app.py

import gradio as gr
import pandas as pd
import xgboost as xgb
import joblib
import os

# ✅ Load the trained XGBoost model
model_path = "xgboost_fraud_model.pkl"  # Ensure this file is uploaded to Hugging Face
if not os.path.exists(model_path):
    raise FileNotFoundError(f"❌ Model file '{model_path}' not found!")

model = joblib.load(model_path)

# ✅ Load the sample transaction dataset
sample_file_path = "Sample_Data_Matching_Model.csv"  # Ensure this file is uploaded to Hugging Face
if not os.path.exists(sample_file_path):
    raise FileNotFoundError(f"❌ Sample data '{sample_file_path}' not found!")

sample_df = pd.read_csv(sample_file_path)

# ✅ Define the required feature columns
REQUIRED_COLS = [
    'index', 'v1', 'v2', 'v3', 'v4', 'v5', 'v6', 'v7', 'v8', 'v9', 'v10', 
    'v11', 'v12', 'v13', 'v14', 'v15', 'v16', 'v17', 'v18', 'v19', 'v20', 
    'v21', 'v22', 'v23', 'v24', 'v25', 'v26', 'v27', 'v28', 'dow', 
    'is_weekend', 'days_between', 'max_days_between', 'monthly_txns_lag_30', 
    'weekly_txns_lag_30', 'cls_lag_30'
]

def predict_fraud(file, use_sample):
    try:
        # Load data based on user input
        if use_sample:
            df = sample_df.copy()  # Use predefined sample data
        elif file is not None:
            df = pd.read_csv(file)  # Load user-uploaded file
        else:
            return "❌ No file provided!", None, None  # Error, No preview, No download

        # ✅ Ensure required columns exist
        missing_cols = [col for col in REQUIRED_COLS if col not in df.columns]
        if missing_cols:
            return f"❌ Error: Missing columns: {', '.join(missing_cols)}", None, None

        df = df[REQUIRED_COLS]  # Keep only necessary columns

        # ✅ Make fraud predictions
        df["fraud_probability"] = model.predict_proba(df)[:, 1]
        df["fraud_prediction"] = (df["fraud_probability"] > 0.5).astype(int)

        # ✅ Select preview (First 5 Transactions)
        preview_df = df[['index', 'fraud_probability', 'fraud_prediction']].head(5)

        # ✅ Save full results for download
        output_csv = "fraud_predictions.csv"
        df.to_csv(output_csv, index=False)

        return None, preview_df, output_csv  # ✅ No error, preview, and download

    except Exception as e:
        return f"❌ Unexpected Error: {str(e)}", None, None  # Ensure other outputs are `None`

# ✅ Gradio App Layout with Styling
with gr.Blocks(css="""
    .gradio-container {max-width: 750px !important; margin: auto;}
    #header {text-align: center; font-size: 1.5em; font-weight: bold; margin-bottom: 10px;}
    #subheader {text-align: center; font-size: 1em; color: gray; margin-bottom: 20px;}
    .gr-box {border-radius: 10px; padding: 15px; background: #f9f9f9;}
""") as app:
    
    gr.HTML("<div id='header'>🚀 Fraud Detection System</div>")
    gr.HTML("<div id='subheader'>Upload a transaction file OR click 'Use Sample Data' to test.</div>")

    with gr.Group():
        with gr.Row():
            file_input = gr.File(label="📂 Upload CSV File")
            use_sample_toggle = gr.Checkbox(label="✅ Use Sample Data", value=False)

    submit_btn = gr.Button("🚀 Predict Fraud")

    with gr.Group():
        error_message = gr.Markdown("")  # ✅ Error Message Output
        preview_output = gr.DataFrame(label="🔍 Fraud Prediction Preview (5 Transactions)")
        download_output = gr.File(label="📥 Download Full Results")

    # ✅ Link button to function
    submit_btn.click(
        predict_fraud,
        inputs=[file_input, use_sample_toggle],
        outputs=[error_message, preview_output, download_output]  # ✅ FIXED: Assign all outputs
    )

app.launch()