Azure OpenAI Proxy (FastAPI)

This Python project implements a proxy server that forwards JSON requests to the Azure OpenAI API, handling Azure AD authentication and providing an interactive API documentation UI via Swagger.

Table of Contents

• Azure OpenAI Proxy (FastAPI)
  • Table of Contents
  • Prerequisites
  • Installation
  • Configuration
  • Usage
  • API Documentation
  • Environment Variables
  • Endpoints
  • Client setting
  • License

Prerequisites

• Python 3.8 or newer
• An Azure subscription with an OpenAI resource deployed
• Valid Azure AD credentials with permissions to access the OpenAI resource

Installation

1. Clone the repository:

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

2. Create and activate a virtual environment (optional but recommended):

   python3 -m venv .venv
   source .venv/bin/activate

3. Install the Python dependencies:

   pip install --upgrade pip
   pip install -r requirements.txt

Configuration

1. Copy the example environment file and set your credentials:

   cp .env-example .env

2. Edit .env and populate the following variables:

   AZURE_OPENAI_ENDPOINT=https://<your-azure-openai-endpoint>.openai.azure.com/
   AZURE_TENANT_ID=<your-azure-tenant-id>
   AZURE_CLIENT_ID=<your-azure-client-id>
   AZURE_CLIENT_SECRET=<your-azure-client-secret>
   PORT=8899
   

Usage

Start the FastAPI server with Uvicorn:

uvicorn main:app --host 0.0.0.0 --port ${PORT:-8899} --reload

The proxy will listen on http://localhost:8899 by default.

API Documentation

Open your browser and navigate to:

• Swagger UI: http://localhost:8899/docs
• ReDoc: http://localhost:8899/redoc

Environment Variables

• AZURE_TENANT_ID: Azure AD tenant ID.
• AZURE_CLIENT_ID: Service principal client ID.
• AZURE_CLIENT_SECRET: Service principal client secret.
• AZURE_OPENAI_ENDPOINT: Base URL of your Azure OpenAI resource.
• PORT: (Optional) HTTP port for Uvicorn, default is 8899.

Endpoints

• GET /check: A simple endpoint to check if the proxy is working.

  • Request: None.
  • Response: JSON object with status and response message.

• POST /v1/chat/completions: Forward a JSON body to the Azure OpenAI chat completion API with optional streaming support.

  • Request: JSON body with parameters compliant with the Azure OpenAI specification, including messages, model, temperature, max_tokens, etc.
  • Response: JSON response from Azure OpenAI, with streaming support if requested.

Client setting

• Example of VSCode Cline

  

  • The API key can be any text

• Open AI URL of VSWizard[https://github.com/dqj1998/VSWizard.git]

  http://localhost:8899/v1/chat/completions

License

Licensed under the MIT License. See LICENSE for details.