Project
BuildingGlucoCore
A diabetes data platform that speaks Tidepool's wire format, so existing clients point at it unchanged.
What it is
Tidepool's platform is a fleet of Go and Node services behind an API gateway, backed by MongoDB. GlucoCore collapses that into one Next.js application on Vercel with Postgres, keeping the endpoint surface and the JSON shapes drop-in compatible, including the parts that are technically deprecated but which every real client still depends on.
It presents two API faces over one store. Uploaders write through Tidepool's upload path or Nightscout's v1 API; displays read through Tidepool's legacy read path or Nightscout's. A record written through one face is readable through the other. It is a projection, not a copy.
It is the hub the other two projects sit around. GlucoLoop syncs into it and reads its version manifest from it, and GlucoCube pairs with it to be configured and managed from the web.
What it does
Tidepool-compatible routes at their exact legacy paths, with every client quirk held in place by a test.
Imports from Dexcom Share, Nightscout sites and anything speaking Tidepool's API, on a schedule per link that backs off when there is nothing new.
A web UI rendered on the server: today, history, sharing with a care circle, connections, devices and account. Every screen works with scripting off.
A glucose forecast, delivered where a display can draw it, along with the reasons it declines to make one.
A SugarCube device registry: QR pairing, the config a display long-polls, a command queue, heartbeat, diagnostics, and wallpaper art per person.
Passkeys and emailed sign-up codes, credentials for linked services encrypted at rest, and row-level security in Postgres.
Part of the same system
One loop, one store, one wall
GlucoLoop runs the loop and writes. GlucoCore keeps everything once. GlucoCube reads it back and puts it where the household can see it.