This middleware application provides an OpenAI-compatible endpoint for Dify instances, allowing you to use OpenAI API clients with Dify's API. It maintains conversation context and supports both streaming and non-streaming responses.

• Node.js >= 18.0.0
• A running Dify instance with API access
• Your Dify API key

1. Clone this repository:

git clone https://github.com/yourusername/dify2openai.git cd dify2openai

2. Install dependencies:

npm install

3. Configure environment variables:

# Copy the example environment file cp .env.example .env # Edit .env with your settings nano .env # or use your preferred editor

Required environment variables:

• DIFY_API_URL: Your Dify API URL (e.g., http://your-dify-instance/v1)
• DIFY_API_KEY: Your Dify API key (found in your Dify application settings)
• PORT: Port number for the middleware server (default: 3000)
• LOG_LEVEL: Logging verbosity level (default: info)
  • error: Only errors
  • warn: Errors and warnings
  • info: Basic operational info (default)
  • debug: Detailed debugging information

1. Start the server:

# Production mode npm start # Development mode with auto-reload npm run dev

2. The middleware will run on http://localhost:3000 (or your configured PORT)

3. Use any OpenAI API client by pointing it to your middleware URL. Examples:

import OpenAI from 'openai'; const openai = new OpenAI({ baseURL: 'http://localhost:3000/v1', apiKey: 'not-needed' // The middleware uses Dify's API key }); const completion = await openai.chat.completions.create({ messages: [{ role: 'user', content: 'Hello!' }], model: 'gpt-3.5-turbo', // Model name is ignored, Dify's configured model is used stream: true // Supports both streaming and non-streaming });

from openai import OpenAI client = OpenAI( base_url="http://localhost:3000/v1", api_key="not-needed" # The middleware uses Dify's API key ) completion = client.chat.completions.create( messages=[{"role": "user", "content": "Hello!"}], model="gpt-3.5-turbo", # Model name is ignored stream=True # Supports both streaming and non-streaming )

curl http://localhost:3000/v1/chat/completions \ -H "Content-Type: application/json" \ -d '{  "model": "gpt-3.5-turbo",  "messages": [{"role": "user", "content": "Hello!"}]  }'

• OpenAI API compatibility:
  • Supports the chat completions endpoint
  • Works with any OpenAI API client
  • Maintains conversation context
• Response formats:
  • Supports both streaming and non-streaming responses
  • Matches OpenAI's response format
• Error handling:
  • Graceful error handling and reporting
  • OpenAI-compatible error responses
• Development friendly:
  • Easy setup with environment variables
  • Development mode with auto-reload
  • Configurable logging levels
  • Detailed debug information

The middleware uses a leveled logging system with timestamps:

[2024-01-23T12:34:56.789Z] [INFO] Dify2OpenAI middleware running on port 3000 

Log levels (from least to most verbose):

1. error: Critical issues that need immediate attention
2. warn: Important issues that don't affect core functionality
3. info: General operational information (default)
4. debug: Detailed information for debugging

Configure the log level in your .env file:

# Set to error, warn, info, or debug LOG_LEVEL=info

Debug logs include:

• Conversation ID tracking
• Message format conversion details
• Request/response information
• Streaming events
• Client connections/disconnections

The middleware exposes a single endpoint that mimics OpenAI's chat completions API:

POST /v1/chat/completions

Request format follows OpenAI's specification:

{ "messages": [ {"role": "user", "content": "Hello!"} ], "model": "gpt-3.5-turbo", // Ignored, uses Dify's model "stream": true, // Optional, defaults to false "user": "user123" // Optional }

• Only supports the chat completions endpoint
• Model selection is ignored (uses Dify's configured model)
• Some OpenAI-specific parameters may be ignored
• Function calling is not supported

The middleware provides a health check endpoint:

GET /health

Returns {"status": "ok"} when the server is running.

The middleware handles various error cases:

• Invalid request format
• Missing/invalid messages
• API errors from Dify
• Network errors
• Server errors

All errors are returned in a format compatible with OpenAI's error responses:

{ "error": { "message": "Error description", "type": "error_type", "status": 400 // HTTP status code } }