[{"data":1,"prerenderedAt":2265},["ShallowReactive",2],{"header-counts":3,"compare-fastgpt-vs-dify":6,"footer-counts":7,"compare-a-fastgpt":10,"compare-b-dify":1233},{"tools":4,"reviews":5},70,14,null,{"tools":4,"reviews":5,"playbooks":8,"news":9},19,13,{"id":11,"title":12,"alternatives":13,"api_compatible":18,"body":20,"category":1143,"chinese_friendly":416,"cover":1144,"description":1145,"domestic":1146,"extension":1147,"faq":6,"free":1146,"github":1103,"languages":1148,"meta":1151,"models":1152,"navigation":412,"notSuitable":1159,"opensource":412,"path":1163,"pillar":1164,"platforms":1165,"priceTable":1169,"pricing":1194,"published":1195,"relatedPlaybooks":1196,"relatedReviews":1198,"score":1200,"self_host":412,"seo":1201,"seoTitle":6,"slug":1202,"sources":1203,"stem":1214,"suitable":1215,"tagline":1221,"tags":1222,"updated":1230,"verdict":1231,"website":1096,"__hash__":1232},"tools\u002Ftools\u002Fagent\u002Fplatform\u002Ffastgpt.md","FastGPT",[14,15,16,17],"agent\u002Fplatform\u002Fdify","agent\u002Fplatform\u002Fcoze","agent\u002Fplatform\u002Fragflow","agent\u002Fplatform\u002Fanythingllm",[19],"openai",{"type":21,"value":22,"toc":1125},"minimark",[23,28,66,78,81,86,89,105,108,134,138,145,178,185,189,249,256,259,262,265,272,318,326,330,361,365,500,503,506,510,518,646,661,664,834,845,880,886,889,965,969,972,992,995,1013,1016,1086,1089,1114,1121],[24,25,27],"h2",{"id":26},"tldr","TL;DR",[29,30,35,55],"div",{"className":31},[32,33,34],"card","p-5","my-4",[36,37,38,42,43,50,51,54],"p",{},[39,40,41],"strong",{},"一句话："," labring 团队开源的 LLM 知识库 RAG 平台，27k+ GitHub star（截至 2026-03 数据，",[44,45,49],"a",{"href":46,"rel":47},"https:\u002F\u002Fcloud.tencent.com\u002Fdeveloper\u002Farticle\u002F2632669",[48],"nofollow","腾讯云 2026-03 教程"," 引用），Apache 2.0 许可证可商用。",[39,52,53],{},"核心优势是 RAG 链路工程做得极细","——问题预处理、混合检索、重排序、上下文组装、答案生成每一步都可视化调参。",[36,56,57,58,61,62,65],{},"最大价值在 ",[39,59,60],{},"国内企业知识库 + 私有部署"," 场景。代价是 ",[39,63,64],{},"配置门槛","：docker 基础 + 网络知识 + 一定运维能力。",[67,68,69],"blockquote",{},[36,70,71,72,77],{},"来源说明：本文基于 fastgpt.io 官方页面、github.com\u002Flabring\u002FFastGPT 仓库、",[44,73,76],{"href":74,"rel":75},"https:\u002F\u002Fwww.nanhuantech.com\u002Fzh\u002Fai-reviews\u002Ffastgpt-2025-review",[48],"南环 AI 2026-05 评测","、腾讯云开发者社区 2026-03 部署教程综合整理。版本迭代较快，命令和价格请以最新官方文档为准。",[24,79,80],{"id":80},"核心特性",[82,83,85],"h3",{"id":84},"知识库管理核心能力","知识库管理（核心能力）",[36,87,88],{},"支持文件类型：",[90,91,92,96,99,102],"ul",{},[93,94,95],"li",{},"文档：PDF \u002F Word \u002F Markdown \u002F TXT \u002F HTML",[93,97,98],{},"表格：Excel \u002F CSV",[93,100,101],{},"网页：URL 抓取 + 定时同步",[93,103,104],{},"API：通过接口推送内容",[36,106,107],{},"处理流程：上传 → 文本切分 → 向量化 → 存储 → 可用于问答。支持：",[90,109,110,116,122,128],{},[93,111,112,115],{},[39,113,114],{},"文件夹分组","：不同主题 \u002F 部门分类",[93,117,118,121],{},[39,119,120],{},"多种分块策略","：默认按段落 \u002F 按 token 数 \u002F 自定义",[93,123,124,127],{},[39,125,126],{},"批量导入","：脚本化大批量同步",[93,129,130,133],{},[39,131,132],{},"定时同步","：网页源自动更新",[82,135,137],{"id":136},"rag-流程编排最强卖点","RAG 流程编排（最强卖点）",[36,139,140,144],{},[44,141,143],{"href":74,"rel":142},[48],"南环 AI 2026 评测"," 总结的 FastGPT RAG 链路：",[146,147,148,154,160,166,172],"ol",{},[93,149,150,153],{},[39,151,152],{},"问题预处理","：改写 \u002F 扩展 \u002F 错词纠正（提升召回率）",[93,155,156,159],{},[39,157,158],{},"检索策略","：语义检索 \u002F 关键词 BM25 \u002F 混合检索，可调相似度阈值",[93,161,162,165],{},[39,163,164],{},"重排序（Rerank）","：对初步检索结果二次排序，提升相关性",[93,167,168,171],{},[39,169,170],{},"上下文组装","：最优 chunk + 问题 → prompt",[93,173,174,177],{},[39,175,176],{},"答案生成","：调大模型基于检索结果回答 + 引用标注",[36,179,180,181,184],{},"每一步都可视化调参，这是 FastGPT 比 Coze \u002F Dify 在 ",[39,182,183],{},"纯知识库 QA 精度","上更高的原因。",[82,186,188],{"id":187},"多模型支持不绑定厂商","多模型支持（不绑定厂商）",[190,191,192,205],"table",{},[193,194,195],"thead",{},[196,197,198,202],"tr",{},[199,200,201],"th",{},"模型类别",[199,203,204],{},"支持",[206,207,208,217,225,233,241],"tbody",{},[196,209,210,214],{},[211,212,213],"td",{},"国产闭源",[211,215,216],{},"豆包 \u002F 通义千问 \u002F 文心一言 \u002F 智谱 GLM \u002F Moonshot Kimi \u002F MiniMax",[196,218,219,222],{},[211,220,221],{},"开源",[211,223,224],{},"LLaMA \u002F Qwen \u002F ChatGLM \u002F DeepSeek 等可自部署",[196,226,227,230],{},[211,228,229],{},"OpenAI 系",[211,231,232],{},"GPT-5 \u002F GPT-5 mini \u002F o3",[196,234,235,238],{},[211,236,237],{},"Claude 系",[211,239,240],{},"Sonnet 4 \u002F Opus 4 \u002F Haiku",[196,242,243,246],{},[211,244,245],{},"嵌入 \u002F 重排",[211,247,248],{},"BGE \u002F m3e \u002F OpenAI text-embedding-3",[36,250,251,252,255],{},"可以在 ",[39,253,254],{},"应用级别","为不同知识库 \u002F 不同场景配置不同模型，做\"低成本 embedding + 高质量 LLM 生成\"组合。",[82,257,258],{"id":258},"工作流与高级编排",[36,260,261],{},"新版本（v4.14.x）支持类似 Dify 的工作流节点编排——条件分支、循环、HTTP 调用、代码节点。能做\"分类 → 路由到不同子知识库 → 不同模型回答\"这类复杂场景。",[82,263,264],{"id":264},"多向量库选择",[36,266,267,271],{},[44,268,270],{"href":46,"rel":269},[48],"腾讯云教程"," 公开的 4 种向量后端：",[190,273,274,284],{},[193,275,276],{},[196,277,278,281],{},[199,279,280],{},"后端",[199,282,283],{},"适用",[206,285,286,294,302,310],{},[196,287,288,291],{},[211,289,290],{},"PgVector",[211,292,293],{},"5000 万索引以下，新手 \u002F 小规模",[196,295,296,299],{},[211,297,298],{},"Milvus",[211,300,301],{},"亿级以上，高性能",[196,303,304,307],{},[211,305,306],{},"Zilliz Cloud",[211,308,309],{},"Milvus 全托管 SaaS",[196,311,312,315],{},[211,313,314],{},"SeekDB \u002F OceanBase",[211,316,317],{},"企业级国产化",[36,319,320,321,325],{},"部署时选对应 ",[322,323,324],"code",{},"docker-compose.{pgvector|milvus|...}.yml","。",[82,327,329],{"id":328},"api-与-mcp","API 与 MCP",[90,331,332,338,344,355],{},[93,333,334,337],{},[39,335,336],{},"对话 API","：流式 \u002F 非流式 HTTP，OpenAI 兼容",[93,339,340,343],{},[39,341,342],{},"知识库检索 API","：单独调检索（不走生成）做 hybrid pipeline",[93,345,346,349,350,354],{},[39,347,348],{},"MCP Server","：3005 端口暴露 MCP SSE 服务，可被 ",[44,351,353],{"href":352},"\u002Fcoding\u002Fcli\u002Fclaude-code.html","Claude Code"," 等客户端直接接入",[93,356,357,360],{},[39,358,359],{},"Webhook","：回调通知",[24,362,364],{"id":363},"部署-10-分钟docker","部署 10 分钟（Docker）",[366,367,372],"pre",{"className":368,"code":369,"language":370,"meta":371,"style":371},"language-bash shiki shiki-themes github-light github-dark","# 克隆代码\ngit clone https:\u002F\u002Fgithub.com\u002Flabring\u002FFastGPT.git\ncd FastGPT\n\n# 切到最新稳定版（参考 GitHub releases）\ngit switch -c 4.14.7.2\n\n# 选向量库版本（个人 \u002F 小规模选 pg）\ncd deploy\u002Fdocker\u002Fcn\nwget https:\u002F\u002Fdoc.fastgpt.cn\u002Fdeploy\u002Fconfig\u002Fconfig.json\n\n# 启动\ndocker-compose -f docker-compose.pg.yml up -d\n\n# 访问 http:\u002F\u002F\u003Cip>:3000，默认账号 root \u002F 1234\n","bash","",[322,373,374,383,397,407,414,420,434,439,445,453,462,467,473,490,494],{"__ignoreMap":371},[375,376,379],"span",{"class":377,"line":378},"line",1,[375,380,382],{"class":381},"sJ8bj","# 克隆代码\n",[375,384,386,390,394],{"class":377,"line":385},2,[375,387,389],{"class":388},"sScJk","git",[375,391,393],{"class":392},"sZZnC"," clone",[375,395,396],{"class":392}," https:\u002F\u002Fgithub.com\u002Flabring\u002FFastGPT.git\n",[375,398,400,404],{"class":377,"line":399},3,[375,401,403],{"class":402},"sj4cs","cd",[375,405,406],{"class":392}," FastGPT\n",[375,408,410],{"class":377,"line":409},4,[375,411,413],{"emptyLinePlaceholder":412},true,"\n",[375,415,417],{"class":377,"line":416},5,[375,418,419],{"class":381},"# 切到最新稳定版（参考 GitHub releases）\n",[375,421,423,425,428,431],{"class":377,"line":422},6,[375,424,389],{"class":388},[375,426,427],{"class":392}," switch",[375,429,430],{"class":402}," -c",[375,432,433],{"class":402}," 4.14.7.2\n",[375,435,437],{"class":377,"line":436},7,[375,438,413],{"emptyLinePlaceholder":412},[375,440,442],{"class":377,"line":441},8,[375,443,444],{"class":381},"# 选向量库版本（个人 \u002F 小规模选 pg）\n",[375,446,448,450],{"class":377,"line":447},9,[375,449,403],{"class":402},[375,451,452],{"class":392}," deploy\u002Fdocker\u002Fcn\n",[375,454,456,459],{"class":377,"line":455},10,[375,457,458],{"class":388},"wget",[375,460,461],{"class":392}," https:\u002F\u002Fdoc.fastgpt.cn\u002Fdeploy\u002Fconfig\u002Fconfig.json\n",[375,463,465],{"class":377,"line":464},11,[375,466,413],{"emptyLinePlaceholder":412},[375,468,470],{"class":377,"line":469},12,[375,471,472],{"class":381},"# 启动\n",[375,474,475,478,481,484,487],{"class":377,"line":9},[375,476,477],{"class":388},"docker-compose",[375,479,480],{"class":402}," -f",[375,482,483],{"class":392}," docker-compose.pg.yml",[375,485,486],{"class":392}," up",[375,488,489],{"class":402}," -d\n",[375,491,492],{"class":377,"line":5},[375,493,413],{"emptyLinePlaceholder":412},[375,495,497],{"class":377,"line":496},15,[375,498,499],{"class":381},"# 访问 http:\u002F\u002F\u003Cip>:3000，默认账号 root \u002F 1234\n",[36,501,502],{},"最低配置：2C4G + 20GB 硬盘 + Docker 28+ + Docker Compose 2.20+。",[36,504,505],{},"进入后台 → 账号 → 模型提供商 → 配置至少 1 个对话模型 + 1 个嵌入模型 → 即可开始建知识库。",[24,507,509],{"id":508},"云版-vs-自托管对比","云版 vs 自托管对比",[36,511,512,517],{},[44,513,516],{"href":514,"rel":515},"https:\u002F\u002Ffastgpt.io\u002Fzh\u002Fprice",[48],"fastgpt.io 官方定价"," 公开数据：",[190,519,520,548],{},[193,521,522],{},[196,523,524,527,530,533,536,539,542,545],{},[199,525,526],{},"套餐",[199,528,529],{},"价格",[199,531,532],{},"AI 积分",[199,534,535],{},"知识库索引",[199,537,538],{},"团队",[199,540,541],{},"Agent",[199,543,544],{},"知识库",[199,546,547],{},"QPM",[206,549,550,576,601,625],{},[196,551,552,555,558,561,564,567,570,573],{},[211,553,554],{},"免费",[211,556,557],{},"¥0",[211,559,560],{},"100",[211,562,563],{},"600",[211,565,566],{},"1",[211,568,569],{},"10",[211,571,572],{},"3",[211,574,575],{},"30",[196,577,578,581,584,587,590,593,596,598],{},[211,579,580],{},"基础",[211,582,583],{},"¥99\u002F月",[211,585,586],{},"4000",[211,588,589],{},"6000",[211,591,592],{},"5",[211,594,595],{},"50",[211,597,575],{},[211,599,600],{},"300",[196,602,603,606,609,612,615,617,620,622],{},[211,604,605],{},"高级",[211,607,608],{},"¥599\u002F月",[211,610,611],{},"25000",[211,613,614],{},"36000",[211,616,595],{},[211,618,619],{},"200",[211,621,560],{},[211,623,624],{},"1500",[196,626,627,630,633,636,638,640,642,644],{},[211,628,629],{},"定制",[211,631,632],{},"议价",[211,634,635],{},"弹性",[211,637,635],{},[211,639,635],{},[211,641,635],{},[211,643,635],{},[211,645,635],{},[36,647,648,651,652,655,656,660],{},[39,649,650],{},"云版适合","：不想运维、量小、要快速上线\n",[39,653,654],{},"自托管适合","：量大（10 