Teneo Developers
Table of Contents
Was this page helpful?

Share to

Teneo Studio

Generative AI

Teneo 7.4 brings generative AI copilot features to be used at development time in order to generate the following:

  • Class Training examples
  • Class Test examples
  • Entity entries
  • User Intent Examples (only Studio Desktop)
  • Answer Variations (only Studio Desktop)

Studio Account page

This work brings changes to the Account page in Studio Desktop and a completely new Account page in Studio Web available by clicking the user name.

Account page

To define the settings to use Generative AI, the user must:

  • Create a new setting selecting LLM Azure OpenAI
  • Give the setting a Name and optionally add a Description
  • Provide the Resource Name, Deployment Id and API Key all of which can be found in Azure as described in the reference documentation

With these details the Setting can now be used in Studio; be aware that more than one LLM Setting can be added to Teneo and therefore the user should activate/deactivate the setting which is needed/wanted for the Generative AI/copilot features in the Studio solution and leave the other(s) deactivated for later use.

With this first iteration, the following Large Language Models are supported:

  • Azure OpenAI
    • gpt 3.5
    • gpt 4.

The needed Settings or credentials are - as already seen with the CLU integration in Teneo 7.3 - of course stored encrypted in Teneo and it is the user of Studio who brings and uploads their own key(s). The user must have the "Modify Account Settings" permission to be able to modify any account settings and it is not possible to read the keys neither via the API nor directly in the User Interface (Studio).

Visit the Account page in the Reference Documentation for more details.

Generate Examples

Once the above settings are defined on the Account page, it is time to generate examples in the Solution; please note that examples can only be generated if the Setting is added to the Teneo Account, i.e., the Generate buttons are only visible when the settings are added.

Classes

For Classes, the developer can generate both Training Examples and Test Examples in both of the Studio frontends, this is done by opening a Class in edit mode and for Studio Web clicking either Generate Training Examples or Generate Test Examples below the lists of each type of examples and in Studio Desktop by selecting either the Training Examples tab or the Test Examples tab and then clicking Generate.

In either of the above cases, Studio opens a window asking the user to select which type of data the LLM should use for the generation: existing examples or a description (which should be added in case it is selected). The user must also specify the number of examples to be generated, the default is 5.

When the new examples are generated, these are highlighted with a yellow border as seen in the below image.

Entity Entries

The Entity entries are generated similarly to the Class examples after opening the Entity in edit mode. In Studio Desktop, the developer finds the Generate button in the top ribbon while in Studio Web it is available in the far right hand side of the window. Again, the developer should select if the entries should be generated based on existing entries or based on a description and also the number of entries to generate.

The entries are then generated and added to the Entries list and highlighted with a yellow border to make them easily recognizable.

User Intents - Desktop Only

Both Positive and Negative examples of User Intent can be generated by the copilot feature allowing to easily bring in new examples which can later be used for testing purposes or to simply improve the coverage of the selected Flow. Note that examples of User Intent can be generated both for Flow Triggers and Transitions and currently only in Studio Desktop.

Ones again, after selecting the requirements for the generation, the new examples are added to the respective list in a pre-selected state and highlighted with a yellow left-side border.

Output Answer Variation - Desktop Only

Teneo Studio Desktop also allows the Generative AI Copilot feature to generate more answers and thereby allows to easily add more answer variations in an easy and simple manner with just a few clicks. Again, the developer must choose if to generate these from existing answers or from a provided description as well as how many answers to generate and once add to the list, the new answers are highlighted with a yellow border.

Teneo Studio Web

Sub-flows and Prompt Flows

The Teneo 7.4 release brings the option to create both Sub-flows and Prompt Flows directly in Studio Web.
Both options are available upon Flow creation allowing the developer to select the wanted Flow type directly in the Create new flow dialogue.

Create New Flow

Others

  • The Plus icon, following directly after a Flow Link node which links to a non-Sub-flow is not available anymore as of Teneo 7.4. The removal of the Plus icon is due to the Flow becoming invalid and the impossibility of running Engine if a Flow Link node is linked to a non-Sub-flow and has post transitions.
  • An issue with headers and columns not being aligned in the overview pages of File Resources, Variables, Scripted Contexts and Class has been fixed.
  • The read-only message is again displayed correctly when clicking to open an Entity or Language Object from the Solution Structure which is already opened by another user.
  • Furthermore, the enabled/disabled status of documents is now also correctly displayed in the Solution structure view of a solution, showing disabled documents in italic with the disabled icon. The enable/disable button is also available in the far right-hand side of a document on hover over.

