Reading a too-weak / too-strong CR out of your CamAPS FX loop data
I'm the father of a kid on CamAPS FX, and something I keep running into: under Auto Mode the loop compensates a too-weak CR by adding basal, so glucose still returns to range and the usual "did it come back to baseline?" check misses it. The signal isn't really the glucose curve — it's the extra basal the loop delivers in the meal window.
I've been trying to make that readable instead of guessing per meal. It works off a Glooko export, so you need CamAPS FX connected to Glooko and then download the data export from there (CGM, bolus and basal). From that, for each slot (breakfast / lunch / dinner), it integrates the loop's extra basal over the postprandial window, puts it next to the meal bolus, and shows whether the CR looks too weak, too strong, or about right — with AGP and the usual consensus metrics (TIR/GMI/CV) alongside for context. Curve shape then hints at what to look at next (pre-bolus timing, dose, fat/protein).
To be clear about limits: this is only a discussion aid for the care team, not a therapy recommendation, and under a hybrid closed loop it's an indication, not proof. It also only holds for CamAPS FX — the whole thing rests on corrections arriving as modulated basal; Control-IQ / Omnipod 5 push part of it as boluses, which would break the logic.
Mostly I'm curious how others handle this: **if you're on CamAPS FX, how do you currently decide a CR needs changing?** What signal do you actually trust?
I built the above into a small open-source tool while working through it (full disclosure, it's mine — AGPL, non-commercial, synthetic demo data included), in case anyone wants to look at the method or try it: https://github.com/peisenh/loop-cr-review