Skip to main content
Deploy your PAS2 application to Hugging Face Spaces for a publicly accessible web interface with automatic persistent storage.

​
Prerequisites

Before deploying to Hugging Face Spaces, ensure you have:
  • A Hugging Face account
  • Mistral AI API key
  • OpenAI API key
  • Your PAS2 repository ready for deployment

​
Deployment steps

1

Create a new Space

  1. Go to Hugging Face Spaces
  2. Click “Create new Space”
  3. Choose a name for your Space
  4. Select Gradio as the SDK
  5. Choose your preferred visibility (Public or Private)
2

Add your repository

Connect your PAS2 repository to the Space:
  • Upload your files directly, or
  • Link to your GitHub repository for automatic syncing
Ensure these files are included:
  • pas2.py (main application code)
  • requirements.txt (dependencies)
  • README.md (optional documentation)
3

Configure API keys as secrets

Set your API keys in the Space’s settings to keep them secure:
  1. Navigate to your Space’s Settings tab
  2. Scroll to the Repository secrets section
  3. Add the following secrets:
HF_MISTRAL_API_KEY=your_mistral_api_key_here
HF_OPENAI_API_KEY=your_openai_api_key_here
The application automatically reads these environment variables from lines pas2.py:44-45. The HF_* prefix is used specifically for Hugging Face Spaces secrets.
4

Enable persistent storage

Hugging Face Spaces automatically provides a /data directory for persistent storage. The PAS2 application uses this for the SQLite feedback database.The persistent storage setup is configured in pas2.py:384-386:
self.data_dir = "/data"
self.db_path = os.path.join(self.data_dir, "feedback.db")
This ensures:
  • User feedback survives Space restarts
  • Statistics are preserved long-term
  • No data loss during inactivity periods
If the /data directory is not accessible, the application will fall back to a temporary directory (pas2.py:423-427). However, data in this fallback location will not persist across restarts.
5

Deploy and verify

  1. Commit your changes to trigger the build
  2. Wait for the Space to build (this may take a few minutes)
  3. Once built, your application will be accessible at: 
    https://huggingface.co/spaces/YOUR_USERNAME/YOUR_SPACE_NAME
  4. Test the application by entering a query and verifying the hallucination detection works

​
Environment variable configuration

The application supports multiple API key configurations with the following priority order:
  1. Explicitly passed parameters
  2. Hugging Face Spaces secrets (HF_MISTRAL_API_KEY, HF_OPENAI_API_KEY)
  3. Standard environment variables (MISTRAL_API_KEY, OPENAI_API_KEY)
From pas2.py:44-45: 
self.mistral_api_key = mistral_api_key or os.environ.get("HF_MISTRAL_API_KEY") or os.environ.get("MISTRAL_API_KEY")
self.openai_api_key = openai_api_key or os.environ.get("HF_OPENAI_API_KEY") or os.environ.get("OPENAI_API_KEY")

​
Required dependencies

Your requirements.txt should include: 
gradio
pandas
numpy
mistralai
openai
pydantic
python-dotenv

​
Data persistence details

The SQLite database stores:
  • User query history
  • Original and paraphrased responses
  • Hallucination detection results
  • User feedback ratings
  • Confidence scores and reasoning
Database initialization occurs automatically on first run (pas2.py:390-420), creating the necessary schema in the /data directory.

​
Monitoring your deployment

After deployment, you can:
  • View real-time logs in the Space’s Logs tab
  • Monitor usage statistics through the built-in dashboard
  • Check feedback data persistence by submitting test queries
  • Review API usage to ensure rate limits are not exceeded
The application logs all operations using Python’s logging module with INFO level. Check the logs if you encounter API initialization errors or database issues.

​
Troubleshooting

​
API keys not found

If you see “API key is required” errors:
  • Verify secrets are set correctly in Space settings
  • Ensure secret names match exactly: HF_MISTRAL_API_KEY and HF_OPENAI_API_KEY
  • Restart the Space after adding secrets

​
Database initialization fails

If the database fails to initialize:
  • Check the logs for permission errors
  • Verify the Space has persistent storage enabled
  • The application will attempt fallback to temporary directory

​
Build failures

If the Space fails to build:
  • Verify requirements.txt includes all dependencies
  • Check for syntax errors in pas2.py
  • Review build logs for specific error messages

Build docs developers (and LLMs) love

Get started for freeTalk to us