[{"data":1,"prerenderedAt":2833},["ShallowReactive",2],{"header-counts":3,"footer-counts":6,"tool-tools\u002Fcoding\u002Flocal\u002Fopen-webui":9,"tool-stats-coding\u002Flocal\u002Fopen-webui":576,"cat-rank-coding-local":579,"tool-related-coding\u002Flocal\u002Fopen-webui":2832},{"tools":4,"reviews":5},65,7,{"tools":4,"reviews":5,"playbooks":7,"news":8},10,8,{"id":10,"title":11,"alternatives":12,"api_compatible":25,"body":26,"category":503,"chinese_friendly":504,"cover":505,"description":506,"domestic":507,"extension":508,"faq":509,"free":507,"github":25,"languages":522,"meta":525,"models":25,"navigation":526,"notSuitable":25,"opensource":526,"path":527,"pillar":528,"platforms":529,"priceTable":535,"pricing":546,"published":547,"relatedPlaybooks":548,"relatedReviews":25,"score":554,"self_host":526,"seo":556,"slug":557,"sources":558,"stem":566,"suitable":25,"tagline":567,"tags":568,"updated":561,"verdict":573,"website":574,"__hash__":575},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Fopen-webui.md","Open WebUI",[13,16,19,22],{"name":14,"url":15},"lobe-chat","\u002Ftools\u002Fcoding\u002Flocal\u002Flobe-chat",{"name":17,"url":18},"cherry-studio","\u002Ftools\u002Fcoding\u002Flocal\u002Fcherry-studio",{"name":20,"url":21},"ollama","\u002Ftools\u002Fcoding\u002Flocal\u002Follama",{"name":23,"url":24},"lm-studio","\u002Ftools\u002Fcoding\u002Flocal\u002Flm-studio",null,{"type":27,"value":28,"toc":488},"minimark",[29,34,38,41,44,109,112,115,119,124,153,158,183,186,221,224,361,364,407,411,434,437,461,464],[30,31,33],"h2",{"id":32},"tldr","TL;DR",[35,36,37],"p",{},"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。",[35,39,40],{},"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。",[30,42,43],{"id":43},"核心能力",[45,46,47,55,61,67,73,79,85,91,97,103],"ul",{},[48,49,50,54],"li",{},[51,52,53],"strong",{},"多模型后端","：Ollama \u002F OpenAI \u002F vLLM \u002F Anthropic \u002F Groq \u002F LocalAI \u002F 任意 OpenAI 兼容",[48,56,57,60],{},[51,58,59],{},"多用户 + RBAC","：注册 \u002F 邀请 \u002F 角色权限 \u002F 工作区隔离",[48,62,63,66],{},[51,64,65],{},"RAG 知识库","：上传文档 \u002F 网址 \u002F SearXNG 联网搜索 → 向量化 → 对话引用",[48,68,69,72],{},[51,70,71],{},"Tools \u002F Functions","：Python 写函数即扩展（联网 \u002F 计算器 \u002F 自定义 API）",[48,74,75,78],{},[51,76,77],{},"mcpo","：MCP-to-OpenAPI 代理，任意 MCP 服务器接进来",[48,80,81,84],{},[51,82,83],{},"oikb","：知识库同步本地文件夹 \u002F GitHub \u002F S3 \u002F Confluence 等 40+ 源",[48,86,87,90],{},[51,88,89],{},"open-terminal \u002F cptr","：给 AI 真实终端 + 文件 + 沙箱执行",[48,92,93,96],{},[51,94,95],{},"图像生成","：Stable Diffusion \u002F DALL·E \u002F 自托管接入",[48,98,99,102],{},[51,100,101],{},"语音输入 \u002F TTS","：内置",[48,104,105,108],{},[51,106,107],{},"企业 LTS","：custom branding + SLA + 长期支持版本（联系销售）",[30,110,111],{"id":111},"价格",[35,113,114],{},"完全免费、MIT 开源、商用免费。Enterprise 提供品牌定制 + SLA + LTS。",[30,116,118],{"id":117},"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）",[35,120,121],{},[51,122,123],{},"亮点：",[45,125,126,134,137,144,147,150],{},[48,127,128,129,133],{},"单条 ",[130,131,132],"code",{},"docker run"," 五分钟上线",[48,135,136],{},"自带的多用户 + 角色权限省去重新搭 Auth",[48,138,139,140,143],{},"RAG 直传 30 个 PDF 后向量化顺利，对话中 ",[130,141,142],{},"#知识库"," 引用准确",[48,145,146],{},"mcpo 把 GitHub MCP 服务器接进来，团队对话里直接 issue \u002F PR 操作",[48,148,149],{},"模型切换流畅，OpenAI + Ollama 并存",[48,151,152],{},"SearXNG 联网搜索给模型实时信息，过时知识截止问题缓解",[35,154,155],{},[51,156,157],{},"踩坑：",[45,159,160,163,170,177,180],{},[48,161,162],{},"Docker 镜像 ~1.5GB，首次拉取偏慢",[48,164,165,166,169],{},"默认 ",[130,167,168],{},"0.0.0.0"," 公网暴露要加 HTTPS + 反代",[48,171,172,173,176],{},"嵌入模型 ",[130,174,175],{},"sentence-transformers"," 中文效果一般，建议换 bge-m3",[48,178,179],{},"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）",[48,181,182],{},"版本升级要看 changelog，部分 minor 含 breaking 改动",[30,184,185],{"id":185},"上手",[187,188,189,195,202,205,208,211,214],"ol",{},[48,190,191,192],{},"装 Docker → ",[130,193,194],{},"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",[48,196,197,198,201],{},"浏览器开 ",[130,199,200],{},"http:\u002F\u002Flocalhost:3000"," → 注册第一个账号（管理员）",[48,203,204],{},"设置 → Connections → 连接 Ollama \u002F 加 OpenAI Key",[48,206,207],{},"Models → Pull \u002F Discover 模型",[48,209,210],{},"Workspaces → 建知识库 → 上传文档",[48,212,213],{},"Tools → 启用 \u002F 写自定义函数",[48,215,216,217,220],{},"生产部署：Nginx 反代 + Let's Encrypt + 备份 ",[130,218,219],{},"\u002Fapp\u002Fbackend\u002Fdata"," volume",[30,222,223],{"id":223},"对比",[225,226,227,248],"table",{},[228,229,230],"thead",{},[231,232,233,237,239,242,245],"tr",{},[234,235,236],"th",{},"维度",[234,238,11],{},[234,240,241],{},"LobeChat",[234,243,244],{},"Cherry Studio",[234,246,247],{},"LM Studio",[249,250,251,268,284,299,313,328,345],"tbody",{},[231,252,253,257,260,263,266],{},[254,255,256],"td",{},"形态",[254,258,259],{},"Docker \u002F 桌面",[254,261,262],{},"Web + 桌面",[254,264,265],{},"桌面",[254,267,265],{},[231,269,270,273,276,279,282],{},[254,271,272],{},"多用户",[254,274,275],{},"✅ 一等",[254,277,278],{},"✅",[254,280,281],{},"无",[254,283,281],{},[231,285,286,289,292,294,296],{},[254,287,288],{},"RAG",[254,290,291],{},"✅ 强 + oikb",[254,293,278],{},[254,295,278],{},[254,297,298],{},"弱",[231,300,301,304,307,309,311],{},[254,302,303],{},"工具 \u002F MCP",[254,305,306],{},"✅ mcpo",[254,308,278],{},[254,310,278],{},[254,312,298],{},[231,314,315,318,321,324,326],{},[254,316,317],{},"自托管",[254,319,320],{},"✅ Docker \u002F K8s",[254,322,323],{},"✅ Docker",[254,325,281],{},[254,327,281],{},[231,329,330,333,336,339,342],{},[254,331,332],{},"GitHub Stars",[254,334,335],{},"126k+",[254,337,338],{},"72k+",[254,340,341],{},"60k+",[254,343,344],{},"–",[231,346,347,350,353,355,358],{},[254,348,349],{},"开源协议",[254,351,352],{},"MIT",[254,354,352],{},[254,356,357],{},"AGPL-3.0",[254,359,360],{},"闭源（免费）",[30,362,363],{"id":363},"避坑",[45,365,366,372,380,389,395,401],{},[48,367,368,371],{},[51,369,370],{},"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",[48,373,374,379],{},[51,375,376,377,220],{},"备份 ",[130,378,219],{},"：知识库 \u002F 用户 \u002F 对话全在里面",[48,381,382,385,386,388],{},[51,383,384],{},"中文 RAG 换嵌入模型","：默认 ",[130,387,175],{}," 中文一般，配 bge-m3 或硅基流动嵌入 API",[48,390,391,394],{},[51,392,393],{},"mcpo 工具范围谨慎","：MCP 给 AI 真实能力，第三方服务器审一遍",[48,396,397,400],{},[51,398,399],{},"后端吞吐看 