Calk asks you to assemble a dish from its parts rather than photograph it. The first entry takes more input, but it makes the calculation explicit.
An image may not reveal how much oil went into a salad or whether a sauce contains sugar. A model has to infer those details, and the resulting estimate cannot show which hidden assumption was wrong.
With a meal template, you choose the base, protein, sauce and other parts. Calk calculates from those choices. Save a dish you repeat and its usual setup is ready next time.
The controls are deliberately practical rather than laboratory-precise. A template may ask about portion size, the relative amount of a filling, or whether an ingredient was present. Ingredients that can move the total substantially get more detailed choices. Others are kept simple so that a normal meal does not turn into a weighing exercise.
Templates also separate the structure of a meal from its name. Search may take you from the name of a particular dish to a broader soup, sandwich or bowl template. You then describe the version you actually ate. If no single template fits, you can assemble the main part in one and record an addition separately; the daily total still contains both.
This matters beyond the current meal. Reusing the same saved composition applies the same assumptions each time, which makes comparisons across days more consistent.
You can always inspect and change a saved setup. The heart stores your usual selections; it does not turn them into a recommendation. Today’s dish can still use a different sauce, portion or side without altering the favorite unless you save the new state.
The result is a diary whose numbers have a visible source. When a daily total looks surprising, Day analysis traces a selected nutrient back to the ingredients that contributed it — down to the sauce you picked this morning.