import streamlit as st
import joblib
import numpy as np

st.set_page_config(page_title="Loan Default Predictor", layout="centered")

st.title("Loan Default Predictor — Give Me Some Credit")
st.write("Enter applicant features below and click **Predict** to get `SeriousDlqin2yrs` prediction.")

@st.cache_resource
def load_model(path="model.pkl"):
    return joblib.load(path)

try:
    model = load_model("model.pkl")  
except Exception as e:
    st.error(f"Unable to load model.pkl — check file path. Error: {e}")
    st.stop()

st.subheader("Applicant features")

col1, col2 = st.columns(2)
with col1:
    RevolvingUtilizationOfUnsecuredLines = st.number_input(
        "Revolving Utilization Of Unsecured Lines", value=0.769413, format="%.6f"
    )
    age = st.number_input("age", value=32, min_value=18, max_value=120, step=1)
    NumberOfTime30_59DaysPastDueNotWorse = st.number_input(
        "Number Of Time 30-59 Days Past Due Not Worse", value=2, step=1
    )
    DebtRatio = st.number_input("DebtRatio", value=0.161342, format="%.6f")
    MonthlyIncome = st.number_input("MonthlyIncome", value=45000.0, format="%.2f")

with col2:
    NumberOfOpenCreditLinesAndLoans = st.number_input(
        "Number Of Open Credit Lines And Loans", value=6.0, step=1.0
    )
    NumberOfTimes90DaysLate = st.number_input(
        "Number Of Times 90 Days Late", value=1.0, step=1.0
    )
    NumberRealEstateLoansOrLines = st.number_input(
        "Number Real Estate Loans Or Lines", value=6.0, step=1.0
    )
    NumberOfTime60_89DaysPastDueNotWorse = st.number_input(
        "Number Of Time 60-89 Days Past Due Not Worse", value=0.0, step=1.0
    )
    NumberOfDependents = st.number_input(
        "Number Of Dependents", value=0.999883, format="%.6f"
    )

col3, col4 = st.columns(2)
with col3:
    TotalPastDue = st.number_input("Total Past Due", value=2.0, step=1.0)
    DebtRatioPerDependent = st.number_input("DebtRatio Per Dependent", value=25.1344)
    UtilizationPerLine = st.number_input("Utilization Per Line", value=0.217371)
with col4:
    IncomeDebtRatio = st.number_input("IncomeDebtRatio", value=3003.393701)
    HasDependents = st.selectbox("HasDependents", options=[0, 1], index=1)
    HighDebtRatio = st.selectbox("HighDebtRatio", options=[0, 1], index=0)
    HighUtilization = st.selectbox("HighUtilization", options=[0, 1], index=0)

feature_order = [
    RevolvingUtilizationOfUnsecuredLines,
    age,
    NumberOfTime30_59DaysPastDueNotWorse,
    DebtRatio,
    MonthlyIncome,
    NumberOfOpenCreditLinesAndLoans,
    NumberOfTimes90DaysLate,
    NumberRealEstateLoansOrLines,
    NumberOfTime60_89DaysPastDueNotWorse,
    NumberOfDependents,
    TotalPastDue,
    DebtRatioPerDependent,
    UtilizationPerLine,
    IncomeDebtRatio,
    HasDependents,
    HighDebtRatio,
    HighUtilization,
]

X_input = np.array(feature_order).reshape(1, -1)

if st.button("Predict"):
    try:
        pred = model.predict(X_input)[0]
    except Exception as e:
        st.error(f"Prediction failed. Check feature order/types. Error: {e}")
        st.stop()

    label_map = {0: "No Default (Good Credit)", 1: "Default (High Risk)"}
    meaning = label_map.get(int(pred), str(pred))

    proba_str = ""
    try:
        proba = model.predict_proba(X_input)[0]
        if len(proba) == 2:
            proba_str = f" — P(Default) = {proba[1]:.4f}"
        else:
            proba_str = " — probabilities: " + ", ".join(f"{p:.4f}" for p in proba)
    except Exception:
        proba_str = ""

    if int(pred) == 1:
        st.error(f"Prediction: {pred} → {meaning}{proba_str}")
    else:
        st.success(f"Prediction: {pred} → {meaning}{proba_str}")

    with st.expander("Show raw inputs and model output"):
        st.write("Input vector (order):", feature_order)
        st.write("Raw prediction:", int(pred))
        if proba_str:
            st.write("Raw probabilities:", proba if 'proba' in locals() else None)

st.caption("Model expects features in a specific order. If predictions seem off, verify the feature order and preprocessing used during training.")