Grovity now supports native semantic catalog search. Upload your product catalog to the Grovity portal, and it’s automatically processed and vectorized — ready for your agents to search as soon as it’s in.
People don’t search by SKU
Customers don’t think in SKUs, categories, or exact field values. They send a photo, mention a color, name a budget, and describe what they want — all in the same message. That’s how people actually search, and it’s what a catalog full of structured fields was never built to handle.
Now Grovity agents can search the way customers already talk: “I’m looking for something similar to this photo, but in black and under $80.”
Multimodal search, three native tools
Once a catalog is vectorized, Grovity exposes it to agents as three native tools — text, semantic (vector), and image — that work independently or together in the same conversation:
catalog_product_image_vector_search— find similar products from a photo or visual reference.catalog_product_text_vector_search— find products by semantic meaning when a customer describes what they want in natural language.catalog_product_text_search— find products by exact names, keywords, or known product terms.
An agent isn’t limited to picking one. It can combine image similarity, semantic intent, and exact filters in the same request — a photo plus “in black” plus “under $80” — and return only what satisfies all three.
What happens after you upload
Uploading a catalog to the portal is the trigger, not the finish line. Behind the scenes, Grovity connects the catalog, processes its text, attributes, and images, and generates embeddings with Jina — the embedding layer we use for semantic retrieval. Those embeddings are indexed for similarity search and exposed to agents as the tools above.
In practice, product information usually arrives fragmented: spec sheets, spreadsheets, SQL and Mongo databases, PDFs, incomplete descriptions, images, loose attributes scattered across systems. Grovity doesn’t just enable multimodal search over a catalog — it runs a process beforehand to understand, enrich, and vectorize that information, building a base complete enough for agents to actually search and recommend well.
Because a good agent doesn’t just need access to data. It needs data prepared to be understood.
Why it matters
Native semantic catalog search means more relevant recommendations, faster product discovery, better buying conversations, less dependency on manual lookup, and more value out of a catalog you already have.
Upload your catalog and see it live in your agents’ conversations, or book a demo.