Can a prompt replace Mealime Pro?

Build a weekly meal planner over your own recipe collection, with the shopping list as the real output.

Recipes come from your own store — see the Crouton entry for the import approach. Each carries a yield, ingredients parsed into quantity, unit and item, and tags you assign.

Planning is a week grid. Assign recipes to days, set the number of people per meal, and the planner scales quantities accordingly.

The list generator is where the care goes. Merge every ingredient across the week, converting units within a system so 500 g and 0.5 kg become one line, and never converting across systems silently. Items that cannot be merged stay as separate lines with their recipe named beside them. Then group by aisle using a mapping file you extend over time, and add a pantry list — staples you always have — that is excluded by default with a one-tap override.

Add a leftovers pass: a recipe whose yield exceeds the people assigned generates a leftover portion, which can be dragged onto a later day and is then excluded from the shopping list. That single mechanic is the difference between a plan that survives a week and one abandoned on Wednesday.

Build one report: which recipes you actually cooked versus planned, over the last three months. Planning is easy and cooking is not, and the gap is worth seeing.

Do not have a model invent recipes. An untested generated recipe is a wasted evening and a wasted shop, and the failure is only visible after you have bought the ingredients.