# === 🔟 Monitoring Dashboard ===
st.markdown("---")
st.subheader("📊 Monitoring Dashboard")

if os.path.exists(LOG_FILE):
    # Load logged data
    logged_data = pd.read_csv(LOG_FILE, parse_dates=['timestamp'])

    # Ensure minimal data exists to generate report
    if len(logged_data) >= 3:
        # Reference = first 50 rows, Current = last 50 rows
        reference_data = logged_data.head(50)
        current_data = logged_data.tail(50)

        # Drop empty columns from both datasets
        reference_data = reference_data.dropna(axis=1, how='all')
        current_data = current_data.dropna(axis=1, how='all')

        # Ensure both datasets have the same columns
        common_columns = reference_data.columns.intersection(current_data.columns)
        reference_data = reference_data[common_columns]
        current_data = current_data[common_columns]

        # Generate Evidently report
        report = Report([
            DataDriftPreset(),
            DataSummaryPreset()
        ],
        include_tests=True)
        
        # Run the report (no return value!)  
        report.run(current_data, reference_data)
       
        # Save the report as an HTML file
        report_file = "./evidently_report.html"
        report.save_html(report_file)
 
        # Render the HTML file in Streamlit
        with open(report_file, "r") as f:
            html_content = f.read()
        st.components.v1.html(html_content, height=800)

        with st.expander("🔎 View Logged Data"):
            st.dataframe(logged_data)
    else:
        st.info("Need at least 3 logged predictions to generate monitoring report.")
else:
    st.info("No predictions have been logged yet. Make some predictions to start monitoring.")