Ollama","：5+ 并发上 vLLM 后端，Ollama 单 worker 会排队",[48,402,403,406],{},[51,404,405],{},"升级前看 changelog","：weekly 更新，偶有 breaking",[30,408,410],{"id":409},"适合-不适合","适合 \u002F 不适合",[45,412,413,416,419,422,425,428,431],{},[48,414,415],{},"✅ 团队 \u002F 家庭 \u002F 公司多人共享 AI 平台",[48,417,418],{},"✅ 要 Web 端访问 \u002F 移动端兼容",[48,420,421],{},"✅ 自托管 \u002F 完全控制数据",[48,423,424],{},"✅ MCP \u002F 工具调用刚需",[48,426,427],{},"❌ 单人桌面体验（用 Cherry Studio）",[48,429,430],{},"❌ 零运维 \u002F 不愿碰 Docker",[48,432,433],{},"❌ iOS 原生 App 主力",[30,435,436],{"id":436},"相关阅读",[45,438,439,445,450,455],{},[48,440,441],{},[442,443,444],"a",{"href":15},"LobeChat 评测",[48,446,447],{},[442,448,449],{"href":18},"Cherry Studio 评测",[48,451,452],{},[442,453,454],{"href":21},"Ollama 评测",[48,456,457],{},[442,458,460],{"href":459},"\u002Fplaybook\u002Fonboarding\u002Frag-pipeline-build","RAG Pipeline 搭建 Playbook",[30,462,463],{"id":463},"来源",[187,465,466,474,481],{},[48,467,468,469],{},"Open WebUI 官方文档 ",[442,470,471],{"href":471,"rel":472},"https:\u002F\u002Fdocs.openwebui.com\u002F",[473],"nofollow",[48,475,476,477],{},"Local AI Master — Open WebUI Setup Guide 2026 ",[442,478,479],{"href":479,"rel":480},"https:\u002F\u002Flocalaimaster.com\u002Fblog\u002Fopen-webui-setup-guide",[473],[48,482,483,484],{},"AIToolDiscovery — Set Up Open-WebUI with Ollama 2026 ",[442,485,486],{"href":486,"rel":487},"https:\u002F\u002Fwww.aitooldiscovery.com\u002Fhow-to\u002Fsetup-open-webui-ollama",[473],{"title":489,"searchDepth":490,"depth":490,"links":491},"",3,[492,494,495,496,497,498,499,500,501,502],{"id":32,"depth":493,"text":33},2,{"id":43,"depth":493,"text":43},{"id":111,"depth":493,"text":111},{"id":117,"depth":493,"text":118},{"id":185,"depth":493,"text":185},{"id":223,"depth":493,"text":223},{"id":363,"depth":493,"text":363},{"id":409,"depth":493,"text":410},{"id":436,"depth":493,"text":436},{"id":463,"depth":493,"text":463},"local",4,"\u002Fimg\u002Ftools\u002Fopen-webui.webp","Open WebUI 真实评测：MIT 开源、自托管 AI 平台，离线优先。Docker 一行起飞、支持 Ollama \u002F OpenAI \u002F vLLM \u002F Anthropic \u002F Groq 等后端，内置 RAG 知识库 + 多用户 + 联网搜索 + 工具调用。GitHub 126k+ stars，事实标准本地 AI 前端。",false,"md",[510,513,516,519],{"q":511,"a":512},"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":514,"a":515},"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":517,"a":518},"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":520,"a":521},"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。",[523,524],"en","zh",{},true,"\u002Ftools\u002Fcoding\u002Flocal\u002Fopen-webui","coding",[530,531,532,533,534],"docker","linux","macos","windows","kubernetes",[536,541],{"plan":537,"price":538,"features":539,"notes":540},"开源版","免费","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","MIT 协议",{"plan":542,"price":543,"features":544,"notes":545},"Enterprise","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","2026-06-19",[549,551],{"name":550,"url":459},"RAG Pipeline 搭建",{"name":552,"url":553},"Claude Code 上手","\u002Fplaybook\u002Fonboarding\u002Fclaude-code-getting-started",{"power":555,"ux":504,"price":555,"cn_support":504,"stability":555},5,{"title":11,"description":506},"coding\u002Flocal\u002Fopen-webui",[559,562,564],{"name":560,"url":471,"accessed":561},"Open WebUI 官方文档","2026-06-24",{"name":563,"url":479,"accessed":561},"Local AI Master — Open WebUI Setup Guide 2026",{"name":565,"url":486,"accessed":561},"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",[503,569,530,570,571,20,572],"self-host","rag","multi-user","open-source","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","8iXO-BuvdCKJCMIXxc6Jlp_MhDAMVBwYJJnhXo6dErw",{"ok":526,"slug":577,"viewCount":578,"clickCount":578,"avgRating":578,"ratingCount":578},"coding%2Flocal%2Fopen-webui",0,[580,1032,1518,2006,2476],{"id":581,"title":244,"alternatives":582,"api_compatible":25,"body":588,"category":503,"chinese_friendly":555,"cover":979,"description":980,"domestic":507,"extension":508,"faq":981,"free":507,"github":25,"languages":994,"meta":995,"models":25,"navigation":526,"notSuitable":25,"opensource":526,"path":18,"pillar":528,"platforms":996,"priceTable":998,"pricing":1006,"published":547,"relatedPlaybooks":1007,"relatedReviews":25,"score":1012,"self_host":526,"seo":1013,"slug":1014,"sources":1015,"stem":1022,"suitable":25,"tagline":1023,"tags":1024,"updated":561,"verdict":1029,"website":1030,"__hash__":1031},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Fcherry-studio.md",[583,584,585,586],{"name":14,"url":15},{"name":23,"url":24},{"name":20,"url":21},{"name":587,"url":527},"open-webui",{"type":27,"value":589,"toc":967},[590,592,595,598,600,644,646,658,664,668,672,689,693,710,712,736,738,862,864,896,898,921,923,942,944],[30,591,33],{"id":32},[35,593,594],{},"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。",[35,596,597],{},"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。",[30,599,43],{"id":43},[45,601,602,608,614,620,626,632,638],{},[48,603,604,607],{},[51,605,606],{},"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",[48,609,610,613],{},[51,611,612],{},"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",[48,615,616,619],{},[51,617,618],{},"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",[48,621,622,625],{},[51,623,624],{},"MCP 协议","：扩展工具调用 \u002F 联网搜索 \u002F 文件操作",[48,627,628,631],{},[51,629,630],{},"数据本地优先","：对话历史本地存储，WebDAV 同步，不上传第三方",[48,633,634,637],{},[51,635,636],{},"多模态","：图片识别 \u002F PDF 阅读 \u002F Markdown + Mermaid + 代码高亮",[48,639,640,643],{},[51,641,642],{},"AI 绘画 + 翻译","：内置主流 SD \u002F DALL·E \u002F 翻译 API 集成",[30,645,111],{"id":111},[45,647,648,653],{},[48,649,650,652],{},[51,651,537],{},"：完全免费，AGPL-3.0",[48,654,655,657],{},[51,656,542],{},"：私有化部署 + 团队协作 + 资源管控，联系销售",[659,660,661],"blockquote",{},[35,662,663],{},"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。",[30,665,667],{"id":666},"实测mac-m2-中型知识库","实测（Mac M2 + 中型知识库）",[35,669,670],{},[51,671,123],{},[45,673,674,677,680,683,686],{},[48,675,676],{},"中文 UI \u002F 文档 \u002F 社区都顶级，零门槛上手",[48,678,679],{},"本地 RAG 拖入 30+ PDF 后向量化 \u003C 2 分钟（用 bge-m3）",[48,681,682],{},"多模型并排回答：让 Claude \u002F GPT \u002F DeepSeek 同回一个问题做比较",[48,684,685],{},"MCP 接 Brave Search + 自定义工具流畅",[48,687,688],{},"WebDAV 同步坚果云 \u002F 阿里云盘，桌面 + 移动设备数据互通",[35,690,691],{},[51,692,157],{},[45,694,695,698,701,704,707],{},[48,696,697],{},"没有 Web 端 \u002F Docker 自托管（要这个用 LobeChat）",[48,699,700],{},"iOS 版尚未发布（roadmap 中）",[48,702,703],{},"大型 PDF（>100 MB）向量化偶有失败，要切小",[48,705,706],{},"助手市场质量参差，要自筛",[48,708,709],{},"模型 API 调用全靠你自己付费，新手要先理解 API Key 概念",[30,711,185],{"id":185},[187,713,714,717,720,727,730,733],{},[48,715,716],{},"cherry-ai.com 下载客户端（或 GitHub releases）",[48,718,719],{},"设置 → 模型服务 → 填 OpenAI \u002F Claude \u002F DeepSeek API Key",[48,721,722,723],{},"（可选）本地：装 Ollama → Cherry Studio 自动识别 endpoint ",[442,724,725],{"href":725,"rel":726},"http:\u002F\u002Flocalhost:11434",[473],[48,728,729],{},"新建知识库 → 拖文件 \u002F 加网址 → 等向量化",[48,731,732],{},"新对话 → 选模型 → 勾知识库 → 提问",[48,734,735],{},"进阶：自定义助手（System Prompt）+ MCP 扩展工具",[30,737,223],{"id":223},[225,739,740,754],{},[228,741,742],{},[231,743,744,746,748,750,752],{},[234,745,236],{},[234,747,244],{},[234,749,241],{},[234,751,247],{},[234,753,11],{},[249,755,756,768,782,796,809,823,838,850],{},[231,757,758,760,762,764,766],{},[254,759,256],{},[254,761,265],{},[254,763,262],{},[254,765,265],{},[254,767,259],{},[231,769,770,772,775,777,780],{},[254,771,606],{},[254,773,774],{},"✅ 云 + 本地",[254,776,774],{},[254,778,779],{},"本地为主",[254,781,774],{},[231,783,784,787,790,792,794],{},[254,785,786],{},"知识库 RAG",[254,788,789],{},"✅ 强",[254,791,789],{},[254,793,298],{},[254,795,278],{},[231,797,798,801,803,805,807],{},[254,799,800],{},"MCP",[254,802,278],{},[254,804,278],{},[254,806,298],{},[254,808,278],{},[231,810,811,814,817,819,821],{},[254,812,813],{},"自托管 \u002F Web",[254,815,816],{},"无 Web",[254,818,323],{},[254,820,281],{},[254,822,323],{},[231,824,825,828,831,833,836],{},[254,826,827],{},"中文",[254,829,830],{},"5\u002F5",[254,832,830],{},[254,834,835],{},"4\u002F5",[254,837,835],{},[231,839,840,842,844,846,848],{},[254,841,349],{},[254,843,357],{},[254,845,352],{},[254,847,360],{},[254,849,352],{},[231,851,852,854,856,858,860],{},[254,853,332],{},[254,855,341],{},[254,857,338],{},[254,859,344],{},[254,861,335],{},[30,863,363],{"id":363},[45,865,866,872,878,884,890],{},[48,867,868,871],{},[51,869,870],{},"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",[48,873,874,877],{},[51,875,876],{},"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",[48,879,880,883],{},[51,881,882],{},"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",[48,885,886,889],{},[51,887,888],{},"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",[48,891,892,895],{},[51,893,894],{},"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",[30,897,410],{"id":409},[45,899,900,903,906,909,912,915,918],{},[48,901,902],{},"✅ 中文用户、AI 重度使用 \u002F 多模型管理",[48,904,905],{},"✅ 需要本地 RAG 知识库",[48,907,908],{},"✅ 关注数据隐私 \u002F 本地存储",[48,910,911],{},"✅ 想用 Ollama \u002F LM Studio 本地模型",[48,913,914],{},"❌ 需要 Web 端 \u002F Docker 自托管",[48,916,917],{},"❌ 团队多人共享 \u002F SSO",[48,919,920],{},"❌ iOS 主力用户",[30,922,436],{"id":436},[45,924,925,929,934,938],{},[48,926,927],{},[442,928,444],{"href":15},[48,930,931],{},[442,932,933],{"href":24},"LM Studio 评测",[48,935,936],{},[442,937,454],{"href":21},[48,939,940],{},[442,941,460],{"href":459},[30,943,463],{"id":463},[187,945,946,953,960],{},[48,947,948,949],{},"Cherry Studio 官网（功能 + 下载）",[442,950,951],{"href":951,"rel":952},"https:\u002F\u002Fwww.cherry-ai.com\u002F",[473],[48,954,955,956],{},"MBLUO Studio — Cherry Studio 评测 2026 ",[442,957,958],{"href":958,"rel":959},"https:\u002F\u002Fmbluostudio.com\u002Ftools\u002Fcherry-studio",[473],[48,961,962,963],{},"Cursor IDE 博客 — Cherry Studio 完全指南（2025-03）",[442,964,965],{"href":965,"rel":966},"https:\u002F\u002Fwww.cursor-ide.com\u002Fblog\u002Fcherry-studio-guide",[473],{"title":489,"searchDepth":490,"depth":490,"links":968},[969,970,971,972,973,974,975,976,977,978],{"id":32,"depth":493,"text":33},{"id":43,"depth":493,"text":43},{"id":111,"depth":493,"text":111},{"id":666,"depth":493,"text":667},{"id":185,"depth":493,"text":185},{"id":223,"depth":493,"text":223},{"id":363,"depth":493,"text":363},{"id":409,"depth":493,"text":410},{"id":436,"depth":493,"text":436},{"id":463,"depth":493,"text":463},"\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。",[982,985,988,991],{"q":983,"a":984},"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":986,"a":987},"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":989,"a":990},"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":992,"a":993},"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。",[524,523],{},[533,532,531,997],"android",[999,1002],{"plan":537,"price":538,"features":1000,"notes":1001},"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":542,"price":1003,"features":1004,"notes":1005},"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",[1008,1009],{"name":550,"url":459},{"name":1010,"url":1011},"Cursor MCP 深度集成","\u002Fplaybook\u002Fonboarding\u002Fcursor-mcp-deep-integration",{"power":504,"ux":555,"price":555,"cn_support":555,"stability":504},{"title":244,"description":980},"coding\u002Flocal\u002Fcherry-studio",[1016,1018,1020],{"name":1017,"url":951,"accessed":561},"Cherry Studio 官网",{"name":1019,"url":958,"accessed":561},"MBLUO Studio — Cherry Studio 评测",{"name":1021,"url":965,"accessed":561},"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",[503,1025,1026,1027,570,572,1028],"desktop","multi-model","knowledge-base","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","iH3iDqpTojLLBznROeYXtARpOLDs1yflKA9EMftZkTk",{"id":1033,"title":247,"alternatives":1034,"api_compatible":25,"body":1039,"category":503,"chinese_friendly":490,"cover":1466,"description":1467,"domestic":507,"extension":508,"faq":1468,"free":507,"github":25,"languages":1481,"meta":1482,"models":25,"navigation":526,"notSuitable":25,"opensource":507,"path":24,"pillar":528,"platforms":1483,"priceTable":1484,"pricing":1492,"published":547,"relatedPlaybooks":1493,"relatedReviews":25,"score":1496,"self_host":526,"seo":1497,"slug":1498,"sources":1499,"stem":1506,"suitable":25,"tagline":1507,"tags":1508,"updated":561,"verdict":1515,"website":1516,"__hash__":1517},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Flm-studio.md",[1035,1036,1037,1038],{"name":20,"url":21},{"name":587,"url":527},{"name":17,"url":18},{"name":14,"url":15},{"type":27,"value":1040,"toc":1454},[1041,1043,1050,1053,1055,1112,1114,1128,1133,1137,1141,1158,1162,1179,1181,1207,1209,1348,1350,1382,1384,1407,1409,1429,1431],[30,1042,33],{"id":32},[35,1044,1045,1046,1049],{},"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 + ",[130,1047,1048],{},"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。",[35,1051,1052],{},"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）。",[30,1054,43],{"id":43},[45,1056,1057,1063,1069,1075,1085,1094,1100,1106],{},[48,1058,1059,1062],{},[51,1060,1061],{},"模型浏览器","：内置 Hugging Face 检索，按 GGUF \u002F MLX \u002F 大小筛选、一键下载",[48,1064,1065,1068],{},[51,1066,1067],{},"聊天界面","：System Prompt \u002F temperature \u002F top-p \u002F context size 可视化调参",[48,1070,1071,1074],{},[51,1072,1073],{},"多模型并存 \u002F 切换","：同时加载多模型在不同会话中比较",[48,1076,1077,1080,1081,1084],{},[51,1078,1079],{},"OpenAI 兼容 Local Server","：",[130,1082,1083],{},"http:\u002F\u002Flocalhost:1234\u002Fv1","，任何 SDK 即接即用",[48,1086,1087,1080,1090,1093],{},[51,1088,1089],{},"Headless \u002F CLI",[130,1091,1092],{},"lms server start --port 1234","，无 GUI 可跑",[48,1095,1096,1099],{},[51,1097,1098],{},"PDF \u002F 文档对话","：内置基础 RAG，丢文件就能聊",[48,1101,1102,1105],{},[51,1103,1104],{},"MLX 原生支持（Mac）","：M1+ 上比 GGUF + Metal 快 30–50%",[48,1107,1108,1111],{},[51,1109,1110],{},"持续批处理","：Codersera 2026 测得 50–90 tok\u002Fs（消费级 GPU + 中等模型）",[30,1113,111],{"id":111},[45,1115,1116,1122],{},[48,1117,1118,1121],{},[51,1119,1120],{},"个人 \u002F 评估","：免费，全功能可用",[48,1123,1124,1127],{},[51,1125,1126],{},"商用","：邮件 \u002F 官网联系 LM Studio 团队",[659,1129,1130],{},[35,1131,1132],{},"模型本身免费（开源权重），LM Studio 不抽水任何 token 费用。",[30,1134,1136],{"id":1135},"实测mac-m2-pro-qwen3-coder-7b-gguf-q4_k_m","实测（Mac M2 Pro + Qwen3-Coder-7B GGUF Q4_K_M）",[35,1138,1139],{},[51,1140,123],{},[45,1142,1143,1146,1149,1152,1155],{},[48,1144,1145],{},"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",[48,1147,1148],{},"加载 7B Q4 模型 \u003C 3 秒，生成 ~75 tok\u002Fs",[48,1150,1151],{},"Local Server 开了 Cursor 直接接 baseURL → 本地代码补全零成本",[48,1153,1154],{},"MLX 版同模型 ~110 tok\u002Fs，差距显著",[48,1156,1157],{},"多窗口加载 2 个模型并排测，调 prompt 直观",[35,1159,1160],{},[51,1161,157],{},[45,1163,1164,1167,1170,1173,1176],{},[48,1165,1166],{},"模型库依赖 Hugging Face，国内访问要镜像 \u002F 代理",[48,1168,1169],{},"GPU 显存吃满后会自动 offload 到 CPU，无提示就慢下来",[48,1171,1172],{},"Headless 模式相对 Ollama 偏新，文档稍少",[48,1174,1175],{},"闭源应用（虽免费），不适合企业合规挂钩",[48,1177,1178],{},"中文 UI 可用但部分菜单仍英文",[30,1180,185],{"id":185},[187,1182,1183,1186,1189,1192,1195,1202],{},[48,1184,1185],{},"lmstudio.ai 下载（Mac \u002F Windows \u002F Linux）",[48,1187,1188],{},"打开 → Discover 标签 → 搜模型（如 qwen3-coder、deepseek-v3 GGUF\u002FMLX）→ Download",[48,1190,1191],{},"Chat 标签 → 选模型 → 调参聊天",[48,1193,1194],{},"Local Server 标签 → Start Server → 默认端口 1234",[48,1196,1197,1198,1201],{},"在你的应用里：",[130,1199,1200],{},"baseURL = \"http:\u002F\u002Flocalhost:1234\u002Fv1\"","，API Key 任意",[48,1203,1204,1205],{},"Headless：",[130,1206,1092],{},[30,1208,223],{"id":223},[225,1210,1211,1227],{},[228,1212,1213],{},[231,1214,1215,1217,1219,1222,1224],{},[234,1216,236],{},[234,1218,247],{},[234,1220,1221],{},"Ollama",[234,1223,11],{},[234,1225,1226],{},"llama.cpp",[249,1228,1229,1245,1261,1276,1291,1305,1319,1332],{},[231,1230,1231,1233,1236,1239,1242],{},[254,1232,256],{},[254,1234,1235],{},"GUI + CLI",[254,1237,1238],{},"CLI Daemon",[254,1240,1241],{},"Docker UI",[254,1243,1244],{},"二进制",[231,1246,1247,1250,1253,1256,1258],{},[254,1248,1249],{},"模型浏览",[254,1251,1252],{},"✅ 内置",[254,1254,1255],{},"CLI pull",[254,1257,281],{},[254,1259,1260],{},"手动",[231,1262,1263,1266,1268,1271,1274],{},[254,1264,1265],{},"参数调优 GUI",[254,1267,278],{},[254,1269,1270],{},"❌",[254,1272,1273],{},"部分",[254,1275,1270],{},[231,1277,1278,1281,1284,1287,1289],{},[254,1279,1280],{},"OpenAI 兼容 API",[254,1282,1283],{},"✅ :1234",[254,1285,1286],{},"✅ :11434",[254,1288,278],{},[254,1290,278],{},[231,1292,1293,1296,1298,1301,1303],{},[254,1294,1295],{},"MLX (Mac)",[254,1297,278],{},[254,1299,1300],{},"✅ 0.19+",[254,1302,344],{},[254,1304,344],{},[231,1306,1307,1310,1312,1314,1316],{},[254,1308,1309],{},"多用户并发",[254,1311,298],{},[254,1313,298],{},[254,1315,278],{},[254,1317,1318],{},"中",[231,1320,1321,1324,1326,1328,1330],{},[254,1322,1323],{},"开源",[254,1325,360],{},[254,1327,352],{},[254,1329,352],{},[254,1331,352],{},[231,1333,1334,1337,1340,1343,1345],{},[254,1335,1336],{},"上手难度",[254,1338,1339],{},"极低",[254,1341,1342],{},"低",[254,1344,1318],{},[254,1346,1347],{},"高",[30,1349,363],{"id":363},[45,1351,1352,1358,1364,1370,1376],{},[48,1353,1354,1357],{},[51,1355,1356],{},"国内下模型走镜像","：HF 直连慢 \u002F 卡，配 HF_ENDPOINT=hf-mirror.com",[48,1359,1360,1363],{},[51,1361,1362],{},"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",[48,1365,1366,1369],{},[51,1367,1368],{},"MLX 优先（Mac M 系列）","：能下 MLX 版就别下 GGUF，速度差距明显",[48,1371,1372,1375],{},[51,1373,1374],{},"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",[48,1377,1378,1381],{},[51,1379,1380],{},"闭源合规要核","：企业内部使用前查 license；商用必须联系官方",[30,1383,410],{"id":409},[45,1385,1386,1389,1392,1395,1398,1401,1404],{},[48,1387,1388],{},"✅ 本地 LLM 入门 \u002F 评估",[48,1390,1391],{},"✅ Mac M 系列用户",[48,1393,1394],{},"✅ 想给 Cursor \u002F Cline 接本地 OpenAI 兼容 endpoint",[48,1396,1397],{},"✅ 需要 GUI 调参 \u002F 模型比较",[48,1399,1400],{},"❌ 多用户并发生产服务",[48,1402,1403],{},"❌ 嵌入式 \u002F 边缘设备",[48,1405,1406],{},"❌ 强合规 \u002F 必须开源审计",[30,1408,436],{"id":436},[45,1410,1411,1415,1420,1424],{},[48,1412,1413],{},[442,1414,454],{"href":21},[48,1416,1417],{},[442,1418,1419],{"href":527},"Open WebUI 评测",[48,1421,1422],{},[442,1423,449],{"href":18},[48,1425,1426],{},[442,1427,1428],{"href":553},"Claude Code 上手 Playbook",[30,1430,463],{"id":463},[187,1432,1433,1440,1447],{},[48,1434,1435,1436],{},"LM Studio 官网 ",[442,1437,1438],{"href":1438,"rel":1439},"https:\u002F\u002Flmstudio.ai\u002F",[473],[48,1441,1442,1443],{},"Codersera — LM Studio Complete Guide 2026 ",[442,1444,1445],{"href":1445,"rel":1446},"https:\u002F\u002Fcodersera.com\u002Fblog\u002Flm-studio-complete-guide-2026\u002F",[473],[48,1448,1449,1450],{},"Codersera — Ollama vs LM Studio vs vLLM vs llama.cpp vs MLX 2026 ",[442,1451,1452],{"href":1452,"rel":1453},"https:\u002F\u002Fcodersera.com\u002Fblog\u002Follama-vs-lm-studio-vs-vllm-vs-llama-cpp-vs-mlx-2026\u002F",[473],{"title":489,"searchDepth":490,"depth":490,"links":1455},[1456,1457,1458,1459,1460,1461,1462,1463,1464,1465],{"id":32,"depth":493,"text":33},{"id":43,"depth":493,"text":43},{"id":111,"depth":493,"text":111},{"id":1135,"depth":493,"text":1136},{"id":185,"depth":493,"text":185},{"id":223,"depth":493,"text":223},{"id":363,"depth":493,"text":363},{"id":409,"depth":493,"text":410},{"id":436,"depth":493,"text":436},{"id":463,"depth":493,"text":463},"\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。",[1469,1472,1475,1478],{"q":1470,"a":1471},"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":1473,"a":1474},"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":1476,"a":1477},"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":1479,"a":1480},"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。",[523,524],{},[533,532,531],[1485,1488],{"plan":1120,"price":538,"features":1486,"notes":1487},"全功能 GUI + Headless API + GGUF\u002FMLX","供个人 \u002F 评估使用",{"plan":1126,"price":1489,"features":1490,"notes":1491},"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",[1494,1495],{"name":550,"url":459},{"name":552,"url":553},{"power":504,"ux":555,"price":555,"cn_support":490,"stability":504},{"title":247,"description":1467},"coding\u002Flocal\u002Flm-studio",[1500,1502,1504],{"name":1501,"url":1438,"accessed":561},"LM Studio 官网",{"name":1503,"url":1445,"accessed":561},"Codersera — LM Studio Complete Guide 2026",{"name":1505,"url":1452,"accessed":561},"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",[503,1509,1510,1511,1512,1513,1514],"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","mFDr4hC-XSmdJmmz0qk02JMEKIemqp0HnyhGIxnoWo0",{"id":1519,"title":241,"alternatives":1520,"api_compatible":25,"body":1525,"category":503,"chinese_friendly":555,"cover":1958,"description":1959,"domestic":507,"extension":508,"faq":1960,"free":507,"github":25,"languages":1973,"meta":1974,"models":25,"navigation":526,"notSuitable":25,"opensource":526,"path":15,"pillar":528,"platforms":1975,"priceTable":1977,"pricing":1984,"published":547,"relatedPlaybooks":1985,"relatedReviews":25,"score":1988,"self_host":526,"seo":1989,"slug":1990,"sources":1991,"stem":1998,"suitable":25,"tagline":1999,"tags":2000,"updated":561,"verdict":2003,"website":2004,"__hash__":2005},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Flobe-chat.md",[1521,1522,1523,1524],{"name":17,"url":18},{"name":587,"url":527},{"name":20,"url":21},{"name":23,"url":24},{"type":27,"value":1526,"toc":1946},[1527,1529,1536,1539,1541,1603,1605,1618,1621,1625,1629,1649,1653,1670,1672,1698,1700,1833,1835,1873,1875,1901,1903,1921,1923],[30,1528,33],{"id":32},[35,1530,1531,1532,1535],{},"LobeChat 是 LobeHub 团队的开源 AI 聊天框架，2023 年发布、GitHub 72k+ stars、MIT 协议。",[51,1533,1534],{},"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。",[35,1537,1538],{},"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）。",[30,1540,43],{"id":43},[45,1542,1543,1548,1554,1559,1565,1571,1576,1581,1587,1593],{},[48,1544,1545,1547],{},[51,1546,606],{},"：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",[48,1549,1550,1553],{},[51,1551,1552],{},"多模型对比","：同 prompt 给多模型并排回答",[48,1555,1556,1558],{},[51,1557,612],{},"：上传 PDF \u002F Word \u002F 网页 → 向量化 → 检索引用",[48,1560,1561,1564],{},[51,1562,1563],{},"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",[48,1566,1567,1570],{},[51,1568,1569],{},"助手市场","：几百个预设 AI 角色，一键导入",[48,1572,1573,1575],{},[51,1574,624],{},"：扩展任意工具能力",[48,1577,1578],{},[51,1579,1580],{},"代码解释器 \u002F 文件上传 \u002F TTS \u002F 多模态",[48,1582,1583,1586],{},[51,1584,1585],{},"Web + 桌面 + Docker","：三形态，数据可完全本地",[48,1588,1589,1592],{},[51,1590,1591],{},"LobeHub Cloud","：官方云托管，免部署",[48,1594,1595,1598,1599,1602],{},[51,1596,1597],{},"快捷指令 \u002F 工作流","：自定义 prompt 模板，",[130,1600,1601],{},"\u002Fpodcast-summary"," 类用法",[30,1604,111],{"id":111},[45,1606,1607,1613],{},[48,1608,1609,1612],{},[51,1610,1611],{},"自托管 \u002F 桌面","：完全免费、MIT 开源",[48,1614,1615,1617],{},[51,1616,1591],{},"：订阅制，云端托管 + 团队协作 + 同步",[35,1619,1620],{},"模型 API 费用按你自己的供应商付费；本地 Ollama \u002F LM Studio 零成本。",[30,1622,1624],{"id":1623},"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）",[35,1626,1627],{},[51,1628,123],{},[45,1630,1631,1634,1637,1640,1643,1646],{},[48,1632,1633],{},"界面颜值是这一类工具里第一档（深色 \u002F 透明 \u002F 现代感）",[48,1635,1636],{},"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",[48,1638,1639],{},"知识库 RAG 上传 50+ PDF 后检索准确，引用片段可视化",[48,1641,1642],{},"助手市场拿来即用——「Code Reviewer」「Translation Polish」节省 prompt 编写",[48,1644,1645],{},"Docker 一键部署，团队 5 人共享流畅",[48,1647,1648],{},"多平台数据同步（Cloud \u002F WebDAV）",[35,1650,1651],{},[51,1652,157],{},[45,1654,1655,1658,1661,1664,1667],{},[48,1656,1657],{},"自托管要熟悉 Docker + 反代 + HTTPS",[48,1659,1660],{},"国内连 OpenAI \u002F Claude 需自带网络方案",[48,1662,1663],{},"Web 版数据存 LobeHub，隐私敏感场景走桌面 \u002F Docker",[48,1665,1666],{},"插件市场质量参差，要自筛",[48,1668,1669],{},"团队多人共享需配 LobeHub Cloud 或自建数据库（Postgres + S3）",[30,1671,185],{"id":185},[187,1673,1674,1677,1683,1686,1689,1692,1695],{},[48,1675,1676],{},"选形态：Web（chat.lobehub.com 注册即用） \u002F 桌面（GitHub Releases 下载） \u002F Docker",[48,1678,1679,1680],{},"Docker：",[130,1681,1682],{},"docker run -d -p 3210:3210 -e OPENAI_API_KEY=sk-xxx --name lobe-chat lobehub\u002Flobe-chat",[48,1684,1685],{},"设置 → AI 服务商 → 添加 OpenAI \u002F Claude \u002F DeepSeek \u002F Ollama",[48,1687,1688],{},"模型选择器测试对话",[48,1690,1691],{},"知识库：拖文件 → 等向量化 → 对话引用",[48,1693,1694],{},"助手市场拉「Code Reviewer」「论文翻译润色」试用",[48,1696,1697],{},"进阶：插件市场启用联网搜索 \u002F 代码执行；MCP 自定义工具",[30,1699,223],{"id":223},[225,1701,1702,1716],{},[228,1703,1704],{},[231,1705,1706,1708,1710,1712,1714],{},[234,1707,236],{},[234,1709,241],{},[234,1711,244],{},[234,1713,11],{},[234,1715,247],{},[249,1717,1718,1730,1743,1755,1768,1784,1796,1809,1821],{},[231,1719,1720,1722,1724,1726,1728],{},[254,1721,256],{},[254,1723,1585],{},[254,1725,265],{},[254,1727,259],{},[254,1729,265],{},[231,1731,1732,1734,1737,1739,1741],{},[254,1733,606],{},[254,1735,1736],{},"✅ 