The story
Challenge, approach, result
Challenge
Parts buyers arrive with a car and a symptom, or an OEM reference — not a lifestyle browse. Generic storefronts bury fitment, conflate language with region, and cannot index or join 'this pad fits the 1.5 dCi but not the 1.2 TCe'.
Approach
We forked the Medusa DTC starter and modeled the catalogue as three custom modules — brand, part-reference, and vehicle fitment — linked to products. A fail-closed store middleware resolves fitment filters; the storefront keeps one Belgian region in the URL and UI language in a locale cookie (fr / nl / en).
Result
A production stack on one VPS: trilingual catalogue, vehicle-aware filtering, and a storefront/admin host split for session isolation — with an end-to-end audit that closed the silent fitment bug and restored build gates before those seams stayed invisible.
Stack