Disabled document in Italic

  • Also, see the description below of an incorrect behavior detected in the Flow graph (both in Studio Web and Desktop) which has been corrected in the 7.4 release.

Teneo Studio Desktop

Bug Fixes and Improvements

The 7.4 release brings various bug fixes and improvements to Teneo Studio Desktop as described below.

Flow Graph
  • In previous Studio version (note that this is applicable to both Desktop and Web), when splitting the path in the Flow graph from the Plus menu of an existing input consuming transition into any kind of node but a full transaction, the transition for the new node was added as non-input consuming. This behavior was not correct and therefore, in Teneo 7.4, the behavior has been corrected and now when splitting path from a transition that starts with a "Give Response" node (i.e., at least one of the transitions leaving the previous node has get input before continuing checked), then the transition of any new node added with the split also have the get input before continuing checked.
    If none of the transitions have get input before continuing checked, and no "Get Response" node exists, then the transition of any new node added with the split of course does not have this option checked.
  • Improvements are performed to the visualization of the Flow graph with a special focus on transition loops to and from the same Output node, and also on correct visualization of drop nodes after import/upgrade from a Teneo 6.2 environment. Enhancements are implemented to ensure correct focus on newly added nodes and on the Transition panel when the developer deselects "Get input before continuing".
  • An extra Plus icon, which in previous Teneo 7.x releases would appear in the Flow editor when deleting all the Matches and After Matches on a transition, has been removed in Teneo 7.4.
  • An issue, related to nodes not being linked in the Flow editor when redoing an action after undoing it, has been fixed.
Solution Explorer - Root Folder
  • Improvements are implemented for selecting/deselecting multiple documents and the solution's root folder in the Explorer view to ensure correct selection of both documents and folders. Also, an issue causing documents to be displayed incorrectly when selecting the solution's root folder and clicking Disabling this folder and all sub-folders is improved to ensure correct visualization of disabled documents.
  • An issue found in Teneo 7.3 related to copy and paste of Entities, Language Objects and Flows in the Solution's root folder has been fixed in Teneo 7.4.
  • The functionality of the Filter in the top ribbon of Teneo Studio is reviewed to ensure correct behavior in scenarios where a solution is closed and another opened.
Log Data Source
  • In previous Teneo versions, when importing an augmenter with a minimum/maximum solution revision set - or because after some time a revision was no longer within the data range - it could happen that the maximum and the minimum values were not in the drop down list and the selected revision would erroneously state No minimum or No maximum although the minimum/maximum revisions were actually set. In Teneo 7.4, this has been reviewed to ensure correct display of the revisions to the developer.
  • Also, an issue detected when performing reload of the Augmenters view due to an augmenter name containing a dot has been been fixed.
Others
  • An issue with the filter in the Recycle Bin when closing and opening the same or another solution has been corrected.
  • If cancelling the edit of a Resource File in previous Teneo 7.x versions and then clicking View, the displayed version was not always the latest version of the Resource File. This has been fixed in the 7.4 release.
  • Comments can now be added when saving classes in Studio Desktop's Class Manager.
  • In the Lexical tab (in the backstage of Teneo Studio), lexical resources are again displayed with the newest (unassigned) version at the top followed by the assigned version. All other versions of the lexical resource can be displayed by clicking the link.
  • Since 7.1, when adding a new Global Script to a solution a number of errors would be printed in the backend logs, but this has now been reviewed and corrected.
  • In Teneo 7.3, an error was erroneously written to the Studio log files when a document was opened/modified/closes several times. For the 7.4 release this has been fixed.
  • A review of use cases where error handling was missing has been performed to ensure correct handling and a better user experience.
  • A potential security issue detected in the Studio API (related to Log Data Source endpoints) has been fixed.
  • Some encountered use cases where the Teneo Studio frontend would crash have been reviewed and fixed to ensure stability of Teneo Studio.