You are on the right path here but I think you are missing the “big players” for lack of a better term. The prediction software available now (open source) is quite good and works with different types of CGMS and pumps. You are really going to want to look at Loop.
Loop basically collects the inputs in the app automatically for insulin if you use a pump. I’m on the Omnipod DASH and Loop works with a few, Omnipod being my favorite. You can also input injections. It can also collect CGMS data automatically from that system. It works with Dexcom and others (I think Libre). You manually input carbs, and you are still gonna do that based on VIBES. After that, you get these magic prediction lines that show you where you are headed. And with the pump, it can add or lower insulin amounts (closed loop mode) to keep you in range. Pretty common to be 75-90% in range!
If you have some command line experience you can rent out some compute and then run it on rented hardware. I am renting an A40 with 48G of vram for $0.40/hr on Vast.ai right now to make cool images. I am still using Big Dream which I think is still based on DALL-E, not DALL-E 2. I couldn't figure out how to run this latest iteration (yet!).
You are on the right path here but I think you are missing the “big players” for lack of a better term. The prediction software available now (open source) is quite good and works with different types of CGMS and pumps. You are really going to want to look at Loop.
Loop basically collects the inputs in the app automatically for insulin if you use a pump. I’m on the Omnipod DASH and Loop works with a few, Omnipod being my favorite. You can also input injections. It can also collect CGMS data automatically from that system. It works with Dexcom and others (I think Libre). You manually input carbs, and you are still gonna do that based on VIBES. After that, you get these magic prediction lines that show you where you are headed. And with the pump, it can add or lower insulin amounts (closed loop mode) to keep you in range. Pretty common to be 75-90% in range!
Check it out:
https://github.com/LoopKit/Loop https://www.loopnlearn.org/