I'm building TabsPilot solo: tabspilot.com. It maps supplier products to an online store's existing category tree.

A supplier might put phone cases, USB hubs and watch straps under “Accessories”. The store needs separate categories, so a simple lookup won't work.

I use JEV by TypeSafe AI. It's fast and cheap per product, keeping bulk processing quick and affordable.

There's also imports, column mapping, background jobs and review. Original columns and row order stay intact for export.

Checking every row would defeat the purpose. Uncertain assignments are grouped for review, and users can fix one product or a whole selection. Job retries must respect manual corrections.

For now its CSV/Excel in, CSV/Excel out. Store integrations are outside the initial scope.

Anyone built something similar? How do you handle review when input data is messy and confidence scores aren't enough?