Skip to content

AI

Comersly

Upload a messy supplier spreadsheet. Get a commerce-ready catalog — brands cleaned, photos attached, storefront path written.

Comersly preview

Overview

Comersly turns messy industrial catalog files into something a storefront can actually use. You drop a CSV or XLSX full of abbreviations, blank photos, and no Dept / Class / Fine — then the pipeline interprets the raw cells, classifies each row from a manufacturer page, writes catalog copy, attaches the manufacturer photo, and exports a commerce-ready file. The supplier sheet stays raw. This is the project that taught me how to ship a real product pipeline, not just a demo.

Features

  • CSV / XLSX upload for the file a supplier actually sends
  • Interpret supplier cells into catalog fields without overwriting the source
  • Classify each row into Dept / Class / Fine from a manufacturer page
  • Confidence scoring — high-confidence rows ship, the rest wait for review
  • Manufacturer photos attached, then one-click CSV or XLSX export

Tech Stack

Next.jsTypeScriptDrizzlePostgreSQLAI

What I Learned

  • How to design a multi-stage import pipeline: upload → interpret → classify → catalog → export
  • Why the raw cell has to stay — SQD is what the supplier typed, Square D is what the catalog sells
  • Building classification that waits on a real manufacturer page instead of guessing
  • Shipping a product people can click through end to end, not just a localhost experiment

Gallery

Comersly gallery
Comersly gallery