app.py

import gradio as gr
import spaces
import pandas as pd
import numpy as np
import os
from datetime import datetime, timedelta
import logging
from core.data import load_data, add_technical_indicators, add_sentiment
from core.model_runner import get_model
from core.plot import plot_forecast, plot_metrics_r2, plot_metrics_errors, plot_metrics_precision_recall, plot_metrics_risk, plot_loss_curve, plot_model_architecture, plot_future_forecast, plot_indicators, plot_signals, plot_backtest
import plotly.io as pio
from core.signals import generate_signals
from config import AVAILABLE_MODELS, DEFAULT_TICKERS, AVAILABLE_TIMEFRAMES, AVAILABLE_INDICATORS
from newsapi import NewsApiClient
from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer

log_path = "/tmp/app_log.txt"
os.makedirs("/tmp", exist_ok=True)
logging.basicConfig(
    level=logging.DEBUG,
    handlers=[
        logging.FileHandler(log_path),
        logging.StreamHandler()
    ],
    format='%(asctime)s - %(levelname)s - %(message)s'
)
analyzer = SentimentIntensityAnalyzer()

def sentiment_analysis(ticker, start_date, end_date, api_key):
    try:
        if not api_key:
            return "No API key provided", None
        newsapi = NewsApiClient(api_key=api_key)
        start = pd.to_datetime(start_date)
        end = pd.to_datetime(end_date)
        articles = newsapi.get_everything(
            q=ticker, from_param=start.strftime("%Y-%m-%d"), to=end.strftime("%Y-%m-%d"), 
            language='en', sort_by='relevancy'
        )
        sentiments = [analyzer.polarity_scores(article["title"])["compound"] for article in articles["articles"]]
        avg_sentiment = np.mean(sentiments) if sentiments else 0.0
        sentiment_text = f"Average sentiment for {ticker}: {avg_sentiment:.2f}"
        return sentiment_text, avg_sentiment
    except Exception as e:
        logging.error(f"Sentiment analysis failed: {str(e)}")
        return f"Sentiment analysis failed: {str(e)}", None

def update_horizon_label(timeframe):
    units = {'1m': 'minutes', '5m': 'minutes', '15m': 'minutes', '30m': 'minutes', 
             '1h': 'hours', '4h': 'hours', '1d': 'days', '1wk': 'weeks'}
    return gr.update(label=f"Horizon ({units.get(timeframe, 'days')})")

@spaces.GPU
def run_dashboard(data_src, ticker, file_upload, timeframe, start_date, end_date, horizon, indicators, 
                 include_sentiment, news_api_key, alpha_api_key, account_size, risk_percent, model, 
                 hidden_units, n_layers, epochs, learning_rate, beta1, beta2, weight_decay, dropout, 
                 window_size, test_split, rsi_mid, macd_sens, adx_thr, sent_thr, vote_buy, vote_sell, 
                 feat_selector, feat_threshold):
    try:
        logging.info(f"Running dashboard for {ticker}, timeframe: {timeframe}, model: {model}")
        start_date = pd.to_datetime(start_date).strftime("%Y-%m-%d")
        end_date = pd.to_datetime(end_date).strftime("%Y-%m-%d")
        
        df = load_data(data_src=data_src, ticker=ticker, start=start_date, end=end_date, 
                       interval=timeframe, file_upload=file_upload, alpha_api_key=alpha_api_key)
        if df.empty:
            logging.error("Failed to load data")
            return [None] * 12 + ["Failed to load data", None, None, None, None, "Failed to load data"]

        df, valid_indicators = add_technical_indicators(df, indicators)
        if include_sentiment and news_api_key:
            df = add_sentiment(df, ticker, news_api_key, start_date, end_date)
        sentiment_text, sentiment_score = sentiment_analysis(ticker, start_date, end_date, news_api_key)

        features = valid_indicators # Use valid_indicators for feature
        target = 'value'
        result = get_model(
            df=df,
            features=features,
            target=target,
            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,
            selector_method=feat_selector,
            importance_threshold=feat_threshold
        )
        if isinstance(result, dict) and result.get("error"):
            logging.error(f"Model training failed: {result['error']})")
            return [None] * 12 + [f"Model training failed: {result['error']}", None, None, None, None, f"Model training failed: {result['error']}"]

