Classification-specific AI text generation – novomind iPIM Documentation

Since version 5.6 it is possible to define taxonomy- resp. category-specific prompts for the AI-supported text generation. For this purpose, you can also define attribute profiles that are used as the basis for the text generation. Using an example, this chapter explains how the AI-supported text generation is configured for a category of the primary taxonomy and for the text type “PLAIN”. Additionally, it is explained how to create an attribute profile and how to specify it as the basis for the text generation in this particular category, whereby this step is optional.

The classification-specific AI text generation has to be activated via the property general.openaitext.classification.dependent which needs to be set to “true”.
The steps regarding the attribute profile are optional. If no attribute profile is specified to the text generation, a default selection of attributes is used. This selection comprises of the attributes that where defined via the property attribute.identifier.textgen as well as the category attributes of the respective products.
In this example, the category properties were created as localized string attributes, so that separate AIHub workflows can created for the various languages. Though, there is also the option to create the category properties as non-localized strings. Then, the differentiation of the languages happens within the AIHub workflow. (prompt_<language>, e.g. prompt_en, prompt_fr etc.)

The classification-specific generation of values also works for the “Attribute generation via AI” action.

Category property for the text type relation

In this step, a category property is created. This category property is used to relate a text type to a classification. In this example, this is the text type “PLAIN” that is related to the primary taxonomy. Later, the category property is used to specify the appropriate AI workflow as the basis for the AI text generation in a particular category.

Create the category property

In this step, the category property is created:

Call up iPIM CC.
Go to the “Configuration” tab of the attribute configuration.

Click on the CREATE button. A new attribute dataset with empty input fields is generated. Only the “ID” field contains a placeholder.

Enter the identifier for the category property in the “Attribute ID” field . The required identifier comprises of a prefix and the identifier of the desired text type which are to be separated by an underline. In this example, this is “OPENAI_CONFIG”* and “PLAIN”.
From the “Attribute type” menu , select the attribute type “Character string, localized”.
From the “Scope” menu, select the attribute scope “Category property”.
From the “Mandatory field” menu, select the option “Optional.” Once all the mandatory fields are filled in, the SAVE button is shown.
Click on the SAVE button to save your input. A success message is shown. The attributes resp. the category property is now created.

*The prefix can be defined via the property general.openaitext.classification.prefix. In this example, the value for the property is “OPENAI_CONFIG”.

Relate the category property to the classification

In this step, the category property is related to the primary taxonomy:

Go to the “Categories” tab.

Click on the ADD button to call up the corresponding dialog box.

From the “Versions” menu , select the primary taxonomy.
Tick the primary taxonomy .
Click on the SAVE button to finalise the classification selection.
The dialog box closes and a success message is shown. The category property is now related to the primay taxonomy.

Category property for the attribute profile relation

In this step, another category property is defined. Also this property is used to relate a text type to a classification. In this example, this is the text type “PLAIN” that is related to the primary taxonomy. Later this category property is used to specify which attribute profile is used as the basis for the AI text generation in a particular category.

Create the category property

In this step, the category property is created:

Go back to the “Configuration” tab of the attribute configuration.

Click on the CREATE button. A new attribute dataset with empty input fields is generated. Only the “ID” field contains a placeholder.

Enter the identifier for the category property in the “Attribute ID” field . The required identifier comprises of a prefix and the identifier of the desired text type which are to be separated by an underline. In this example, this is “AI_TEXT_GEN_PROFILE”* and “PLAIN”.
From the “Attribute type” menu , select the attribute type “Character string, localized”.
From the “Scope” menu, select the attribute scope “Category property”.
From the “Mandatory field” menu, select the option “Optional.” Once all the mandatory fields are filled in, the SAVE button is shown.
Click on the SAVE button to save your input. A success message is shown. The attributes resp. the category property is now created.

*The prefix can be defined via the property general.openaitext.classification.profile. In this example, the value for the property is “AI_TEXT_GEN_PROFILE”.

Relate the category property to the classification

In this step, the category property is related to the primary taxonomy:

Go to the “Categories” tab.

Click on the ADD button to call up the corresponding dialog box.

From the “Versions” menu , select the primary taxonomy.
Tick the primary taxonomy .
Click on the SAVE button to finalise the classification selection.
The dialog box closes and a success message is shown. The category property is now related to the primay taxonomy.

