app.py

import gradio as gr
import pandas as pd
from core.data import load_data
from core.model_runner import get_model
from core.plot import plot_forecast, plot_metrics_precision, plot_metrics_risk, plot_loss_curve, plot_future_forecast, plot_model_architecture
from config import AVAILABLE_MODELS, DEFAULT_TICKERS


def main_interface():
    with gr.Blocks(theme=gr.themes.Soft()) as app:
        gr.Markdown("# 📈 AI Forecasting Studio")

        with gr.Row():
            with gr.Column(scale=1):
                data_src = gr.Radio(["Yahoo Finance", "Upload CSV"], label="Data Source", value="Yahoo Finance")
                ticker = gr.Dropdown(choices=DEFAULT_TICKERS, label="Ticker", value="BTC-USD")
                file_upload = gr.File(label="Upload CSV", visible=False, file_types=[".csv"])
                start_date = gr.Textbox(label="Start Date (YYYY-MM-DD)", value="2022-01-01")
                end_date = gr.Textbox(label="End Date (YYYY-MM-DD)", value="2023-12-31")
                horizon = gr.Slider(1, 15, step=1, label="Forecast Days", value=1)

                gr.Markdown("## ⚙️ Model Settings")
                model = gr.Dropdown(choices=AVAILABLE_MODELS, label="Model", value="LSTM")
                hidden_units = gr.Slider(8, 512, label="Hidden Units", value=64)
                n_layers = gr.Slider(1, 5, step=1, label="# Hidden Layers", value=2)
                epochs = gr.Slider(10, 300, label="Epochs", value=100)
                learning_rate = gr.Slider(1e-5, 0.01, label="Learning Rate", value=0.001)
                beta1 = gr.Slider(0.8, 0.95, label="AdamW Beta1", value=0.9, step=0.01)
                beta2 = gr.Slider(0.9, 0.999, label="AdamW Beta2", value=0.999, step=0.001)
                weight_decay = gr.Slider(0.0, 0.1, label="Weight Decay", value=0.01, step=0.001)
                dropout = gr.Slider(0.0, 0.3, label="Drop Out", value=0.2)
                window_size = gr.Slider(5, 90, label="Window Size", value=30)
                test_split = gr.Slider(0.05, 0.5, label="Test Split", value=0.2)
                scheduler_factor = gr.Slider(0.1, 0.9, label="Scheduler Factor (LR Reduction)", value=0.5, step=0.1)

                run_btn = gr.Button("🚀 Train & Predict")
                status = gr.Textbox(label="Status", interactive=False, lines=2)

            with gr.Column(scale=2):
                backtest_plot = gr.Plot(label="📊 Backtesting: Actual vs Forecast")
                future_plot = gr.Plot(label="🔮 Future Forecast with Actuals")
                future_table = gr.Dataframe(label="📋 Future Predictions")
                precision_plot = gr.Plot(label="📉 Precision Metrics (Model Accuracy: R² (%), Explained Variance (%), MDA (%))")
                risk_plot = gr.Plot(label="📉 Risk Metrics (Error Magnitude: RMSE, MAE, MAPE (%), MASE)")
                loss_plot = gr.Plot(label="📈 Training Loss Curve")
                architecture_plot = gr.Plot(label="🧠 Model Architecture")

        def run_pipeline(data_src, ticker, file_upload, start_date, end_date, horizon, model,
                        hidden_units, n_layers, epochs, learning_rate, beta1, beta2, weight_decay,
                        dropout, window_size, test_split, scheduler_factor):
            try:
                pd.to_datetime(start_date)
                pd.to_datetime(end_date)

                source_key = "csv" if data_src == "Upload CSV" else "yahoo"
                main_df, future_df = load_data(data_src=source_key, ticker=ticker, file_upload=file_upload, 
                                             start=start_date, end=end_date, horizon=horizon)
                if main_df is None or main_df.empty:
                    return None, None, None, None, None, None, None, "❌ Failed to load data. Please check input."

                result = get_model(
                    df=main_df,
                    future_df=future_df,
                    model_name=model,
                    horizon=horizon,
                    hidden_units=hidden_units,
                    n_layers=n_layers,
                    epochs=epochs,
                    learning_rate=learning_rate,
                    beta1=beta1,
                    beta2=beta2,
                    weight_decay=weight_decay,
                    dropout=dropout,
                    window_size=window_size,
                    test_split=test_split,
                    scheduler_factor=scheduler_factor
                )
                forecast_plot = plot_forecast(result)
                future_plot = plot_future_forecast(main_df, result, future_df)
                precision_plot = plot_metrics_precision(result)
                risk_plot = plot_metrics_risk(result)
                loss_plot = plot_loss_curve(result)
                architecture_plot = plot_model_architecture(result)

                msg = "✅ Done."
                if "latest_prediction" in result:
                    last_date = main_df['Date'].iloc[-1]
                    future_dates = pd.date_range(start=last_date + pd.Timedelta(days=1), periods=horizon, freq='B')
                    future_data = {'Date': future_dates, 'Predicted Value': result["latest_prediction"]}
                    if not future_df.empty and "future_actuals" in result:
                        future_data['Actual Value'] = result["future_actuals"] + [None] * (horizon - len(result["future_actuals"]))
                    future_df_out = pd.DataFrame(future_data)
                    msg += f" Next predicted value(s): {[f'{val:.2f}' for val in result['latest_prediction']]}"
                else:
                    future_df_out = pd.DataFrame()

                return forecast_plot, future_plot, future_df_out, precision_plot, risk_plot, loss_plot, architecture_plot, msg
            except Exception as e:
                return None, None, None, None, None, None, None, f"❌ Error: {str(e)}"

        run_btn.click(
            fn=run_pipeline,
            inputs=[
                data_src, ticker, file_upload,
                start_date, end_date, horizon, model,
                hidden_units, n_layers, epochs, learning_rate,
                beta1, beta2, weight_decay, dropout, window_size, test_split, scheduler_factor
            ],
            outputs=[backtest_plot, future_plot, future_table, precision_plot, risk_plot, loss_plot, architecture_plot, status]
        )

        def toggle_file(src):
            return gr.update(visible=(src == "Upload CSV"))

        data_src.change(fn=toggle_file, inputs=[data_src], outputs=[file_upload])

    return app


if __name__ == '__main__':
    main_interface().launch()