rss_processor.py

import os
import feedparser
from huggingface_hub import HfApi, login
from langchain.vectorstores import Chroma
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.docstore.document import Document
import shutil
import logging

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

# Hugging Face setup
HF_API_TOKEN = os.getenv("DEMO_HF_API_TOKEN", "YOUR_HF_API_TOKEN")
REPO_ID = "broadfield-dev/news-rag-db"
LOCAL_DB_DIR = "chroma_db"

# Explicitly login to Hugging Face Hub
login(token=HF_API_TOKEN)
hf_api = HfApi()

# RSS feeds
RSS_FEEDS = [
    "https://www.sciencedaily.com/rss/top/science.xml",
    "https://www.horoscope.com/us/horoscopes/general/rss/horoscope-rss.aspx",
    "http://rss.cnn.com/rss/cnn_allpolitics.rss",
    "https://phys.org/rss-feed/physics-news/",
    "https://www.spaceweatherlive.com/en/news/rss",
    "https://weather.com/feeds/rss",
    "https://www.wired.com/feed/rss",
    "https://www.nasa.gov/rss/dyn/breaking_news.rss",
    "https://www.nationalgeographic.com/feed/",
    "https://www.nature.com/nature.rss",
    "https://www.scientificamerican.com/rss/",
    "https://www.newscientist.com/feed/home/",
    "https://www.livescience.com/feeds/all",
    "https://astrostyle.com/feed/",
    "https://www.vogue.com/feed/rss",
    "https://feeds.bbci.co.uk/news/politics/rss.xml",
    "https://www.reuters.com/arc/outboundfeeds/newsletter-politics/?outputType=xml",
    "https://www.politico.com/rss/politics.xml",
    "https://thehill.com/feed/",
    "https://www.aps.org/publications/apsnews/updates/rss.cfm",
    "https://www.quantamagazine.org/feed/",
    "https://www.sciencedaily.com/rss/matter_energy/physics.xml",
    "https://physicsworld.com/feed/",
    "https://www.swpc.noaa.gov/rss.xml",
    "https://feeds.bbci.co.uk/weather/feeds/rss/5day/world/",
    "https://www.weather.gov/rss",
    "https://www.foxweather.com/rss",
    "https://techcrunch.com/feed/",
    "https://arstechnica.com/feed/",
    "https://gizmodo.com/rss",
    "https://www.theverge.com/rss/index.xml",
    "https://www.space.com/feeds/all",
    "https://www.universetoday.com/feed/",
    "https://skyandtelescope.org/feed/",
    "https://www.esa.int/rss",
    "https://www.smithsonianmag.com/rss/",
    "https://www.popsci.com/rss.xml",
    "https://www.discovermagazine.com/rss",
    "https://www.atlasobscura.com/feeds/latest"
]

# Embedding model and vector DB
embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
vector_db = Chroma(persist_directory=LOCAL_DB_DIR, embedding_function=embedding_model)

def fetch_rss_feeds():
    articles = []
    seen_articles = set()  # Track unique articles by title, link, and description
    for feed_url in RSS_FEEDS:
        try:
            logger.info(f"Fetching feed: {feed_url}")
            feed = feedparser.parse(feed_url)
            if feed.bozo:
                logger.warning(f"Failed to parse {feed_url}: {feed.bozo_exception}")
                continue
            unique_count = 0
            for entry in feed.entries[:5]:
                title = entry.get("title", "No Title")
                link = entry.get("link", "")
                description = entry.get("summary", entry.get("description", "No Description"))
                # Create a unique key for deduplication (title, link, and description for stricter uniqueness)
                article_key = f"{title}|{link}|{description[:50]}"  # Use first 50 chars of description to avoid overly long keys
                if article_key not in seen_articles:
                    seen_articles.add(article_key)
                    unique_count += 1
                    image = entry.get("media_content", [{}])[0].get("url") or entry.get("media_thumbnail", [{}])[0].get("url") or ""
                    articles.append({
                        "title": title,
                        "link": link,
                        "description": description,
                        "published": entry.get("published", "Unknown Date"),
                        "category": categorize_feed(feed_url),
                        "image": image if image else "",
                    })
            logger.info(f"Processed {unique_count} unique entries from {feed_url}")
        except Exception as e:
            logger.error(f"Error fetching {feed_url}: {e}")
    return articles

def categorize_feed(url):
    if "sciencedaily" in url or "phys.org" in url:
        return "Science & Physics"
    elif "horoscope" in url:
        return "Astrology"
    elif "politics" in url:
        return "Politics"
    elif "spaceweather" in url or "nasa" in url:
        return "Solar & Space"
    elif "weather" in url:
        return "Earth Weather"
    else:
        return "Cool Stuff"

def process_and_store_articles(articles):
    documents = []
    seen_docs = set()  # Additional de-duplication at DB level
    for article in articles:
        try:
            key = f"{article['title']}|{article['link']}|{article['description'][:50]}"
            if key not in seen_docs:
                seen_docs.add(key)
                metadata = {
                    "title": article["title"] or "No Title",
                    "link": article["link"] or "",
                    "original_description": article["description"] or "No Description",
                    "published": article["published"] or "Unknown Date",
                    "category": article["category"] or "Uncategorized",
                    "image": article["image"] or "",
                }
                doc = Document(
                    page_content=article["description"] or "No Description",
                    metadata=metadata
                )
                documents.append(doc)
        except Exception as e:
            logger.error(f"Error processing article {article['title']}: {e}")
    try:
        vector_db.add_documents(documents)
        vector_db.persist()
        logger.info("Vector DB persisted")
    except Exception as e:
        logger.error(f"Error adding documents to vector DB: {e}")
    upload_to_hf_hub()

def upload_to_hf_hub():
    if os.path.exists(LOCAL_DB_DIR):
        try:
            hf_api.create_repo(repo_id=REPO_ID, repo_type="dataset", exist_ok=True, token=HF_API_TOKEN)
            logger.info(f"Repository {REPO_ID} created or exists.")
        except Exception as e:
            logger.error(f"Error creating repo: {e}")
            return
        for root, _, files in os.walk(LOCAL_DB_DIR):
            for file in files:
                local_path = os.path.join(root, file)
                remote_path = os.path.relpath(local_path, LOCAL_DB_DIR)
                try:
                    hf_api.upload_file(
                        path_or_fileobj=local_path,
                        path_in_repo=remote_path,
                        repo_id=REPO_ID,
                        repo_type="dataset",
                        token=HF_API_TOKEN
                    )
                    logger.info(f"Uploaded {file} to {REPO_ID}")
                except Exception as e:
                    logger.error(f"Error uploading file {file}: {e}")
        logger.info(f"Database uploaded to: {REPO_ID}")