app.py

import streamlit as st
import requests
from transformers import pipeline
import pandas as pd
from datetime import datetime, timedelta, timezone
import calendar

# --- Imports for Yahoo Scraper ---
import feedparser
from urllib.parse import quote
from bs4 import BeautifulSoup

# --- 1. การตั้งค่าหน้าและ API Key ---
st.set_page_config(page_title="📰 News Theme Analysis", layout="wide")

API_KEY = "88bc396d4eab4be494a4b86ec842db47"
NEWS_API_URL = "https://newsapi.org/v2/everything"

# --- 2. โหลดโมเดล (อัปเกรดแล้ว) ---
@st.cache_resource
def load_model():
    try:
        print("Loading model: MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli")
        classifier = pipeline("zero-shot-classification", model="MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli")
        print("Model loaded successfully!")
        return classifier
    except Exception as e:
        print(f"Error loading model: {e}")
        return None

classifier = load_model()
theme_labels = ["Earnings & Finance", "Product & Innovation", "Legal & Regulation", "Partnership & Deals", "Stock Price Movement", "Executive Changes"]

# --- 3. ฟังก์ชันดึงข่าว (วิธีที่ 1: NewsAPI) ---
@st.cache_data(ttl=3600)
def fetch_news_from_api(topic, api_key):
    """
    ดึงข่าวโดยใช้ NewsAPI.org (Reuters, Bloomberg...)
    """
    headers = {'X-Api-Key': api_key}
    from_date = (datetime.now() - timedelta(days=7)).strftime('%Y-%m-%d')
    trusted_sources = "reuters,bloomberg,the-wall-street-journal,marketwatch,cnbc"

    params = {
        'q': topic,
        'sortBy': 'publishedAt',
        'pageSize': 20,
        'from': from_date,
        'sources': trusted_sources
    }
    
    try:
        response = requests.get(NEWS_API_URL, headers=headers, params=params)
        response.raise_for_status()
        articles = response.json().get('articles', [])
        
        news_list = []
        for article in articles:
            if article.get('title') == "[Removed]":
                continue
            news_list.append({
                'title': article.get('title'),
                'summary': article.get('description', 'No description available.'),
                'link': article.get('url'),
                'published': article.get('publishedAt')
            })
        return news_list
    except Exception as e:
        st.error(f"NewsAPI Error: {e}")
        return []

# --- 4. ฟังก์ชันดึงข่าว (วิธีที่ 2: Yahoo Finance RSS) ---
def _parse_feed(url, max_articles=10):
    """
    Helper function for Yahoo Scraper
    """
    news_list = []
    try:
        cutoff_date_utc = datetime.now(timezone.utc) - timedelta(days=7)
        feed = feedparser.parse(url, agent='Mozilla/5.0')

        for entry in feed.entries:
            published_struct = entry.get('published_parsed')
            if not published_struct:
                continue

            try:
                article_timestamp = calendar.timegm(published_struct)
                article_date_utc = datetime.fromtimestamp(article_timestamp, tz=timezone.utc)
            except Exception:
                continue

            if article_date_utc >= cutoff_date_utc:
                raw_summary_html = entry.get('summary', '')
                soup = BeautifulSoup(raw_summary_html, 'html.parser')
                clean_summary = soup.get_text()

                news_list.append({
                    'title': entry.get('title', 'No title'),
                    'link': entry.get('link', '').split('&url=')[-1],
                    'summary': clean_summary[:300],
                    'published': article_date_utc.isoformat()
                })
                if len(news_list) >= max_articles:
                    break
    except Exception as e:
        print(f"Feedparser Error: {e}")
    return news_list

@st.cache_data(ttl=3600)
def search_yahoo_news(keyword, max_articles=10):
    """
    ดึงข่าวจาก Yahoo Finance (ผ่าน Google RSS)
    """
    try:
        safe_keyword = quote(keyword)
        url = f"https://news.google.com/rss/search?q={safe_keyword}+site:finance.yahoo.com&hl=en-US&gl=US&ceid=US:en"
        return _parse_feed(url, max_articles)
    except Exception as e:
        st.error(f"Yahoo Search Error: {e}")
        return []

# --- 5. ฟังก์ชันวิเคราะห์ ---
def analyze_themes(news_list, model, labels):
    results = []
    if not model:
        st.error("Model failed to load. Cannot analyze themes.")
        return pd.DataFrame()
        
    for news in news_list:
        title = news.get('title', '')
        if not title:
            continue
        try:
            analysis = model(title, candidate_labels=labels)
            news['theme'] = analysis['labels'][0]
            news['confidence'] = analysis['scores'][0]
            results.append(news)
        except Exception as e:
            print(f"Error analyzing title '{title}': {e}")
            
    return pd.DataFrame(results)

# --- 6. UI ---
st.title("📰 News Theme Analyzer")
st.markdown("Analyzes news from trusted sources (Reuters, Bloomberg, Yahoo...)")

topic = st.text_input("Enter a keyword or stock symbol:", "NVIDIA")

if st.button("Analyze News"):
    if not topic:
        st.warning("Please enter a topic.")
    elif not classifier:
         st.error("Model is not available. Please check logs.")
    else:
        with st.spinner(f"Fetching and analyzing news for '{topic}'..."):
            
            news_api = fetch_news_from_api(topic, API_KEY)
            news_yahoo = search_yahoo_news(topic)
            
            all_news = news_api + news_yahoo
            seen_titles = set()
            unique_news_items = []
            
            for item in all_news:
                if item['title'] not in seen_titles:
                    unique_news_items.append(item)
                    seen_titles.add(item['title'])
            
            if unique_news_items:
                df = analyze_themes(unique_news_items, classifier, theme_labels)
                
                if df.empty:
                    st.warning("Analysis failed, though news was fetched. Check logs.")
                else:
                    st.subheader("📊 Theme Distribution Summary")
                    theme_counts = df['theme'].value_counts()
                    total_articles = len(df)
                    
                    summary_data = []
                    for theme, count in theme_counts.items():
                        percentage = (count / total_articles) * 100
                        summary_data.append({
                            "Theme": theme,
                            "Percentage": f"{percentage:.1f}%",
                            "Articles": f"{count} / {total_articles}"
                        })
                    
                    st.dataframe(pd.DataFrame(summary_data), use_container_width=True)
                    st.markdown("---")

                    st.subheader(f"📰 Analysis Results (Found {len(df)} unique articles)")
                    
                    for index, row in df.iterrows():
                        st.markdown(f"### [{row['title']}]({row['link']})")
                        try:
                            date_str = pd.to_datetime(row['published']).strftime("%Y-%m-%d %H:%M")
                        except:
                            date_str = "N/A"
                        st.markdown(f"**Theme:** {row['theme']} (Confidence: {row['confidence']:.2f}) \n**Published:** {date_str}")
                        st.markdown(f"**Summary:** {row['summary']}")
                        st.markdown("---")
            else:
                st.error("No news found from any source.")