Catalog Intelligence MVP

Next.js + GPT-4o + guarded demo API

Retail catalog ingestion, normalization, and AI-assisted taxonomy review.

Feeds parsed

6.2k

rows staged

Entity groups

3

canonical candidates

Pending review

4

operator decisions

1. Ingest feeds

Nord Market affiliate feed

4,182 rows · 6 min ago

parsed

New color names and size aliases detected

GlowShelf retail API

1,218 rows · 14 min ago

normalized

Variant IDs stable; taxonomy changed for SPF category

TrailBox partner CSV

772 rows · 31 min ago

needs-review

Duplicate SKU prefixes across region feeds

Stage a source row
2. Normalize entities
Entity-resolution watchlist

1 cross-retailer duplicate group needs canonical product review before auto-publish.

3. Run GPT-4o mapping

Selected raw row

AeroFlex Womens Ribbed Matching Gym Set Blue Small

Brand
AeroFlex
Retailer
Nord Market
Price
$78
Match group
active-aeroflex-ribbed-set

Server route keeps the key private, caps output tokens, and limits demo requests per browser/IP.

Choose a product row and run the analysis. The result will show taxonomy mapping, duplicate risk, ETL rules, and reviewer routing.

Operator decision

Current state: pending

Review queue
Audit trail

demo_initialized

5:38:30 PM

Seed feeds and canonical review queue loaded.

Trust boundary

Raw feed, normalized product, AI suggestion, and reviewer decision are stored as separate records.

Production path

Replace demo arrays with Postgres tables, scheduled ETL jobs, affiliate API adapters, and audit logs.

First milestone

Ship one retailer adapter, taxonomy schema, duplicate review queue, and guarded AI mapping endpoint.