Method: projects.generateProductText

Generates a set of candidate text completions/responses from product information.

HTTP request

POST https://productstudio.googleapis.com/v1/{parent=projects/*}:generateProductText

The URL uses gRPC Transcoding syntax.

Path parameters

Parameters
parent

string

Required. Project to make a call. Must refer to a caller's project. Format: projects/{project-id}.

Request body

The request body contains data with the following structure:

JSON representation 
{
  "titleExamples": [
    {
      object (TitleExample)
    }
  ],
  "productInfo": {
    object (ProductInfo)
  },
  "outputSpec": {
    object (OutputSpec)
  }
}
Fields
titleExamples[]

object (TitleExample)

Optional. Provide some hand-crafted examples of title improvements that are unique to your usecase. TiDe is a general tool that handles multiple product categories, but your brand identity may require custom functionality. Feel free to specify that here.
productInfo

object (ProductInfo)

Required. Available information about the product. Used to inform the genAI models.
outputSpec

object (OutputSpec)

Optional. Configuration parameters that directly influence what content is generated, and how that content is rendered in the final response.

Response body

Text generated for a provided product.

If successful, the response body contains data with the following structure:

JSON representation 
{
  "attributes": {
    string: string,
    ...
  },
  "title": {
    object (ProductTextGenerationSuggestion)
  },
  "description": {
    object (ProductTextGenerationSuggestion)
  },
  "metadata": {
    object (ProductTextGenerationMetadata)
  }
}
Fields
attributes

map (key: string, value: string)

Any other generated attributes

An object containing a list of "key": value pairs. Example: { "name": "wrench", "mass": "1.3kg", "count": "3" }.
title

object (ProductTextGenerationSuggestion)

Generated title.
description

object (ProductTextGenerationSuggestion)

Generated description.
metadata

object (ProductTextGenerationMetadata)

Additional info that clients may want to audit surrounding the generation.

ProductInfo

Available information about the product. Used to inform the genAI models.

JSON representation 
{
  "productAttributes": {
    string: string,
    ...
  },
  "productImage": {
    object (Image)
  }
}
Fields
productAttributes

map (key: string, value: string)

Required. A mapping of all available product attributes. This may include title, description, brand, gender, color, size, etc.

An object containing a list of "key": value pairs. Example: { "name": "wrench", "mass": "1.3kg", "count": "3" }.
productImage

object (Image)

Optional. Image associated with the product.

OutputSpec

Configuration parameters that directly influence what content is generated, and how that content is rendered in the final response.

JSON representation 
{
  "workflowId": string,
  "tone": string,
  "editorialChanges": string
}
Fields
workflowId

string

Optional. The workflow to execute for the provided product data. Workflows may populate the response's title, description, or both.
tone

string

Optional. The desired output tone for the generated text. e.g. serious, casual.
editorialChanges

string

Optional. Any editorial changes for the generated product data. e.g. replace Small with "S", do not modify color if already present.

TitleExample

A hand-crafted example of a product title improvement. These examples are provided to the AI to improve its quality.

JSON representation 
{
  "productInfo": {
    string: string,
    ...
  },
  "finalProductInfo": {
    string: string,
    ...
  },
  "titleFormat": string,
  "category": string
}
Fields
productInfo

map (key: string, value: string)

Required. A map containing all existing product information. e.g., {"title": "dress", "description": "A red dress", "brand": "Dresses4All"} Any information that you might use to populate your product feed.

An object containing a list of "key": value pairs. Example: { "name": "wrench", "mass": "1.3kg", "count": "3" }.
finalProductInfo

map (key: string, value: string)

Required. A map in the same format as productInfo but with all improvements included. e.g., {"brand": "Dresses4All", "product": "dress", "color": "red", ...} For now, order these attributes in the order that you want them to appear in the final title. e.g., the above will become: Dresses4All dress | red

An object containing a list of "key": value pairs. Example: { "name": "wrench", "mass": "1.3kg", "count": "3" }.
titleFormat

string

Required. The attributes or approximate attributes that make up the title. e.g., title "Google GShoe M" titleFormat can be "brand | product | size".
category

string

Required. The product's category. This helps the AI understand when certain examples are more relevant than others.

ProductTextGenerationSuggestion

Text generated for a product, optionally including its quality score.

JSON representation 
{
  "text": string,
  "score": number,
  "changeSummary": string
}
Fields
text

string

The text generated
score

number

The quality score associated with the generation. Heuristic implemented according to the feedgen team's implementation styles.
changeSummary

string

A brief summarization of all the changes that have been made.

ProductTextGenerationMetadata

Wrapper data type for any metadata associated with text generation.

JSON representation 
{
  "metadata": {
    object
  }
}
Fields
metadata

object (Struct format)

Metadata is a pretty loose concept. The data is modeled as a map here to indicate that there is no guaranteed structure to the output past a simple K:V association. The first use-case is to track words added/removed/changed in generations.