from fastapi import FastAPI
from pydantic import BaseModel
import pandas as pd
import joblib
import os
import requests

MODEL_URL  = "https://huggingface.co/adeshjain/model1/resolve/main/model_j.joblib"
SCALER_URL = "https://huggingface.co/adeshjain/model1/resolve/main/scaler_j.joblib"

def download_file(url, filename):
    path = os.path.join("/tmp", filename)
    if not os.path.exists(path):
        r = requests.get(url)
        r.raise_for_status()
        with open(path, "wb") as f:
            f.write(r.content)
    return path

model  = joblib.load(download_file(MODEL_URL,  "model_j.joblib"))
scaler = joblib.load(download_file(SCALER_URL, "scaler_j.joblib"))

class ClaimData(BaseModel):
    OPAnnualReimbursementAmt: float
    OPAnnualDeductibleAmt:   float
    DeductibleAmtPaid:       float
    claim: float
    period: float
    phy_same: float
    Gender_1: float
    Gender_2: float
    RenalDiseaseIndicator: float
    age: float
    alife: float
    Provider: float
    NoOfMonths_PartACov: float
    NoOfMonths_PartBCov: float
    ChronicCond_Alzheimer: float
    ChronicCond_KidneyDisease: float
    ChronicCond_Cancer: float
    ChronicCond_ObstrPulmonary: float
    ChronicCond_Depression: float
    ChronicCond_Diabetes: float
    ChronicCond_IschemicHeart: float
    ChronicCond_stroke: float
    IPAnnualReimbursementAmt: float

app = FastAPI(title="Medicare Fraud Prediction API")

@app.post("/predict")
def predict_fraud(data: ClaimData):
    # Convert input to DataFrame
    df = pd.DataFrame([data.dict()])

    # Scale only the relevant columns
    columns_to_scale = ['OPAnnualReimbursementAmt','OPAnnualDeductibleAmt','DeductibleAmtPaid']
    df[columns_to_scale] = scaler.transform(df[columns_to_scale])

    # Predict
    pred = model.predict(df)[0]
    result = "Fraud" if pred == 1 else "Not Fraud"
    
    return {"prediction": result}

@app.get("/")
def root():
    return {"message": "working"}