Product intelligence · proof of concept
Describe the piece.
We'll find it.
Understood as
Nothing matched. Try removing a filter above.
01
Understand
An LLM reads the request and splits it into hard facts (type, size, seats) and the look you're after.
02
Filter
The facts become SQL filters on AI-normalised attributes: millimetres, colours, materials, shapes.
03
Rank by meaning
Every product has a vector embedding in Postgres (pgvector). What's left is ranked by how close it is to the look.
Coming in the full build
- Designer portal with projects, rooms and shortlists
- Supplier feed imports (CSV, Excel, API) with originals retained
- One-click push to Shopify
- CAD, STEP and spec-sheet files per product
- Search by image