A commerce search engine
Search that understands what shoppers mean.
SearchSmith is the engine running the box below — hybrid lexical + semantic ranking with typo tolerance, answers cited from the store's own guides, stock-aware results, and a merchant console behind it all. Every keystroke here hits the live engine, searching a fictional coffee-supply demo store.
Waking the widget…
Results, filters, guide answers and reviews are live from the engine. Products open in the full demo store; every product, brand, price and review in the demo catalog is fictional.
Light where it counts
The first-paint search JavaScript is under 7 KB gzipped, with size budgets enforced as a deploy gate — measured against ~102 KB for Boost's widget on the live production store where this engine replaced it.
Answers with citations
Ask a question and the Ask tab answers from the store's own guides with numbered citations you can check — never from thin air. Try the water-temperature chip above.
White-label by design
Shoppers only ever see the store's brand — here that's Stonelark. The theme, accent and corners you
just switched are the same one-click Appearance settings every merchant gets.
Browse the full demo store →
Stonelark is a complete storefront running SearchSmith end to end — collections, product pages, recommendations, back-in-stock capture, cart.
Open the merchant dashboard →
Analytics, merchandising rules, A/B tests, synonyms, review moderation — open to visitors, no sign-in needed. The sandbox resets itself every 6 hours.
See it measured vs Boost & Algolia →
A feature-by-feature scoreboard against both incumbents — including the places where they still lead.