newNews

A curated industry news feed for a ~50-person team at Microsoft working on Azure / Azure Government data-platform products (Power Query Online, HDInsight, OneLake, Service Assist, Microsoft Fabric, Data Factory, Dataflows Gen2, Pipelines, etc.).

This is a proof of concept: a static Next.js site hosted on GitHub Pages, fed by a GitHub Actions cron job that fetches news from BBC, The Guardian, Hacker News, and Reddit, scores each item against role + product keyword profiles, and commits a public/data/feed.json artifact the client reads.

Public news only. No Microsoft-confidential content. No internal sources.

Architecture

GitHub Action (every 20 min)
  └─ scripts/fetch-feed.ts
       ├─ BBC RSS (tech/world/business)
       ├─ Guardian API (or RSS fallback)
       ├─ Hacker News Firebase API
       └─ Reddit multireddit JSON
  └─ score + tag with role / product keywords
  └─ commit public/data/feed.json (only if changed)
       │
       ▼
Static Next.js export → GitHub Pages
  └─ Role picker (Manager / Dev / SRE / PM)
  └─ Product channel toggles
  └─ Filtered, ranked feed (cached in localStorage)

Local development

npm install
npm run fetch-feed     # generates public/data/feed.json
npm run dev            # http://localhost:3000
npm test               # unit tests for scoring
npm run build          # static export to ./out

Tuning curation

Keyword maps live in lib/curation/keywords.ts. Edit them directly — no code changes required. The scoring formula is in lib/curation/score.ts.

Each card shows tag chips (and an optional debug breakdown) so you can see why an article surfaced and tune accordingly.

Deployment

1. Push to main — deploy-pages.yml builds and publishes the site.
2. In repo Settings → Pages, set source to GitHub Actions.
3. (Optional) Add a GUARDIAN_API_KEY repo secret for richer Guardian data (otherwise we fall back to Guardian RSS).
4. The refresh-feed.yml workflow runs every 20 minutes and commits an updated feed.json when content changes.

Sources & rate limiting

Source	Endpoint	Auth
BBC	RSS	none
Guardian	content.guardianapis.com/search	free key (optional)
Hacker News	hacker-news.firebaseio.com/v0	none
Reddit	reddit.com/r/<multireddit>/top.json	none

All fetches happen server-side in the Action — clients only read the prebuilt JSON, so per-user rate limits don't apply.