Fetching metadata from the HF Docker repository...
-
Refactor Gemini to official SDK + lazy singleton + report caching
-
Deploy Breathe — Respiratory Disease Prediction System
Full-stack healthcare application with:
- Next.js 16 frontend (glassmorphism UI, dark-mode-free, responsive)
- Python ML backend (TensorFlow CNN + LightGBM) via Flask on port 5001
- Audio disease prediction (5 classes: Bronchial, Asthma, COPD, Healthy, Pneumonia)
- Asthma risk assessment (10 clinical features)
- Environmental monitoring (reverse geocoding + AQI/weather)
- AI Doctor: Google Gemini-powered clinical decision support report with PDF export
- Demo audio library (persistent Prisma + SQLite CRUD)
- ML models stored via Git LFS
Docker configuration:
- Single container running Next.js (port 7860) + Python ML service (port 5001)
- Python 3.11 + Node.js 20 base image
- Standalone Next.js build for minimal footprint
- Startup script handles DB init, sample seeding, and dual-service launch
Required HF Space Secrets:
- GOOGLE_API_KEY (for AI Doctor)
- DATABASE_URL (optional, defaults to file:/tmp/breathe.db)