万+ 日问答）、数据敏感、要深度定制——按 ",[44,657,659],{"href":74,"rel":658},[48],"南环评测"," 估算：\"日均 10 万次问答的企业场景，商业 SaaS 年费数十万，自建 FastGPT + 开源模型只需数万硬件投入\"",[24,662,663],{"id":663},"与同类怎么选",[190,665,666,693],{},[193,667,668],{},[196,669,670,673,675,681,687,690],{},[199,671,672],{},"维度",[199,674,12],{},[199,676,677],{},[44,678,680],{"href":679},"\u002Fagent\u002Fplatform\u002Fdify.html","Dify",[199,682,683],{},[44,684,686],{"href":685},"\u002Fagent\u002Fplatform\u002Fcoze.html","Coze",[199,688,689],{},"RAGFlow",[199,691,692],{},"AnythingLLM",[206,694,695,715,732,751,769,785,802,817],{},[196,696,697,700,703,706,709,712],{},[211,698,699],{},"核心定位",[211,701,702],{},"知识库 QA",[211,704,705],{},"综合 LLMOps",[211,707,708],{},"Bot + 工作流",[211,710,711],{},"文档解析+RAG",[211,713,714],{},"桌面级 KB",[196,716,717,719,722,724,727,729],{},[211,718,221],{},[211,720,721],{},"✅ Apache 2.0",[211,723,721],{},[211,725,726],{},"❌",[211,728,721],{},[211,730,731],{},"✅ MIT",[196,733,734,737,740,743,746,749],{},[211,735,736],{},"私有部署",[211,738,739],{},"★★★★★ docker",[211,741,742],{},"★★★★★",[211,744,745],{},"⚠️ 仅企业版",[211,747,748],{},"★★★★☆",[211,750,742],{},[196,752,753,756,759,761,764,767],{},[211,754,755],{},"RAG 深度",[211,757,758],{},"★★★★★ 最细",[211,760,748],{},[211,762,763],{},"★★★☆☆",[211,765,766],{},"★★★★★ 文档解析最强",[211,768,763],{},[196,770,771,774,776,778,780,782],{},[211,772,773],{},"工作流",[211,775,748],{},[211,777,742],{},[211,779,748],{},[211,781,763],{},[211,783,784],{},"★★☆☆☆",[196,786,787,790,793,795,798,800],{},[211,788,789],{},"上手",[211,791,792],{},"★★★☆☆ 需 docker",[211,794,748],{},[211,796,797],{},"★★★★★ 最简单",[211,799,763],{},[211,801,748],{},[196,803,804,807,809,811,813,815],{},[211,805,806],{},"中文优化",[211,808,742],{},[211,810,748],{},[211,812,742],{},[211,814,748],{},[211,816,763],{},[196,818,819,822,825,827,829,832],{},[211,820,821],{},"多平台发布",[211,823,824],{},"⚠️ API 为主",[211,826,748],{},[211,828,742],{},[211,830,831],{},"⚠️",[211,833,831],{},[36,835,836,839,840,844],{},[39,837,838],{},"怎么选","（综合 ",[44,841,843],{"href":74,"rel":842},[48],"南环 AI 评测","）：",[90,846,847,853,861,868,874],{},[93,848,849,852],{},[39,850,851],{},"核心需求是 RAG 精度"," → FastGPT",[93,854,855,858,859],{},[39,856,857],{},"需要丰富插件 + 复杂工作流 + 多平台发布"," → ",[44,860,680],{"href":679},[93,862,863,858,866],{},[39,864,865],{},"零代码、快速发布到飞书 \u002F 微信",[44,867,686],{"href":685},[93,869,870,873],{},[39,871,872],{},"文档解析（含 OCR \u002F 表格 \u002F 公式）是瓶颈"," → RAGFlow",[93,875,876,879],{},[39,877,878],{},"桌面 \u002F 单机使用"," → AnythingLLM",[36,881,882,885],{},[39,883,884],{},"很多企业同时用","：FastGPT 做知识库底座 + Coze 做前端 Bot 发布 \u002F 工作流编排。",[24,887,888],{"id":888},"避坑清单",[90,890,891,905,914,920,930,942,948,954],{},[93,892,893,896,897,900,901,904],{},[39,894,895],{},"docker-compose 镜像 tag 不一致","：",[44,898,270],{"href":46,"rel":899},[48]," 实测的坑——某些版本编排文件的 image tag 与最新 release 不一致，启动报\"镜像找不到\"，手动改 ",[322,902,903],{},"image:"," 行为正确版本即可",[93,906,907,910,911,913],{},[39,908,909],{},"3000 端口冲突","：默认占用 3000（主服务）\u002F 9000（S3 \u002F MinIO）\u002F 3005（MCP）；改 ",[322,912,477],{}," 的 ports 映射端口",[93,915,916,919],{},[39,917,918],{},"PostgreSQL pgvector 不够用就换 Milvus","：单库索引超 5000 万时 pgvector 查询性能下降，切 Milvus",[93,921,922,925,926,929],{},[39,923,924],{},"向量库选错代价大","：先评估索引量再选向量后端，迁移要重新 embedding 整库，按 ",[44,927,659],{"href":74,"rel":928},[48],"：\"新手 \u002F 小规模 PgVector，中大规模 Milvus，企业 \u002F 国产 OceanBase\"",[93,931,932,896,935,938,939],{},[39,933,934],{},"MinIO 默认密码",[322,936,937],{},"minioadmin\u002Fminioadmin","，",[39,940,941],{},"部署到公网前必须改",[93,943,944,947],{},[39,945,946],{},"分段策略影响巨大","：默认分段对长法律 \u002F 医疗文档不友好，需调\"按章节\"或\"自定义\"",[93,949,950,953],{},[39,951,952],{},"嵌入模型 ≠ 对话模型","：经常有人只配 GPT-4 没配 embedding 模型，知识库无法索引——必须同时配两类",[93,955,956,959,960,964],{},[39,957,958],{},"云版 AI 积分会过期","：未用完不能跨月累积（按 ",[44,961,963],{"href":514,"rel":962},[48],"fastgpt.io 定价 FAQ","）",[24,966,968],{"id":967},"适合-不适合","适合 \u002F 不适合",[36,970,971],{},"✅ 适合：",[90,973,974,977,980,983,986,989],{},[93,975,976],{},"企业内部知识库（员工手册 \u002F 制度 \u002F 流程）",[93,978,979],{},"产品 FAQ \u002F 用户手册问答",[93,981,982],{},"医疗 \u002F 法律 \u002F 金融垂直领域知识系统",[93,984,985],{},"数据严格不出网 + Apache 2.0 商用",[93,987,988],{},"有 docker 运维基础的技术团队",[93,990,991],{},"需要把 RAG 当后端服务的开发者（API 接入业务系统）",[36,993,994],{},"❌ 不适合：",[90,996,997,1002,1007,1010],{},[93,998,999,1000,964],{},"完全非技术用户（去 ",[44,1001,686],{"href":685},[93,1003,1004,1005,964],{},"主要需求是工作流 + 插件集成（去 ",[44,1006,680],{"href":679},[93,1008,1009],{},"文档解析 \u002F OCR 是首要痛点（RAGFlow）",[93,1011,1012],{},"不想自己运维 + 量很小（FastGPT 