README.md

Deploying the Eval backend to Hugging Face Spaces (Docker)

Prerequisites

• A Hugging Face account and a new Space (set type to Docker)
• Your OpenAI API key (if using OpenAI evaluation)

Files included

• app.py: Flask API exposing endpoints under /api
• models.py, utils.py, evaluator.py, chunking_utils.py, config.py
• requirements.txt: minimal dependencies for CPU deployment
• Dockerfile: builds and runs the Flask API on port 7860

Environment variables

• OPENAI_API_KEY: set in the Space Secrets if Config.EVALUATION_MODEL is 'openai'
• FLASK_SECRET_KEY (optional)
• DATABASE_URL (optional; defaults to sqlite:///evalai_new.db)

Build and run (locally)

docker build -t eval-backend ./Eval
docker run -p 7860:7860 -e OPENAI_API_KEY=YOUR_KEY eval-backend

Deploy on Spaces

1. Create a new Space, select Docker as the SDK
2. Upload the contents of the Eval/ folder to the Space root
3. In the Space Settings, add a Secret named OPENAI_API_KEY
4. Spaces will build using the Dockerfile and expose the API at / on port 7860

API quick check

curl -s https://<your-space>.hf.space/ | jq