app.py

import os
import requests
import joblib
import logging
import zipfile
import pandas as pd
import numpy as np
import warnings
from flask import Flask, request, jsonify
from flask_cors import CORS

# Suppress sklearn warnings
warnings.filterwarnings('ignore', category=UserWarning, module='sklearn')

# Logging setup
logging.basicConfig(level=logging.INFO)

# Load URLs from environment
MODEL_URLS = {
    "DIABETES_MODEL": os.getenv("DIABETES_MODEL_URL"),
    "SCALER": os.getenv("SCALER_URL"),
    "MULTI_MODEL": os.getenv("MULTI_MODEL_URL"),
}

# Model ZIP filenames
MODEL_PATHS = {
    "DIABETES_MODEL": "finaliseddiabetes_model.zip",
    "SCALER": "finalisedscaler.zip",
    "MULTI_MODEL": "nodiabetes.zip",
}

# Extracted model file names
EXTRACTED_MODELS = {
    "DIABETES_MODEL": "finaliseddiabetes_model.joblib",
    "SCALER": "finalisedscaler.joblib",
    "MULTI_MODEL": "nodiabetes.joblib",
}

TMP_DIR = "/tmp"

app = Flask(__name__)
CORS(app, supports_credentials=True)

@app.route('/')
def index():
    return """
    <h1>Welcome to the Diabetes Health Predictor API 👋</h1>
    <p>This Hugging Face Space provides health risk predictions including diabetes, hypertension, stroke, and cardiovascular conditions.</p>
    <p>Use the <code>/predict</code> endpoint via POST request to get started with your health insights!</p>
    """

def download_model(url, zip_filename):
    zip_path = os.path.join(TMP_DIR, zip_filename)
    if not url:
        logging.error(f"URL for {zip_filename} is missing!")
        return False
    try:
        response = requests.get(url, allow_redirects=True)
        if response.status_code == 200:
            with open(zip_path, 'wb') as f:
                f.write(response.content)
            logging.info(f"Downloaded {zip_filename} successfully.")
            return True
        else:
            logging.error(f"Failed to download {zip_filename}. HTTP Status: {response.status_code}")
            return False
    except Exception as e:
        logging.error(f"Error downloading {zip_filename}: {e}")
        return False

def extract_if_needed(zip_filename, extracted_filename):
    zip_path = os.path.join(TMP_DIR, zip_filename)
    extracted_path = os.path.join(TMP_DIR, extracted_filename)
    if os.path.exists(extracted_path):
        logging.info(f"{extracted_filename} already exists. Skipping extraction.")
        return True
    if not os.path.exists(zip_path):
        logging.error(f"Zip file missing: {zip_path}")
        return False
    try:
        with zipfile.ZipFile(zip_path, 'r') as zip_ref:
            zip_ref.extractall(TMP_DIR)
        logging.info(f"Extracted {zip_filename} to {TMP_DIR}")
        return True
    except Exception as e:
        logging.error(f"Error extracting {zip_filename}: {e}")
        return False

def load_model(model_filename):
    model_path = os.path.join(TMP_DIR, model_filename)
    if not os.path.exists(model_path):
        logging.error(f"Model file not found: {model_path}")
        return None
    try:
        model = joblib.load(model_path)
        logging.info(f"Loaded {model_filename} successfully.")
        return model
    except Exception as e:
        logging.error(f"Error loading {model_filename}: {e}")
        return None

def initialize_models():
    models = {}
    for model_key, zip_filename in MODEL_PATHS.items():
        extracted_filename = EXTRACTED_MODELS[model_key]
        url = MODEL_URLS.get(model_key)

        zip_path = os.path.join(TMP_DIR, zip_filename)
        if not os.path.exists(zip_path):
            if not download_model(url, zip_filename):
                continue

        if not extract_if_needed(zip_filename, extracted_filename):
            continue

        models[model_key] = load_model(extracted_filename)

    return models

models = initialize_models()

FEATURE_ORDER = [
    'Pregnancies', 'Glucose', 'BloodPressure', 'Insulin',
    'BMI', 'DiabetesPedigreeFunction', 'Age'
]

def validate_input(value, input_type=float, min_value=0, max_value=None):
    try:
        value = input_type(value)
        if value < min_value:
            return None
        if max_value is not None and value > max_value:
            return None
        return value
    except (ValueError, TypeError):
        return None

def validate_blood_pressure(systolic, diastolic):
    systolic = validate_input(systolic, float, 0, 300)
    diastolic = validate_input(diastolic, float, 0, 200)
    if systolic is None or diastolic is None:
        return None, None
    return systolic, diastolic

def validate_gender(gender):
    if isinstance(gender, str) and gender.lower() in ['male', 'female']:
        return 1 if gender.lower() == 'male' else 0
    return None