云免费版起步即可）",[24,1014,1015],{"id":1015},"相关阅读",[90,1017,1018,1031,1052,1075],{},[93,1019,1020,1021,1023,1024,1026,1027],{},"同类对比：",[44,1022,680],{"href":679}," \u002F ",[44,1025,686],{"href":685}," \u002F RAGFlow \u002F AnythingLLM \u002F ",[44,1028,1030],{"href":1029},"\u002Fagent\u002Fplatform\u002Fn8n.html","n8n",[93,1032,1033,1034,1023,1038,1023,1042,1023,1045,1023,1048],{},"概念：",[44,1035,1037],{"href":1036},"\u002Fwiki\u002Frag.html","RAG",[44,1039,1041],{"href":1040},"\u002Fwiki\u002Fembedding.html","Embedding",[44,1043,1044],{"href":1040},"Vector Database",[44,1046,1047],{"href":1036},"Reranker",[44,1049,1051],{"href":1050},"\u002Fwiki\u002Fai-agent.html","AI Agent",[93,1053,1054,1055,1023,1059,1023,1063,1023,1067,1023,1071],{},"模型：",[44,1056,1058],{"href":1057},"\u002Fmodels\u002Fdeepseek-v3.html","DeepSeek-V3",[44,1060,1062],{"href":1061},"\u002Fmodels\u002Fqwen-3.html","Qwen3",[44,1064,1066],{"href":1065},"\u002Fmodels\u002Fglm-5.2.html","GLM-5.2",[44,1068,1070],{"href":1069},"\u002Fmodels\u002Fkimi-k2.html","Kimi K2",[44,1072,1074],{"href":1073},"\u002Fmodels\u002Fdoubao-1-5-pro.html","豆包 Doubao",[93,1076,1077,1078,1023,1082],{},"进阶：",[44,1079,1081],{"href":1080},"\u002Fwiki\u002Fcontext-engineering.html","Context Engineering",[44,1083,1085],{"href":1084},"\u002Fwiki\u002Fprompt-engineering.html","Prompt Engineering",[24,1087,1088],{"id":1088},"来源",[90,1090,1091,1098,1105,1111],{},[93,1092,1093,1094],{},"官网：",[44,1095,1096],{"href":1096,"rel":1097},"https:\u002F\u002Ffastgpt.io",[48],[93,1099,1100,1101],{},"GitHub：",[44,1102,1103],{"href":1103,"rel":1104},"https:\u002F\u002Fgithub.com\u002Flabring\u002FFastGPT",[48],[93,1106,1107,1108],{},"定价：",[44,1109,514],{"href":514,"rel":1110},[48],[93,1112,1113],{},"第三方评测：南环 AI \u002F 腾讯云开发者社区 \u002F 飞书 AGI 掘金知识库",[36,1115,1116,1117,1120],{},"本卡片由 AIHO 编辑部根据官方公开资料与第三方评测整理。所有事实点均标注来源；如发现价格 \u002F 命令 \u002F 功能与最新官方信息不一致，请通过 ",[44,1118,1119],{"href":1119},"\u002Fsubmit"," 反馈。",[1122,1123,1124],"style",{},"html pre.shiki code .sJ8bj, html code.shiki .sJ8bj{--shiki-default:#6A737D;--shiki-dark:#6A737D}html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"title":371,"searchDepth":399,"depth":399,"links":1126},[1127,1128,1136,1137,1138,1139,1140,1141,1142],{"id":26,"depth":385,"text":27},{"id":80,"depth":385,"text":80,"children":1129},[1130,1131,1132,1133,1134,1135],{"id":84,"depth":399,"text":85},{"id":136,"depth":399,"text":137},{"id":187,"depth":399,"text":188},{"id":258,"depth":399,"text":258},{"id":264,"depth":399,"text":264},{"id":328,"depth":399,"text":329},{"id":363,"depth":385,"text":364},{"id":508,"depth":385,"text":509},{"id":663,"depth":385,"text":663},{"id":888,"depth":385,"text":888},{"id":967,"depth":385,"text":968},{"id":1015,"depth":385,"text":1015},{"id":1088,"depth":385,"text":1088},"platform","\u002Fimg\u002Ftools\u002Ffastgpt.webp","FastGPT 真实评测：开源 LLM 知识库 RAG 平台，labring 团队出品，27k+ GitHub star。一键 docker-compose 部署、RAG 流程编排可视化、多向量库支持。AIHO 编辑部基于官方文档与社区资料整理，含与 Dify\u002FCoze 对比、避坑指南。",false,"md",[1149,1150],"zh","en",{},[1153,1154,1155,1156,1157,1158],"deepseek-v3","qwen-max","doubao-pro","gpt-4o","claude-sonnet-4","kimi",[1160,1161,1162],"完全零代码 \u002F 不懂 docker 的用户（去 Coze）","Bot 多平台一键发布场景（Coze 强项）","插件 \u002F 工作流复杂集成（去 Dify）","\u002Ftools\u002Fagent\u002Fplatform\u002Ffastgpt","agent",[1166,1167,1168],"windows","macos","linux",[1170,1175,1180,1185,1190],{"plan":1171,"price":554,"limit":1172,"cn_pay":1173,"note":1174},"Self-host 开源","全功能 + 全数据本地","—","Apache 2.0 可商用",{"plan":1176,"price":1177,"limit":1178,"cn_pay":1173,"note":1179},"云免费版","¥0\u002F月","100 AI 积分 + 600 索引 + 3 知识库","试水",{"plan":1181,"price":583,"limit":1182,"cn_pay":1183,"note":1184},"云基础版","4000 积分 + 6000 索引 + 50 Agent","✅ 微信\u002F支付宝","中小团队 SaaS",{"plan":1186,"price":608,"limit":1187,"cn_pay":1188,"note":1189},"云高级版","25000 积分 + 36000 索引 + 50 成员 + 200 Agent + 1500 QPM","✅","企业级生产",{"plan":1191,"price":632,"limit":1192,"cn_pay":1188,"note":1193},"云定制版","弹性资源 + 深度技术支持 + 专属客户经理","中大型企业","自托管开源免费 \u002F 云版 ¥0-¥599\u002F月","2026-06-18",[1197],"onboarding\u002Ffastgpt-getting-started",[1199],"fastgpt-deep-review",{"power":409,"ux":409,"price":416,"cn_support":416,"stability":409},{"title":12,"description":1145},"agent\u002Fplatform\u002Ffastgpt",[1204,1206,1208,1210,1212],{"title":1205,"url":1096},"FastGPT 官网",{"title":1207,"url":1103},"FastGPT GitHub",{"title":1209,"url":514},"FastGPT 定价页",{"title":1211,"url":74},"FastGPT 2025 测评（南环 AI）",{"title":1213,"url":46},"FastGPT 部署教程（腾讯云）","tools\u002Fagent\u002Fplatform\u002Ffastgpt",[1216,1217,1218,1219,1220],"企业内部知识库（员工手册、规章、流程）","产品文档智能问答（FAQ \u002F 用户手册）","垂直领域知识库（医疗、法律、金融）","数据严格不出网的合规场景","需要精细 RAG 流程编排（重排序、混合检索、阈值调节）","开源知识库问答系统，国内私有部署友好",[1223,1224,1225,1226,1227,1228,1229],"agent-platform","opensource","self-host","rag","china","knowledge-base","labring","2026-06-24","国内企业知识库私有化首选。RAG 召回工程做得很细，可视化调试好用，docker-compose 一键部署。