Product · 2025 – present
AI product content & image search
LLM-generated product content for SEO, image search, and a Gemini + MCP workflow that proposes catalogue fixes for a person to approve.
Marketplace catalogues are uneven: one vendor writes careful descriptions, the next uploads a title and a photo. We built AI tooling around that gap.
- Content and SEO. LLM-generated and LLM-improved product content, with image search so customers can find products by picture.
- Enhance with AI. One action in the seller portal writes or improves a product’s name and description in English and Arabic, following naming rules set per category. An overnight job enriches the listings that are still thin.
- A score for every listing. Each product gets a content-quality score, so the weakest listings are fixed first.
- Catalogue enhancement with a human in the loop. Gemini reads marketplace catalogue data through MCP and proposes improvements. Merchants, vendors, moderators and vendor-success managers review the suggestions; only approved changes go to Magento, and only through the existing Marketplace API, never directly.
- Readable by AI engines. The storefront publishes an llms.txt so assistants can read the catalogue.