PROVIDER - Visual LLM app and workflow platform

Dify

Dify is a visual LLM app and workflow platform assembling apps on 1 canvas, bundling RAG and knowledge bases with hundreds of model integrations, self-hostable via 1 Docker Compose stack and shipping logs plus annotation; visual limits and high-concurrency tuning are the gaps.

AI Platform China 2026.11.19
4.3pts

Detailed Notes

Pros 5 items

Drag-and-drop workflow orchestration lets non-engineers assemble an app on 1 canvas
Ships a built-in RAG pipeline and knowledge base, so documents become Q&A in 3 steps
Connects to hundreds of models including 1 class of OpenAI-compatible self-hosted endpoints
Self-hosts the community edition with 1 Docker Compose stack on an intranet
Includes logs, annotation and operations features to iterate on 1 application continuously

Cons 4 items

Complex business logic hits visual limits and needs code nodes
High concurrency requires extra deployment and performance work
Community and enterprise editions differ in capability
Upgrades involve database migrations and must follow the guide

Service Overview

Dify is an open source LLM application platform that combines prompt orchestration, RAG, agents and operations in a visual workbench.

In the AI platform category it is the LLMOps platform, letting product and business teams ship AI apps without writing front ends.

Key Strengths

  • Visual orchestration: one canvas
  • Built-in RAG: knowledge in 3 steps
  • Broad model support: hundreds of endpoints
  • Self-hostable: 1 Compose stack
  • Operations loop: logs and annotations

Product Ecosystem

Platform capabilities

Workflow and chatflow orchestration, knowledge bases and retrieval pipelines, agents and tool calling, prompt management, logs and annotation plus API publishing and permissions.

Integrations

Supports OpenAI-compatible endpoints and mainstream model services, callable from your own systems via API.

Typical scenarios

Enterprise knowledge Q&A, support and ticket assistants, content generation pipelines and internal AI tools.

Limitations

  • Expression ceiling: code nodes needed
  • Performance: concurrency needs design
  • Edition drift: community vs enterprise
  • Upgrade cost: database migrations

Use Cases

  • Fast AI app delivery (★★★★★): least manual effort
  • Enterprise knowledge bases (★★★★★): RAG out of the box
  • Private intranet deployment (★★★★☆): community edition
  • Deep customisation needs (★★★☆☆): drop to code

Pricing

Item Price Notes
Community $0 Open source self-hosting, bring servers
Professional Monthly subscription Cloud edition and advanced features
Enterprise Contact sales Private deployment and support

FAQ

Can it run offline?

Yes, the community edition supports offline Docker Compose deployment.

Does it support self-hosted models?

Yes, OpenAI-compatible endpoints such as vLLM or Ollama can be attached.

How is it different from LangChain?

Dify is a productised visual platform; LangChain is a code-level framework.

Will upgrades lose data?

Run the documented migration scripts and back up the database first.

Code Examples

# Start the community edition self-hosted
git clone https://github.com/langgenius/dify.git
cd dify/docker && cp .env.example .env && docker compose up -d
# Call a published app through the API
curl -X POST 'http://localhost/v1/chat-messages' \
  -H "Authorization: Bearer APP_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"inputs":{},"query":"How do I request an invoice?","user":"u-1","response_mode":"blocking"}'

Community & Resources

Visit Dify Website → https://dify.ai/
Visit Website ↗