The 2 farm record pages
Field polygons and sensor settings describe the account’s configuration: the shapes the satellite pipeline averages inside, and each node’s crop, phenology dates, last spray date and degree-day accumulator. Alerts and advisories is the account’s audit trail instead: every threshold alert that went out, and every irrigation, spray-window, frost and vegetation-anomaly advisory the platform ever generated, sent or held, with the reason attached either way.
What the phenology dates do
The potato risk models are gated by crop stage. Common scab is only evaluated inside a window after tuber initiation; late blight severity accumulates starting at emergence; a haulm-kill date closes the season outright. All of those dates come from sensor settings, so a settings row left empty means the models simply run without their phenology gates, and say as much in their own metadata.
Why the advisory log keeps what was not sent
An advisory gets written whether or not it went out, with the reason recorded when it was held: quiet hours, no phone on the account, cooldown, not prioritised, dry run, send failed. Even an account with no phone numbers on file, the zero-notification setting, still builds up a full advisory record it can read here; the notifications view in the dashboard shows the same rows.
Quiet hours and digests
Alerts raised between 22:00 and 06:00 East Africa Time are held back and delivered together at 06:00 as 1 digest. 1 notification per metric per day is the hard ceiling, so a reading hovering right at a bound never fires twice.