Dify

Documentation Index

Fetch the complete documentation index at: https://docs.laozhang.ai/llms.txt

Use this file to discover all available pages before exploring further.

​

Platform Introduction

• 🎨 Visual Workflow Editor - Build complex AI applications via drag-and-drop
• 🔌 Multi-model Support - Support for OpenAI, Anthropic, Google and other providers
• 📝 Prompt Engineering - Built-in powerful prompt template system
• 💬 Conversation Management - Complete conversation history and context management
• 🔄 RAG Support - Integrated retrieval-augmented generation capability
• 🚀 Quick Deployment - Support for API and Web applications
• 📊 Data Analysis - Comprehensive application usage analysis

​

Quick Integration

​

Step 1: Log in to Dify Platform

​

Step 2: Add Model Provider

1

Open Settings

Click the user icon in the upper right corner, select “Settings”

2

Select Model Provider

In the left sidebar, click “Model Provider”

3

Add OpenAI Compatible Provider

Find “OpenAI-API-compatible” in the model provider list, click “Add”

4

Fill In Configuration

Fill in the following configuration:

• Provider Name: Laozhang API (custom name)
• API Base URL: https://api.laozhang.ai/v1
• API Key: Enter your Laozhang API key
• Model Type: Select models you need (completion/chat/embedding)

5

Test Connection

Click “Test” button to ensure connection is successful

6

Save Configuration

Click “Save” button to complete configuration

​

Step 3: Select Model

• Creating new applications
• Configuring workflows
• Setting Agent models

​

Supported Models

​

Text Generation Models (Chat/Completion)

​

Embedding Models

​

Application Scenarios

​

1. Intelligent Customer Service Chatbot

• 24/7 automated responses
• Multi-turn context conversation
• Knowledge base integration
• Intent recognition and routing

• Model: gpt-3.5-turbo (economical and fast)
• RAG: Enable knowledge base retrieval
• Temperature: 0.7
• Max Tokens: 1000

​

2. Document Analysis Assistant

• Automatic document summarization
• Key information extraction
• Q&A based on documents
• Multi-document comparison

• Model: claude-sonnet-4 (long context support)
• RAG: Enable vector database
• Temperature: 0.3 (more accurate)
• Max Tokens: 2000

​

3. Code Generation Tool

• Automatic code generation
• Code explanation and annotation
• Code review
• Bug fixing suggestions

• Model: gpt-4-turbo (strong code capability)
• Temperature: 0.2 (more deterministic)
• Max Tokens: 2000

​

4. Content Creation Platform

• Article writing
• Marketing copy
• Product descriptions
• SEO optimization

• Model: gpt-4-turbo or claude-sonnet-4
• Temperature: 0.8 (more creative)
• Max Tokens: 2000

​

Advanced Features

​

RAG (Retrieval Augmented Generation)

1. Create Knowledge Base
   • Upload documents (PDF, Word, Markdown, etc.)
   • Automatic chunking and vectorization
   • Choose appropriate embedding model
2. Configure Retrieval Strategy
   • Set retrieval quantity (Top K)
   • Set similarity threshold
   • Choose reranking strategy
3. Integration into Application
   • Associate knowledge base in workflow
   • Configure retrieval parameters
   • Test retrieval effectiveness

​

Workflow Orchestration

User Input
    ↓
Intent Recognition (GPT-3.5)
    ↓
Knowledge Base Retrieval (Embedding)
    ↓
Answer Generation (GPT-4)
    ↓
Quality Check (GPT-3.5)
    ↓
Return Result

​

Agent Capabilities

• Tool Calling - Call external APIs
• Multi-step Reasoning - Decompose complex tasks
• Self-reflection - Evaluate and improve outputs
• Memory Management - Maintain long-term memory

​

API Integration

​

Python Example

import requests

# Dify API endpoint
api_url = "https://your-dify-instance.com/v1/chat-messages"

# Request configuration
headers = {
    "Authorization": "Bearer Your Dify API key",
    "Content-Type": "application/json"
}

data = {
    "inputs": {},
    "query": "Hello, please introduce yourself",
    "response_mode": "blocking",
    "user": "user-123"
}

