from fastapi import FastAPI
from pydantic import BaseModel
import pickle
import pandas as pd
import numpy as np
import uvicorn
import os
from sklearn.preprocessing import StandardScaler
import joblib



app = FastAPI(title="API")


"""We load a machine learning model and a scaler that help us make predictions based on data."""
model = joblib.load('model.pkl',mmap_mode='r')
scaler = joblib.load('scaler.pkl',mmap_mode='r')


def predict(df, endpoint='simple'):
    # Scaling
    scaled_df = scaler.transform(df)

    # Prediction
    prediction = model.predict_proba(scaled_df)
    highest_proba = prediction.max(axis=1)

    predicted_labels = ["Patient does not have sepsis" if i == 0 else "Patient has Sepsis" for i in highest_proba]
    response = []
    for label, proba in zip(predicted_labels, highest_proba):
        output = {
            "prediction": label,
            "probability of prediction": str(round(proba * 100)) + '%'
        }
        response.append(output)
    return response




class Patient(BaseModel):
    Blood_Work_R1: float
    Blood_Pressure: float
    Blood_Work_R3: float
    BMI: float
    Blood_Work_R4: float
    Patient_age: int




@app.get("/")
def root():
    return {"API": "This is an API for sepsis prediction."}

# Prediction endpoint (Where we will input our features)
@app.post("/predict")
def predict_sepsis(patient: Patient):

    # Make prediction
    data = pd.DataFrame(patient.dict(), index=[0])
    scaled_data = scaler.transform(data)
    parsed = predict(df=scaled_data)
    return {"output": parsed}


if __name__ == "__main__":
    os.environ["DEBUG"] = "True"  # Enable debug mode
    uvicorn.run("main:app", reload=True)