Dify开源平台，用于构建和编排具备内置 RAG 与 Agent 工作流的 LLM 应用。

5.0 (5)

审阅者 Daniel Nikulshyn·更新 2026年5月

Open Source LLM Platform RAG Agents Workflow Builder Self-Hosted Developer Tools Observability

概览

Dify 是一个开源的开发平台，旨在简化团队构建、部署和管理由大型语言模型驱动的应用程序的方式。它结合了可视化工作流构建器、提示工程工具和检索增强生成（RAG）管道，使开发者能够在无需拼接多个服务的情况下，从原型快速迭代到生产环境。平台支持多种模型提供商，内置用于工具使用和多步推理的 Agent 框架，并提供可观测性功能，以监控使用量、成本和质量。由于可自行部署，Dify 适合需要对数据、基础设施和合规性进行控制的组织，同时仍能受益于现代 LLMOps 工具链。典型的使用场景包括内部知识助理、客服机器人、内容生成流水线，以及需要将私有数据与商业或开源模型结合的定制 AI 产品。

主要功能

可视化 LLM 工作流构建器
检索增强生成（RAG）管道
具备工具集成的 Agent 框架
提示词管理与版本控制
多模型提供商支持
使用分析与可观测性

价格

模型: Free
分类: 杰与拉常系安全目分
评分: 5.0 / 5 (5)

使用场景

构建基于 RAG 的知识助理

利用内置的检索增强生成（RAG）管道和知识库工具，创建能够基于内部文档回答问题的聊天机器人。

可视化原型设计并部署 LLM 应用

在可视化构建器中设计提示词和多步 LLM 工作流，随后无需集成多个独立服务即可从原型直接进入生产。

编排多步 AI Agent

利用具备工具集成的 Agent 框架，构建在多步骤中进行推理并调用外部工具完成复杂任务的助理。

自行托管 LLM 应用以满足合规要求

在自有基础设施上部署 Dify，保持对数据的控制并满足合规需求，同时仍可使用多种 LLM 提供商。

优点 & 缺点

优点

开源且支持自行部署
可视化工作流与提示编排
内置 RAG 与知识库工具
支持众多 LLM 提供商和模型
活跃社区和频繁更新

缺点

自行部署需要技术搭建和维护
高级功能有学习曲线
部分企业功能需付费订阅

评测

5.0

5 个评分的平均值。

登录以留下评测。

Camille Laurent

May 3, 2026

Skeptical, then convinced

I went in skeptical — most tools in this space overpromise. It actually delivers on agent framework with tool integrations, and visual workflow and prompt orchestration caught me off guard. Self-hosting requires technical setup and maintenance is why this isn't a perfect score, still, I'd recommend giving it a real trial.

Esther Adeyemi

Mar 14, 2026

Solid for our team

We rolled this out across the team last quarter and open-source with self-hosting options. Usage analytics and observability fits neatly into how we already work, and usage analytics and observability removed a step we used to do by hand. Self-hosting requires technical setup and maintenance, which is the main caveat, but it has held up under daily use.

Pierre Dubois

Dec 9, 2025

Does the job

Pretty happy overall. Multi-model provider support just works and active community and frequent updates. Self-hosting requires technical setup and maintenance can be annoying, but no dealbreakers — I'd recommend it to a friend without hesitating.

Nadia Petrova

Jul 24, 2025

Skeptical, then convinced

I went in skeptical — most tools in this space overpromise. It actually delivers on prompt management and versioning, and built-in RAG and knowledge base tools caught me off guard. Advanced features have a learning curve is why this isn't a perfect score, still, I'd recommend giving it a real trial.

Liam O’Connor

Jun 13, 2025

Skeptical, then convinced

I went in skeptical — most tools in this space overpromise. It actually delivers on usage analytics and observability, and supports many LLM providers and models caught me off guard. Advanced features have a learning curve is why this isn't a perfect score, still, I'd recommend giving it a real trial.

问答

What are common use cases for Dify, and how steep is the learning curve?

Typical use cases include internal knowledge assistants and customer-facing applications built on RAG and agent workflows. Basic prototyping is approachable via the visual builder, but advanced features like agent tool use, prompt versioning, and observability have a learning curve.

Asked by Diego Fernández · Sep 18, 2025

Which LLM providers and models does Dify support?

Dify offers multi-model provider support, allowing you to connect a wide range of LLM providers and switch between models within the same workflows. This flexibility is useful for comparing outputs, optimizing costs, or meeting provider-specific compliance requirements.

Asked by Carlos Mendoza · Aug 21, 2025

Can I self-host Dify, and what trade-offs come with that?

Yes, Dify is open-source and supports self-hosting, which gives you control over data, infrastructure, and compliance. The trade-off is that self-hosting requires technical setup and ongoing maintenance, so teams without DevOps capacity may prefer a managed deployment.

Asked by Camille Laurent · Jul 19, 2025