
Parastore is an open-source (MIT) retail simulation where LLM-powered synthetic consumers walk through a 3D virtual store, browse shelves, and make purchase decisions. Each consumer follows one of 12 behavioral patterns with grammar-constrained actions, randomized context (mood, budget, company), and impulse-buy logic triggered by what they see along their route. Validated against real POS data with 0.955 Spearman correlation. Python/FastAPI + React/Three.js. Any LLM backend.
Parastore is an open-source retail simulation tool that utilizes LLM-powered synthetic consumers to navigate a 3D virtual store and make purchase decisions based on predefined behavioral patterns. It is built with Python/FastAPI and React/Three.js, and has been validated against real POS data, achieving a Spearman correlation of 0.955.