main.py

# FILE: main.py
# Install requirements first: 
# pip install fastapi uvicorn scikit-learn joblib numpy

from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
import joblib
import numpy as np

# 1. Initialize the App
app = FastAPI(title="Diabetes Prediction API", version="1.0.0")

# Enable CORS for Flutter app
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# 2. Load the Trained Model
# (Ensure the .pkl file is in the same folder)
model = joblib.load('diabetes_model_professional.pkl')

# 3. Define the Data Structure (Input Validation)
class PatientData(BaseModel):
    pregnancies: int
    glucose: float
    blood_pressure: float
    skin_thickness: float
    insulin: float
    bmi: float
    pedigree: float
    age: int

@app.get("/")
def home():
    return {"status": "online", "message": "Diabetes Prediction API is running"}

@app.post("/predict")
def predict_diabetes(data: PatientData):
    # Prepare data for the model
    features = [[
        data.pregnancies, 
        data.glucose, 
        data.blood_pressure, 
        data.skin_thickness, 
        data.insulin, 
        data.bmi, 
        data.pedigree, 
        data.age
    ]]
    
    # Get the raw probability (0 to 1)
    probability = model.predict_proba(features)[0][1]
    
    # APPLY THE PROFESSIONAL THRESHOLD (0.35)
    # We use the lower threshold we discovered in your analysis to minimize missed cases
    threshold = 0.35
    is_diabetic = probability >= threshold
    
    return {
        "prediction": "Diabetic" if is_diabetic else "Healthy",
        "risk_score": round(float(probability) * 100, 2), # Return as percentage
        "risk_level": "High" if probability > 0.6 else "Moderate" if is_diabetic else "Low",
        "alert": "Patient requires further clinical screening." if is_diabetic else "Patient appears healthy."
    }

# To start: uvicorn main:app --reload --host 0.0.0.0