生态插件不如 Dify 丰富。","RC9i5Gg2wX0bUcrC3Lfq5pHAeWLGi9KP2E_gZTefpe4",{"id":1234,"title":680,"alternatives":1235,"api_compatible":6,"body":1238,"category":1143,"chinese_friendly":409,"cover":2217,"description":2218,"domestic":1146,"extension":1147,"faq":6,"free":1146,"github":1520,"languages":2219,"meta":2221,"models":6,"navigation":412,"notSuitable":6,"opensource":412,"path":2222,"pillar":1164,"platforms":2223,"priceTable":2224,"pricing":2241,"published":1195,"relatedPlaybooks":6,"relatedReviews":6,"score":2242,"self_host":412,"seo":2243,"seoTitle":2244,"slug":14,"sources":2245,"stem":2257,"suitable":6,"tagline":2258,"tags":2259,"updated":1230,"verdict":2263,"website":2164,"__hash__":2264},"tools\u002Ftools\u002Fagent\u002Fplatform\u002Fdify.md",[15,1202,1236,1237],"agent\u002Fplatform\u002Fn8n","agent\u002Fplatform\u002Flangflow",{"type":21,"value":1239,"toc":2195},[1240,1242,1260,1265,1267,1271,1274,1342,1345,1349,1352,1366,1383,1387,1390,1401,1405,1413,1424,1428,1431,1434,1438,1447,1505,1512,1516,1524,1573,1576,1596,1600,1603,1663,1669,1673,1757,1760,1789,1791,1921,1938,1976,1978,2050,2052,2054,2074,2076,2105,2107,2156,2158,2187,2192],[24,1241,27],{"id":26},[29,1243,1245,1250],{"className":1244},[32,33,34],[36,1246,1247,1249],{},[39,1248,41],{}," Dify 是开源 LLMOps 平台的事实标准。GitHub 13 万 star、累计 100 万+ 生产 app（据 chatforest.com 2026 评测引用 Dify 官方数据），把\"可视化工作流编排 + RAG 知识库 + Agent + MCP 协议\"打包成一个 Docker Compose 能跑起来的东西。",[36,1251,1252,1253,1256,1257,1259],{},"最大价值是 ",[39,1254,1255],{},"完全开源 + 模型不挑食","——同一个工作流里同时调 OpenAI、Anthropic、Ollama 本地、DeepSeek、Qwen 都行。代价是部署比 ",[44,1258,686],{"href":685}," 折腾，新手得读 1-2 小时文档。",[67,1261,1262],{},[36,1263,1264],{},"来源说明：本文基于 docs.dify.ai 官方文档、langgenius\u002Fdify GitHub 仓库、第三方评测（besthub.dev \u002F chatforest.com \u002F joshuaopolko.com \u002F zhihu 知名专栏）综合归纳。版本号会变，部署要求请以官方最新文档为准。",[24,1266,80],{"id":80},[82,1268,1270],{"id":1269},"可视化工作流chatflow-workflow","可视化工作流（Chatflow + Workflow）",[36,1272,1273],{},"Dify 把 LLM 应用拆成两种\"应用类型\"：",[190,1275,1276,1289],{},[193,1277,1278],{},[196,1279,1280,1283,1286],{},[199,1281,1282],{},"类型",[199,1284,1285],{},"适合场景",[199,1287,1288],{},"编排范式",[206,1290,1291,1304,1316,1329],{},[196,1292,1293,1298,1301],{},[211,1294,1295],{},[39,1296,1297],{},"Chatbot",[211,1299,1300],{},"简单对话机器人",[211,1302,1303],{},"prompt + tools",[196,1305,1306,1310,1313],{},[211,1307,1308],{},[39,1309,541],{},[211,1311,1312],{},"自主多步任务",[211,1314,1315],{},"ReAct \u002F Function Calling",[196,1317,1318,1323,1326],{},[211,1319,1320],{},[39,1321,1322],{},"Chatflow",[211,1324,1325],{},"对话型工作流（多轮 + 分支）",[211,1327,1328],{},"节点 DAG，带聊天上下文",[196,1330,1331,1336,1339],{},[211,1332,1333],{},[39,1334,1335],{},"Workflow",[211,1337,1338],{},"单次输入→输出（API 模式）",[211,1340,1341],{},"节点 DAG，无对话状态",[36,1343,1344],{},"节点类型覆盖：LLM、知识检索、HTTP 请求、代码执行（Python \u002F JS）、条件分支、迭代、变量聚合、参数提取、问题分类——满足\"用拖拽实现可观测的 LLM pipeline\"。",[82,1346,1348],{"id":1347},"rag-知识库","RAG 知识库",[36,1350,1351],{},"内置完整 RAG 链路：",[146,1353,1354,1357,1360,1363],{},[93,1355,1356],{},"上传文档（PDF \u002F Word \u002F Markdown \u002F 网页）",[93,1358,1359],{},"自动分块 + embedding（可配置分段策略和 embedding 模型）",[93,1361,1362],{},"混合检索（向量 + 全文 + 重排）",[93,1364,1365],{},"引用溯源（回答末尾自动附原文片段）",[36,1367,1368,1369,1374,1375,1378,1379,1382],{},"注意：根据 ",[44,1370,1373],{"href":1371,"rel":1372},"https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F1887141987838309480",[48],"知乎 LLM 实战笔记 2025-03 对比"," 的实测，Dify ",[39,1376,1377],{},"社区版默认是基础语义检索","，企业版才解锁多路召回 + 重排。RAG 极致精度场景仍推荐 ",[44,1380,12],{"href":1381},"\u002Fagent\u002Fplatform\u002Ffastgpt.html","（实测准确率高 10+ 个百分点），Dify 胜在工作流而非纯 RAG。",[82,1384,1386],{"id":1385},"模型生态40-提供商","模型生态：40+ 提供商",[36,1388,1389],{},"Dify 通过插件市场接入主流模型——OpenAI、Anthropic、Google Gemini、Azure、AWS Bedrock、Cohere、xAI、DeepSeek、Qwen、智谱、文心、豆包、月之暗面、Ollama、LM Studio、Replicate、Together AI、OpenRouter……几乎你能数出来的 LLM 提供商都在。",[36,1391,1392,1393,1395,1396,1400],{},"国产模型原生支持（不像 ",[44,1394,12],{"href":1381}," 需要 ",[44,1397,1399],{"href":1398},"\u002Fcoding\u002Fapi\u002Fone-api.html","OneAPI"," 中转），是 Dify 在国内 toB 场景流行的关键。",[82,1402,1404],{"id":1403},"mcp-协议支持","MCP 协议支持",[36,1406,1407,1408,1412],{},"Dify 较早接入了 ",[44,1409,1411],{"href":1410},"\u002Fwiki\u002Fmcp.html","MCP（Model Context Protocol）","，工作流可以直接调 MCP Server 暴露的 tools。意味着你可以让 Dify 工作流：",[90,1414,1415,1418,1421],{},[93,1416,1417],{},"通过 MCP 调本地 PostgreSQL \u002F SQLite",[93,1419,1420],{},"通过 MCP 调 GitHub \u002F Slack \u002F Linear",[93,1422,1423],{},"通过 MCP 调自家内部系统（写一个 MCP Server 即可）",[82,1425,1427],{"id":1426},"api-first","API-first",[36,1429,1430],{},"每个 app 自动暴露 REST API，参数和返回结构自动生成 OpenAPI Schema。