The problem
Exact-match search loses shoppers who don't know your vocabulary
Classic keyword search fails the moment a shopper's words don't match your product titles. Semantic search compares meaning, not spelling — it is built as an optional strategy inside the storefront's existing search fallback chain, so nothing else about your search has to change.
Similar products
A 'you might also like' that's actually related
Because every product gets an embedding vector, "similar products" becomes a real nearest-neighbor query instead of a same-category shuffle.
Keep exploring AI Semantic Search.
Explore the full architecture- pgvector
- Product embeddings are stored and queried as native Postgres vectors — no separate vector database to run or pay for.
- Embedding provider choice
- Generate embeddings with Gemini or OpenAI — the same multi-provider pattern as the rest of the AI plugin family.
- Storefront fallback chain
- Wires into the existing search and recommendation flow as a strategy, not a replacement — the storefront needs zero changes to use it.
Customize: AI Semantic Search
Selected Product
AI Semantic SearchEmbedding-based semantic product search and 'similar products' — understands what a shopper means, not just the exact words they typed, on top of pgvector.
$29