README.md

metadata

title: Pneumonia Space
emoji: 🫁
colorFrom: blue
colorTo: green
sdk: docker
app_port: 5000
pinned: false
license: mit

🫁 Pneumonia Risk Assessment API

AI-powered API for assessing pneumonia risk from respiratory audio recordings.

🚀 Features

• HeAR Model Integration: Uses Google's Health Acoustic Representations model
• Risk Scoring: Provides probability-based risk assessment (not diagnostic)
• Fallback System: Uses librosa-based features if HeAR model unavailable
• REST API: Simple Flask endpoint for audio file uploads

🔑 Setup

Hugging Face Authentication

The HeAR model requires Hugging Face authentication. Set your token as an environment variable:

export HF_TOKEN="your_huggingface_token_here"

Or login using the CLI:

huggingface-cli login

Get your token from: https://huggingface.co/settings/tokens

Running Locally

# Install dependencies
pip install -r requirements.txt

# Run the application
python app.py

Using Docker

# Build the image
docker build -t pneumonia-api .

# Run with HF token
docker run -p 5000:5000 -e HF_TOKEN="your_token" pneumonia-api

📡 API Usage

Endpoint: POST /predict_pneumonia

Request: Multipart form data with audio_file

Response:

{
  "filename": "recording.wav",
  "pneumonia_risk_score": 0.7234,
  "risk_level": "High",
  "note": "This is an AI assessment, not a medical diagnosis. Consult a healthcare professional."
}

Example with curl:

curl -X POST -F "audio_file=@recording.wav" http://localhost:5000/predict_pneumonia

⚠️ Disclaimer

This tool provides risk assessment scores, not medical diagnoses. Always consult healthcare professionals for medical decisions.

📄 License

MIT License

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference