intellicia-public / parastore

What is parastore?

Parastore is an isometric 3D retail store simulator where you draw a store layout, generate AI-powered customer personas, and watch synthetic shoppers navigate your aisles in real time. The backend runs Python with FastAPI, while the frontend uses React and Three.js for the 3D visualization. You define shelves, fridges, counters, and entry points on a grid editor, assign product categories, then hit play to see personas walk in, browse, buy, and leave. The simulation outputs metrics like conversion rate, dwell time, and per-rack engagement.

Why is it gaining traction?

The hook is the LLM-driven behavior simulation. Instead of hardcoding customer rules, Parastore generates personas that think and act based on trade-area analysis and individual profiles. The 3D visualization makes the simulation tangible—you watch heatmaps form as customers traverse the store. The project also publishes correlation benchmarks against real convenience store data, showing 0.955 Spearman correlation by category. For developers curious about LLM agent architectures in physical spaces, this is a working end-to-end example.

Who should use this?

Retail planners exploring layout variants, researchers prototyping consumer behavior models, and developers building LLM-driven simulations will find the most value. If you need to A/B test aisle configurations or evaluate product placement without physical experimentation, this provides a starting framework. Teams wanting validated forecasting should look elsewhere—this is explicitly a prototype, not a production tool.

Verdict

With only 29 stars and a credibility score of 0.699999988079071%, Parastore is early-stage and unproven for real-world deployment. The documentation is solid for a prototype, and the architecture is clean enough to extend. Use it as a reference implementation or experimental sandbox, not as-is for business decisions.