# Send request
response = requests.post(api_url, headers=headers, json=data)
result = response.json()

print(result["answer"])

​

JavaScript Example

const apiUrl = 'https://your-dify-instance.com/v1/chat-messages';

const requestData = {
  inputs: {},
  query: 'Hello, please introduce yourself',
  response_mode: 'blocking',
  user: 'user-123'
};

fetch(apiUrl, {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer Your Dify API key',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify(requestData)
})
.then(response => response.json())
.then(data => {
  console.log(data.answer);
});

​

Streaming Response

import requests

url = "https://your-dify-instance.com/v1/chat-messages"
headers = {
    "Authorization": "Bearer Your Dify API key",
    "Content-Type": "application/json"
}
data = {
    "inputs": {},
    "query": "Write a short story",
    "response_mode": "streaming",
    "user": "user-123"
}

response = requests.post(url, headers=headers, json=data, stream=True)

for line in response.iter_lines():
    if line:
        print(line.decode('utf-8'))

​

Deployment Guide

​

Docker Deployment

# Clone repository
git clone https://github.com/langgenius/dify.git
cd dify/docker

# Copy configuration file
cp .env.example .env

# Edit .env file, add Laozhang API configuration
# OPENAI_API_BASE=https://api.laozhang.ai/v1
# OPENAI_API_KEY=Your Laozhang API key

# Start services
docker-compose up -d

​

Environment Variables

# Database configuration
DB_USERNAME=postgres
DB_PASSWORD=your_password
DB_HOST=db
DB_PORT=5432
DB_DATABASE=dify

# Redis configuration
REDIS_HOST=redis
REDIS_PORT=6379
REDIS_PASSWORD=your_redis_password

# LLM configuration
OPENAI_API_BASE=https://api.laozhang.ai/v1
OPENAI_API_KEY=Your Laozhang API key

# Vector database configuration
VECTOR_STORE=qdrant
QDRANT_URL=http://qdrant:6333

​

Troubleshooting

​

Connection Issues

1. Check if API Base URL is correct: https://api.laozhang.ai/v1
2. Verify API Key validity
3. Ensure network connection is normal
4. Check firewall settings

​

Model Unavailable

1. Verify model ID is correct
2. Check account balance
3. Confirm model is in service scope
4. Try other models

​

RAG Performance Issues

1. Optimize document chunking strategy
2. Adjust retrieval parameters (Top K, similarity threshold)
3. Try different embedding models
4. Add more relevant documents

​

High Latency

1. Switch to a faster model (e.g., GPT-3.5 Turbo)
2. Reduce Max Tokens setting
3. Optimize prompt length
4. Enable caching mechanism

​

Best Practices

​

1. Prompt Engineering

❌ Bad example:
"Write an article"

✅ Good example:
"You are a professional tech blogger. Please write an article about AI applications, requirements:
- Word count: 800-1000 words
- Audience: Technical professionals
- Style: Professional yet approachable
- Include: 3 practical case studies
- End with: Actionable suggestions"

​

2. Cost Optimization

• Use lighter models for simple tasks (GPT-3.5 Turbo)
• Set reasonable Max Tokens limits
• Enable result caching
• Batch process requests
• Regularly review usage

​

3. Security Configuration

• Use environment variables to store API keys
• Enable API rate limiting
• Implement user authentication
• Audit logs
• Data encryption

​

4. Performance Optimization

• Enable caching mechanism
• Use CDN to distribute static resources
• Optimize database queries
• Implement load balancing
• Monitor system performance

​

Further Learning

​

Official Resources

• Dify Official Documentation
• Dify GitHub Repository
• Dify Community Forum

​

Learning Resources

• Dify Quick Start Guide
• Dify Best Practices
• Dify API Documentation

​

Community Support

• Dify Discord Community
• Dify GitHub Discussions