Skip to content

Back to the catalog

Dify
github.comrepo

Dify

Dify is an open-source LLM application development platform combining a visual workflow builder, RAG pipeline, agent capabilities (with MCP server and tool integration), model management across hundreds of providers, and LLMOps observability. It offers both a hosted cloud service and self-hosted Community/Enterprise editions deployable via Docker Compose, aiming to take LLM apps from prototype to production with a backend-as-a-service API layer.

TL;DR Open-source platform for building LLM apps with visual workflows, RAG, agents, and model management.
Open original ↗
Document

What the summary omits

The summary names the platform's scope but leaves out the concrete adoption facts an engineer needs before committing: the licence, the install path, the minimum hardware, and which observability backends are actually wired in.

Licence (check before building on it)

The repository is not plain Apache 2.0. It is licensed under the Dify Open Source License, based on Apache 2.0 with additional conditions. Anyone planning to embed, resell, or offer Dify as a hosted service should read those additional conditions rather than assuming standard Apache permissions.

How to run it

Self-hosting is via Docker Compose. Minimum requirements stated: CPU ≥ 2 cores, RAM ≥ 4 GiB, and Docker Compose v2.24.0 or later.

cd dify
cd docker
cp .env.example .env
docker compose up -d

Then open http://localhost/install to run initialization. Source-based deployment has a separate guide for contributors.

Details the summary generalizes

  • Observability is not generic "LLMOps" — it integrates specific backends: Opik, Langfuse, and Arize Phoenix.
  • Agents run in a per-agent sandbox where they execute commands, install software, and handle files; the agent can be described in natural language and self-build, then be wired tools from the Dify Marketplace, MCP servers, or your own APIs, and deployed as a chat app or as a workflow step.
  • RAG ships out-of-box text extraction for PDFs, PPTs, and other common formats.
  • Model coverage spans GPT, Mistral, Llama3, and any OpenAI-API-compatible model across dozens of inference providers plus self-hosted.

Editions

  • Cloud: zero-setup hosted, sandbox plan includes 200 free GPT-4 calls.
  • Community: self-hosted, the Docker path above.
  • Enterprise: adds SSO, RBAC, and support SLAs — gated behind a sales contact form, not self-serve.

Security issues should go to security@dify.ai, not GitHub issues.

Organisations named here

  • Langfuse
  • Llama3
  • Mistral

Technologies named here

  • Dify
  • GPT-4
  • Mistral
  • Docker Compose
  • Llama 3
  • Opik
  • Langfuse
  • Arize Phoenix
  • Dify Marketplace
  • MCP servers
  • Model Context Protocol

Techniques named here

  • Workflow
  • RAG Pipeline
  • Prompt IDE
  • Agent
  • LLMOps

Related in llm-apps