集成到自家产品里不需要写包装代码，给前端 \u002F 微信小程序 \u002F 飞书机器人调用都方便。",[24,1432,1433],{"id":1433},"价格与运行成本",[82,1435,1437],{"id":1436},"云版difyai","云版（dify.ai）",[36,1439,1440,1441,1446],{},"根据 ",[44,1442,1445],{"href":1443,"rel":1444},"https:\u002F\u002Fwww.tooljunction.io\u002Fai-tools\u002Fdify-ai",[48],"tooljunction.io 2026 评测"," 引用的官方定价：",[190,1448,1449,1460],{},[193,1450,1451],{},[196,1452,1453,1455,1457],{},[199,1454,526],{},[199,1456,529],{},[199,1458,1459],{},"主要限制",[206,1461,1462,1472,1483,1494],{},[196,1463,1464,1467,1469],{},[211,1465,1466],{},"Sandbox",[211,1468,554],{},[211,1470,1471],{},"200 次模型调用，1 app，5MB 知识库",[196,1473,1474,1477,1480],{},[211,1475,1476],{},"Professional",[211,1478,1479],{},"$59\u002F月起",[211,1481,1482],{},"5000 调用\u002F月，多 app，50MB 知识库",[196,1484,1485,1488,1491],{},[211,1486,1487],{},"Team",[211,1489,1490],{},"$159\u002F月起",[211,1492,1493],{},"团队协作、SSO",[196,1495,1496,1499,1502],{},[211,1497,1498],{},"Enterprise",[211,1500,1501],{},"联系销售",[211,1503,1504],{},"定制 SLA、私有云",[36,1506,1507,1508,1511],{},"注意：云版价格只是 Dify 平台费，",[39,1509,1510],{},"模型 API 费用另算","（自带 OpenAI \u002F Anthropic key）。",[82,1513,1515],{"id":1514},"自托管推荐","自托管（推荐）",[36,1517,1518,1523],{},[44,1519,1522],{"href":1520,"rel":1521},"https:\u002F\u002Fgithub.com\u002Flanggenius\u002Fdify",[48],"官方 GitHub 仓库"," 提供 Docker Compose 部署，社区版完全免费可商用：",[366,1525,1527],{"className":368,"code":1526,"language":370,"meta":371,"style":371},"git clone https:\u002F\u002Fgithub.com\u002Flanggenius\u002Fdify.git\ncd dify\u002Fdocker\ncp .env.example .env\ndocker compose up -d\n# 默认 http:\u002F\u002Flocalhost \u002F 端口可在 .env 调整\n",[322,1528,1529,1538,1545,1556,1568],{"__ignoreMap":371},[375,1530,1531,1533,1535],{"class":377,"line":378},[375,1532,389],{"class":388},[375,1534,393],{"class":392},[375,1536,1537],{"class":392}," https:\u002F\u002Fgithub.com\u002Flanggenius\u002Fdify.git\n",[375,1539,1540,1542],{"class":377,"line":385},[375,1541,403],{"class":402},[375,1543,1544],{"class":392}," dify\u002Fdocker\n",[375,1546,1547,1550,1553],{"class":377,"line":399},[375,1548,1549],{"class":388},"cp",[375,1551,1552],{"class":392}," .env.example",[375,1554,1555],{"class":392}," .env\n",[375,1557,1558,1561,1564,1566],{"class":377,"line":409},[375,1559,1560],{"class":388},"docker",[375,1562,1563],{"class":392}," compose",[375,1565,486],{"class":392},[375,1567,489],{"class":402},[375,1569,1570],{"class":377,"line":416},[375,1571,1572],{"class":381},"# 默认 http:\u002F\u002Flocalhost \u002F 端口可在 .env 调整\n",[36,1574,1575],{},"硬件门槛（社区共识，非官方硬性要求）：",[90,1577,1578,1584,1590],{},[93,1579,1580,1583],{},[39,1581,1582],{},"最低","：2 核 4G，纯外接 API 模式",[93,1585,1586,1589],{},[39,1587,1588],{},"推荐","：4 核 8G + 至少 30GB 磁盘（向量数据 + 文件存储）",[93,1591,1592,1595],{},[39,1593,1594],{},"企业","：8 核 16G+，单机日活上千",[82,1597,1599],{"id":1598},"真实-tco","真实 TCO",[36,1601,1602],{},"按一家中小团队 3 年场景估算（基于上面引用的多份评测交叉对比）：",[190,1604,1605,1618],{},[193,1606,1607],{},[196,1608,1609,1612,1615],{},[199,1610,1611],{},"成本项",[199,1613,1614],{},"云版 Professional",[199,1616,1617],{},"自托管",[206,1619,1620,1631,1641,1652],{},[196,1621,1622,1625,1628],{},[211,1623,1624],{},"平台费",[211,1626,1627],{},"~$2,100（3 年）",[211,1629,1630],{},"$0",[196,1632,1633,1636,1638],{},[211,1634,1635],{},"服务器",[211,1637,1630],{},[211,1639,1640],{},"~$50\u002F月 × 36 = $1,800",[196,1642,1643,1646,1649],{},[211,1644,1645],{},"模型 API",[211,1647,1648],{},"与下同",[211,1650,1651],{},"与上同",[196,1653,1654,1657,1660],{},[211,1655,1656],{},"运维人力",[211,1658,1659],{},"0",[211,1661,1662],{},"约 0.2 人月",[36,1664,1665,1668],{},[39,1666,1667],{},"结论","：日活 \u003C 100 用云版省心；> 500 或数据敏感场景自托管 ROI 更好。",[24,1670,1672],{"id":1671},"上手-10-分钟","上手 10 分钟",[366,1674,1676],{"className":368,"code":1675,"language":370,"meta":371,"style":371},"# 1. 自托管（社区版）\ngit clone https:\u002F\u002Fgithub.com\u002Flanggenius\u002Fdify.git\ncd dify\u002Fdocker\ncp .env.example .env\ndocker compose up -d\n\n# 2. 浏览器打开 http:\u002F\u002Flocalhost\n#    首次会让你创建 admin 账号\n\n# 3. 进入\"设置 → 模型供应商\"，配置 OpenAI \u002F 国产模型 API key\n\n# 4. 在主界面\"创建空白应用\"，选 Chatflow 或 Workflow\n# 5. 拖入\"开始 → LLM → 结束\"节点试一下基础 prompt\n# 6. 满意了点右上\"发布\"，自动生成 API endpoint\n",[322,1677,1678,1683,1691,1697,1705,1715,1719,1724,1729,1733,1738,1742,1747,1752],{"__ignoreMap":371},[375,1679,1680],{"class":377,"line":378},[375,1681,1682],{"class":381},"# 1. 自托管（社区版）\n",[375,1684,1685,1687,1689],{"class":377,"line":385},[375,1686,389],{"class":388},[375,1688,393],{"class":392},[375,1690,1537],{"class":392},[375,1692,1693,1695],{"class":377,"line":399},[375,1694,403],{"class":402},[375,1696,1544],{"class":392},[375,1698,1699,1701,1703],{"class":377,"line":409},[375,1700,1549],{"class":388},[375,1702,1552],{"class":392},[375,1704,1555],{"class":392},[375,1706,1707,1709,1711,1713],{"class":377,"line":416},[375,1708,1560],{"class":388},[375,1710,1563],{"class":392},[375,1712,486],{"class":392},[375,1714,489],{"class":402},[375,1716,1717],{"class":377,"line":422},[375,1718,413],{"emptyLinePlaceholder":412},[375,1720,1721],{"class":377,"line":436},[375,1722,1723],{"class":381},"# 2. 浏览器打开 http:\u002F\u002Flocalhost\n",[375,1725,1726],{"class":377,"line":441},[375,1727,1728],{"class":381},"#    首次会让你创建 admin 账号\n",[375,1730,1731],{"class":377,"line":447},[375,1732,413],{"emptyLinePlaceholder":412},[375,1734,1735],{"class":377,"line":455},[375,1736,1737],{"class":381},"# 3. 