import pandas as pd
import pickle
import gradio as gr

# =========================
# Load Model & Encoders
# =========================

with open("bank_dt_model.pkl", "rb") as f:
    model = pickle.load(f)

with open("le_dict.pkl", "rb") as f:
    le_dict = pickle.load(f)

categorical_cols = [
    'job', 'marital', 'education', 'default',
    'housing', 'loan', 'contact', 'month', 'poutcome'
]

feature_order = [
    'age', 'job', 'marital', 'education', 'default',
    'balance', 'housing', 'loan', 'contact', 'day',
    'month', 'duration', 'campaign', 'pdays',
    'previous', 'poutcome'
]

# =========================
# Prediction Function
# =========================

def predict(
    age, job, marital, education, default,
    balance, housing, loan, contact, day,
    month, duration, campaign, pdays,
    previous, poutcome
):
    try:
        input_data = {
            'age': int(age),
            'job': le_dict['job'].transform([job])[0],
            'marital': le_dict['marital'].transform([marital])[0],
            'education': le_dict['education'].transform([education])[0],
            'default': le_dict['default'].transform([default])[0],
            'balance': int(balance),
            'housing': le_dict['housing'].transform([housing])[0],
            'loan': le_dict['loan'].transform([loan])[0],
            'contact': le_dict['contact'].transform([contact])[0],
            'day': int(day),
            'month': le_dict['month'].transform([month])[0],
            'duration': int(duration),
            'campaign': int(campaign),
            'pdays': int(pdays),
            'previous': int(previous),
            'poutcome': le_dict['poutcome'].transform([poutcome])[0],
        }

        df_input = pd.DataFrame([input_data], columns=feature_order)

        pred = model.predict(df_input)[0]
        confidence = model.predict_proba(df_input)[0][pred]

        message = (
            "✅ Likely to subscribe. Recommend follow-up."
            if pred == 1 else
            "❌ Unlikely to subscribe."
        )

        return message, round(float(confidence), 3)

    except Exception as e:
        return f"Error: {str(e)}", None


# =========================
# Gradio UI
# =========================

inputs = [
    gr.Number(label="Age"),
    gr.Dropdown(le_dict['job'].classes_.tolist(), label="Job"),
    gr.Dropdown(le_dict['marital'].classes_.tolist(), label="Marital Status"),
    gr.Dropdown(le_dict['education'].classes_.tolist(), label="Education"),
    gr.Dropdown(le_dict['default'].classes_.tolist(), label="Credit Default"),
    gr.Number(label="Account Balance"),
    gr.Dropdown(le_dict['housing'].classes_.tolist(), label="Housing Loan"),
    gr.Dropdown(le_dict['loan'].classes_.tolist(), label="Personal Loan"),
    gr.Dropdown(le_dict['contact'].classes_.tolist(), label="Contact Type"),
    gr.Number(label="Day of Month"),
    gr.Dropdown(le_dict['month'].classes_.tolist(), label="Month"),
    gr.Number(label="Call Duration (seconds)"),
    gr.Number(label="Campaign Contacts"),
    gr.Number(label="Days Since Last Contact"),
    gr.Number(label="Previous Contacts"),
    gr.Dropdown(le_dict['poutcome'].classes_.tolist(), label="Previous Campaign Outcome"),
]

outputs = [
    gr.Textbox(label="Prediction"),
    gr.Number(label="Confidence")
]

app = gr.Interface(
    fn=predict,
    inputs=inputs,
    outputs=outputs,
    title="🏦 Bank Term Deposit Predictor",
    description="Decision Tree model deployed using Hugging Face Spaces"
)

app.launch()