MCP & Gọi công cụ

Dify: Collaborative Platform for Agentic Workflows and RAG Applications

langgenius/dify

Dify is an LLM application development platform for visually building workflows, RAG pipelines, and tool-using agents. It combines model management, prompt development, observability, and application APIs in a collaborative workspace that can be used as a hosted service or self-hosted.

★ 151,2KSố sao
⑂ 23,9KFork
925Vấn đề đang mở
TypeScriptNgôn ngữ
NOASSERTIONGiấy phép
Q@project.QualityScoreĐiểm biên tập

Tổng quan

The project is designed to move AI applications from prototyping to production without replacing the underlying development stack. Its primary language is TypeScript, and its interface covers workflow composition, prompt testing, document retrieval, agent configuration, model integration, monitoring, and backend APIs. Deployment options described by the repository include Dify Cloud, self-hosted Community Edition, and enterprise-oriented offerings.

Tính năng chính

  • Visual canvas for building and testing AI workflows
  • RAG pipeline covering document ingestion, extraction, and retrieval
  • Text extraction support for PDFs, presentations, and other common document formats
  • Agents based on LLM Function Calling or ReAct
  • More than 50 built-in agent tools, with support for custom tools
  • Prompt IDE for prompt authoring and model comparison
  • Integration with proprietary, open-source, hosted, and self-hosted model providers
  • LLMOps capabilities for application logs, performance analysis, annotations, and iterative improvement
  • Observability integrations including Opik, Langfuse, and Arize Phoenix
  • APIs for integrating Dify applications with external business logic
  • Cloud, VPC, and self-hosted deployment positioning

Yêu cầu, cài đặt và bắt đầu nhanh

Clone or otherwise obtain the repository, then run the documented Docker Compose quick start: `cd dify`, `cd docker`, `cp .env.example .env`, and `docker compose up -d`. Docker and Docker Compose v2.24.0 or later must already be installed. For configuration changes, edit `docker/.env`; defaults are in `docker/.env.example`, while optional variables are organized under `docker/envs/`. Source-based development instructions are separate from this quick start.

Cách sử dụng

After Docker Compose starts the services, open http://localhost/install and complete initialization. Build applications in the dashboard by selecting models, defining prompts or visual workflows, configuring datasets and retrieval where needed, and adding agents or tools. Dify applications can be consumed through their corresponding APIs. Hosted access is also available through Dify Cloud.

Khả năng tương thích của mô hình và trường hợp sử dụng

The README states support for hundreds of proprietary and open-source LLMs from dozens of inference providers, including GPT, Mistral, Llama 3, self-hosted solutions, and models exposed through an OpenAI API-compatible interface. Claude and OpenAI also appear in the repository topics. Exact versions, feature parity, credentials, and provider-specific constraints are not stated in the repository metadata. MCP appears as a repository topic, but the supplied README excerpt does not describe its implementation or compatibility details.

Ghi chú về giấy phép và rủi ro

The README says the repository uses the Dify Open Source License, based on Apache 2.0 with additional conditions. GitHub metadata reports the SPDX value as NOASSERTION, and no license URL is stated in the repository metadata. Users should review the repository's LICENSE file before adoption, redistribution, hosted use, or commercial deployment.

ToolAI metadata-only listing: repository metadata is public; editorial content and manual verification are still pending.

Phát hành và bảo trì

No release version, changelog entry, or release notes were supplied in the repository record.

MemPalace

mempalace/mempalace

★ 58KPython

litellm

berriai/litellm

★ 56,7KPython

QwenPaw

agentscope-ai/qwenpaw

★ 33,9KPython

Composio

composiohq/composio

★ 29,5KTypeScript