Skip to content
1337skills 1337skills

Dify Cheat Sheet

Overview

Dify is an open-source platform for building LLM-powered applications. It provides a visual workflow editor for orchestrating complex AI pipelines, built-in RAG engine, model management across multiple providers, and tools for prompt engineering. Dify supports chatbots, agents, text generators, and workflow-based applications out of the box.

The platform abstracts away the complexity of integrating LLMs, vector databases, and external tools. It offers a web-based studio for non-developers while exposing comprehensive APIs for programmatic control. Dify supports OpenAI, Anthropic, local models via Ollama, and dozens of other providers through a unified interface.

Installation

git clone https://github.com/langgenius/dify.git
cd dify/docker
cp .env.example .env
docker compose up -d

Access the web UI at http://localhost/install to set up the admin account.

From Source

git clone https://github.com/langgenius/dify.git
cd dify

# Start backend
cd api
cp .env.example .env
pip install -r requirements.txt
flask db upgrade
flask run --host 0.0.0.0 --port 5001

# Start frontend (new terminal)
cd web
npm install
npm run build
npm run start

Environment Variables

# .env key settings
SECRET_KEY=your-secret-key
OPENAI_API_KEY=sk-...
CONSOLE_WEB_URL=http://localhost:3000
SERVICE_API_URL=http://localhost:5001
DB_USERNAME=postgres
DB_PASSWORD=difypassword
DB_HOST=db
DB_PORT=5432
DB_DATABASE=dify
REDIS_HOST=redis
REDIS_PORT=6379
VECTOR_STORE=weaviate

Core Concepts

Application Types

TypeDescriptionUse Case
ChatbotConversational AI with memoryCustomer support, Q&A
AgentAutonomous tool-using AIResearch, task execution
Text GeneratorSingle prompt-responseContent creation, summarization
WorkflowMulti-step orchestrationComplex data pipelines

API Usage

# Create a chat message
curl -X POST 'http://localhost/v1/chat-messages' \
  -H 'Authorization: Bearer app-YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "inputs": {},
    "query": "What is RAG?",
    "response_mode": "streaming",
    "conversation_id": "",
    "user": "user-123"
  }'

# Upload a file for RAG
curl -X POST 'http://localhost/v1/files/upload' \
  -H 'Authorization: Bearer app-YOUR_API_KEY' \
  -F 'file=@document.pdf' \
  -F 'user=user-123'

# Get conversation history
curl -X GET 'http://localhost/v1/messages?conversation_id=conv-123&user=user-123' \
  -H 'Authorization: Bearer app-YOUR_API_KEY'

# Send feedback
curl -X POST 'http://localhost/v1/messages/msg-id/feedbacks' \
  -H 'Authorization: Bearer app-YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{"rating": "like", "user": "user-123"}'

Knowledge Base (RAG)

Creating a Knowledge Base

# Create knowledge base via API
curl -X POST 'http://localhost/v1/datasets' \
  -H 'Authorization: Bearer dataset-YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "name": "Product Documentation",
    "description": "Company product docs",
    "indexing_technique": "high_quality",
    "permission": "only_me"
  }'

# Add document to knowledge base
curl -X POST 'http://localhost/v1/datasets/{dataset_id}/document/create_by_text' \
  -H 'Authorization: Bearer dataset-YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "name": "product-guide.txt",
    "text": "Your document content here...",
    "indexing_technique": "high_quality",
    "process_rule": {
      "mode": "automatic"
    }
  }'

# Upload file to knowledge base
curl -X POST 'http://localhost/v1/datasets/{dataset_id}/document/create_by_file' \
  -H 'Authorization: Bearer dataset-YOUR_API_KEY' \
  -F 'file=@manual.pdf' \
  -F 'data={"indexing_technique":"high_quality","process_rule":{"mode":"automatic"}}'

Retrieval Settings

ParameterDescriptionDefault
search_methodkeyword_search, semantic_search, hybrid_searchsemantic_search
top_kNumber of chunks to retrieve3
score_thresholdMinimum relevance score (0-1)0.5
reranking_modelModel for reranking resultsNone

