This started with a straightforward frustration. The historical data sitting in Glooko, particularly from the Omnipod 5, is far richer than anything the manufacturer's own app surfaces in a useful way. The question eventually became: what happens if you hand that data to a large language model and ask it to behave like an endocrinologist?
Omni-Endo AI is a locally run tool that extracts your Glooko data, covering bolus records, blood sugar readings, CGM trends and settings changes, and restructures it into a detailed prompt ready to paste into whichever AI chat tool you already use. ChatGPT, Claude, Gemini, or anything else. There are no API keys, no subscription fees bolted on top, and no data passing through any server other than Glooko's own. The handoff is deliberately manual, which sounds like a limitation but is actually the point. You stay in control of what goes where.
The AI persona built into the prompt is configured to be direct rather than reassuring. Diabetes management does not particularly benefit from being told everything looks fine when it does not, so the analysis is designed to identify where bolus timing is off, where corrections are being overdone, and where basal rates are quietly failing. It will ask follow-up questions before drawing conclusions, which in practice makes the conversation feel less like querying a spreadsheet and more like a consultation with someone who has actually read your notes beforehand.
The whole thing runs in Docker, requires no coding knowledge to set up (I have written detailed instructions for you to get it up and running), and keeps your credentials entirely on your own machine. It is a clinical audit tool built out of personal necessity, and it has genuinely changed how I prepare for consultations.