进入\"设置 → 模型供应商\"，配置 OpenAI \u002F 国产模型 API key\n",[375,1739,1740],{"class":377,"line":464},[375,1741,413],{"emptyLinePlaceholder":412},[375,1743,1744],{"class":377,"line":469},[375,1745,1746],{"class":381},"# 4. 在主界面\"创建空白应用\"，选 Chatflow 或 Workflow\n",[375,1748,1749],{"class":377,"line":9},[375,1750,1751],{"class":381},"# 5. 拖入\"开始 → LLM → 结束\"节点试一下基础 prompt\n",[375,1753,1754],{"class":377,"line":5},[375,1755,1756],{"class":381},"# 6. 满意了点右上\"发布\"，自动生成 API endpoint\n",[24,1758,1759],{"id":1759},"国内使用注意事项",[146,1761,1762,1768,1774,1780],{},[93,1763,1764,1767],{},[39,1765,1766],{},"云版 dify.ai 直连国内访问稳定但需要付款","——支持国际信用卡 \u002F Stripe",[93,1769,1770,1773],{},[39,1771,1772],{},"自托管 + 国产模型"," = 完全国内闭环，是 Dify 在国内最大优势",[93,1775,1776,1779],{},[39,1777,1778],{},"Docker 镜像拉取","：国内可能慢，建议配 Docker registry 镜像（阿里云 \u002F 网易）",[93,1781,1782,1785,1786,1788],{},[39,1783,1784],{},"数据合规","：完全自托管时，数据零外泄；某些金融 \u002F 政府客户因此从 ",[44,1787,686],{"href":685}," 迁到 Dify",[24,1790,663],{"id":663},[190,1792,1793,1813],{},[193,1794,1795],{},[196,1796,1797,1799,1801,1805,1809],{},[199,1798,672],{},[199,1800,680],{},[199,1802,1803],{},[44,1804,686],{"href":685},[199,1806,1807],{},[44,1808,12],{"href":1381},[199,1810,1811],{},[44,1812,1030],{"href":1029},[206,1814,1815,1828,1840,1854,1867,1880,1894,1907],{},[196,1816,1817,1819,1821,1823,1825],{},[211,1818,221],{},[211,1820,1188],{},[211,1822,726],{},[211,1824,1188],{},[211,1826,1827],{},"✅（fair-code）",[196,1829,1830,1832,1834,1836,1838],{},[211,1831,736],{},[211,1833,1188],{},[211,1835,726],{},[211,1837,1188],{},[211,1839,1188],{},[196,1841,1842,1845,1847,1850,1852],{},[211,1843,1844],{},"上手难度",[211,1846,763],{},[211,1848,1849],{},"★★☆☆☆ 最简单",[211,1851,763],{},[211,1853,748],{},[196,1855,1856,1859,1861,1863,1865],{},[211,1857,1858],{},"工作流编排",[211,1860,742],{},[211,1862,748],{},[211,1864,763],{},[211,1866,742],{},[196,1868,1869,1872,1874,1876,1878],{},[211,1870,1871],{},"RAG 精度",[211,1873,748],{},[211,1875,763],{},[211,1877,742],{},[211,1879,784],{},[196,1881,1882,1885,1887,1889,1892],{},[211,1883,1884],{},"模型生态",[211,1886,742],{},[211,1888,748],{},[211,1890,1891],{},"★★★☆☆（OneAPI 中转）",[211,1893,748],{},[196,1895,1896,1899,1901,1903,1905],{},[211,1897,1898],{},"中文场景",[211,1900,748],{},[211,1902,742],{},[211,1904,748],{},[211,1906,763],{},[196,1908,1909,1912,1914,1917,1919],{},[211,1910,1911],{},"字节生态绑定",[211,1913,726],{},[211,1915,1916],{},"✅（飞书\u002F抖音深度集成）",[211,1918,726],{},[211,1920,726],{},[36,1922,1923,1925,1926,1931,1932,1937],{},[39,1924,838],{},"（基于 ",[44,1927,1930],{"href":1928,"rel":1929},"https:\u002F\u002Fwww.besthub.dev\u002Farticles\u002Fcoze-vs-dify-vs-fastgpt-which-ai-agent-platform-fits-your-needs-fa59cf97b798",[48],"BestHub 2025-07"," 和 ",[44,1933,1936],{"href":1934,"rel":1935},"https:\u002F\u002Fwww.cnblogs.com\u002Fuulucias\u002Fp\u002F19449008",[48],"博客园 2026-01"," 两份选型指南综合）：",[90,1939,1940,1946,1953,1960,1967],{},[93,1941,1942,1945],{},[39,1943,1944],{},"数据必须不出内网 + 工作流复杂"," → Dify",[93,1947,1948,858,1951],{},[39,1949,1950],{},"个人 \u002F 小团队 \u002F 快速原型 + 字节生态",[44,1952,686],{"href":685},[93,1954,1955,858,1958],{},[39,1956,1957],{},"核心场景就是企业知识库 QA",[44,1959,12],{"href":1381},[93,1961,1962,858,1965],{},[39,1963,1964],{},"重点是连接外部 SaaS（Slack \u002F Notion \u002F 数据库）",[44,1966,1030],{"href":1029},[93,1968,1969,858,1972],{},[39,1970,1971],{},"要画图式表达 LangChain pipeline",[44,1973,1975],{"href":1974},"\u002Fagent\u002Fplatform\u002Flangflow.html","Langflow",[24,1977,888],{"id":888},[90,1979,1980,1986,2002,2013,2026,2032,2038,2044],{},[93,1981,1982,1985],{},[39,1983,1984],{},"社区版与企业版差距比想象大","：多路召回 \u002F 重排序 \u002F 单点登录 \u002F 审计日志都在企业版。社区版做生产前心里要有数。",[93,1987,1988,896,1994,1997,1998,2001],{},[39,1989,1990,1993],{},[322,1991,1992],{},".env"," 文件改完忘 restart",[322,1995,1996],{},"docker compose down && up -d","，不是 ",[322,1999,2000],{},"restart","——后者不重新加载 env。",[93,2003,2004,896,2007,2012],{},[39,2005,2006],{},"大版本升级会破坏数据库 schema",[44,2008,2011],{"href":2009,"rel":2010},"https:\u002F\u002Fdocs.dify.ai\u002Fzh-hans",[48],"官方升级文档"," 有详细 migration 步骤，跨大版本（如 0.x → 1.x）务必先备份 PostgreSQL 卷。生产环境强烈建议跑 staging 完整验证后再升。",[93,2014,2015,2018,2019,2021,2022,2025],{},[39,2016,2017],{},"RAG 文件大小社区版默认 15MB","：根据上述知乎实测，超过会失败。改 ",[322,2020,1992],{}," 的 ",[322,2023,2024],{},"UPLOAD_FILE_SIZE_LIMIT"," 并重启容器。",[93,2027,2028,2031],{},[39,2029,2030],{},"代码节点的 Sandbox 性能差","：内置代码执行节点跑在隔离容器里启动慢、内存小。