Shopify · Headless · API

Customers search. Cerno understands.

Replace your store's search bar with one that reads like a person. It stays keyword-fast for exact queries, understands natural language for everything else, and needs nothing configured to start returning better results.

Exact queries never touch the model. A SKU, a brand, or a product name hits the keyword index directly and comes back in milliseconds, so the fast path stays fast no matter how the rest of the query is phrased.

Vague queries fall through to semantic search. Cerno reads scent notes, materials, use cases, and intent from each description at index time, so "something light for summer evenings" lands on the right products without a rule behind it.

Misspellings, plurals, and regional spellings resolve on their own. There is no synonym list to maintain, no stopword file to tune, and nothing that silently breaks when the catalogue changes.

Vague in. Right products out.

What customers type

Vague memory

Stack of broken chocolate squares beside an amber apothecary bottle on cocoa powder

“something that smells like chocolate”

No product is named chocolate. Semantic search matches scent notes in descriptions and returns the right candles anyway.

Exact match

Navy knit running shoe suspended above a deep blue backdrop

“wool runner mizzle navy 42”

Exact queries never touch the model. They hit the keyword index and come back in milliseconds, like search is supposed to.

Full sentence

Olive and orange trail running shoe kicking up a burst of water

“a gift for my dad who runs in the rain”

Intent, audience, and use case parsed from one sentence — waterproof running gear, ranked by fit.

Typo

White pump bottle and cream jar covered in water droplets

“moisturiser for sensitiv skin”

Misspellings and regional spellings resolve without a synonym list. Typo tolerance is table stakes; Cerno treats it that way.

Constraint

Stack of folded linen shirts in sage and oat on a blue surface

“linen shirt under 50 dollars”

Price, size, and availability constraints are extracted from the query and applied as filters — not matched as keywords.

Built to answer every query

The AI-native search bar

  • Keyword tier

    Exact queries like SKUs and product names hit a regional index and return in milliseconds.

    Instant
  • Semantic tier

    Vague and conversational queries fall through to semantic search over embeddings.

    AI
  • Merchandising rules

    Boost products, hide products, pin results for a query. The model works inside your rules.

    Control
  • Catalog sync

    Catalog changes index in minutes, not after a nightly rebuild. Price and stock stay current.

    Real-time
  • Search analytics

    See which queries return nothing or get no clicks, then fix them with a rule instead of a guess.

    Insight
Learn more →
Card catalog cabinet in spotlight, one drawer open with blue-tabbed index cards

The search bar is where customers tell you exactly what they want. Cerno is built to listen.

Built for how you sell.

Use cases

01

Shelves of folded knitwear in a color gradient from cream to blue

D2C brands

Your customers search in their own words. Cerno matches products, not just titles, with no synonym lists to maintain.

Learn more →

02

Stacked kraft shipping boxes in a bright warehouse aisle

Large catalogs

Thousands of SKUs stay searchable. Keyword queries hit a regional index; vague ones fall through to semantic search.

Learn more →

03

Figure ascending a concrete escalator, motion-blurred

Headless stores

One typed SDK against the same search API the widget uses. Bring your own frontend, keep the same results.

Learn more →

Find the answer first.

The usual questions are here. The last row searches our docs for you, faster than waiting on a reply.

Both, and the customer never picks. Exact queries like SKUs and product names hit a regional keyword index and return in milliseconds. Vague or conversational queries fall through to semantic search over embeddings. One bar, routed per query.
Shopify through an embedded app that installs the widget and keeps the catalog in sync. Headless and custom storefronts through the public API and the typed SDK, @ribban/cerno, which returns the same results the widget does.
No. Matching is on meaning, not title strings, so sneaker, trainer and kicks find the same shoes and misspellings resolve on their own. You can still add a synonym rule when you want a specific mapping.
Yes. Boost products, hide products, pin a result for a query. Rules apply on top of the model's ranking, never underneath it, and take effect immediately. A copilot in the dashboard turns a plain-language instruction into rules you can review.
On Shopify, usually the same day. Install the app, the catalog syncs, the widget attaches to your theme's existing search input. New and changed products are searchable in minutes, not after a nightly rebuild.
Every plan starts with 14 free days, no card. Growth is 999 kr a month for a single storefront and Scale is 2 999 kr a month for large catalogs and several markets. Overage is capped by a spend limit you control, and search never switches off because of billing.

Live demo

Not a demo catalog. Yours.

Cerno indexes your public catalog in about a minute. No signup, nothing installed. Then search it the way your customers do: vague, misspelled, in their own words.

Paste your store URL