        signals_df, trades_df, equity_df = generate_signals(df, result)
        if signals_df.empty:
            logging.error("Failed to generate signals")
            return [None] * 12 + ["Failed to generate signals", None, None, None, None, "Failed to generate signals"]

        chart_plot = plot_indicators(df, ticker)
        signals_plot = plot_signals(signals_df, ticker)
        backtest_plot = plot_backtest(equity_df, trades_df, ticker)


        future_plot = plot_future_forecast(df, result, indicators)
        future_table = pd.DataFrame({
            "Date": [df.index[-1] + timedelta(days=i+1) for i in range(horizon)],
            "Prediction": result["latest_prediction"]
        })
        signals_table = signals_df.reset_index()[["Date", "Price", "Signal", "Position_Size", "Stop_Loss", "Take_Profit", "Equity"]]
        r2_plot = plot_metrics_r2(result)
        error_plot = plot_metrics_errors(result)
        precision_recall_plot = plot_metrics_precision_recall(result)
        risk_plot = plot_metrics_risk(result)
        loss_plot = plot_loss_curve(result)
        architecture_plot = plot_model_architecture(result)

        signals_csv = f"signals_{ticker}.csv"
        signals_df.to_csv(signals_csv)



        predictions_csv = f"predictions_{ticker}.csv"
        pd.DataFrame({
            "Actual": result["actual"],
            "Forecast": result["forecast"]
        }).to_csv(predictions_csv)
        chart_png = f"chart_{ticker}.png"
        pio.write_image(chart_plot, chart_png, format='png')

        with open(log_path, 'r') as log_file:
            log_output = log_file.read()

        logging.info("Dashboard run completed successfully")
        return [
            chart_plot, sentiment_text, signals_table, backtest_plot, future_plot, future_table,
            r2_plot, error_plot, precision_recall_plot, risk_plot, loss_plot, architecture_plot,
            "Dashboard generated successfully", chart_png, signals_csv, predictions_csv, signals_plot, log_output
        ]
    except Exception as e:
        logging.error(f"Dashboard error: {str(e)}")
        return [None] * 12 + [f"Error: {str(e)}", None, None, None, None, f"Error: {str(e)}"]

def main_interface():
    try:
        with gr.Blocks(title="Market Prediction Pro", theme=gr.themes.Default()) as app:
            gr.Markdown("# Market Prediction Pro")
            with gr.Row():
                with gr.Column(scale=1):
                    data_src = gr.Dropdown(["yahoo", "csv"], label="Data Source", value="yahoo")
                    ticker = gr.Dropdown(DEFAULT_TICKERS, label="Ticker", value="AAPL")
                    file_upload = gr.File(label="Upload CSV", visible=False)
                    timeframe = gr.Dropdown(AVAILABLE_TIMEFRAMES, label="Timeframe", value="1d")
                    start_date = gr.Textbox("2020-01-01", label="Start Date (YYYY-MM-DD)")
                    end_date = gr.Textbox("2023-01-01", label="End Date (YYYY-MM-DD)")
                    horizon = gr.Slider(1, 30, step=1, label="Horizon (days)", value=1)
                    indicators = gr.CheckboxGroup(AVAILABLE_INDICATORS, label="Technical Indicators", value=["rsi", "macd", "bbands"])
                    include_sentiment = gr.Checkbox(label="Include Sentiment Analysis", value=False)
                    news_api_key = gr.Textbox(label="News API Key", visible=False, type="password")
                    alpha_api_key = gr.Textbox(label="Alpha Vantage API Key", type="password")
                    account_size = gr.Slider(1000, 100000, step=1000, label="Account Size ($)", value=10000)
                    risk_percent = gr.Slider(0.01, 0.1, step=0.01, label="Risk per Trade (%)", value=0.01)
                with gr.Column(scale=1):
                    model = gr.Dropdown(AVAILABLE_MODELS, label="Model", value="LSTM")
                    hidden_units = gr.Slider(16, 256, step=16, label="Hidden Units", value=64)
                    n_layers = gr.Slider(1, 4, step=1, label="Layers", value=1)
                    epochs = gr.Slider(10, 100, step=10, label="Epochs", value=50)
                    learning_rate = gr.Slider(0.0001, 0.01, step=0.0001, label="Learning Rate", value=0.001)
                    beta1 = gr.Slider(0.8, 0.99, step=0.01, label="Beta 1", value=0.9)
                    beta2 = gr.Slider(0.9, 0.999, step=0.001, label="Beta 2", value=0.999)
                    weight_decay = gr.Slider(0.0, 0.1, step=0.01, label="Weight Decay", value=0.01)
                    dropout = gr.Slider(0.0, 0.5, step=0.05, label="Dropout", value=0.2)
                    window_size = gr.Slider(5, 60, step=5, label="Window Size", value=30)
                    test_split = gr.Slider(0.1, 0.5, step=0.05, label="Test Split", value=0.2)
                    rsi_mid = gr.Slider(30, 70, step=5, label="RSI Middle", value=50)
                    macd_sens = gr.Slider(0.0, 0.5, step=0.05, label="MACD Sensitivity", value=0.0)
                    adx_thr = gr.Slider(10, 50, step=5, label="ADX Threshold", value=20)
                    sent_thr = gr.Slider(0.0, 0.5, step=0.05, label="Sentiment Threshold", value=0.1)
                    vote_buy = gr.Slider(1, 5, step=1, label="Vote Buy Threshold", value=2)
                    vote_sell = gr.Slider(-5, -1, step=1, label="Vote Sell Threshold", value=-2)
                    feat_selector = gr.Dropdown(["RandomForest", "PCA"], label="Feature Selector", value="RandomForest")
                    feat_threshold = gr.Slider(0.0, 0.5, step=0.05, label="Feature Importance Threshold", value=0.0)
        
