vielhuber/extrablatt

A simple news aggregator in php.

2.4.3 2026-08-14 07:07 UTC

Requires

• php: >=8.3
• ext-gd: *
• ext-pdo_sqlite: *
• vielhuber/aihelper: ^4.1.2

MIT

Published on Packagist.org on 2026-08-14 07:08 UTC

README

📰 extrablatt 📰

a simple news aggregator in politically charged times. pulls articles from configurable rss feeds (plus reddit, hacker news and x via cookie-authenticated scrape), stores them in sqlite, detects paywalls, fetches thumbnails, categorises through an llm, and opens single articles through an archive.ph proxy with mobile-friendly css rewrites. installable as a progressive web app.

installation

mkdir extrablatt
cd extrablatt
composer require vielhuber/extrablatt
./vendor/bin/extrablatt-init

after install, edit:

• .data/config.json: papers (see schema below)
• .data/.env: AI_API_KEY / AUTH_PASSWORD / AI_PROVIDER / AI_MODEL
• .data/cookies/: drop cookie exports per host into
• .data/database.sqlite: restore database (optional)

config.json schema

{
    "papers": {
        "<paper-key>": {
            "url": "https://example.com",
            "label": "Display Name",
            "rss": "https://example.com/feed.xml",
            "default_image": "https://example.com/fallback.png",
            "stub_markers": ["Subscribe to read", "Premium content"]
        }
    }
}

• default_image (optional): fallback thumbnail when the RSS item carries no image.
• stub_markers (optional): substrings present in the archive.ph snapshot of a PLUS article when it's only a teaser, so the snapshot is dropped instead of surfaced as if it were the full text.
• following (optional, medium://home only): explicit list of followed handles (@username) and publication slugs whose RSS feeds get aggregated. Without it the list is discovered from the cookie-authenticated /following page, which Cloudflare challenges from datacenter IPs — so shared hosting needs this key.
• Special rss schemes: reddit://home, x://home and hackernews://best activate the dedicated scrapers in place of XML parsing; medium://home aggregates the followed authors' RSS feeds; ct://archiv indexes the complete c't article archive (needs no login, backfills a bounded number of issues per scrape, rating = editorial page count).

the Watch tab reads the Google Health API (daily rollups of the paired watch). set GOOGLE_HEALTH_CLIENT_ID / GOOGLE_HEALTH_CLIENT_SECRET in .env, register the site root as the OAuth redirect URI, publish the cloud project (publishing status Testing makes google revoke the refresh token every 7 days), then visit /?health=connect once.

categories, AI defaults (temperature, timeout, max_tries), and the archive fulltext minimum (8000 chars) are hardcoded in the package.

usage

php -S 127.0.0.1:8080 -t .

cron

0 6,18 * * * curl -s 'https://your-host/?scrape=1&key=<AUTH_PASSWORD>' >/dev/null

backup

zip -r backup.zip .data