Qwen API to OpenAI Standard Proxy

This project provides a lightweight, single-file proxy server designed to run on Deno. It translates standard OpenAI API requests into the proprietary format used by chat.qwen.ai and transforms the responses back into the standard OpenAI format.

This allows you to use OpenAI-compatible clients with the Qwen (Tongyi Qianwen) chat service.

✨ Features

• OpenAI Compatibility: Acts as a drop-in replacement for the OpenAI API base URL.
• Request Conversion: Translates OpenAI chat completion requests to the Qwen format.
• Stream Transformation: Converts Qwen's Server-Sent Events (SSE) stream to the OpenAI format in real-time.
• Model Variants: Automatically creates special model variants like qwen-max-thinking and qwen-max-search based on the upstream model capabilities.
• Token Rotation: Supports multiple upstream API_KEYs and rotates through them for each request.
• Authentication: Secure your proxy endpoint with an OPENAI_API_KEY.
• Zero Dependencies (Deno): Runs as a single script on Deno Deploy or locally without needing npm install.

🚀 Deployment (Deno Deploy)

1. Create a Deno Deploy Project:

   • Go to Deno Deploy and create a new "Playground" project.
   • Copy the entire content of main.ts and paste it into the editor.

2. Set Environment Variables: In your Deno Deploy project's "Settings" > "Environment Variables" section, add the following:

   • OPENAI_API_KEY: (Recommended) Your secret key for clients to access this proxy. (e.g., sk-my-secret-key-12345)
   • API_KEY: Your Qwen account token(s). You can provide multiple tokens separated by commas for rotation. (e.g., ey...abc,ey...def)
   • SSXMOD_ITNA: The special cookie value required for the upstream API.

3. Run: The script will be deployed and run automatically. Your endpoint URL will be provided by Deno Deploy.

💻 Local Usage

1. Save the file as main.ts.

2. Set environment variables in your terminal:

   export OPENAI_API_KEY="your_secret_proxy_key"
   export API_KEY="your_qwen_token"
   export SSXMOD_ITNA="your_cookie_value"

3. Run the script:

   deno run --allow-net --allow-env main.ts

   The server will start on http://localhost:8000.

⚙️ Configuration

The server is configured via the following environment variables:

Variable	Description	Required	Example
OPENAI_API_KEY	A secret Bearer token to protect your proxy endpoint. If not set, the proxy will be open to the public.	No	sk-my-secret-key-12345
API_KEY	Your Qwen account token(s) for the upstream API. Separate multiple keys with a comma for rotation.	Yes	ey...abc,ey...def
SSXMOD_ITNA	The required ssxmod_itna cookie value from chat.qwen.ai.	Yes	mqUxRDBD...DYAEDBYD74G+DDeDixGm...

🔌 API Endpoints

• GET /v1/models
  • Retrieves a list of available Qwen models, including special variants like -thinking, -search, and -image.
• POST /v1/chat/completions
  • The main endpoint for chat. It accepts standard OpenAI chat completion requests and supports streaming responses.