EvalCart

About

EvalCart is a fake grocery store with a frozen catalog of about 2,000 items and a deterministic scorer for testing shopping agents. After an agent builds a cart, it receives a receipt scored on completeness, constraints such as budget, taste fit, pantry and leftovers, availability, collaboration, and efficiency. A cart can contain every expected item and still fail if it goes over budget, which shows failures that a simple item match would miss. No real orders or money are involved, and a leaderboard compares submitted runs.