Skip to main content
The More Like This endpoint finds products similar to a known source product. It is product-anchored: pass a product URL or indexed UUID, and StoreRouter derives the query from product enrichment, style, tags, audience, and optional facet constraints. Use POST /v1/products/more-like-this when you already have a product and want related items. Use POST /v1/products/search when the user starts from a keyword, facets, or an arbitrary price range.

Basic similar-products request

Scope to specific catalogs

Omit catalogs to search across all active crawled catalogs. Pass catalogs, a list of 1–100 catalog keys, to restrict results to those catalogs — for example, a one-entry list when a related-products rail should stay within the same merchant. The source product itself may live outside the list:

Choose relative price behavior

price_preference is relative to the source product’s currentPrice:

Add facets, exclusions, and bounds

More Like This takes the same caller controls as catalog search:
  • facets lists filters a product must match. A facet you send replaces the generated facet of the same name, and it is never relaxed.
  • exclude_facets removes products that match any listed value.
  • price_min / price_max bound price. They combine with price_preference by keeping the tighter bound.
  • brand_quality_min / brand_quality_max bound brand quality, from 1 (very poor) to 6 (exceptional).
Each facet name may appear only once in facets or exclude_facets.
facets replaces include_facets. A request that still sends include_facets gets a 422 whose message names facets. Upgrade to octogen-ai-sdk 0.8.0, @octogen-ai/sdk 0.6.0, or @storerouter/cli 0.4.0. Older releases send include_facets.

Relaxation

More Like This builds its query from facets StoreRouter generates from the source product: category, gender, color, and attributes. When those facets leave page 1 short, relax retries with generated facets dropped in stages, least trusted first. It returns the first stage that fills the page. relax defaults to true on More Like This. Search defaults to false. Both routes follow the same rule: what you send is never silently dropped, but StoreRouter may drop its own inferences.
  • On search, the generated facets come from the shopper’s words in q, so dropping them silently would ignore what the caller said.
  • On More Like This there is no caller text. The generated facets are StoreRouter’s inference from the source product.
Your facets, exclude_facets, and price and brand-quality bounds are never dropped. When relaxation runs depends on the source: Send relax: false to keep every generated facet. A relaxed page’s nextCursor continues the same relaxed results. Sending that cursor with relax: false returns 400.

Debug the derived query

Set debug: true to inspect the public, server-derived query fields used for retrieval:
The response adds effectiveQuery with fields such as text, retrievalEmbeddingColumns, facets, priceMin, priceMax, and limit. With relax: true (the default), it also carries relaxation, in the same shape search uses:
  • rounds: stages applied to produce this page, 0 if none;
  • declinedRounds: stages tried that did not fill the page;
  • droppedFacets: the generated facets dropped.
Use this for diagnostics and logging, not as a stable replacement for the endpoint itself.

Paginate similar products

More Like This uses the same cursor model as search. When a response contains nextCursor, pass it back as cursor with the same source and filters:

Going further

For five worked recommendation strategies — same-look, attribute matching, broad discovery, attribute-retrieval ranked by style, and price-bounded similarity — with real requests and live results, see the strategy recipes guide.