def calculate_diabetes_pedigree(family_history, first_degree=0, second_degree=0):
    if not family_history:
        return 0.0
    genetic_contribution = (first_degree * 0.5) + (second_degree * 0.25)
    return min(genetic_contribution, 1.0)

def get_multi_condition_predictions(model, df):
    try:
        predictions = model.predict(df)[0]
        probs_list = model.predict_proba(df)
        return {
            'hypertension': bool(predictions[0]),
            'cardiovascular': float(probs_list[1][0][1]),
            'stroke': float(probs_list[2][0][1]),
            'diabetes': float(probs_list[3][0][1])
        }
    except Exception as e:
        logging.error(f"Error in multi-condition prediction: {str(e)}")
        return None

def get_diabetes_prediction(model, df):
    try:
        prediction = model.predict(df)[0]
        probability = float(model.predict_proba(df)[0][1] * 100)
        return 'Diabetes' if prediction else 'No Diabetes', probability
    except Exception as e:
        logging.error(f"Error in diabetes prediction: {str(e)}")
        return None, 0.0

@app.route('/health', methods=['GET'])
def health_check():
    return jsonify({'status': 'healthy', 'message': 'Service is running'})

@app.route('/predict', methods=['POST'])
def predict_health():
    try:
        data = request.get_json()
        logging.info(f"Received data: {data}")
        if not data:
            return jsonify({'status': 'error', 'error': 'Invalid JSON payload'}), 400

        gender = validate_gender(data.get('gender'))
        if gender is None:
            return jsonify({'status': 'error', 'error': 'Invalid gender value. Must be \"male\" or \"female\"'}), 400

        systolic, diastolic = validate_blood_pressure(data.get('systolic'), data.get('diastolic'))
        if systolic is None or diastolic is None:
            return jsonify({'status': 'error', 'error': 'Invalid blood pressure values'}), 400

        age = validate_input(data.get('age'), float, 0, 120)
        glucose = validate_input(data.get('glucose'), float, 0, 1000)
        bmi = validate_input(data.get('bmi'), float, 0, 100)

        if any(v is None for v in [age, glucose, bmi]):
            return jsonify({'status': 'error', 'error': 'Invalid values for age, glucose, or BMI'}), 400

        use_multi_condition = systolic < 90 or diastolic < 60

        if use_multi_condition:
            df_multi = pd.DataFrame([{
                'Age': age,
                'Gender': gender,
                'Systolic_bp': systolic,
                'Diastolic_bp': diastolic,
                'Glucose': glucose,
                'BMI': bmi
            }])

            results = get_multi_condition_predictions(models['MULTI_MODEL'], df_multi)
            if results is None:
                return jsonify({'status': 'error', 'error': 'Error in multi-condition prediction'}), 500

            return jsonify({
                'status': 'success',
                'model': 'multi-condition',
                'predictions': {
                    'hypertension': results['hypertension'],
                    'cardiovascular_risk': results['cardiovascular'],
                    'stroke_risk': results['stroke'],
                    'diabetes_risk': results['diabetes']
                }
            })

        pregnancies = validate_input(data.get('pregnancies', 0 if gender == 1 else None), float, 0, 20)
        insulin = validate_input(data.get('insulin'), float, 0, 1000)

        family_history = data.get('family_history', False)
        first_degree = validate_input(data.get('first_degree_relatives', 0), float, 0, 10)
        second_degree = validate_input(data.get('second_degree_relatives', 0), float, 0, 20)

        diabetes_pedigree = calculate_diabetes_pedigree(
            family_history,
            first_degree if first_degree is not None else 0,
            second_degree if second_degree is not None else 0
        )

        if any(v is None for v in [pregnancies, insulin]):
            return jsonify({'status': 'error', 'error': 'Invalid values for pregnancies or insulin'}), 400

        df_diabetes = pd.DataFrame([{
            'Pregnancies': pregnancies,
            'Glucose': glucose,
            'BloodPressure': systolic,
            'Insulin': insulin,
            'BMI': bmi,
            'DiabetesPedigreeFunction': diabetes_pedigree,
            'Age': age
        }])

        df_diabetes = df_diabetes[FEATURE_ORDER]
        df_scaled = models['SCALER'].transform(df_diabetes)
        prediction, probability = get_diabetes_prediction(models['DIABETES_MODEL'], df_scaled)

        return jsonify({
            'status': 'success',
            'model': 'diabetes',
            'prediction': prediction,
            'probability': probability,
            'risk_level': 'HIGH' if probability > 70 else 'MODERATE' if probability > 40 else 'LOW'
        })

    except Exception as e:
        logging.error(f"Error: {e}")
        return jsonify({'status': 'error', 'error': str(e)}), 500

if __name__ == '__main__':
    app.run(host="0.0.0.0", port=7860)