Activate the text type

In this step, the text type is activated for the AI text generation. In this example, this is the text type “PLAIN”. In order to activate the text type, the type property AiGeneratedText (REST-API: AI generated Text) needs to be set to “true”:

Go to the “Types” area of the “System setup” section in iPIM CC.

From the “Select type” menu, select the option “TEXT_TYPE”. The existing text types are shown in the table.

Choose the text type “PLAIN” from the table. Optional: Enter a search term in the search field to filter the list of text types by a search term.
Click on the button to call up the dialog box for selecting type properties.

Choose the type property AiGeneratedText (REST-API: AI generated Text) from the list. Possibly enter a search term in the search field to filter the list of type properties by a search term.
Transfer the type property from the left column to the right column to select it. To do so, click on the type property .
Click on the ACCEPT button to finalise the type property selection.
The dialog box closes and the type property is shown in the table.

In the table row of the text type “PLAIN”, double-click in the cell of the column of the type property .
Select the option YES from the menu.
Click on the SAVE button to finalise the configuration of the type property.
A success message is shown. The text type is now activated for the AI text generation.

Create an attribute profile (optional)

In this step, the attribute profile is created. The attributes from this profile are used for the AI text generation. In this example, a selection of category attributes is assembled. These attributes belong to the “Kitchen devices” category of the primary taxonomy:

Go to the “Profiles” area of the “Attributes” section in iPIM CC.

Click on the CREATE button to call up the dialog box for creating attribute profiles.

Enter the identifier for the attribute profile in the “Identifier” field .
Activate the “User” radio button .
Select the appropriate user* from the menu .
Click on thebutton to proceed to the next slider.

Activate the “Category attributes” radio button .
Click on thebutton to proceed to the next slider.

From the “Versions” menu , select the primary taxonomy. The category directory of the primary taxonomy is shown below the menu.
Choose the category from the category directory. Optional: Enter a search term in the search field to filter the list of categories by a search term.
Click on the category to select it.
Click on thebutton to proceed to the next slider.

Choose the desired attributes from the list of available attributes. Optional: Enter a search term in the search field to filter the list of attributes by a search term.
Transfer the first attribute from the left column to the right column to select it. To do so, click on the attribute.
Transfer the rest of the desired attributes to the right column.
Click on the ACCEPT button to finalise the the creation of the attribute profile.
The dialog box closes and a success message is shown. The attribute profile is now created.

The user selected at this point needs to be assigned to the property general.openaitext.profile.userid as its value.

Configure the category properties

In this example, the appropriate AIHub workflow is specified for the text generation in the category “Kitchen devices” of the primary taxonomy. For this purpose, the AIHub workflow is assigned to the category property “OPENAI_CONFIG_PLAIN” that was created before. Additionally, the attribute profile that was created is assigned to the category as the basis for the text generation, whereby this step is optional*.

Call up the iPIM Portal.
Go to the “Taxonomies” app.

From the upper menu in the “Taxonomy search” card, select the primary taxonomy. The category directory of the taxonomy is shown in the lower part of the card.
Choose the category from the category directory. Optional: Enter a search term in the search field to filter the list of categories by a search term.
Click on the category to select it.
Select the “Taxonomy properties” tab .
Click on the ADD PROPERTIES button to call the dialog box for category properties.

Choose the category properties “OPENAI_CONFIG_PLAIN” and “AI_TEXT_GEN_PROFILE_PLAIN” from the list. Optional: Enter a search term in the search field to filter the list of category properties by a search term.
Transfer the category properties from the left column to the right column to select them. To do so, click on each of the category properties .
Click on the ACCEPT button to finalise the selection of the category properties.
The dialog box closes and the category properties are shown in the table.

In the table row of the category property “OPENAI_CONFIG_PLAIN”, double-click in the cell of the “Value” column .
Enter the identifier of the AIHub workflow. In this example, this is “create_product_description”.
In the table row of the category property “AI_TEXT_GEN_PROFILE_PLAIN”, double-click in the cell of the “Value” column .
Enter the identifier of the attribute profile (ai-att-profile).
Click on the SAVE button to to save your settings.
A success message is shown. The category properties are now configured.

*If no attribute profile is specified to the text generation, a default selection of attributes is used. This selection comprises of the attributes that where defined via the property attribute.identifier.textgen as well as the category attributes of the respective products.