Add a medicine, edit it, or switch it on/off. A medicine that is on shows to beekeepers in the treatment screen; switch it off to retire it — it stays on record (recoverable) and past treatments that used it are never broken. Where allowed: tick the countries a product is authorised in and the treatment screen only offers it there; leave all unticked and it shows everywhere (how every existing row behaves).

Are the three services the app stands on answering right now? Each card runs a real live check from your browser and has its own Check again button. A green card means “answering”, not “everything is perfect”: the server’s own error logs can’t be read from here yet, and that card says so honestly.

How big the platform is and how fast it’s growing. Sizes are judged against the current plan (the Supabase free plan: 500 MB of database, 1 GB of file storage) and a card turns amber once 80% is used. When a size can’t be read it says so plainly; it never guesses a percentage.

Do the records themselves hang together: no broken links, no lost files, no records nobody can see. These are whole-platform counts and no beekeeper names are shown. Anything not “Clear” says how urgent it is and exactly what to do about it.

The shared lists the app ships with: clubs, medicines, catalogue and templates. This is a presence check, not a fault list; the adding and editing happens in the Reference Data section, and the buttons below take you straight there.

Who changed what, and where the audit trail has gaps. The counts are live; the coverage card is a known fact about what’s wired, not a live measurement.

Recent user-consent changes (the consent compliance trail):

The badge is verified health: green shows only where a real live call proved it — the Supabase cards by the page itself answering, the keyed services (Stripe, email, AI, webhooks) by a real test call to the provider from the server. Operator intent (in use / planned) is your plan for a service, shown separately — it never turns a badge green.

Connect stores a key in the server vault and tests it at once; Test now re-runs the real provider call; Disconnect deletes the key. Manage edits intent, mode, on/off and notes (with a reason, never secrets) — it can never set a badge green by hand. Every action lands in the Audit Log.

The AI features (photograph a paper record, voice notes, receipt reading, plant ID) all go through one server with a hard spend cap. Right now they run in mock mode: sample drafts only, no real calls, nothing spent. Real paid drafting stays OFF until you turn it on here, and even then it is capped per month. Every change needs a reason and lands in the Audit Log.

Stripe presence is read from profiles (no secret read); storage is read straight from the buckets. Test/live mode is whatever the operator set — the app can't read the Stripe secret to confirm it.

Add a country or region that is not in the list yet. It is created as a draft and is never published on create: fill in its regulator, notifiable diseases and record rules on its card below, then publish it deliberately once you have checked an approved source. Both a reason and an approved-source note are required, and the create is written to the Audit Log.

Each card is one jurisdiction. Edit changes the four display fields and the source note; Publish / Unpublish moves it between draft and published. A Needs review badge means the source itself marks that jurisdiction’s notifiable-disease or medicine-record duty as unconfirmed — verify it before publishing. Legal and disease wording stays verbatim from an approved source.

The app’s reference catalogues (forage months, app flags and more) live in catalog_items. Here you can see every row and publish or unpublish it — only published rows are read by the app. Every change needs a reason and lands in the Audit Log. Editing a row’s structured content (plant lists, options) is a separate, later increment; the safe lever here is which rows are live.

Our own Tropilaelaps presence map, refreshed by hand from the official EU source (EURL Bee Health). Add or update one region at a time: pick where it stands, the date the figures were current, and an approved-source note. A new region is saved as a draft and only shows to keepers once you publish it. Every change needs a reason and lands in the Audit Log. Keepers are told plainly the map is refreshed by hand and may lag the official figures.

Our own AFB / EFB foulbrood map, refreshed by hand from the official figures (National Bee Unit / BeeBase for Great Britain, DAERA for Northern Ireland, DAFM for the Republic of Ireland). Add or update one region at a time: set where AFB and EFB stand, which period the figures cover, and an approved-source note. Foulbrood really does occur here, so nothing is seeded and “None recorded” means no cases were recorded in that period, not that an area is clear. A new region is saved as a draft and only shows to keepers once you publish it. Every change needs a reason and lands in the Audit Log.

The world Asian hornet map on the Threat Watch page. Add or update one country or area at a time: where it stands, its map position, the date you checked, an approved-source note, and (for a keeper’s home country) the official report address and map links. A new place is saved as a draft and only shows to keepers once you publish it. Every change needs a reason and lands in the Audit Log. Keepers are told the map is refreshed by hand and may lag.

Four gates, in order. Each row says what it is, whose move it is, and what done looks like. No pill here turns green until the thing itself is actually done.