生产高频用建议改成 HTTP 节点调外部服务。",[93,2033,2034,2037],{},[39,2035,2036],{},"工作流\"迭代节点\"循环上限","：默认 10 次，复杂 ReAct agent 容易撞天花板，需要在节点设置里调高。",[93,2039,2040,2043],{},[39,2041,2042],{},"Dify Plugin 系统是新东西","：1.0 后引入的 Plugin 体系替代了原来的 Tools\u002FModels 配置方式，老教程可能已过时——以最新官方文档为准。",[93,2045,2046,2049],{},[39,2047,2048],{},"国内 Docker 拉取镜像慢","：先配国内 registry，否则首次 pull 可能要 30+ 分钟。",[24,2051,968],{"id":967},[36,2053,971],{},[90,2055,2056,2059,2062,2065,2068,2071],{},[93,2057,2058],{},"中大型企业 LLM 中台建设",[93,2060,2061],{},"需要私有化部署（金融 \u002F 医疗 \u002F 政府）",[93,2063,2064],{},"想做\"AI 工作流即产品\"的开发团队",[93,2066,2067],{},"同时需要 RAG + Agent + Workflow 三件套",[93,2069,2070],{},"想用国产模型 + 国际模型混合编排",[93,2072,2073],{},"已经接受 Docker + 一定运维投入",[36,2075,994],{},[90,2077,2078,2084,2090,2093,2099],{},[93,2079,2080,2081,2083],{},"纯个人玩家做对话机器人（",[44,2082,686],{"href":685}," 更快）",[93,2085,2086,2087,2089],{},"只想做企业知识库 QA（",[44,2088,12],{"href":1381}," RAG 更专）",[93,2091,2092],{},"团队完全没运维能力（云版还行，自托管会踩坑）",[93,2094,2095,2096,2098],{},"需要深度对接字节飞书 \u002F 抖音（",[44,2097,686],{"href":685}," 原生）",[93,2100,2101,2102,2104],{},"工作流核心是连接 100+ SaaS（",[44,2103,1030],{"href":1029}," 节点更全）",[24,2106,1015],{"id":1015},[90,2108,2109,2119,2133,2148],{},[93,2110,1020,2111,1023,2113,1023,2115,1023,2117],{},[44,2112,686],{"href":685},[44,2114,12],{"href":1381},[44,2116,1030],{"href":1029},[44,2118,1975],{"href":1974},[93,2120,2121,2122,1023,2124,1023,2126,1023,2129],{},"概念基础：",[44,2123,1051],{"href":1050},[44,2125,1037],{"href":1036},[44,2127,2128],{"href":1410},"MCP",[44,2130,2132],{"href":2131},"\u002Fwiki\u002Ffunction-calling.html","Function Calling",[93,2134,2135,2136,1023,2140,1023,2144,1023,2146],{},"模型选型：",[44,2137,2139],{"href":2138},"\u002Fmodels\u002Fgpt-5.html","GPT-5",[44,2141,2143],{"href":2142},"\u002Fmodels\u002Fclaude-sonnet-4.html","Claude Sonnet 4",[44,2145,1058],{"href":1057},[44,2147,1066],{"href":1065},[93,2149,1077,2150,1023,2154],{},[44,2151,2153],{"href":2152},"\u002Fwiki\u002Ffine-tuning-vs-rag.html","Fine-tuning vs RAG",[44,2155,1081],{"href":1080},[24,2157,1088],{"id":1088},[90,2159,2160,2166,2172,2177,2184],{},[93,2161,1093,2162],{},[44,2163,2164],{"href":2164,"rel":2165},"https:\u002F\u002Fdify.ai",[48],[93,2167,2168,2169],{},"中文文档：",[44,2170,2009],{"href":2009,"rel":2171},[48],[93,2173,1100,2174],{},[44,2175,1520],{"href":1520,"rel":2176},[48],[93,2178,2179,2180],{},"官方定价：",[44,2181,2182],{"href":2182,"rel":2183},"https:\u002F\u002Fdify.ai\u002Fpricing",[48],[93,2185,2186],{},"第三方评测：tooljunction.io \u002F chatforest.com \u002F besthub.dev \u002F joshuaopolko.com \u002F 知乎 LLM 实战笔记",[36,2188,2189,2190,1120],{},"本卡片由 AIHO 编辑部根据官方公开资料与第三方评测整理。所有事实点均标注来源；如发现版本号 \u002F 价格 \u002F 功能与最新官方信息不一致，请通过 ",[44,2191,1119],{"href":1119},[1122,2193,2194],{},"html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html pre.shiki code .sJ8bj, html code.shiki .sJ8bj{--shiki-default:#6A737D;--shiki-dark:#6A737D}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"title":371,"searchDepth":399,"depth":399,"links":2196},[2197,2198,2205,2210,2211,2212,2213,2214,2215,2216],{"id":26,"depth":385,"text":27},{"id":80,"depth":385,"text":80,"children":2199},[2200,2201,2202,2203,2204],{"id":1269,"depth":399,"text":1270},{"id":1347,"depth":399,"text":1348},{"id":1385,"depth":399,"text":1386},{"id":1403,"depth":399,"text":1404},{"id":1426,"depth":399,"text":1427},{"id":1433,"depth":385,"text":1433,"children":2206},[2207,2208,2209],{"id":1436,"depth":399,"text":1437},{"id":1514,"depth":399,"text":1515},{"id":1598,"depth":399,"text":1599},{"id":1671,"depth":385,"text":1672},{"id":1759,"depth":385,"text":1759},{"id":663,"depth":385,"text":663},{"id":888,"depth":385,"text":888},{"id":967,"depth":385,"text":968},{"id":1015,"depth":385,"text":1015},{"id":1088,"depth":385,"text":1088},"\u002Fimg\u002Ftools\u002Fdify.webp","Dify 2026 真实评测：开源 LLMOps 与 AI Agent 平台，集工作流编排、RAG 知识库、Agent、MCP 和多模型接入于一体。本文对比 Coze、FastGPT、n8n，整理自托管部署、云版价格、适合团队和避坑建议。",[1149,1150,2220],"ja",{},"\u002Ftools\u002Fagent\u002Fplatform\u002Fdify",[1166,1167,1168,1560],[2225,2229,2233,2237],{"plan":2226,"price":1630,"features":2227,"notes":2228},"Self-hosted（开源版）","Docker 一键部署 + 全部核心功能（工作流 \u002F RAG \u002F Agent \u002F MCP）+ 接任意模型 API","私有部署 \u002F 完全免费 \u002F Apache 2.0",{"plan":2230,"price":1630,"features":2231,"notes":2232},"Cloud Sandbox（免费云）","官方托管试水档，含基础调用配额","免运维 \u002F 试水 POC",{"plan":2234,"price":1479,"features":2235,"notes":2236},"Cloud Professional","更高调用额度 + 团队协作 + 商用支持","商用云首选",{"plan":2238,"price":2239,"features":2240,"notes":1501},"Cloud Team \u002F Enterprise","Custom","更大配额 + SLA + 私有部署支持 + 合规","云版 SaaS（免费档 \u002F Professional $59\u002F月起） + 开源自托管完全免费",{"power":416,"ux":409,"price":416,"cn_support":409,"stability":409},{"title":680,"description":2218},"Dify 评测 2026：开源 LLMOps 与 AI Agent 平台，自托管指南",[2246,2248,2250,2252,2254],{"title":2247,"url":2009},"Dify 官方文档（中文）",{"title":2249,"url":1520},"Dify GitHub",{"title":2251,"url":2182},"Dify 官方定价",{"title":2253,"url":1928},"Coze vs Dify vs FastGPT 选型",{"title":2255,"url":2256},"Dify Self-Hosted Guide 2026","https:\u002F\u002Fjoshuaopolko.com\u002Fdify-self-hosted-guide","tools\u002Fagent\u002Fplatform\u002Fdify","开源 LLMOps 平台，私有部署 Agent 首选",[1223,1224,1225,1226,2260,2261,2262],"workflow","llmops","mcp","想私有部署、想接全球任意模型，Dify 是答案。比 Coze 工程化、上手陡一点；比 FastGPT 工作流强、RAG 略弱。","0LsPnncTjEa2rm_jLZvwR6L0Zv1hFNDKVRmDyloL7uo",1783006599848]