Mellitus Diabetes Predictor

Diabetes Predictor (DP) builds short‑term CGM forecasts and helps visualize trends. Each user gets a private instance with its own Nightscout connection, dashboards, and settings.
Slots reset after 7 days (except excluded users) so new testers can register and try it.

Disclaimer: DIY project for experimentation. Not a medical device. Do not rely on these outputs for treatment decisions without your own clinical judgement. Any changes to Loop/AAPS/therapy settings are at your own risk.

Register (Free Test Slot)

Checking free slots...
One active slot per email and per Nightscout URL. Use Nightscout base URL (https://example.com) and a real read token/secret (example: reader-ab12cd34ef56...), not just "reader" or a name.

Grafana Examples

How It Works (Technical)

DP trains a sequence model (LSTM/GRU) on your recent Nightscout data to predict BG 30–120 minutes ahead, and exposes live metrics (BG/IOB/COB, prediction curves, error histograms, ISF/CR/basal suggestions) in Grafana. It can retrain periodically and uses actual outcomes to track forecast error over time.

Build Notes