Search client

VisualSearchRequestResource

VisualSearchRequestResource models the visual search payload used inside SearchRequest.visualSearch for image/text-driven similarity searches. This resource is defined in src/models/search/visual-search-request-resource.ts.

Class structure

VisualSearchResource (base) - shared visual-search fields (color, type, imageEntityId, text)
VisualSearchRequestResource (request) - extends base and adds image (base64 image hash/string)

Property reference

TS property	Serialized JSON key	Type	Description
`color`	`color`	`string`	Selected color filter for visual search context.
`type`	`type`	`VisualSearchType`	Strategy used to combine vector and text signals.
`imageEntityId`	`imageEntityId`	`number`	ID of the asset used as the visual reference image.
`text`	`text`	`string`	Additional text query prompt to influence results.
`image`	`image`	`string`	Base64 hash/string of uploaded image used as search input.

VisualSearchType values

type VisualSearchType =
  | "vector"
  | "vector+text"
  | "text"
  | "vectorOrText"
  | "vectorAndText";

In practical terms:

vector - image/vector similarity only
text - text intent only
vector+text - combined relevance
vectorOrText - broader matching (either signal can match)
vectorAndText - stricter matching (both signals should match)

Example use cases

Use case: Code
Find visually similar assets from an uploaded image:

new VisualSearchRequestResource({
    type: "vector",
    image: "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA...",
});

Hybrid search: image + text intent + color hintTypical use case - a user uploads a reference image and adds text to bias results toward a specific style.:

const visualSearch = new VisualSearchRequestResource({
  type: "vector+text",
  image: "data:image/jpeg;base64,/9j/4AAQSk...",
  text: "minimalist white sneaker",
  color: "white",
});

Use an existing asset as the reference imageTypical use case - user starts from an already indexed asset in the DAM instead of uploading a new file.:

const visualSearch = new VisualSearchRequestResource({
  type: "vectorAndText",
  imageEntityId: 123456,
  text: "outdoor product lifestyle",
});

Text-first fallback for AI search box behaviorTypical use case - no image available, but still using the same visual-search request shape for text-driven discovery.:

const visualSearch = new VisualSearchRequestResource({
  type: "text",
  text: "blue packaging with geometric pattern"
});

If you have suggestions for improving this article, let us know!