Create app.py

app.py

@@ -0,0 +1,113 @@
+# app.py
+import yfinance as yf
+import pandas as pd
+import numpy as np
+from datetime import datetime, timedelta
+import gradio as gr
+import matplotlib.pyplot as plt
+from io import BytesIO
+from statsmodels.tsa.arima.model import ARIMA
+from mcp import Client
+from langchain.prompts import PromptTemplate
+from langchain.chains import LLMChain
+from langchain.llms import OpenAI  # Replace with local MCP-compatible LLM if needed
+import re
+
+# ---------------- MCP Setup ----------------
+client = Client(api_key="YOUR_GROQ_API_KEY")  # optional if MCP uses Groq API
+
+# ---------------- Helper Functions ----------------
+def fetch_metal_data(ticker="GC=F", period="1mo", interval="1d"):
+    """Fetch historical data for a metal."""
+    df = yf.download(ticker, period=period, interval=interval)
+    df = df.reset_index()
+    df = df[['Date', 'Close']]
+    df.rename(columns={'Close': 'Price'}, inplace=True)
+    return df
+
+def predict_next_days(df, days=5):
+    """Predict next 'days' prices using ARIMA."""
+    try:
+        model = ARIMA(df['Price'], order=(2,1,2))
+        model_fit = model.fit()
+        forecast = model_fit.forecast(steps=days)
+        forecast_dates = [df['Date'].iloc[-1] + timedelta(days=i+1) for i in range(days)]
+        return pd.DataFrame({"Date": forecast_dates, "Predicted_Price": forecast})
+    except Exception as e:
+        return pd.DataFrame({"Date": [], "Predicted_Price": []})
+
+def plot_forecast(df, forecast_df, metal):
+    plt.figure(figsize=(10,5))
+    plt.plot(df['Date'], df['Price'], label="Historical")
+    plt.plot(forecast_df['Date'], forecast_df['Predicted_Price'], label="Forecast", linestyle="--")
+    plt.title(f"{metal.capitalize()} Price Forecast")
+    plt.xlabel("Date")
+    plt.ylabel("Price")
+    plt.legend()
+    buf = BytesIO()
+    plt.savefig(buf, format='png')
+    buf.seek(0)
+    plt.close()
+    return buf
+
+# ---------------- LangChain Prompt ----------------
+prompt_template = """
+Extract the metal and number of forecast days from user query.
+User Query: "{query}"
+Return as JSON with keys: "metal" ("gold" or "silver") and "days" (integer).
+"""
+
+prompt = PromptTemplate(
+    input_variables=["query"],
+    template=prompt_template
+)
+llm = OpenAI(temperature=0)  # Replace with MCP-compatible LLM if needed
+chain = LLMChain(llm=llm, prompt=prompt)
+
+def parse_query(query):
+    # Use LangChain to parse
+    try:
+        response = chain.run(query)
+        # Extract values from JSON-like string
+        metal = re.search(r'"metal"\s*:\s*"(\w+)"', response).group(1).lower()
+        days = int(re.search(r'"days"\s*:\s*(\d+)', response).group(1))
+        days = min(max(days, 1), 7)  # limit 1-7 days
+        return metal, days
+    except:
+        # fallback
+        return "gold", 3
+
+# ---------------- MCP Wrapped Function ----------------
+@client.wrap
+def metal_forecast_agent(query: str):
+    metal, days = parse_query(query)
+    ticker_map = {"gold": "GC=F", "silver": "SI=F"}
+    ticker = ticker_map.get(metal, "GC=F")
+    df = fetch_metal_data(ticker=ticker)
+    forecast_df = predict_next_days(df, days=days)
+    latest_price = float(df['Price'].iloc[-1])
+    plot_buf = plot_forecast(df, forecast_df, metal)
+    return {
+        "metal": metal,
+        "latest_price": latest_price,
+        "forecast": forecast_df.to_dict(orient="records"),
+        "plot_buf": plot_buf
+    }
+
+# ---------------- Gradio Interface ----------------
+def gr_metal_agent(query):
+    result = metal_forecast_agent(query)
+    latest_text = f"Latest {result['metal'].capitalize()} Price: {result['latest_price']:.2f}"
+    forecast_df = pd.DataFrame(result['forecast'])
+    return latest_text, forecast_df, result['plot_buf']
+
+iface = gr.Interface(
+    fn=gr_metal_agent,
+    inputs=gr.Textbox(label="Ask about metal prices, e.g., 'gold next 5 days'"),
+    outputs=[gr.Textbox(label="Latest Price"), gr.Dataframe(label="Forecast"), gr.Image(label="Forecast Plot")],
+    title="Metal Price Predictor (MCP + LangChain)",
+    description="Ask naturally for gold/silver forecast and get live + predicted prices."
+)
+
+if __name__ == "__main__":
+    iface.launch()