At a large university, events are listed in separate systems. At Alabama, the official calendar, department pages, student organisations' social accounts and athletics each list their own. A student who wants to know what is happening tonight has to check several places, and many do not bother.
UA Today pulls events from 6 sources on a schedule and shows them on one page. The hardest part of the build was detecting a source that stops sending events.
Combining the feeds
Fetching and normalising six feeds took a day of work. The feeds use different date formats, describe locations differently, and disagree on whether a recurring event is one record or forty.
Detecting a source that stops without an error
A source that returns an error is easy to catch, because you see the error and fix it.
A source that returns an empty list, with a 200 status and a well-formed response, is harder to catch. Somebody renames a query parameter over winter break. The feed still answers with nothing, your site shows fewer events, and nobody notices because a broken feed looks like a quiet week on campus.
Most of the bugs I wrote about this year fail the same way: a canonical tag telling Google to skip nine pages, and eleven agents that were never being invoked while the runner exited zero. Nothing raises an error because the failure is missing data.
Floor checks
The fix is to assert a minimum event count. Each source declares how many events it should plausibly produce, and returning fewer is a failure even when the request succeeded.
const SOURCES = [
{ name: "official", floor: 40 },
{ name: "student org", floor: 20 },
// ...
];The floor is loose enough to pass a slow week and still catch a source that returns a 200 status with nothing in it. A health endpoint reports each source with its count and its response time, and the morning check reads that endpoint to see whether the job produced events.
Choosing the number is a judgement call. A floor set too high fires on every quiet week, people learn to ignore it, and the alert that matters gets ignored with the rest.
Moderation and the unofficial notice
The first is a moderation queue. Anyone can submit an event, and nothing appears until a person reviews it. For a site with a university's name near it, the cost of one bad listing is much higher than the cost of a slower pipeline.
The second is a notice in the masthead that the site is unofficial. It goes on the first screen. It says the site is a student project, it is not affiliated with the university, and event details are collected automatically and can be wrong.
What I would do differently
I would write the floor checks first. I added them only after a source went quiet, and until then I was trusting a status report I had never tested.
Students use an events site only if they trust its listings, and a student who finds one wrong listing may stop coming back. That is why most of the work in the build went into getting the listings right.
Part of the guide AI websites and getting found by AI search.