Dify#
Dify is an open-source LLM application development platform with a visual orchestration interface that lets developers build AI applications without writing much code.
What It Can Do#
1. Build AI Apps Visually#
- Chat assistants: configure conversation flows via drag-and-drop, with support for multiple LLMs (OpenAI, Claude, local models)
- Workflow orchestration: a visual DAG editor that chains LLM, code, HTTP, conditional-branch, and other nodes
- Agents: configure tool calling so the LLM can plan and execute tasks autonomously
graph LR
A[User input] --> B[Intent detection]
B --> C{Branch}
C -->|Query| D[Knowledge base retrieval]
C -->|Action| E[Tool call]
D --> F[LLM generation]
E --> F
F --> G[Output]
2. Knowledge Base / RAG#
- Upload documents (PDF, Word, Markdown, web pages)
- Automatic chunking, embedding, and storage
- Relevant content is retrieved automatically during conversations to improve answer accuracy
3. Multi-Model Management#
| Supported Models | Examples |
|---|---|
| Commercial APIs | OpenAI, Claude, Gemini, Tongyi Qianwen |
| Open-source models | LLaMA, Qwen, ChatGLM (via Ollama/vLLM) |
| Embedding | text-embedding-3, bge, m3e |
| Rerank | Cohere, bge-reranker |
4. One-Click Deployment#
- Generates a WebApp once built (shareable link)
- Provides an API (REST, can be integrated into your own apps)
- Can be embedded in web pages (iframe / JS SDK)
5. Other Capabilities#
- Prompt template management: version control, A/B testing
- Conversation logs and annotation: review user conversations and annotate them for fine-tuning
- Team collaboration: share apps and knowledge bases across a team
- Self-hosting: one-command Docker deployment, data stays on-premises
Use Cases#
| Scenario | Description |
|---|---|
| Enterprise knowledge base Q&A | Upload internal documents for employee self-service lookup |
| Customer service bot | Connect product docs to answer user questions automatically |
| Internal tools | Orchestrate workflows to automate tasks |
| AI app prototyping | Validate ideas quickly without building from scratch |
Quick Start#
# Docker deployment
git clone https://github.com/langgenius/dify.git
cd dify/docker
docker compose up -d
# Visit http://localhost/install to complete initialization
Website: dify.ai | Repository: github.com/langgenius/dify