Update app.py

app.py

@@ -5,121 +5,165 @@ import plotly.graph_objects as go
 from plotly.subplots import make_subplots
 import numpy as np
 import yfinance as yf
-from timesfm import TimesFm
+from timesfm import TimesFm, TimesFmHparams, TimesFmCheckpoint
 
-# Initialize Google TimesFM with updated API arguments
-# Note: 'context_len' changed to 'context_length', etc.
+# --- 1. FIXED INITIALIZATION ---
+# The new API requires 'hparams' and 'checkpoint' objects, not flat arguments.
 tfm = TimesFm(
-    context_length=512,
-    horizon_length=128,
-    input_patch_len=32,
-    output_patch_len=128,
-    num_layers=20,
-    model_dims=1280,
-    backend="cpu",
+    hparams=TimesFmHparams(
+        backend="cpu",
+        per_core_batch_size=32,
+        horizon_len=128,
+        context_len=512,
+        num_layers=20,
+        model_dims=1280,
+    ),
+    checkpoint=TimesFmCheckpoint(
+        huggingface_repo_id="google/timesfm-1.0-200m"
+    ),
 )
 
-tfm.load_from_checkpoint(repo_id="google/timesfm-1.0-200m")
-
-def get_tradingview_plot(df, forecast_df, ticker):
-    # Create subplots: Chart on top (80%), Volume on bottom (20%)
+def get_financial_plot(df, forecast_df, ticker, is_backtest=False):
+    # Create Professional 2-Row Chart (Price & Volume)
     fig = make_subplots(rows=2, cols=1, shared_xaxes=True, 
-                        vertical_spacing=0.03, subplot_titles=(f'{ticker} Analysis', ''), 
-                        row_width=[0.2, 0.8])
+                        vertical_spacing=0.05, row_heights=[0.75, 0.25])
 
-    # 1. Main Price Line
+    # A. Historical Data (Blue)
+    # If backtesting, we only show 'training' data in blue to prove the AI didn't see the rest
+    display_df = df[:-30] if is_backtest else df
+    
     fig.add_trace(go.Scatter(
-        x=df.index, y=df['Close'],
-        name='Historical Close',
+        x=display_df.index, y=display_df['Close'],
+        name='Historical Price',
         line=dict(color='#2962FF', width=2)
     ), row=1, col=1)
 
-    # 2. TimesFM Forecast (Dotted Projection)
+    # B. The "Truth" (Grey Line - Only for Backtesting)
+    if is_backtest:
+        truth_df = df[-30:]
+        fig.add_trace(go.Scatter(
+            x=truth_df.index, y=truth_df['Close'],
+            name='Actual Market Move',
+            line=dict(color='#787b86', width=2, dash='dot')
+        ), row=1, col=1)
+
+    # C. AI Forecast (Red)
+    # Connect the forecast line to the last historical point for a seamless look
+    last_hist_date = display_df.index[-1]
+    last_hist_val = display_df['Close'].iloc[-1]
+    
+    # Prepend the last historical point to the forecast data
+    fc_dates = [last_hist_date] + list(forecast_df['ds'])
+    fc_vals = [last_hist_val] + list(forecast_df['timesfm'])
+    
     fig.add_trace(go.Scatter(
-        x=forecast_df['ds'], y=forecast_df['timesfm'],
-        name='AI Forecast',
-        line=dict(color='#F23645', width=3, dash='dot')
+        x=fc_dates, y=fc_vals,
+        name='AI Prediction',
+        line=dict(color='#F23645', width=3)
     ), row=1, col=1)
 
-    # 3. Volume Profile (Bars)
+    # D. Volume (Bars)
     fig.add_trace(go.Bar(
         x=df.index, y=df['Volume'],
         name='Volume',
-        marker_color='#26a69a',
-        opacity=0.5
+        marker_color='rgba(38, 166, 154, 0.5)'
     ), row=2, col=1)
 
-    # Professional Dark Theme Styling
+    # TradingView-Style Dark Theme
     fig.update_layout(
         template='plotly_dark',
-        hovermode="x unified",
         paper_bgcolor='#131722',
         plot_bgcolor='#131722',
-        showlegend=True,
-        margin=dict(l=10, r=10, t=50, b=10),
-        legend=dict(orientation="h", yanchor="bottom", y=1.02, xanchor="right", x=1)
+        margin=dict(l=10, r=10, t=40, b=10),
+        legend=dict(orientation="h", y=1.02, x=0),
+        hovermode="x unified"
     )
-    
-    fig.update_xaxes(showgrid=True, gridcolor='#2a2e39', rangeslider_visible=False)
-    fig.update_yaxes(showgrid=True, gridcolor='#2a2e39', side='right')
+    fig.update_yaxes(gridcolor='#2a2e39', side='right')
+    fig.update_xaxes(gridcolor='#2a2e39')
     
     return fig
 
-def run_analysis(ticker, horizon):
+def run_analysis(ticker, horizon, mode):
     try:
-        df = yf.download(ticker, period="1y")
-        if df.empty: return None, "⚠️ Invalid Ticker"
+        # Fetch Data
+        df = yf.download(ticker, period="2y") # Get more data for stability
+        if df.empty: return None, "⚠️ Ticker Not Found"
+        
+        # --- MODE LOGIC ---
+        if mode == "Backtest (Reality Check)":
+            # Hide the last 30 days from the AI
+            train_df = df[:-30]
+            horizon = 30 # Fixed horizon for backtest comparison
+        else:
+            # Use full data for real future prediction
+            train_df = df
         
