import streamlit as st
import joblib
import pandas as pd
import numpy as np

MODEL_PATH = 'src/xgb_grade_prediction.joblib' 

# The 5 features used for training
FEATURES = ["G1", "G2", "studytime", "failures", "absences"]

@st.cache_resource
def load_xgb_model():
    try:
        model = joblib.load(MODEL_PATH)
        return model
    except Exception as e:
        st.error(f"Error loading the XGBoost model. Ensure '{MODEL_PATH}' is uploaded and the 'xgboost' library is installed. Error: {e}")
        return None

def predict_grade(model, input_data):
    
    input_df = pd.DataFrame([input_data])[FEATURES]
    
    prediction = model.predict(input_df)
    
    return round(float(prediction[0]))

st.set_page_config(page_title="Student Grade Predictor", layout="centered")
st.title("🎓 Student Final Grade (G3) Prediction")
st.markdown("Enter student performance metrics to predict the final grade (G3).")

model = load_xgb_model()

if model is not None:
    st.sidebar.header("Student Metrics (0-20 Scale)")

    # --- INPUT WIDGETS ---
    # G1 and G2 (First and Second Period Grades)
    g1 = st.sidebar.slider("Period 1 Grade (G1):", min_value=0, max_value=20, value=10)
    g2 = st.sidebar.slider("Period 2 Grade (G2):", min_value=0, max_value=20, value=11)
    
    # Study Time (Categorical in original dataset, often 1-4)
    studytime = st.sidebar.slider("Study Time (Hours/Week Index):", min_value=1, max_value=4, value=2)
    
    # Failures (Past class failures)
    failures = st.sidebar.slider("Past Class Failures:", min_value=0, max_value=4, value=0)
    
    # Absences (Number of school absences)
    absences = st.sidebar.number_input("Absences (Days):", min_value=0, max_value=93, value=5)

    # Collect inputs
    input_data = {
        "G1": g1,
        "G2": g2,
        "studytime": studytime,
        "failures": failures,
        "absences": absences
    }

    st.subheader("Current Input Summary:")
    st.dataframe(pd.DataFrame([input_data]), hide_index=True)

    if st.button("Predict Final Grade (G3)"):
        with st.spinner('Calculating prediction...'):
            predicted_g3 = predict_grade(model, input_data)
            
            st.success("Prediction Successful!")
            st.markdown("### Predicted Final Grade (G3):")
            st.markdown(f"**{predicted_g3} / 20**")
            
            if predicted_g3 >= 10:
                st.balloons()