Save a meal you eat often

You realize you've been typing « 30 g whey, banana, 40 g oats » into your post-workout meal log for three weeks. Time to save it as a unit so the next time you can just say « my usual shake » and be done.

What the agent needs to hear

Behind your sentence, the agent works backwards to two pieces: the meal’s name and the items in it (each with a food and a quantity). The name is what you’ll say when you want to log this meal in the future — pick something you’ll actually use (« post-workout shake », « Tuesday breakfast », « hotel-room lunch »).

Items get listed the same way they would in a regular meal log: food name + quantity. « 30 g whey, 1 banana, 40 g oats » resolves three items, each with its own quantity and unit. The agent shows you a card with the meal name and its items and macros computed, and waits for a nod before saving.

All at once vs item by item

Most of the time you save the whole meal in one sentence and the agent splits it into items. If the meal is more complex or you want to think about each item, building it up incrementally also works: « create a saved meal called post-workout shake » opens an empty meal; subsequent sentences add items one at a time (« add 30 g whey », « add a banana »). Either path lands at the same saved structure.

If a quantity is fuzzy (« a banana » without a weight), the agent uses the catalog’s typical serving for that food. « A banana » resolves to roughly 120 g of fruit; « a scoop of whey » to whatever your default scoop size is. Naming a specific weight overrides the default: « a 100 g banana » sticks at 100 g.

What this saved meal enables

Once saved, the meal lives in your library next to the food catalog. Future logging gets the shortcut: « log my post-workout shake as lunch » expands to a meal log with all three items at the saved quantities, no per-item re-typing. The agent reads the saved meal’s macros and rolls them into the day’s totals in one shot.

Saved meals differ from recipes in a small but real way. A saved meal is a fixed combination of foods you eat as a unit (the shake is always whey + banana + oats; you don’t divide it into servings). A recipe is something you cook in batches and portion out — there’s per-serving math involved. If you find yourself dividing the meal into halves or thirds, you probably want a recipe, not a saved meal.

When the agent gets it wrong

The card is the moment to catch a missing ingredient or wrong quantity. If an item is off, name the food: « the oats were 50, not 40 ». If the agent matched an item to the wrong food (« whey » resolved to the wrong brand in your library), say which one you meant: « the unflavored whey, not the chocolate ».

If the agent saved with a wrong name — you said « post-workout shake » but it stored as « Saved Meal 1 » — push back: « rename it to « post-workout shake » ». And if you forgot an item, add it after: « also include 5 g creatine », and the meal updates.

What makes the saved meal worth keeping

Three things decide whether this saved meal is useful tomorrow: the ingredient list is complete (a forgotten item means every future log of this meal is missing it), the quantities match what you actually eat (not a planned amount, the real one), and the name is one you’ll recognize and type quickly (the whole point of saving is the shortcut). The system reads saved meals as one-shot lookups during meal logging — once it’s wrong, every future shortcut log carries the same error forward. Save the meal the way you actually eat it, name it the way you’ll say it, and log fast from then on.