            run_btn = gr.Button("Run Analysis")
        
            with gr.Tabs():
                with gr.TabItem("Price & Indicators"):
                    chart_plot = gr.Plot(label="📈 Price and Indicators")
                with gr.TabItem("Signals"):
                    signals_plot = gr.Plot(label="📈 Trading Signals")
                    signals_table = gr.DataFrame(label="📅 Signals")
                with gr.TabItem("Sentiment"):
                    sentiment_text = gr.Textbox(label="Sentiment Analysis")
                with gr.TabItem("Forecast"):
                    backtest_plot = gr.Plot(label="📈 Backtest Results")
                    future_plot = gr.Plot(label="📈 Future Forecast")
                    future_table = gr.DataFrame(label="📅 Future Predictions")
                with gr.TabItem("Metrics"):
                    r2_plot = gr.Plot(label="📊 R² & MAPE")
                    error_plot = gr.Plot(label="📊 Error Metrics")
                    precision_recall_plot = gr.Plot(label="📊 Precision & Recall")
                    risk_plot = gr.Plot(label="📊 Risk Metrics")
                    loss_plot = gr.Plot(label="📈 Training Loss Curve")
                    architecture_plot = gr.Plot(label="🧠 Model Architecture")
                with gr.TabItem("Export"):
                    status = gr.Textbox(label="Status")
                    export_chart = gr.File(label="Export Chart (PNG)")
                    export_signals = gr.File(label="Export Signals (CSV)")
                    export_predictions = gr.File(label="Export Predictions (CSV)")
                    log_output = gr.Textbox(label="Debug Logs", lines=10)

            data_src.change(fn=lambda src: gr.update(visible=(src == "csv")), inputs=[data_src], outputs=[file_upload])
            include_sentiment.change(fn=lambda sent: gr.update(visible=sent), inputs=[include_sentiment], outputs=[news_api_key])
            timeframe.change(fn=update_horizon_label, inputs=[timeframe], outputs=[horizon])

            run_btn.click(
                fn=run_dashboard,
                inputs=[data_src, ticker, file_upload, timeframe, start_date, end_date, horizon, indicators, 
                        include_sentiment, news_api_key, alpha_api_key, account_size, risk_percent, model, 
                        hidden_units, n_layers, epochs, learning_rate, beta1, beta2, weight_decay, dropout, 
                        window_size, test_split, rsi_mid, macd_sens, adx_thr, sent_thr, vote_buy, vote_sell, 
                        feat_selector, feat_threshold],
                outputs=[chart_plot, sentiment_text, signals_table, backtest_plot, future_plot, future_table,
                         r2_plot, error_plot, precision_recall_plot, risk_plot, loss_plot, architecture_plot,
                         status, export_chart, export_signals, export_predictions, signals_plot, log_output]
            )
    except Exception as e:
        logging.error(f"Error in main_interface: {str(e)}")
        raise
    return app

if __name__ == "__main__":
    main_interface().launch(server_name="0.0.0.0", server_port=7860, share=False)