OpenAI API Router with OpenRouter Backend

This project is an API router that intercepts multimodal requests (Image/Video), even when the multimodal data is a link in the text body of the request. It intelligently processes requests that include image or video links, logs all activity, tracks costs, and routes requests to appropriate models via OpenRouter.

Installation Guide

1. Clone the repository:

   git clone https://github.com/dsetareh/OpenRouter-MultiModal-Proxy
   cd OpenRouter-MultiModal-Proxy

2. Configure environment variables:

   • Copy .env.example to .env in the root of the project:

     cp .env.example .env

   • Edit the .env file and add your OPENROUTER_API_KEY and any other configurations. The docker-compose.yml file is set up to pass these variables to the container.

3. Build and run the application using Docker Compose: Open your terminal in the project root directory and run:

   docker-compose up -d --build

• The application will be available at http://<docker_host>:8000.

Core Features

• OpenAI API Compatibility:
  • Supports POST /v1/chat/completions for chat-based interactions.
  • Supports POST /v1/completions for legacy text completion.
• Intelligent Media Processing:
  • Detects image and video URLs within request payloads.
  • Processes images (downloads and converts to base64 if necessary).
  • Processes videos by downloading, extracting audio, transcribing audio (using a configurable faster-whisper model and device), and extracting keyframes.
  • Transforms requests with media content into a format suitable for multimodal models on OpenRouter.
• OpenRouter Integration:
  • Routes requests to specified or default models on OpenRouter.
  • Utilizes OPENROUTER_API_KEY for authentication.
  • Allows configuration of HTTP-Referer and X-Title for OpenRouter requests.
• Comprehensive Logging:
  • Logs full request and response bodies in structured JSON format to router.log.json.
  • Includes timestamps, unique request IDs, endpoint details, and processing durations.
  • Implements log rotation.
• Cost Tracking & Analytics:
  • Stores detailed request information in a local SQLite database (.db).
  • Tracks model usage, tokens, costs (as reported by OpenRouter), latencies, and potential errors.
• Web UI for Tracking:
  • Provides a web interface at /ui/tracking to view and filter logged API requests.
  • Allows pagination, sorting, and searching of request data.

Supported API Endpoints

OpenAI

• POST /v1/chat/completions
• POST /v1/completions

Internal

• GET /ui/tracking_data
• GET /ui/tracking