app.py

from fastapi import FastAPI, Request
from fastapi.templating import Jinja2Templates
from pydantic import BaseModel
import pickle
import uvicorn
import numpy as np
import pandas as pd
from fastapi.middleware.cors import CORSMiddleware

# Load Model
model = pickle.load(open("model.pkl", "rb"))

app = FastAPI()

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


templates = Jinja2Templates(directory="templates")

# Define Request Model
class InsuranceInput(BaseModel):
    age: int
    sex: str
    bmi: float
    children: int
    smoker: str
    region: str

# Preprocess Inputs
region_map = {'Northwest': 0, 'Northeast': 1, 'Southeast': 2, 'Southwest': 3}
sex_map = {'Male': 0, 'Female': 1}
smoker_map = {'Yes': 1, 'No': 0}

@app.get("/")
def home(request: Request):
    return templates.TemplateResponse("index.html", {"request": request})

@app.post("/predict")
def predict_charges(input_data: InsuranceInput):
    # Convert input to DataFrame following your test code logic
    new_data = pd.DataFrame({
        'age': [input_data.age],
        'sex': [input_data.sex],
        'bmi': [input_data.bmi],
        'children': [input_data.children],
        'smoker': [input_data.smoker],
        'region': [input_data.region]
    })
    
    # Apply preprocessing
    new_data['smoker'] = new_data['smoker'].map(smoker_map)
    new_data.drop(columns=['sex', 'region'], inplace=True)
    
    # Predict
    prediction = model.predict(new_data)[0]
    return {"predicted_charges": round(float(prediction), 2)}

if __name__ == "__main__":
    uvicorn.run(app, port=8000)