：默认无 HTTPS，必上反代 + 强密码 + 速率限制",[39,2299,2300,2305],{},[42,2301,2302,2303,2178],{},"备份 ",[635,2304,2177],{},"：知识库 \u002F 用户 \u002F 对话全在里面",[39,2307,2308,2311,2312,2314],{},[42,2309,2310],{},"中文 RAG 换嵌入模型","：默认 ",[635,2313,2135],{}," 中文一般，配 bge-m3 或硅基流动嵌入 API",[39,2316,2317,2320],{},[42,2318,2319],{},"mcpo 工具范围谨慎","：MCP 给 AI 真实能力，第三方服务器审一遍",[39,2322,2323,2326],{},[42,2324,2325],{},"后端吞吐看 Ollama","：5+ 并发上 vLLM 后端，Ollama 单 worker 会排队",[39,2328,2329,2332],{},[42,2330,2331],{},"升级前看 changelog","：weekly 更新，偶有 breaking",[21,2334,375],{"id":374},[36,2336,2337,2340,2343,2346,2349,2352,2355],{},[39,2338,2339],{},"✅ 团队 \u002F 家庭 \u002F 公司多人共享 AI 平台",[39,2341,2342],{},"✅ 要 Web 端访问 \u002F 移动端兼容",[39,2344,2345],{},"✅ 自托管 \u002F 完全控制数据",[39,2347,2348],{},"✅ MCP \u002F 工具调用刚需",[39,2350,2351],{},"❌ 单人桌面体验（用 Cherry Studio）",[39,2353,2354],{},"❌ 零运维 \u002F 不愿碰 Docker",[39,2356,2357],{},"❌ iOS 原生 App 主力",[21,2359,401],{"id":401},[36,2361,2362,2366,2370,2374],{},[39,2363,2364],{},[169,2365,409],{"href":408},[39,2367,2368],{},[169,2369,948],{"href":494},[39,2371,2372],{},[169,2373,421],{"href":420},[39,2375,2376],{},[169,2377,427],{"href":426},[21,2379,430],{"id":430},[157,2381,2382,2389,2396],{},[39,2383,2384,2385],{},"Open WebUI 官方文档 ",[169,2386,2387],{"href":2387,"rel":2388},"https:\u002F\u002Fdocs.openwebui.com\u002F",[173],[39,2390,2391,2392],{},"Local AI Master — Open WebUI Setup Guide 2026 ",[169,2393,2394],{"href":2394,"rel":2395},"https:\u002F\u002Flocalaimaster.com\u002Fblog\u002Fopen-webui-setup-guide",[173],[39,2397,2398,2399],{},"AIToolDiscovery — Set Up Open-WebUI with Ollama 2026 ",[169,2400,2401],{"href":2401,"rel":2402},"https:\u002F\u002Fwww.aitooldiscovery.com\u002Fhow-to\u002Fsetup-open-webui-ollama",[173],{"title":455,"searchDepth":456,"depth":456,"links":2404},[2405,2406,2407,2408,2409,2410,2411,2412,2413,2414],{"id":23,"depth":459,"text":24},{"id":34,"depth":459,"text":34},{"id":84,"depth":459,"text":84},{"id":2081,"depth":459,"text":2082},{"id":155,"depth":459,"text":155},{"id":185,"depth":459,"text":185},{"id":339,"depth":459,"text":339},{"id":374,"depth":459,"text":375},{"id":401,"depth":459,"text":401},{"id":430,"depth":459,"text":430},"\u002Fimg\u002Ftools\u002Fopen-webui.webp","Open WebUI 2026 真实评测：MIT 开源、自托管 ChatGPT 替代和 Ollama Web 前端。支持 Docker 一行部署、Ollama\u002FOpenAI\u002FvLLM 多后端、RAG 知识库、多用户、联网搜索、工具调用和 MCP-to-OpenAPI，适合团队私有 AI 平台。",[2418,2421,2424,2427],{"q":2419,"a":2420},"Docker 一行命令真的够用吗？","够。`docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:\u002Fapp\u002Fbackend\u002Fdata --name open-webui --restart always ghcr.io\u002Fopen-webui\u002Fopen-webui:main`，5 分钟可上线、能多人注册、能接 Ollama \u002F OpenAI。生产再加反代 + HTTPS + 备份。",{"q":2422,"a":2423},"支持哪些模型后端？","Ollama（首选）+ 任何 OpenAI 兼容 endpoint：OpenAI 官方 \u002F Anthropic（OpenAI 兼容代理）\u002F vLLM \u002F Groq \u002F LocalAI \u002F 自建 baseURL。可同时配多个，对话中切换。",{"q":2425,"a":2426},"RAG \u002F 知识库怎么做？","内置：上传 PDF \u002F DOCX \u002F TXT、网址抓取、SearXNG 联网搜索 → 自动向量化 → 在对话中 `#` 引用知识库。配套 oikb 项目可同步本地文件夹 \u002F GitHub \u002F S3 \u002F Confluence 等 40+ 数据源。",{"q":2428,"a":2429},"MCP 怎么接？","通过 mcpo（官方的 MCP-to-OpenAPI 代理）把任意 MCP 服务器暴露成 OpenAPI 工具，再在 Open WebUI 注册即可。无需写 glue code。","https:\u002F\u002Fgithub.com\u002Fopen-webui\u002Fopen-webui",[491,490],{},[1019,499,498,497,2434],"kubernetes",[2436,2438],{"plan":91,"price":503,"features":2437,"notes":1023},"全功能 \u002F 多用户 \u002F RAG \u002F Tools \u002F 联网搜索 \u002F MCP-OpenAPI 代理 \u002F Docker \u002F K8s",{"plan":97,"price":2439,"features":2440,"notes":2441},"咨询","Custom branding \u002F SLA \u002F LTS 长期支持版本","邮件官方","完全免费（MIT 开源） \u002F Enterprise SLA 联系",[513,1030],{"power":470,"ux":516,"price":470,"cn_support":516,"stability":470},{"title":209,"description":2416},"Open WebUI 评测 2026：自托管 ChatGPT 替代，Ollama 前端部署指南",[2448,2450,2452],{"name":2449,"url":2387,"accessed":523},"Open WebUI 官方文档",{"name":2451,"url":2394,"accessed":523},"Local AI Master — Open WebUI Setup Guide 2026",{"name":2453,"url":2401,"accessed":523},"AIToolDiscovery — Open-WebUI with Ollama 2026","tools\u002Fcoding\u002Flocal\u002Fopen-webui","自托管的 ChatGPT 替代：Ollama \u002F OpenAI 兼容、多用户、RAG、126k+ GitHub stars",[469,2457,1019,534,2458,2459,535],"self-host","multi-user","ollama","自托管多用户 AI 前端的事实标准。团队 \u002F 家庭 \u002F 公司部署一份共享，多模型聚合 + RAG + 工具调用全有。单机 \u002F 桌面体验首选 Cherry Studio \u002F LobeChat。","https:\u002F\u002Fdocs.openwebui.com","5Y-zy_XMSVV_gsS0aa6V7zk268r2TvaIcssIp-HXM_U",1785660641072]