         # Format for TimesFM
         input_df = pd.DataFrame({
             'unique_id': [ticker],
-            'ds': df.index,
-            'y': df['Close'].values
+            'ds': train_df.index,
+            'y': train_df['Close'].values
         })
 
-        # Execute Forecast
+        # Run Forecast
         forecast_df, _ = tfm.forecast_on_df(
             inputs=input_df,
             freq="D", 
-            value_name="y"
+            value_name="y",
+            forecast_context_len=512 # Explicitly use context
         )
-
         forecast_df = forecast_df.head(horizon)
         
-        # Calculate Signals
-        current_price = df['Close'].iloc[-1]
-        final_pred = forecast_df['timesfm'].iloc[-1]
-        change_pct = ((final_pred - current_price) / current_price) * 100
-        
-        signal = "STRONG BUY" if change_pct > 5 else "BUY" if change_pct > 0 else "SELL"
-        signal_color = "#00ff88" if "BUY" in signal else "#ff4444"
-        
-        fig = get_tradingview_plot(df, forecast_df, ticker)
+        # Generate Signal (Only for Future Mode)
+        if mode != "Backtest (Reality Check)":
+            start_price = train_df['Close'].iloc[-1]
+            end_price = forecast_df['timesfm'].iloc[-1]
+            pct_change = ((end_price - start_price) / start_price) * 100
+            
+            color = "#00ff88" if pct_change > 0 else "#ff4444"
+            direction = "BULLISH" if pct_change > 0 else "BEARISH"
+            signal_html = f"<h3 style='color: {color}; margin: 0;'>{direction} ({pct_change:+.2f}%)</h3>"
+        else:
+            # Calculate Accuracy Score for Backtest
+            real_end = df['Close'].iloc[-1]
+            pred_end = forecast_df['timesfm'].iloc[-1]
+            error = abs(real_end - pred_end) / real_end * 100
+            accuracy = 100 - error
+            signal_html = f"<h3 style='color: #FFD700; margin: 0;'>AI Accuracy: {accuracy:.1f}%</h3>"
+
+        fig = get_financial_plot(df, forecast_df, ticker, is_backtest=(mode == "Backtest (Reality Check)"))
         
-        html_signal = f"<h2 style='color: {signal_color}; text-align: center;'>{signal} ({change_pct:+.2f}%)</h2>"
+        return fig, signal_html
         
-        return fig, html_signal
     except Exception as e:
-        return None, f"Error: {str(e)}"
+        return None, f"<span style='color:red'>Error: {str(e)}</span>"
 
-# Professional Layout
-with gr.Blocks(theme=gr.themes.Default(), css=".gradio-container {background-color: #131722; color: white;}") as demo:
-    gr.HTML("<h1 style='text-align: center; color: #2962FF; font-family: sans-serif;'>G-TIMES QUANT TERMINAL</h1>")
+# UI Layout
+with gr.Blocks(theme=gr.themes.Default(), css=".gradio-container {background-color: #000000}") as demo:
+    gr.HTML("""
+    <div style='background-color: #131722; padding: 15px; border-bottom: 2px solid #2962FF;'>
+        <h2 style='color: white; margin:0; text-align: center; font-family: sans-serif;'>
+            G-TIMES <span style='color: #2962FF;'>PRO TERMINAL</span>
+        </h2>
+    </div>
+    """)
     
     with gr.Row():
-        with gr.Column(scale=1):
-            with gr.Box():
-                ticker_input = gr.Textbox(label="Ticker Symbol", value="NVDA")
-                horizon_slider = gr.Slider(5, 128, value=30, label="Forecast Horizon (Days)")
-                analyze_btn = gr.Button("RUN QUANT ANALYSIS", variant="primary")
+        with gr.Column(scale=1, min_width=300):
+            with gr.Group():
+                ticker_in = gr.Textbox(label="SYMBOL", value="BTC-USD")
+                mode_in = gr.Radio(["Future Forecast", "Backtest (Reality Check)"], 
+                                 label="ANALYSIS MODE", value="Future Forecast")
+                days_in = gr.Slider(5, 128, value=14, label="Horizon (Days)")
+                
+                btn = gr.Button("EXECUTE STRATEGY", variant="primary")
             
-            output_signal = gr.HTML(label="Market Signal")
-            gr.Markdown("---")
-            gr.Markdown("### Terminal Info\n- **Model**: Google TimesFM-1.0\n- **Backend**: CPU PyTorch\n- **Freq**: Daily Close")
+            gr.HTML("<br>")
+            result_box = gr.HTML(label="Signal")
             
         with gr.Column(scale=4):
-            plot_output = gr.Plot()
+            plot_out = gr.Plot(label="Technical Chart")
 
-    analyze_btn.click(run_analysis, [ticker_input, horizon_slider], [plot_output, output_signal])
+    btn.click(run_analysis, [ticker_in, days_in, mode_in], [plot_out, result_box])
 
 demo.launch()