Configuration

Model Provider Setup

# Supported model providers
providers:
  - openai        # GPT-4, GPT-3.5
  - anthropic     # Claude 3/4
  - azure_openai  # Azure-hosted OpenAI
  - huggingface   # Open-source models
  - ollama        # Local models
  - xinference    # Local model serving
  - tongyi        # Alibaba Cloud
  - zhipuai       # GLM models
  - minimax       # MiniMax models

Docker Compose Customization

# docker-compose.override.yml
services:
  api:
    environment:
      - LOG_LEVEL=DEBUG
      - CELERY_BROKER_URL=redis://redis:6379/1
    deploy:
      resources:
        limits:
          memory: 4G

  worker:
    environment:
      - CELERY_WORKER_AMOUNT=4

  web:
    ports:
      - "3000:3000"

  weaviate:
    environment:
      - PERSISTENCE_DATA_PATH=/var/lib/weaviate
    volumes:
      - weaviate_data:/var/lib/weaviate

Nginx Reverse Proxy

server {
    listen 80;
    server_name dify.example.com;

    location / {
        proxy_pass http://localhost:3000;
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
    }

    location /v1 {
        proxy_pass http://localhost:5001/v1;
        proxy_set_header Host $host;
        proxy_read_timeout 300s;
    }

    location /console/api {
        proxy_pass http://localhost:5001/console/api;
        proxy_set_header Host $host;
    }
}

Advanced Usage

Workflow Nodes

Node TypePurpose
LLMCall language model
Knowledge RetrievalQuery RAG knowledge base
CodeExecute Python/JavaScript
HTTP RequestCall external APIs
ConditionBranch logic
Variable AggregatorCombine variables
Template TransformJinja2 text templates
Question ClassifierRoute by intent
IterationLoop over arrays

Custom Tools

# Register an external tool via OpenAPI schema
curl -X POST 'http://localhost/console/api/workspaces/current/tool-provider/api/create' \
  -H 'Authorization: Bearer YOUR_TOKEN' \
  -H 'Content-Type: application/json' \
  -d '{
    "schema_type": "openapi",
    "schema": "openapi: 3.0.0\ninfo:\n  title: Weather\n  version: 1.0\npaths:\n  /weather:\n    get:\n      summary: Get weather\n      parameters:\n        - name: city\n          in: query\n          required: true\n          schema:\n            type: string",
    "credentials": {
      "api_key": "your-weather-api-key"
    }
  }'

Webhook Integration

import requests

DIFY_API_KEY = "app-YOUR_API_KEY"
BASE_URL = "http://localhost/v1"

def send_message(query, conversation_id=""):
    response = requests.post(
        f"{BASE_URL}/chat-messages",
        headers={"Authorization": f"Bearer {DIFY_API_KEY}"},
        json={
            "inputs": {},
            "query": query,
            "response_mode": "blocking",
            "conversation_id": conversation_id,
            "user": "webhook-user"
        }
    )
    return response.json()

result = send_message("Summarize latest sales data")
print(result["answer"])

Troubleshooting

IssueSolution
Docker containers fail to startCheck .env file, ensure ports 80, 5432, 6379 are free
Model provider connection errorVerify API keys in Settings > Model Providers
Knowledge base indexing stuckCheck worker logs: docker compose logs worker
Out of memory during embeddingReduce batch size or increase Docker memory limits
SSO login not workingVerify CONSOLE_WEB_URL matches your actual URL
Slow response timesEnable model load balancing, check Redis connection
File upload failsCheck UPLOAD_FILE_SIZE_LIMIT env var (default 15MB)
Database migration errorsRun flask db upgrade manually in the API container
# View logs for debugging
docker compose logs -f api
docker compose logs -f worker
docker compose logs -f web

# Restart all services
docker compose restart

# Reset database (WARNING: destroys data)
docker compose down -v
docker compose up -d