খবরঘর — Bangladesh News Hub

A news discovery portal: pick a newspaper, browse its latest headlines from your chosen sections, search, and open the original article on the publisher's site. Currently: Prothom Alo (Bangla), The Daily Star, The Guardian and The New York Times (English).

npm install
npm start        # http://localhost:3000
npm test

Architecture

server/
  newspapers/        one config file per paper + index.js registry   <- the only place to touch to add a paper
  lib/rss.js         RSS 2.0 parsing + normalisation into a common Article shape
  lib/text.js        sanitising (tags stripped, entities decoded, only https URLs kept)
  lib/service.js     fetch all feeds of a paper, merge, de-duplicate, cache (5 min), serve stale on failure
  sections.js        canonical sections (Bangladesh, Politics, World, …) with Bangla/English labels
  index.js           Express: GET /api/newspapers, GET /api/news/:id[?refresh=1], static files, CSP
  lib/images.js      og:image lookup for feeds without images (cached, throttled)
public/              no-build ES-module frontend (components/card.js = NewsCard, components/views.js, app.js)

The browser only talks to /api/* — newspaper sites are fetched server-side (no CORS issues, one shared cache, one place to sanitise). No API keys are needed or present.

Adding a newspaper

1. Create server/newspapers/ittefaq.js:

export default {
  id: 'ittefaq', name: 'Ittefaq', nativeName: 'ইত্তেফাক', language: 'bn',
  monogram: 'ই', brandColor: '#1b7f4b', logo: null, homepage: 'https://www.ittefaq.com.bd',
  feeds: [
    { url: 'https://…/politics/rss.xml', section: 'politics' },   // section fixed per feed, or…
    { url: 'https://…/rss.xml' },                                  // …derived via sectionByPath (see prothom-alo.js)
  ],
  // optional: sectionByPath
};

2. Add it to the newspapers array in server/newspapers/index.js.

The header tab, section navigation, search, caching, refresh and error handling all pick it up automatically. A paper whose feed isn't standard RSS 2.0 would need a small custom parser added to lib/rss.js.

Behaviour notes

• Landing page: the default All sources tab merges every newspaper into one newest-first list (GET /api/news returns all papers in one request; one failing paper doesn't hide the rest). Category chips (Bangladesh, Politics, World, Business, Opinion) filter it, and the By source toggle groups it per newspaper. Cards show source, category, exact time and a 1–2 sentence summary.
• Switching is client-side; loaded papers are kept in memory, so switching back is instant and revalidated in the background if older than 5 minutes. URL hash (#/daily-star/sports) is shareable.
• Refresh button forces a refetch (server throttles to once per 30 s per paper); the page also refreshes itself every 5 min while visible. If a fetch fails but older data exists, the older data is shown with a notice and Retry; with no data, a full error state with Retry.
• Search filters the selected paper's loaded articles (title, summary, category), Bangla and English.
• Thumbnails: each card shows the article's own image, taken from the feed (media:content / enclosure; the widest one) or, for The Daily Star (no images in its RSS), from the article's og:image tag (fetched politely: 3 at a time, cached 24 h, only for stories shown). Images are hot-linked from the publisher (referrerpolicy="no-referrer"), never copied or stored; a branded placeholder with the source initials is used when an image is missing or fails to load. They remain the publishers' copyright, so keep the link-back to the original article.
• Focused sections: each paper has a sections allowlist in its config (Prothom Alo: politics + economy; The Daily Star: Bangladesh + opinion; The Guardian: Bangladesh, world, business, opinion; The New York Times: world, business, opinion). Everything else is dropped on the server, so the feed, search and Telegram alerts only cover those sections. Stories are sorted newest first and de-duplicated; up to 30 per section are kept (maxPerSection). Prothom Alo combines its RSS feed (summaries) with its official news sitemap (news_sitemap.xml, advertised in its robots.txt; ~600 recent stories, no summaries), so its sections have 20–30 stories each. Telegram alerts ignore stories older than 6 hours.
• Only titles/summaries are shown; nothing of article bodies is stored or displayed.

Telegram alerts

While the server runs, it checks every paper every 3 minutes and sends each newly published headline (title, section, source, link; text only) to a Telegram chat.

1. In Telegram open @BotFather → /newbot → copy the bot token.
2. Open your new bot, press Start and send "hi".
3. Copy .env.example to .env, paste the token as TELEGRAM_BOT_TOKEN=.
4. Run npm run telegram:chatid, paste the printed id as TELEGRAM_CHAT_ID= in .env.
5. Run npm run telegram:test (you should get a message), then restart npm start.

Notes: the first run only records current stories (no flood of old news); at most 8 alerts per paper per round; a paper can opt out with notify: false in its config (The Guardian and The New York Times do); seen stories are remembered in data/seen.json so restarts don't repeat alerts. Alerts only run while the server is running. .env is git-ignored and never sent to the browser.