Method: projects.locations.catalogs.servingConfigs.predict

Makes a recommendation prediction.

HTTP request

POST https://retail.googleapis.com/v2alpha/{placement=projects/*/locations/*/catalogs/*/servingConfigs/*}:predict

The URL uses gRPC Transcoding syntax.

Path parameters

Parameters
placement

string

Required. Full resource name of the format: {placement=projects/*/locations/global/catalogs/default_catalog/servingConfigs/*} or {placement=projects/*/locations/global/catalogs/default_catalog/placements/*}. We recommend using the servingConfigs resource. placements is a legacy resource. The ID of the Recommendations AI serving config or placement. Before you can request predictions from your model, you must create at least one serving config or placement for it. For more information, see Manage serving configs.

The full list of available serving configs can be seen at https://console.cloud.google.com/ai/retail/catalogs/default_catalog/configs

Request body

The request body contains data with the following structure:

JSON representation 
{
  "userEvent": {
    object (UserEvent)
  },
  "pageSize": integer,
  "pageToken": string,
  "filter": string,
  "validateOnly": boolean,
  "params": {
    string: value,
    ...
  },
  "labels": {
    string: string,
    ...
  }
}
Fields
userEvent

object (UserEvent)

Required. Context about the user, what they are looking at and what action they took to trigger the predict request. Note that this user event detail won't be ingested to userEvent logs. Thus, a separate userEvent write request is required for event logging.

Don't set UserEvent.visitor_id or UserInfo.user_id to the same fixed ID for different users. If you are trying to receive non-personalized recommendations (not recommended; this can negatively impact model performance), instead set UserEvent.visitor_id to a random unique ID and leave UserInfo.user_id unset.
pageSize

integer

Maximum number of results to return. Set this property to the number of prediction results needed. If zero, the service will choose a reasonable default. The maximum allowed value is 100. Values above 100 will be coerced to 100.
pageToken
(deprecated)

string

This field is not used; leave it unset.
filter

string

Filter for restricting prediction results with a length limit of 5,000 characters. Accepts values for tags and the filterOutOfStockItems flag.

  • Tag expressions. Restricts predictions to products that match all of the specified tags. Boolean operators OR and NOT are supported if the expression is enclosed in parentheses, and must be separated from the tag values by a space. -"tagA" is also supported and is equivalent to NOT "tagA". Tag values must be double quoted UTF-8 encoded strings with a size limit of 1,000 characters.

Note: "Recently viewed" models don't support tag filtering at the moment.

  • filterOutOfStockItems. Restricts predictions to products that do not have a stockState value of OUT_OF_STOCK.

Examples:

  • tag=("Red" OR "Blue") tag="New-Arrival" tag=(NOT "promotional")
  • filterOutOfStockItems tag=(-"promotional")
  • filterOutOfStockItems

If your filter blocks all prediction results, the API will return no results. If instead you want empty result sets to return generic (unfiltered) popular products, set strictFiltering to False in PredictRequest.params. Note that the API will never return items with storageStatus of "EXPIRED" or "DELETED" regardless of filter choices.

If filterSyntaxV2 is set to true under the params field, then attribute-based expressions are expected instead of the above described tag-based syntax. Examples:

  • (colors: ANY("Red", "Blue")) AND NOT (categories: ANY("Phones"))
  • (availability: ANY("IN_STOCK")) AND (colors: ANY("Red") OR categories: ANY("Phones"))

For more information, see Filter recommendations.
validateOnly

boolean

Use validate only mode for this prediction query. If set to true, a dummy model will be used that returns arbitrary products. Note that the validate only mode should only be used for testing the API, or if the model is not ready.
params

map (key: string, value: value (Value format))

Additional domain specific parameters for the predictions.

Allowed values:

  • returnProduct: Boolean. If set to true, the associated product object will be returned in the results.metadata field in the prediction response.
  • returnScore: Boolean. If set to true, the prediction 'score' corresponding to each returned product will be set in the results.metadata field in the prediction response. The given 'score' indicates the probability of a product being clicked/purchased given the user's context and history.
  • strictFiltering: Boolean. True by default. If set to false, the service will return generic (unfiltered) popular products instead of empty if your filter blocks all prediction results.
  • priceRerankLevel: String. Default empty. If set to be non-empty, then it needs to be one of {'no-price-reranking', 'low-price-reranking', 'medium-price-reranking', 'high-price-reranking'}. This gives request-level control and adjusts prediction results based on product price.
  • diversityLevel: String. Default empty. If set to be non-empty, then it needs to be one of {'no-diversity', 'low-diversity', 'medium-diversity', 'high-diversity', 'auto-diversity'}. This gives request-level control and adjusts prediction results based on product category.
  • filterSyntaxV2: Boolean. False by default. If set to true, the filter field is interpreteted according to the new, attribute-based syntax.
labels

map (key: string, value: string)

The labels applied to a resource must meet the following requirements:

  • Each resource can have multiple labels, up to a maximum of 64.
  • Each label must be a key-value pair.
  • Keys have a minimum length of 1 character and a maximum length of 63 characters and cannot be empty. Values can be empty and have a maximum length of 63 characters.
  • Keys and values can contain only lowercase letters, numeric characters, underscores, and dashes. All characters must use UTF-8 encoding, and international characters are allowed.
  • The key portion of a label must be unique. However, you can use the same key with multiple resources.
  • Keys must start with a lowercase letter or international character.

See Google Cloud Document for more details.

Response body

If successful, the response body contains an instance of PredictResponse.

Authorization scopes

Requires the following OAuth scope:

  • https://www.googleapis.com/auth/cloud-platform

For more information, see the Authentication Overview.

IAM Permissions

Requires one of the following IAM permissions on the placement resource, depending on the resource type:

  • retail.placements.predict
  • retail.servingConfigs.predict

For more information, see the IAM documentation.