---
title: Digital Doctors Assistant ML API
emoji: 🏥
colorFrom: blue
colorTo: purple
sdk: docker
pinned: false
license: mit
---

# Digital Doctors Assistant ML API

A FastAPI-based machine learning service that provides two prediction models for healthcare applications:

1. **Risk Assessment Model** - Predicts patient risk levels (Low/Medium/High) based on vitals and health metrics
2. **Treatment Outcome Model** - Predicts treatment success probability based on patient data and medication

## Features

- RESTful API endpoints for ML predictions
- Interactive web interface for easy testing
- ONNX Runtime for fast inference
- Pre-trained models hosted on HuggingFace
- Docker support for easy deployment

## Quick Start

### Using HuggingFace Space

Simply visit the Space URL and use the web interface to make predictions.

### Run Locally with Docker

```bash
docker run -it -p 7860:7860 \
  -e HUGGINGFACE_TOKEN="your_token_here" \
  registry.hf.space/your-username-your-space-name:latest
```

### Local Setup (Without Docker)

1. Install dependencies:
```bash
pip install -r requirements.txt
```

2. Set your HuggingFace token:
```bash
# Windows PowerShell
$env:HUGGINGFACE_TOKEN="your_token_here"

# Windows CMD
set HUGGINGFACE_TOKEN=your_token_here

# Linux/Mac
export HUGGINGFACE_TOKEN=your_token_here
```

3. Run the server:
```bash
uvicorn ml:app --host 0.0.0.0 --port 7860
```

4. Open your browser:
- Web Interface: http://localhost:7860
- API Docs: http://localhost:7860/docs
- Health Check: http://localhost:7860/health

## API Endpoints

### POST /predict/risk
Predict patient risk level based on health metrics.

**Request Body:**
```json
{
  "age": 45,
  "bmi": 28.5,
  "systolic_bp": 140,
  "diastolic_bp": 90,
  "chronic_conditions": "diabetes,hypertension",
  "severity_score": 7.5
}
```

**Response:**
```json
{
  "success": true,
  "model": "risk_assessment",
  "prediction": "High",
  "confidence": 0.85,
  "probabilities": {
    "Low": 0.05,
    "Medium": 0.10,
    "High": 0.85
  }
}
```

### POST /predict/treatment
Predict treatment outcome success probability.

**Request Body:**
```json
{
  "patient_age": 55,
  "severity_score": 6.5,
  "compliance_rate": 0.85,
  "medication": "Metformin",
  "condition": "Diabetes Type 2"
}
```

**Response:**
```json
{
  "success": true,
  "model": "treatment_outcome",
  "prediction": 1,
  "success_probability": 78.5,
  "confidence": 0.78,
  "probabilities": {
    "failure": 0.22,
    "success": 0.78
  }
}
```

## Supported Medications

- Paracetamol
- Ibuprofen
- Amoxicillin
- Ciprofloxacin
- Metformin
- Lisinopril
- Amlodipine
- Omeprazole

## Supported Conditions

- Common Cold
- Influenza
- Pneumonia
- Bronchitis
- Hypertension
- Diabetes Type 2
- Migraine
- Gastroenteritis

## Environment Variables

- `HUGGINGFACE_TOKEN` - Required. Your HuggingFace access token for downloading models

## Technology Stack

- FastAPI - Web framework
- ONNX Runtime - ML inference
- NumPy - Numerical computing
- HuggingFace Hub - Model hosting
- Pydantic - Data validation
- Uvicorn - ASGI server

## License

MIT