80+",[254,1738,278],{},[254,1740,278],{},[254,1742,779],{},[231,1744,1745,1747,1749,1751,1753],{},[254,1746,1552],{},[254,1748,275],{},[254,1750,278],{},[254,1752,298],{},[254,1754,298],{},[231,1756,1757,1759,1761,1763,1766],{},[254,1758,786],{},[254,1760,278],{},[254,1762,278],{},[254,1764,1765],{},"✅ + oikb",[254,1767,298],{},[231,1769,1770,1773,1776,1779,1782],{},[254,1771,1772],{},"插件 \u002F 助手市场",[254,1774,1775],{},"✅ 丰富",[254,1777,1778],{},"300+ 助手",[254,1780,1781],{},"Tools",[254,1783,298],{},[231,1785,1786,1788,1790,1792,1794],{},[254,1787,800],{},[254,1789,278],{},[254,1791,278],{},[254,1793,306],{},[254,1795,298],{},[231,1797,1798,1800,1803,1805,1807],{},[254,1799,272],{},[254,1801,1802],{},"配 Cloud \u002F 自建",[254,1804,281],{},[254,1806,275],{},[254,1808,281],{},[231,1810,1811,1813,1815,1817,1819],{},[254,1812,332],{},[254,1814,338],{},[254,1816,341],{},[254,1818,335],{},[254,1820,344],{},[231,1822,1823,1825,1827,1829,1831],{},[254,1824,349],{},[254,1826,352],{},[254,1828,357],{},[254,1830,352],{},[254,1832,360],{},[30,1834,363],{"id":363},[45,1836,1837,1843,1849,1855,1861,1867],{},[48,1838,1839,1842],{},[51,1840,1841],{},"Web 版数据不本地","：隐私敏感选桌面或 Docker 自托管",[48,1844,1845,1848],{},[51,1846,1847],{},"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",[48,1850,1851,1854],{},[51,1852,1853],{},"Docker 自托管暴露公网","：上反代 + HTTPS + Auth + 备份数据库",[48,1856,1857,1860],{},[51,1858,1859],{},"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",[48,1862,1863,1866],{},[51,1864,1865],{},"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",[48,1868,1869,1872],{},[51,1870,1871],{},"同步选 Cloud vs WebDAV","：团队多端走 LobeHub Cloud；个人多设备 WebDAV 即可",[30,1874,410],{"id":409},[45,1876,1877,1880,1883,1886,1889,1892,1895,1898],{},[48,1878,1879],{},"✅ Web + 桌面双形态需求",[48,1881,1882],{},"✅ Docker 自托管 \u002F 团队共享",[48,1884,1885],{},"✅ 多模型对比 \u002F 选型",[48,1887,1888],{},"✅ 中文重度用户",[48,1890,1891],{},"✅ 助手市场 \u002F 插件生态用户",[48,1893,1894],{},"❌ 强企业 RBAC + 多租户（Open WebUI 更完善）",[48,1896,1897],{},"❌ 只要桌面 + 数据完全本地（Cherry Studio 同样优秀）",[48,1899,1900],{},"❌ 完全不会碰 Docker",[30,1902,436],{"id":436},[45,1904,1905,1909,1913,1917],{},[48,1906,1907],{},[442,1908,449],{"href":18},[48,1910,1911],{},[442,1912,1419],{"href":527},[48,1914,1915],{},[442,1916,454],{"href":21},[48,1918,1919],{},[442,1920,460],{"href":459},[30,1922,463],{"id":463},[187,1924,1925,1932,1939],{},[48,1926,1927,1928],{},"LobeChat GitHub 仓库（72k+ stars，MIT）",[442,1929,1930],{"href":1930,"rel":1931},"https:\u002F\u002Fgithub.com\u002Flobehub\u002Flobe-chat",[473],[48,1933,1934,1935],{},"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）",[442,1936,1937],{"href":1937,"rel":1938},"https:\u002F\u002Fcloud.tencent.com\u002Fdeveloper\u002Farticle\u002F2622150",[473],[48,1940,1941,1942],{},"Ofox.ai — LobeChat 完全配置指南 2026（2026-04-17）",[442,1943,1944],{"href":1944,"rel":1945},"https:\u002F\u002Fofox.ai\u002Fzh\u002Fblog\u002Flobechat-api-configuration-guide-2026",[473],{"title":489,"searchDepth":490,"depth":490,"links":1947},[1948,1949,1950,1951,1952,1953,1954,1955,1956,1957],{"id":32,"depth":493,"text":33},{"id":43,"depth":493,"text":43},{"id":111,"depth":493,"text":111},{"id":1623,"depth":493,"text":1624},{"id":185,"depth":493,"text":185},{"id":223,"depth":493,"text":223},{"id":363,"depth":493,"text":363},{"id":409,"depth":493,"text":410},{"id":436,"depth":493,"text":436},{"id":463,"depth":493,"text":463},"\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。",[1961,1964,1967,1970],{"q":1962,"a":1963},"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":1965,"a":1966},"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":1968,"a":1969},"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":1971,"a":1972},"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。",[524,523],{},[1976,533,532,531,530],"web",[1978,1980],{"plan":1611,"price":538,"features":1979,"notes":540},"全功能 \u002F 80+ 模型 \u002F 知识库 \u002F 插件 \u002F 助手市场",{"plan":1591,"price":1981,"features":1982,"notes":1983},"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",[1986,1987],{"name":550,"url":459},{"name":552,"url":553},{"power":555,"ux":555,"price":555,"cn_support":555,"stability":504},{"title":241,"description":1959},"coding\u002Flocal\u002Flobe-chat",[1992,1994,1996],{"name":1993,"url":1930,"accessed":561},"LobeChat GitHub",{"name":1995,"url":1937,"accessed":561},"腾讯云开发者社区 — Lobe Chat 终极桌面客户端",{"name":1997,"url":1944,"accessed":561},"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",[503,1976,1025,1026,570,2001,2002,572],"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","qiaN3oNudbNSX66toalN5tVx-_meCcKL6o6rq4RdAdI",{"id":2007,"title":1221,"alternatives":2008,"api_compatible":25,"body":2013,"category":503,"chinese_friendly":490,"cover":2431,"description":2432,"domestic":507,"extension":508,"faq":2433,"free":507,"github":25,"languages":2446,"meta":2447,"models":25,"navigation":526,"notSuitable":25,"opensource":526,"path":21,"pillar":528,"platforms":2448,"priceTable":2449,"pricing":2453,"published":547,"relatedPlaybooks":2454,"relatedReviews":25,"score":2457,"self_host":526,"seo":2458,"slug":2459,"sources":2460,"stem":2466,"suitable":25,"tagline":2467,"tags":2468,"updated":561,"verdict":2473,"website":2474,"__hash__":2475},"tools\u002Ftools\u002Fcoding\u002Flocal\u002Follama.md",[2009,2010,2011,2012],{"name":23,"url":24},{"name":587,"url":527},{"name":17,"url":18},{"name":14,"url":15},{"type":27,"value":2014,"toc":2419},[2015,2017,2024,2027,2029,2097,2099,2102,2106,2110,2130,2134,2164,2166,2200,2202,2317,2319,2351,2353,2376,2378,2396,2398],[30,2016,33],{"id":32},[35,2018,2019,2020,2023],{},"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，",[130,2021,2022],{},"ollama pull"," 一键拉。",[35,2025,2026],{},"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）。",[30,2028,43],{"id":43},[45,2030,2031,2037,2045,2051,2058,2073,2079,2085,2091],{},[48,2032,2033,2036],{},[51,2034,2035],{},"后台 Daemon","：开机自启，应用调用零延迟",[48,2038,2039,1080,2042],{},[51,2040,2041],{},"CLI",[130,2043,2044],{},"ollama pull \u002F run \u002F list \u002F show \u002F create \u002F serve",[48,2046,2047,2050],{},[51,2048,2049],{},"Modelfile","：类 Dockerfile 注册任意 GGUF，配 SYSTEM \u002F PARAMETER \u002F TEMPLATE",[48,2052,2053,1080,2055],{},[51,2054,1280],{},[130,2056,2057],{},"http:\u002F\u002Flocalhost:11434\u002Fv1\u002Fchat\u002Fcompletions",[48,2059,2060,1080,2063,2066,2067,2066,2070],{},[51,2061,2062],{},"原生 API",[130,2064,2065],{},"\u002Fapi\u002Fchat","、",[130,2068,2069],{},"\u002Fapi\u002Fgenerate",[130,2071,2072],{},"\u002Fapi\u002Fembeddings",[48,2074,2075,2078],{},[51,2076,2077],{},"模型库","：官方注册表内置 Llama \u002F Qwen \u002F DeepSeek \u002F Gemma \u002F Mistral \u002F GPT-OSS 等",[48,2080,2081,2084],{},[51,2082,2083],{},"MLX 加速（Mac）","：0.19+ 起 M 系列自动用 MLX",[48,2086,2087,2090],{},[51,2088,2089],{},"量化","：默认 Q4_K_M、支持 Q5 \u002F Q8 \u002F FP16",[48,2092,2093,2096],{},[51,2094,2095],{},"跨平台","：Win \u002F Mac \u002F Linux 安装包，Docker 官方镜像",[30,2098,111],{"id":111},[35,2100,2101],{},"完全免费、MIT 开源、商用免费。",[30,2103,2105],{"id":2104},"实测m2-pro-qwen3-coder-7b-q4","实测（M2 Pro + Qwen3-Coder-7B Q4）",[35,2107,2108],{},[51,2109,123],{},[45,2111,2112,2118,2121,2124,2127],{},[48,2113,2114,2117],{},[130,2115,2116],{},"ollama run qwen3-coder:7b"," 一行起飞，3 秒进交互",[48,2119,2120],{},"REST API 配 Cursor \u002F Cline \u002F Continue 几乎全工具开箱即用",[48,2122,2123],{},"Modelfile 写自定义编码助手（low temperature + system prompt + 16K context）几分钟搞定",[48,2125,2126],{},"多模型并存，按需切换，内存占用合理",[48,2128,2129],{},"Mac M 系列 MLX 后比旧 GGUF 模式快显著",[35,2131,2132],{},[51,2133,157],{},[45,2135,2136,2145,2151,2158,2161],{},[48,2137,165,2138,2141,2142],{},[130,2139,2140],{},"num_ctx"," 偏小（2048），跑长上下文要在 Modelfile 加 ",[130,2143,2144],{},"PARAMETER num_ctx 16384",[48,2146,2147,2148],{},"模型默认走 0.0.0.0:11434 ↔ Docker 容器互访要 ",[130,2149,2150],{},"--add-host=host.docker.internal:host-gateway",[48,2152,2153,2154,2157],{},"国内 ",[130,2155,2156],{},"ollama.com\u002Flibrary"," 下载偶有慢，可手动 HF 下 GGUF + Modelfile 自建",[48,2159,2160],{},"多用户并发吞吐显著低于 vLLM",[48,2162,2163],{},"没有 GUI，模型浏览 \u002F 参数面板要走 LM Studio \u002F Open WebUI 配合",[30,2165,185],{"id":185},[187,2167,2168,2174,2180,2185,2191,2197],{},[48,2169,2170,2173],{},[130,2171,2172],{},"curl -fsSL https:\u002F\u002Follama.ai\u002Finstall.sh | sh","（Mac \u002F Linux）；Windows winget",[48,2175,2176,2179],{},[130,2177,2178],{},"ollama pull qwen3-coder:7b","（按需换模型）",[48,2181,2182,2184],{},[130,2183,2116],{}," 直接聊",[48,2186,2187,2188],{},"应用接入：baseURL = ",[130,2189,2190],{},"http:\u002F\u002Flocalhost:11434\u002Fv1",[48,2192,2193,2194],{},"自定义：写 Modelfile → ",[130,2195,2196],{},"ollama create my-coder -f Modelfile",[48,2198,2199],{},"进阶：装 Open WebUI 做前端 \u002F 多人共享",[30,2201,223],{"id":223},[225,2203,2204,2219],{},[228,2205,2206],{},[231,2207,2208,2210,2212,2214,2217],{},[234,2209,236],{},[234,2211,1221],{},[234,2213,247],{},[234,2215,2216],{},"vLLM",[234,2218,1226],{},[249,2220,2221,2237,2249,2262,2275,2291,2303],{},[231,2222,2223,2225,2228,2231,2234],{},[254,2224,256],{},[254,2226,2227],{},"CLI + Daemon",[254,2229,2230],{},"GUI + Headless",[254,2232,2233],{},"Python Server",[254,2235,2236],{},"C++ 二进制",[231,2238,2239,2241,2243,2245,2247],{},[254,2240,185],{},[254,2242,1339],{},[254,2244,1339],{},[254,2246,1318],{},[254,2248,1347],{},[231,2250,2251,2253,2255,2258,2260],{},[254,2252,1249],{},[254,2254,2041],{},[254,2256,2257],{},"✅ GUI",[254,2259,281],{},[254,2261,281],{},[231,2263,2264,2267,2269,2271,2273],{},[254,2265,2266],{},"OpenAI 兼容",[254,2268,1286],{},[254,2270,1283],{},[254,2272,278],{},[254,2274,278],{},[231,2276,2277,2280,2283,2286,2289],{},[254,2278,2279],{},"多用户吞吐",[254,2281,2282],{},"弱（~40 tok\u002Fs）",[254,2284,2285],{},"中（50–90）",[254,2287,2288],{},"强（800–12500）",[254,2290,1318],{},[231,2292,2293,2295,2297,2299,2301],{},[254,2294,1295],{},[254,2296,1300],{},[254,2298,278],{},[254,2300,1273],{},[254,2302,344],{},[231,2304,2305,2307,2309,2312,2315],{},[254,2306,1323],{},[254,2308,352],{},[254,2310,2311],{},"闭源",[254,2313,2314],{},"Apache 2.0",[254,2316,352],{},[30,2318,363],{"id":363},[45,2320,2321,2327,2333,2339,2345],{},[48,2322,2323,2326],{},[51,2324,2325],{},"num_ctx 一定要设","：默认 2K 太小，跑代码 \u002F 长文档要 16K+",[48,2328,2329,2332],{},[51,2330,2331],{},"Modelfile 模板别漏 TEMPLATE","：错的 chat template 会让模型输出乱码 \u002F 不停",[48,2334,2335,2338],{},[51,2336,2337],{},"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×",[48,2340,2341,2344],{},[51,2342,2343],{},"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",[48,2346,2347,2350],{},[51,2348,2349],{},"Mac 让它自动用 MLX","：升 0.19+；不要手动强制 GGUF + Metal",[30,2352,410],{"id":409},[45,2354,2355,2358,2361,2364,2367,2370,2373],{},[48,2356,2357],{},"✅ 应用 \u002F IDE 接本地模型（Cursor \u002F Cline \u002F Continue）",[48,2359,2360],{},"✅ 个人 \u002F 评估 \u002F 脚本自动化",[48,2362,2363],{},"✅ Modelfile 自定义系统 prompt + 参数",[48,2365,2366],{},"✅ Mac M 系列 MLX 用户",[48,2368,2369],{},"❌ 多用户并发生产服务（用 vLLM）",[48,2371,2372],{},"❌ GUI 调参 \u002F 模型浏览（配 LM Studio \u002F Open WebUI）",[48,2374,2375],{},"❌ 极致单卡吞吐研究（直接 llama.cpp \u002F vLLM）",[30,2377,436],{"id":436},[45,2379,2380,2384,2388,2392],{},[48,2381,2382],{},[442,2383,933],{"href":24},[48,2385,2386],{},[442,2387,1419],{"href":527},[48,2389,2390],{},[442,2391,449],{"href":18},[48,2393,2394],{},[442,2395,460],{"href":459},[30,2397,463],{"id":463},[187,2399,2400,2407,2414],{},[48,2401,2402,2403],{},"Markaicode — Import GGUF Models into Ollama 2026（2026-05-15）",[442,2404,2405],{"href":2405,"rel":2406},"https:\u002F\u002Fmarkaicode.com\u002Fimport-gguf-models-ollama-guide",[473],[48,2408,2409,2410],{},"ComputingForGeeks — Ollama Models Cheat Sheet 2026 ",[442,2411,2412],{"href":2412,"rel":2413},"https:\u002F\u002Fcomputingforgeeks.com\u002Follama-models-cheat-sheet",[473],[48,2415,1449,2416],{},[442,2417,1452],{"href":1452,"rel":2418},[473],{"title":489,"searchDepth":490,"depth":490,"links":2420},[2421,2422,2423,2424,2425,2426,2427,2428,2429,2430],{"id":32,"depth":493,"text":33},{"id":43,"depth":493,"text":43},{"id":111,"depth":493,"text":111},{"id":2104,"depth":493,"text":2105},{"id":185,"depth":493,"text":185},{"id":223,"depth":493,"text":223},{"id":363,"depth":493,"text":363},{"id":409,"depth":493,"text":410},{"id":436,"depth":493,"text":436},{"id":463,"depth":493,"text":463},"\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。",[2434,2437,2440,2443],{"q":2435,"a":2436},"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":2438,"a":2439},"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":2441,"a":2442},"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":2444,"a":2445},"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。",[523],{},[533,532,531,530],[2450],{"plan":537,"price":538,"features":2451,"notes":2452},"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）",[2455,2456],{"name":550,"url":459},{"name":552,"url":553},{"power":504,"ux":504,"price":555,"cn_support":490,"stability":555},{"title":1221,"description":2432},"coding\u002Flocal\u002Follama",[2461,2463,2465],{"name":2462,"url":2405,"accessed":561},"Markaicode — Import GGUF 2026",{"name":2464,"url":2412,"accessed":561},"ComputingForGeeks — Ollama Cheat Sheet 2026",{"name":1505,"url":1452,"accessed":561},"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",[503,2469,2470,2471,2472,1510,1511,1514,572],"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","V4PvNLB8lbjAzlhHpWFHSyKzvb328rvWx3nckggFlD8",{"id":10,"title":11,"alternatives":2477,"api_compatible":25,"body":2482,"category":503,"chinese_friendly":504,"cover":505,"description":506,"domestic":507,"extension":508,"faq":2811,"free":507,"github":25,"languages":2816,"meta":2817,"models":25,"navigation":526,"notSuitable":25,"opensource":526,"path":527,"pillar":528,"platforms":2818,"priceTable":2819,"pricing":546,"published":547,"relatedPlaybooks":2822,"relatedReviews":25,"score":2825,"self_host":526,"seo":2826,"slug":557,"sources":2827,"stem":566,"suitable":25,"tagline":567,"tags":2831,"updated":561,"verdict":573,"website":574,"__hash__":575},[2478,2479,2480,2481],{"name":14,"url":15},{"name":17,"url":18},{"name":20,"url":21},{"name":23,"url":24},{"type":27,"value":2483,"toc":2799},[2484,2486,2488,2490,2492,2534,2536,2538,2540,2544,2562,2566,2582,2584,2606,2608,2710,2712,2742,2744,2760,2762,2780,2782],[30,2485,33],{"id":32},[35,2487,37],{},[35,2489,40],{},[30,2491,43],{"id":43},[45,2493,2494,2498,2502,2506,2510,2514,2518,2522,2526,2530],{},[48,2495,2496,54],{},[51,2497,53],{},[48,2499,2500,60],{},[51,2501,59],{},[48,2503,2504,66],{},[51,2505,65],{},[48,2507,2508,72],{},[51,2509,71],{},[48,2511,2512,78],{},[51,2513,77],{},[48,2515,2516,84],{},[51,2517,83],{},[48,2519,2520,90],{},[51,2521,89],{},[48,2523,2524,96],{},[51,2525,95],{},[48,2527,2528,102],{},[51,2529,101],{},[48,2531,2532,108],{},[51,2533,107],{},[30,2535,111],{"id":111},[35,2537,114],{},[30,2539,118],{"id":117},[35,2541,2542],{},[51,2543,123],{},[45,2545,2546,2550,2552,2556,2558,2560],{},[48,2547,128,2548,133],{},[130,2549,132],{},[48,2551,136],{},[48,2553,139,2554,143],{},[130,2555,142],{},[48,2557,146],{},[48,2559,149],{},[48,2561,152],{},[35,2563,2564],{},[51,2565,157],{},[45,2567,2568,2570,2574,2578,2580],{},[48,2569,162],{},[48,2571,165,2572,169],{},[130,2573,168],{},[48,2575,172,2576,176],{},[130,2577,175],{},[48,2579,179],{},[48,2581,182],{},[30,2583,185],{"id":185},[187,2585,2586,2590,2594,2596,2598,2600,2602],{},[48,2587,191,2588],{},[130,2589,194],{},[48,2591,197,2592,201],{},[130,2593,200],{},[48,2595,204],{},[48,2597,207],{},[48,2599,210],{},[48,2601,213],{},[48,2603,216,2604,220],{},[130,2605,219],{},[30,2607,223],{"id":223},[225,2609,2610,2624],{},[228,2611,2612],{},[231,2613,2614,2616,2618,2620,2622],{},[234,2615,236],{},[234,2617,11],{},[234,2619,241],{},[234,2621,244],{},[234,2623,247],{},[249,2625,2626,2638,2650,2662,2674,2686,2698],{},[231,2627,2628,2630,2632,2634,2636],{},[254,2629,256],{},[254,2631,259],{},[254,2633,262],{},[254,2635,265],{},[254,2637,265],{},[231,2639,2640,2642,2644,2646,2648],{},[254,2641,272],{},[254,2643,275],{},[254,2645,278],{},[254,2647,281],{},[254,2649,281],{},[231,2651,2652,2654,2656,2658,2660],{},[254,2653,288],{},[254,2655,291],{},[254,2657,278],{},[254,2659,278],{},[254,2661,298],{},[231,2663,2664,2666,2668,2670,2672],{},[254,2665,303],{},[254,2667,306],{},[254,2669,278],{},[254,2671,278],{},[254,2673,298],{},[231,2675,2676,2678,2680,2682,2684],{},[254,2677,317],{},[254,2679,320],{},[254,2681,323],{},[254,2683,281],{},[254,2685,281],{},[231,2687,2688,2690,2692,2694,2696],{},[254,2689,332],{},[254,2691,335],{},[254,2693,338],{},[254,2695,341],{},[254,2697,344],{},[231,2699,2700,2702,2704,2706,2708],{},[254,2701,349],{},[254,2703,352],{},[254,2705,352],{},[254,2707,357],{},[254,2709,360],{},[30,2711,363],{"id":363},[45,2713,2714,2718,2724,2730,2734,2738],{},[48,2715,2716,371],{},[51,2717,370],{},[48,2719,2720,379],{},[51,2721,376,2722,220],{},[130,2723,219],{},[48,2725,2726,385,2728,388],{},[51,2727,384],{},[130,2729,175],{},[48,2731,2732,394],{},[51,2733,393],{},[48,2735,2736,400],{},[51,2737,399],{},[48,2739,2740,406],{},[51,2741,405],{},[30,2743,410],{"id":409},[45,2745,2746,2748,2750,2752,2754,2756,2758],{},[48,2747,415],{},[48,2749,418],{},[48,2751,421],{},[48,2753,424],{},[48,2755,427],{},[48,2757,430],{},[48,2759,433],{},[30,2761,436],{"id":436},[45,2763,2764,2768,2772,2776],{},[48,2765,2766],{},[442,2767,444],{"href":15},[48,2769,2770],{},[442,2771,449],{"href":18},[48,2773,2774],{},[442,2775,454],{"href":21},[48,2777,2778],{},[442,2779,460],{"href":459},[30,2781,463],{"id":463},[187,2783,2784,2789,2794],{},[48,2785,468,2786],{},[442,2787,471],{"href":471,"rel":2788},[473],[48,2790,476,2791],{},[442,2792,479],{"href":479,"rel":2793},[473],[48,2795,483,2796],{},[442,2797,486],{"href":486,"rel":2798},[473],{"title":489,"searchDepth":490,"depth":490,"links":2800},[2801,2802,2803,2804,2805,2806,2807,2808,2809,2810],{"id":32,"depth":493,"text":33},{"id":43,"depth":493,"text":43},{"id":111,"depth":493,"text":111},{"id":117,"depth":493,"text":118},{"id":185,"depth":493,"text":185},{"id":223,"depth":493,"text":223},{"id":363,"depth":493,"text":363},{"id":409,"depth":493,"text":410},{"id":436,"depth":493,"text":436},{"id":463,"depth":493,"text":463},[2812,2813,2814,2815],{"q":511,"a":512},{"q":514,"a":515},{"q":517,"a":518},{"q":520,"a":521},[523,524],{},[530,531,532,533,534],[2820,2821],{"plan":537,"price":538,"features":539,"notes":540},{"plan":542,"price":543,"features":544,"notes":545},[2823,2824],{"name":550,"url":459},{"name":552,"url":553},{"power":555,"ux":504,"price":555,"cn_support":504,"stability":555},{"title":11,"description":506},[2828,2829,2830],{"name":560,"url":471,"accessed":561},{"name":563,"url":479,"accessed":561},{"name":565,"url":486,"accessed":561},[503,569,530,570,571,20,572],[],1782316490689]