Calculate personalized fit scores for published products belonging to the authenticated vendor, based on a customer's skin analysis data. Optionally pass an array of SKUs to score only specific products; if omitted, all published products are scored. Uses a hybrid scoring approach combining embedding similarity (75% weight) with algorithmic skin type matching (25% weight). Products containing customer allergens score 0.

Requires a completed skin scan batch. The customer's skin profile embedding is cached per scan batch for 24 hours, so subsequent calls are fast.

Authentication: Vendor Bearer token required. The vendor is derived from the authenticated context — no vendor_id parameter needed.