LLM-based virtual store simulator 'Parastore' released

·2026.05.27 17:53

Key point

An open-source simulation framework has been released in which LLM-based synthetic consumers explore a 3D virtual store and make purchasing decisions.

Details

Parastore is an open-source (MIT license) framework that uses LLM-based Synthetic Consumers to simulate real store environments.

Virtual consumers walk around the 3D store, browse the shelves, and make purchasing decisions. They follow a total of 12 behavior patterns, with random context such as mood, budget, and affiliation, as well as impulse-buying logic applied.

The simulation has been validated, recording a Spearman correlation coefficient of 0.955 with actual POS (point-of-sale) data.

The tech stack is built on Python/FastAPI and React/Three.js, and it can integrate with any LLM backend. Users can leverage it to test store layouts or research agent behavior.

open-source llm-agents retail-simulation synthetic-consumers behavioral-ai

Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.