app.py

import gradio as gr
import numpy as np
import tensorflow as tf
import joblib
import yfinance as yf
from datetime import datetime, timedelta

# Load model and scaler
model = tf.keras.models.load_model("stock_price_lstm_model.keras")
scaler = joblib.load("scaler.save")

# Mapping of dropdown labels to Yahoo Finance tickers
ticker_map = {
    "TCS": "TCS.NS",
    "INFY": "INFY.NS",
    "RELIANCE": "RELIANCE.NS"
}

def get_last_60_closing_prices(stock, end_date_str):
    try:
        end_date = datetime.strptime(end_date_str, "%Y-%m-%d")
        start_date = end_date - timedelta(days=100)  # buffer for weekends/holidays
        ticker = ticker_map[stock]
        data = yf.download(ticker, start=start_date.strftime('%Y-%m-%d'), end=end_date_str)
        if data.shape[0] < 60:
            return None
        return data['Close'][-60:].values
    except Exception as e:
        print(e)
        return None

def predict_next_day_price(stock, last_date):
    closing_prices = get_last_60_closing_prices(stock, last_date)
    if closing_prices is None:
        return "❌ Could not fetch enough data. Try a different date closer to market activity."

    # Scale and reshape for LSTM input
    scaled_input = scaler.transform(closing_prices.reshape(-1, 1))
    X_test = np.reshape(scaled_input, (1, 60, 1))
    pred_scaled = model.predict(X_test)
    predicted_price = scaler.inverse_transform(pred_scaled)
    
    return f"📈 Predicted next day price for {stock}: ₹{predicted_price[0][0]:.2f}"

# Gradio UI
with gr.Blocks() as demo:
    gr.Markdown("# 📊 Stock Price Predictor using LSTM")
    gr.Markdown("Select a stock and input the last available date (YYYY-MM-DD). The model will fetch the previous 60 days of data and predict the next day's price.")
    
    stock_dropdown = gr.Dropdown(choices=["TCS", "INFY", "RELIANCE"], label="Select Stock")
    date_input = gr.Textbox(label="Enter last known date (YYYY-MM-DD)", placeholder="Example: 2025-06-10")
    output_text = gr.Textbox(label="Prediction Result")

    predict_btn = gr.Button("Predict Price")
    predict_btn.click(fn=predict_next_day_price, inputs=[stock_dropdown, date_input], outputs=output_text)

demo.launch()