peachy-naomi/naija-news-hub

# 🇳🇬 Naija News Hub A live, full-stack Nigerian news aggregator — built end-to-end with a scheduled ETL pipeline, an ML-powered categorization layer, a FastAPI backend, and a modern Next.js frontend with infinite scroll, search, and a dark/light theme. Naija News Hub pulls headlines from five major Nigerian publishers, cleans and enriches them (real photos, article categories, view-based popularity), and serves them through a searchable, filterable web app that refreshes every 10 minutes — all without ever rewriting or republishing full article content. Every story links straight back to its original publisher. --- ## ✨ Features - **Live aggregation** from 5 sources — Punch, Vanguard, Premium Times, Daily Post, and PM News — refreshed every 10 minutes via Apache Airflow - **Automatic deduplication** — the same story is never inserted twice, even across repeated runs - **Real per-article images** — extracted from each publisher's actual article page (`og:image` / `twitter:image` meta tags), not just generic feed thumbnails - **ML-based topic categorization** — a zero-shot NLI classifier (`valhalla/distilbart-mnli-12-1`) tags every article into Politics, Sports, Business, Entertainment, Technology, Health, Crime, or World with no manual labeling required - **Popularity tracking** — view counts recorded on click-through, surfaced in a homepage carousel of the most-read stories - **Full-text search & multi-dimensional filtering** — by keyword, source, category, and date (year → month → day) - **Archive browsing** — explore any day, month, or year of coverage - **Resilient scraping** — automatic fallback through direct requests → residential proxy (Oxylabs) → graceful skip, so one blocked source never takes down the whole pipeline - **Dark / light theme toggle** with a custom editorial design system (glassmorphism, animated live indicator, no generic "AI dark mode" look) - **Memory-optimized infinite scroll** — windowed rendering (TanStack Virtual) keeps the DOM lig …