flavr.ai

Flavr.ai builds meal plans that balance dietary restrictions, nutrition targets, ingredient preferences, and variety across a full week. It is a constraint problem rather than a recommendation problem, and that distinction changes the whole approach to generating them.

A plan is only valid if every meal in it holds together, and picking meals independently produces a week that is individually reasonable and collectively wrong. Restrictions and preferences are modelled separately because the two of them fail in quite different ways.

An allergy must never be violated, while a dislike can be traded away, and that distinction lets the generator relax the right constraints instead of returning nothing at all. Firebase handles persistence, and generation runs in a separately versioned Express service.