Sources

Where each board's meetings come from and when the collector last reached it. A source that has not succeeded recently is showing stale data, not an empty calendar.

What is on file

The table above is how collection is going; this is what came of it. A meeting can be on file with no agenda behind it, and an agenda can be on file that nobody has read yet, so these are counted separately rather than rolled into one number.

Collecting means this server fetches the source on a schedule. A source is only called Failing after two attempts in a row have failed, or a day has passed with no success at all. A single timeout on a small-town web server is ordinary; it is reported under the row rather than in the color.

Summarization queue

Agendas are summarized by a separate worker that authenticates against this server's API. Nothing here is re-read once its agenda has been summarized; a queued job means the agenda is new or was republished.

The queue holds agendas only, and an agenda is read before its meeting. Minutes and captions arrive afterwards and are the second pass: minutes are where a motion and a vote count are recorded, and captions are the only record of what was actually said and who spoke. Neither is queued, so what is left of them is counted here instead.

Property cross-reference

Agenda items name properties by parcel number or street address. Each one is looked up against the North Carolina OneMap statewide parcel layer, which every county publishes into. A reference that finds nothing is ordinary: land is merged, split, and renumbered between the agenda and the parcel file, and some counties publish thinner records than others.

Housekeeping

Work the server does for itself. Almost none of it is on a fixed schedule, so a date here can look old and be right: a job runs when something makes it due. The search index is rebuilt in full only when the shape of a search row changes, and every meeting is reindexed as it is written. Most passes find nothing to do.