Create app.py

app.py

@@ -0,0 +1,78 @@
+import gradio as gr
+import pandas as pd
+import torch
+import matplotlib.pyplot as plt
+import numpy as np
+import yfinance as yf
+from chronos import Chronos2Pipeline
+
+# 1. Load the Chronos-2 Model
+# Using 'small' for faster performance on Free Tier CPUs
+device = "cuda" if torch.cuda.is_available() else "cpu"
+pipeline = Chronos2Pipeline.from_pretrained(
+    "amazon/chronos-2",
+    device_map=device,
+    torch_dtype=torch.float32,
+)
+
+def forecast_stock(ticker, forecast_days):
+    # 2. Fetch Historical Data
+    data = yf.download(ticker, period="1y")
+    if data.empty:
+        return None, "Error: Ticker not found."
+    
+    # Extract closing prices
+    df = data[['Close']].reset_index()
+    context = torch.tensor(df['Close'].values)
+    
+    # 3. Perform Inference
+    # Chronos produces probabilistic forecasts (multiple samples)
+    forecast = pipeline.predict(context, forecast_days) # shape: [1, num_samples, forecast_days]
+    
+    # Calculate quantiles for the "dotted" area
+    # 0.5 is the median (main line), 0.1 and 0.9 create the confidence interval
+    low, median, high = np.quantile(forecast[0].numpy(), [0.1, 0.5, 0.9], axis=0)
+    
+    # 4. Plotting
+    plt.figure(figsize=(12, 6))
+    
+    # Previous Data (Solid Line)
+    history_indices = np.arange(len(context))
+    plt.plot(history_indices, context, color='royalblue', label="Historical Price", linewidth=2)
+    
+    # Forecast Data (Dotted Section)
+    forecast_indices = np.arange(len(context), len(context) + forecast_days)
+    
+    # Median Forecast (Dashed)
+    plt.plot(forecast_indices, median, color='tomato', linestyle='--', label="Median Forecast")
+    
+    # Shaded Uncertainty Area
+    plt.fill_between(forecast_indices, low, high, color='tomato', alpha=0.2, label="80% Confidence Interval")
+    
+    plt.title(f"Chronos-2 Forecast for {ticker}", fontsize=14)
+    plt.xlabel("Days (Relative)")
+    plt.ylabel("Price (USD)")
+    plt.legend()
+    plt.grid(True, alpha=0.3)
+    
+    return plt.gcf()
+
+# 5. Build Gradio Interface
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 📈 Amazon Chronos-2 Financial Forecaster")
+    gr.Markdown("Input a stock ticker (e.g., AAPL, BTC-USD) to see the AI predict future trends.")
+    
+    with gr.Row():
+        ticker_input = gr.Textbox(label="Stock Ticker", value="AAPL")
+        days_input = gr.Slider(minimum=5, maximum=60, step=1, label="Forecast Horizon (Days)", value=30)
+    
+    forecast_btn = gr.Button("Generate Forecast", variant="primary")
+    plot_output = gr.Plot(label="Time Series Forecast")
+
+    forecast_btn.click(
+        fn=forecast_stock, 
+        inputs=[ticker_input, days_input], 
+        outputs=plot_output
+    )
+
+demo.launch()