TrailLite 22
$129
22L, taped seams, 780g. Match 0.88 — weight and weather fit the brief. DEMO DATA.
AI Shopping Assistant
A DEMO / CONCEPT shopping helper. Catalog cards below are DEMO DATA, not a real store, inventory, or checkout.
Project overview
The project maps a sentence (“lightweight, waterproof, under $150”) onto filters and a shortlist. Nothing here is a live Shopify or warehouse feed.
Problem being explored
Shoppers describe a trip in language. Catalogs are filters, SKUs, and stock flags. Those two shapes rarely meet without a layer in between.
Example user scenario
Example request: a 3-day hike, a waterproof pack, a $150 cap. The demo shows three fictional products with match notes. Treat prices and scores as DEMO DATA.
Proposed workflow
- CUSTOMER REQUEST
- INTENT ANALYSIS
- PRODUCT SEARCH
- FILTERING
- AI COMPARISON
- RECOMMENDATIONS
Architecture explanation
Customer request → intent analysis → product search → filtering → comparison → recommendations. Production would call a catalog API and an inventory API before ranking.
Key features
- Intent parsed into constraints (budget, use, features)
- Catalog filters instead of invented SKUs
- Side-by-side comparison on DEMO DATA
- A match note that can say “no fit”
- A planned catalog and inventory API boundary
DEMO DATA
Example request: I need a lightweight waterproof backpack under $150 for a 3-day hiking trip.
RidgePack 28
$149
28L, rain cover, 910g. Match 0.81 — more volume, still under budget. DEMO DATA.
Coast Roll 20
$118
20L roll-top, 740g. Match 0.76 — lightest, smaller capacity. DEMO DATA.
Demo interface
Product cards are fictional. Try Demo does not query a store. Recommendations are illustrative copy.
FAQ
Is this connected to a real shop?
Product cards are DEMO DATA: fictional items used to show matching and comparison. There is no live catalog, warehouse, or checkout on this page. A real project would query your catalog and inventory APIs.
Can this work with Shopify or WooCommerce?
A production assistant can be integrated through those platforms’ APIs — catalog, stock, and optionally authenticated order status. Checkout stays on the store. That integration is not running on this page.
What if nothing in the catalog fits?
The assistant should say so and avoid inventing a SKU. Honest “no fit” is part of the product, not a failure of the UI. Filters and stock checks in a working system make that answer reliable.
Are the match scores a trained model?
Scores on this page are labeled DEMO DATA to show how a ranking might appear in the interface. A production ranker would use your catalog fields, availability, and rules you approve. Treat demo numbers as illustration only.
Limitations
Three fictional products. No checkout, inventory, or payment.
What production would require
A real catalog API, availability, pricing rules, disclosure of why a SKU was picked, and no invented products.