Items belong to your active organization (switch it in the header). Upload isolated garment photos — images go to S3, item records live in Postgres, and outfit retrieval uses multimodal embeddings in Qdrant plus Qwen3-VL-Reranker. Digitize runs Gemini vision, then updates the row and vector index.
Add items
Add a product from source URL, or use image upload if you do not have a product page.
Add category
Paste a category page URL, choose audience, and enter a retail category name (Pants, Jackets, …). Limit caps how many products the scrape collects. Embed selected then fetches missing product details and stores fashionSigLIP vectors in Qdrant, 20 products per request.
Add custom category
Same scrape, but its products keep an isolated catalog group named exactly as you type — they will not merge into Pants, Jackets, or the other retail buckets.
Scraped categories
Search the products embedded with Embed selected by text, by a reference photo of a worn look, or both. Hebrew queries are translated to English first.
A photo alone finds look-alikes; add words to narrow it down, e.g. “the jacket”.
Loading…