app.py

import streamlit as st
import requests
import re
import torch
from transformers import BertTokenizer, BertForSequenceClassification

# Load tokenizer and model
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
model_path = '/Users/kartikrathi/Documents/News_sentiment_analysis/model'
model = BertForSequenceClassification.from_pretrained(model_path)

# Function to analyze sentiment
def analyze_sentiment(text):
    # Tokenize inputs
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)
    
    # Perform inference
    with torch.no_grad():
        outputs = model(**inputs)
    
    # Get predicted logits
    logits = outputs.logits
    
    # Determine predicted sentiment
    predicted_class_id = torch.argmax(logits, dim=1).item()
    sentiment = {0: "positive", 1: "negative", 2: "neutral"}
    return sentiment[predicted_class_id]

# Function to fetch news using an API
def fetch_stock_news(symbol):
    api_token = '6679177763bcd9.48465511'  # Replace with your actual API token
    url = f'https://eodhd.com/api/news?s={symbol}&offset=0&limit=2&api_token={api_token}&fmt=json'
    
    try:
        response = requests.get(url)
        response.raise_for_status()  # Raise exception for bad status codes
        data = response.json()
        
        news_list = []
        for article in data:
            title = article.get('title', 'Title not available')
            content = article.get('content', 'Content not available')
            url = article.get('url', 'URL not available')
            date = article.get('date', 'Date not available')
            
            # Remove unwanted sections from content
            cleaned_content = remove_sections(content)
            
            news_list.append({'title': title, 'content': cleaned_content, 'url': url, 'date': date})
        
        return news_list
    
    except requests.exceptions.RequestException as e:
        print(f"Error fetching news for {symbol}: {e}")
        return None

# Function to remove unwanted sections
def remove_sections(content):
    # Patterns to remove
    patterns = [
        r'About\s+.*?[\n\r]+',  # 'About {stock}' section
        r'Safe\s+Harbor.*',  # 'Safe Harbor' section
        r'Story\s+continues.*',  # 'Story continues'
        r'Visit\s+www\.infosys\.com.*',  # 'Visit www.infosys.com...'
        r'Logo',  # 'Logo'
    ]
    
    for pattern in patterns:
        content = re.sub(pattern, '', content, flags=re.IGNORECASE)
    
    return content.strip()

# Streamlit app with multi-page support
def main():
    st.sidebar.title('Navigation')
    page = st.sidebar.radio("Go to", ('Home', 'Portfolio & News'))

    if page == 'Home':
        st.title("Financial News Sentiment Analysis")
        st.markdown("""
            This app performs sentiment analysis on financial news using FinBERT model.
            """)

        # Input text box for user input
        user_input = st.text_area("Enter your financial news text here:", height=200)

        # Perform sentiment analysis when user clicks the button
        if st.button("Analyze"):
            if user_input:
                sentiment = analyze_sentiment(user_input)
                if sentiment == "positive":
                    st.markdown(f"<p style='color:green;font-size:40px;font-weight:bold'>{sentiment}</p>", unsafe_allow_html=True)
                elif sentiment == "negative":
                    st.markdown(f"<p style='color:red;font-size:40px;font-weight:bold'>{sentiment}</p>", unsafe_allow_html=True)
                elif sentiment == "neutral":
                    st.markdown(f"<p style='color:blue;font-size:40px;font-weight:bold'>{sentiment}</p>", unsafe_allow_html=True)
            else:
                st.warning("Please enter some text.")

    elif page == 'Portfolio & News':
        st.title('Portfolio & News')

        # Sidebar to manage portfolio and fetch news
        st.sidebar.subheader('Manage Portfolio & Fetch News')
        st.sidebar.info("Enter your portfolio/company names here to fetch news.")

        # Input fields for portfolio management and news fetching
        portfolio = st.sidebar.text_area("Enter your portfolio/company names (one per line):", height=200)
        
        if st.sidebar.button("Save"):
            # Add ".NSE" suffix to each company name
            shares = portfolio.split("\n")
            shares = [share.strip() + ".NSE" for share in shares if share.strip()]
            st.sidebar.markdown("**Shares**")
            st.sidebar.markdown("\n".join(shares))

        # Button to fetch news and perform sentiment analysis
        if st.sidebar.button("Fetch News & Analyze Sentiment"):
            if portfolio:
                companies = portfolio.split("\n")
                companies = [company.strip() + ".NSE" for company in companies if company.strip()]
                for company in companies:
                    st.subheader(f"Latest News for {company}:")
                    news = fetch_stock_news(company)

                    if news:
                        for article in news:
                            title = article['title']
                            content = article['content']
                            url = article['url']
                            date = article['date']
                            st.markdown(f"**Title:** [{title}]({url}) <span style='color:green;font-size:12px;'>({date})</span>", unsafe_allow_html=True)
                            
                            # Display a truncated version of the content
                            truncated_content = content[:200] + "..."
                            st.markdown(f"**Content:** {truncated_content}")
                            
                            # Expander for full content
                            with st.expander("Expand more"):
                                st.markdown(f"{content}")

                            st.markdown("---")

                            # Analyze sentiment of news content (assuming content is available)
                            if content:
                                sentiment = analyze_sentiment(content)
                                if sentiment == "positive":
                                    sentiment_color = "green"
                                elif sentiment == "negative":
                                    sentiment_color = "red"
                                elif sentiment == "neutral":
                                    sentiment_color = "blue"
                                
                                st.markdown(f"**Sentiment:** <span style='color:{sentiment_color};font-weight:bold'>{sentiment}</span>", unsafe_allow_html=True)
                                st.markdown("---")
                    else:
                        st.warning(f"No news articles found for {company}.")
            else:
                st.warning("Please enter portfolio/company names to fetch news.")

if __name__ == '__main__':
    main()