Ursula AI

(Since version 1.19)

 

The Ursula AI powers all implementations within the SAWSConnector suite, leveraging your selected AI language models to deliver its advanced functionalities. Interestingly, the AI shares its name with the owl representing the Owl Cloud Services, symbolizing wisdom and innovation.

Since the version 1.19, ChatGPT could be used by setting the API Key in the Contentserv options menu.

Since the version 1.20, this API Key option was removed and replaced by the AI Control Hub enabling you to use multiple AIs interchangeably.

 

How To Activate and Assign a Supported AI in the AI Control Hub

1) Go to the SAWSConnector Suite.

2) Open Subconfigurations.

3) Select AI Control Hub.

 

The AI Control Hub will open.

4) Activate your intended AIs by setting the API Key (Contentserv Credential) and if necessary the URL.

Please note: All changes will be saved immediately.

 

 

After activating your AI the AI Purpose Model Assignment menu will appear below. Its main purpose is the assignment of models for the purposes they fit best. Upon activation it will align the purposes itself with the recommended models.

The image below shows an example of how the menu it could look like after choosing ChatGPT. You can, of course, choose multiple AIs.

Please note: As long as the recommended assignment is activated the AI Control Hub will download and update the assigned AIs and models on a daily basis.

 

Defining Your Own Assignments

You can always choose other AIs and models for the assignments by clicking on the AI box or typing in another model in the Model box.

This will however deactivate the automated updates for the user-defined assignment, since it is now overwritten by the user.

 

Also, a button will appear beneath the menu:

By clicking on ‘Revert to recommended assignment‘ the AI Control Hub will return into recommended assignment mapping state and re-select the best fitting AIs and models for the specific purposes.

 

AI Control Hub

 

AI Activation

Here you can activate an AI.

Please note: For the Azure ChatGPT you have to enter an URL. For example: https://your-services.openai.azure.com/openai/deployments/your-model

 

AI Purpose Model Assignment

Allows you to assign an AI and a subsequent model to a specific purpose.

These are the following:

 

Master

The Master purpose will always be used when there is no AI selected for the subordinated purposes.

 

Transformation

This means the alteration of values.

For example: Altering and updating descriptions.

Used by: AI Value Transformation Format

 

Translation

This means the translation of existing texts and values into other languages.

Used by: AI Value Translation Format

 

Assignment

This means to assign and sort items into selected trees and leafs.

For example: Automatically sorting the items from the inbox into categories.

Used by: AI Product Tree Organizer Active Job, Mapping Sales Channels via the Ursula AI Active Job

(We recommend selecting a model for huge but simple tasks here, for example gpt-4o-mini, since its token costs are low and high logical intelligence is not needed for these simple assignments.)

 

Population

This means the population of Contentserv item (products or DAM files) attributes. This is escpecially useful to fill missing attributes based on existing item data.

For example: Descriptions.

Used by: Populate Attributes via Ursula AI Active Job, Generate Descriptions with Ursula AI Active Job

(The selected model should support image processing (gpt-4o or pixtral), so it populate values based on the item image(s).)

 

As you can see, our AI features are grouped into different purposes. Since different AI models might be more fit for different purposes.

For example: The Assignment features process huge product lists into each other, hence a model with less logic intelligence might be more recommended here (e.g. gpt-4o-mini).

Ursula AI Assignments

(Since version 1.21.14)

The complexity of AI models has grown over the years. AIs can be provided with components like vector stores or glossaries to work more in the confinements of your company or goal. Some support options such as temperature or top-p to limit its creativity.

To manage your AIs across every AI feature we introduced the AI management in the Owl Cloud Services. Here, you can manage your AI connections and configurations per assignment. This means that multiple configurations can be used for one assignment where previously only one was supported.

For example: One ChatGPT assignment with vector store ID “A” for one transformation and another ChatGPT assignment with vector store ID “B” and higher creativity for another transformation.

The Ursula AI via the Owl Cloud Services is the recommended and future-proof way to use your AI features connector-wide, since the cloud services are updated faster to react to the ever-changing world of AI development and the flexibility of this feature is unmatched.

To do so select the Ursula AI in the AI Control Hub.

If you have an assignment configured in the Owl Cloud Services, you will be able to choose from your assignments in the AI feature of the connector. Here I am seeing a selection of my transformation assignments since I am in the AI transformation format. If no custom assignment was configured, this selection would be hidden and the default assignment for that purpose (here transformation) would be used.

Usage Metrics

The Ursula AI also comes with usage metrics which show you the token usage of your AI features across your system.

Was this article helpful?

6 0

How can we improve this article?

Please submit the reason for your vote so that we can improve the article.

Name

Email

Submit