DietAgent builds a meal plan in seconds. It looks simple, but there’s a deliberate choice behind it: the parts that need precision follow rules, and the parts that need judgment go to the AI.
Precision on one side, judgment on the other
Calorie, macro and water targets aren’t a place for creativity. Those numbers come from clear rules and always give the same result. The AI steps in where it’s good: picking food that fits each person’s routine and tastes.
Everything gets checked before it shows up in the app
Every AI suggestion goes through a check before it shows up in the app. If a meal drifts off target, it gets adjusted. Nobody sees numbers that don’t add up.
Swapping a meal without wrecking the day
When someone swaps a dish, the app looks for another one that fits what’s left of the day. The swap is quick and the plan still holds.
What I learned
AI is great at suggesting and bad at math. Keeping those two jobs apart made the app more reliable and cheaper to run.