Storacle

Logo
Dashboard
Inventory Page
Forecasting Page

Inspiration

During our research for project ideas, our team learned that global retail loses over one trillion dollars every single year to stockouts and overstocking. Our goal is to reduce that waste. We provide a system that retailers and restaurants can utilize to track inventory and effectively restock on supplies.

What it does

Storacle is a retail inventory app that takes one shelf photo in the morning (AM) and one at end of day (EOD). It uses Gemini Vision to count products in each photo, saves those snapshots, and computes daily usage (AM − EOD). From that history it forecasts demand, predicts when a store will run out of stock, and suggests when and how much to reorder.

How we built it

Frontend: Next.js (TypeScript, Tailwind CSS), Shadcn MCP (npx shadcn@latest mcp init --client cursor)
Backend: FastAPI (Python)
Computer Vision: Google Gemini API
Database: SQLite (SQLAlchemy ORM)
Forecasting: Darts

Challenges we ran into

Choosing how to store data
Getting Gemini API to work with free tier
Managing dependencies across teammates without containers
Parsing data from ML model (dates wouldn't format as intended from docs)

Accomplishments that we're proud of

We are proud of making a working computer vision pipeline with many industry standard technologies within tight time constraints.

What we learned

Clear efficient prompting
Learning new technologies under pressure
Efficient version control to avoid merge conflicts
AGILE methodologies

What's next for Storacle

Pro tier allowing users to add more POS data, like seasonal trends, directly to forecasting model
Direct ability to reorder with affiliated suppliers