New Aggregator

A minimal news aggregator: React frontend + OpenServerless Python actions + Redis. Fetches RSS/Atom feeds, deduplicates, stores in Redis, and lists articles through an HTTP action.

Actions

• news/fetch — fetch enabled sources from packages/news/fetch/sources.json, normalize, deduplicate (Redis sorted set, SHA-256 of URL as id), store in Redis. Uses conditional HTTP GET (If-Modified-Since / If-None-Match) so unchanged feeds return 304 and are not re-downloaded. Per-feed failures are isolated (counted in errors, never abort the run). Returns {"added", "existing", "errors"}.
• news/list — read newest articles first. Params: limit (default 20, max 50), offset, source (substring filter), q (title+summary search).

Sources

Sources are configured in packages/news/fetch/sources.json (manually editable, no database table):

[
  {"id": "bbc", "name": "BBC", "url": "https://feeds.bbci.co.uk/news/world/rss.xml", "enabled": true}
]

Set "enabled": false to disable a source. Add a new object to add a source. The watcher redeploys news/fetch automatically when the file is saved.

Scheduled refresh

news/fetch is scheduled with the OpenServerless scheduler via a cron annotation on the action (no custom scheduler, no workers, no extra infrastructure). The action stays independently invokable for testing.

Enable/refresh the schedule (every 10 minutes) with:

ops -wsk action update news/fetch -a cron "*/10 * * * *"

Verify it is set:

ops -wsk action get news/fetch | grep cron

The annotation survives normal ops ide devel watcher redeploys (the OpenWhisk action update flow preserves existing annotations). To change the interval, re-run the command with a different cron expression. The scheduler must be enabled on the cluster (ops config enable --cron); repeated runs do not re-store duplicates thanks to ZADD ... nx, and unchanged feeds are skipped via 304 to minimize external HTTP traffic.