Skip to main content
Version: latest

Use Console Skill Builder

This is a guide to defining Skills for SENSA Fabric using the Skill Builder in Fabric Console.

note

Console Skill Builder can be used only to build Skills that package Models. Use other Skill building methods to build a Skill that does NOT encapsulate a Model.

Build a Skill​

After a model artifact is prepared, you may package it into a Skill using the Console Skill Builder or using the CLI cortex generate skill process.

To build a Skill from the Fabric Console:

  1. Login to Fabric Console and select your Project Context.

  2. Click Agents in the left menu.

  3. Click the Skills tab at the top of the right panel.

  4. A list of Skills is displayed.

  5. Click Add Skill at the top right to start the Skill Builder wizard.

  6. Enter a brief descriptive Skill Title .

  7. Review the auto-generated Skill Name.

    You may toggle the auto-generated switch to off and edit the name or enter a unique URL-friendly name. Names must be alphanumeric, beginning with a letter and ending with a letter or number. Dashes and underscores are allowed in the middle; no other special characters can be used. (Names must be 20 characters or fewer)

  8. Enter a Description of the Skill that will assist collaborators.

  9. Select a Skill Encapsulation method. ("Encapsulate a model" is selected by default).

    To define a Skill that does not encapsulate a model use the CLI cortex generate skill method.

  10. Click Next at the bottom of the right panel.

  11. Specify the Scope by selecting either "Current Project" or "All My Projects".

  12. Click Next at the bottom of the right panel.

  13. Select a Model for the Skill. Filter the Model list by selecting one or more tags in the top field. Enter some or all of the Model Title to further filter the list. (NOTE: Only Models that are "PUBLISHED" can be selected.)

  14. Click Next at the bottom of the right panel.

  15. Select the Model Version (or run artifact) to use in the Skill.

  16. Click Next at the bottom of the right panel.

  17. Select a Skill Template from the drop down.

    Currently, a selection of Skill templates is available to be imported, like "Batch", "Bulk-batch Prediction", "Batch Prediction", and "Online Prediction"

    To use a JDBC connection use this template: cdata-jdbc-connector-template

    Additional templates are available in the SENSA Fabric Examples repo.

  18. Review the Properties from the template you select to apply to the Skill.

  19. To modify the template properties you may:

  • Click + Add New Property below the property list. A row opens in the Properties list where you enter the key (Property name), the data type, and (optionally) a default value. (Skill Properties list)

  • Click Import Podspec to open a window where you may paste a JSON code block that defines the Skill Properties. The table is updated when you click Validate and Save.

    Properties that do not serve as primary keys may be edited in the table. To delete or duplicate Properties select the desired action from the kebab menu at the end of the Property row.

  1. Click Next at the bottom of the right panel.

  2. (Optional) Select an Input Message Type. This selection determines the default Input Parameters.

  3. Enter an Input Title for the the Skill's Input Parameters.

  4. The Input Name may be autogenerated or entered (See Skill Name above)

  5. To add Input Parameters click + Add New Property below the Input Parameters table. A row opens where you enter the parameter name and data type, and check the box if the Parameter is required.

    Parameters may be edited in the table. To delete or duplicate Parameters select the desired action from the kebab menu at the end of the Property row.

    note

    The model_artifact key and model_id properties are configurable as Skill properties to enable model encapsulation.

  6. Click Next at the bottom of the right panel.

  7. (Optional) Click Add Input Parameters to add another set of Input Parameters, OR ...

  8. Click Next at the bottom of the right panel.

  9. (Optional) Select an Output Message Type. This selection determines the default Output Parameters. (By default the Output Message Type defaults to the one that corresponds to the Input Message type you selected.)

  10. Enter an Output Title for the the Skill's Output Parameters.

  11. The Output Name may be autogenerated or entered (See Skill Name above)

  12. To add Output Parameters click + Add New Property below the Output Parameters table. A row opens where you enter the Parameter name and data type, and check the box if the Parameter is required.

    Parameters may be edited in the table. To delete or duplicate Parameters select the desired action from the kebab menu at the end of the Property row.

  13. Click Next at the bottom of the right panel.

  14. (Optional) Click Add Output Parameters to add another set of Output Parameters, OR ...

  15. Click Save at the bottom of the right panel.

The Skill is saved and a read-only view of the Skill configuration is displayed or an error message is displayed in toast at the top, and the Skill is not saved.

A saved Skill is added to the Skills list and to the Agent Composer Skill selection modal in an UNDEPLOYED status. Use the CLI to deploy the Skill.

From the Skill List you may view the Skill metadata by clicking the kebab action menu at the end of the row in the Skill table.

note
  • Skills cannot not be updated, deployed, or undeployed from the Skill Builder in Fabric Console. Use the Fabric CLI for these features.
  • Skills cannot be deleted.

Example Experiments Skill​

The Skill in this example has two distinct actions (predict and train) with models that support each action. ("Test" is a payload file)