app.py

import os
import yfinance as yf
import pandas as pd
from newsapi import NewsApiClient
from transformers import pipeline
import tensorflow as tf
from tensorflow import keras
from sklearn.preprocessing import MinMaxScaler
import numpy as np
from datetime import datetime, timedelta
import alpaca_trade_api as tradeapi
from langchain_community.llms import HuggingFaceEndpoint
import logging

# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Load environment variables with fallback
NEWSAPI_KEY = os.getenv('NEWSAPI_KEY', 'your_newsapi_key')
ALPACA_API_KEY = os.getenv('ALPACA_API_KEY', 'your_alpaca_api_key')
ALPACA_SECRET_KEY = os.getenv('ALPACA_SECRET_KEY', 'your_alpaca_secret_key')
APCA_API_KEY_ID = os.getenv('APCA_API_KEY_ID', ALPACA_API_KEY)
APCA_API_SECRET_KEY = os.getenv('APCA_API_SECRET_KEY', ALPACA_SECRET_KEY)

# Check if all necessary keys are available
if not all([NEWSAPI_KEY, APCA_API_KEY_ID, APCA_API_SECRET_KEY]):
    raise ValueError("Ensure all API keys and secret keys are set as environment variables.")

# Initialize NewsAPI client
newsapi = NewsApiClient(api_key=NEWSAPI_KEY)
# Initialize the chatbot
HF_TOKEN = os.getenv("HF_TOKEN")
llm = HuggingFaceEndpoint(
    #repo_id="google/gemma-2b-it",
    #repo_id="abhinand/gemma-2b-tamil",
    repo_id="mrm8488/distilroberta-finetuned-financial-news-sentiment-analysis",
    #repo_id="google/gemma-2b",
    
    task="financial-analysis",
    max_new_tokens=512,
    top_k=5,
    temperature=0.2,
    repetition_penalty=1.03,
    timeout=300,
    huggingfacehub_api_token=HF_TOKEN
)
# Initialize Alpaca Trade API client
alpaca_api = tradeapi.REST(APCA_API_KEY_ID, APCA_API_SECRET_KEY, base_url='https://paper-api.alpaca.markets')

def collect_market_data(ticker):
    data = yf.download(ticker, start=(datetime.now() - timedelta(days=365)).strftime('%Y-%m-%d'), end=datetime.now().strftime('%Y-%m-%d'))
    data.to_csv(f'{ticker}_market_data.csv')
    logger.info(f'Market data for {ticker} collected successfully.')

def collect_news_data(query, from_date, to_date):
    all_articles = newsapi.get_everything(q=query, from_param=from_date, to=to_date, language='en', sort_by='relevancy')
    if all_articles['status'] == 'ok':
        articles_df = pd.DataFrame(all_articles['articles'])
        articles_df.to_csv('news_data.csv')
        logger.info(f'News data for {query} collected successfully.')
    else:
        logger.error(f'Error collecting news data: {all_articles["message"]}')

def perform_sentiment_analysis():
    sentiment_pipeline = pipeline("sentiment-analysis")
    try:
        news_data = pd.read_csv('news_data.csv')
        news_data['sentiment'] = news_data['description'].apply(lambda x: sentiment_pipeline(x)[0]['label'] if pd.notna(x) else 'NEUTRAL')
        news_data.to_csv('sentiment_data.csv', index=False)
        logger.info('Sentiment analysis performed successfully.')
    except Exception as e:
        logger.error(f'Error performing sentiment analysis: {e}')

def train_price_prediction_model(ticker):
    data = pd.read_csv(f'{ticker}_market_data.csv')
    data = data[['Date', 'Close']].set_index('Date')
    scaler = MinMaxScaler(feature_range=(0, 1))
    scaled_data = scaler.fit_transform(data)

    X = []
    y = []

    for i in range(60, len(scaled_data)):
        X.append(scaled_data[i-60:i, 0])
        y.append(scaled_data[i, 0])

    X = np.array(X)
    y = np.array(y)
    X = np.reshape(X, (X.shape[0], X.shape[1], 1))

    model = keras.Sequential([
        keras.layers.LSTM(50, return_sequences=True, input_shape=(X.shape[1], 1)),
        keras.layers.LSTM(50, return_sequences=False),
        keras.layers.Dense(25),
        keras.layers.Dense(1)
    ])

    model.compile(optimizer='adam', loss='mean_squared_error')
    model.fit(X, y, batch_size=1, epochs=1)

    model.save(f'{ticker}_price_prediction_model.h5')
    logger.info('Price prediction model trained successfully.')

def make_trade_decision(ticker):
    model = keras.models.load_model(f'{ticker}_price_prediction_model.h5')
    data = pd.read_csv(f'{ticker}_market_data.csv')
    last_60_days = data['Close'].tail(60).values
    last_60_days_scaled = MinMaxScaler(feature_range=(0, 1)).fit_transform(last_60_days.reshape(-1, 1))

    X_test = []
    X_test.append(last_60_days_scaled)
    X_test = np.array(X_test)
    X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], 1))

    predicted_price = model.predict(X_test)
    predicted_price = MinMaxScaler(feature_range=(0, 1)).inverse_transform(predicted_price)

    current_price = yf.download(ticker, period='1d')['Close'].values[0]

    if predicted_price > current_price:
        alpaca_api.submit_order(
            symbol=ticker,
            qty=1,
            side='buy',
            type='market',
            time_in_force='gtc'
        )
        logger.info(f'Bought 1 share of {ticker}')
    else:
        alpaca_api.submit_order(
            symbol=ticker,
            qty=1,
            side='sell',
            type='market',
            time_in_force='gtc'
        )
        logger.info(f'Sold 1 share of {ticker}')

if __name__ == "__main__":
    TICKER = 'AAPL'
    QUERY = 'Apple Inc'
    FROM_DATE = (datetime.now() - timedelta(days=30)).strftime('%Y-%m-%d')
    TO_DATE = datetime.now().strftime('%Y-%m-%d')

    collect_market_data(TICKER)
    collect_news_data(QUERY, FROM_DATE, TO_DATE)
    perform_sentiment_analysis()
    train_price_prediction_model(TICKER)
    make_trade_decision(TICKER)