...
Drag and drop the Text prompt from the palette onto the canvas.
On the properties, change the step name from Prompt to a user-friendly name to identify and define what the step does. This improves the readability of the skill.
You could create a new prompt by:
Toggle to ‘New Prompt’.
Add the ‘Message’ Luma should ask the user.
Provide a name for the 'attribute’ to save the user’s response.
Select the ‘datatype' type of information expected.
Confirm if the prompt is mandatory. This means the users must provide the information to go ahead with the skill.
Add the default value.
Click on Create to create the attribute. This is optional. If not clicked, the attribute will be created automatically when Skill is saved.
To add more details to the attribute, such as NLP extraction method and validations, before saving it, click on the Advanced Configuration under the create button.
Now you can pass global attributes to send in web widget context. This is applicable only for web widget.
You could also use an existing attribute.
This enables you to add an existing global attribute to the skill and use the prompt added to the attribute.
Toggle the switch to Choose existing Prompt.
Select the attribute. This loads the Message field with the message configured in the selected attribute.
You change the details by overriding the attribute.
You could click on the pencil icon to update the message. Once done, click on the green tick to save changes.
Alternatively, click on the attribute name and make the required changes on the attribute override pop-up screen.
For more details on attribute override, refer to Building Conversation Flow | Attribute Override.
The Text prompt is now ready.
...
Drag and drop the Text Message from the palette onto the canvas.
On the properties, change the step name from Message to a user-friendly name to identify and define what the step does. This improves the readability of the skill.
Add the ‘Message’ to be displayed.
The Text Message is now ready.
Now you can pass global attributes from text message control to send in web widget context. This is applicable only for web widget.
Message-Quick Replies
Quick Reply Message offers a selection to the end user. Using Quick Reply Messages, skill developers can build personalized conversations. Here you can define the message to be displayed, and the Options to be presented, providing a range of subsequent actions that can be taken.
...
Drag and drop the Quick Reply prompt from the palette onto the canvas.
On the properties, change the step name from Message to a user-friendly name to identify and define what the step does. This improves the readability of the skill.
Add the ‘Message’ to be displayed. This is optional.
Add as many options as required. You could add buttons that trigger another skill or conversation using ‘+Options' or open a URL using '+URL’
Click on the tag icon to change the value. This is the action taken when the option is selected. For more information, refer to Building Conversation Flow | Show Message.
Now you can pass global attributes from message quick reply to send in web widget context. This is applicable only for web widget.
The Quick Reply Message is now ready.
...
Drag and drop the Card prompt from the palette onto the canvas.
On the properties, change the step name from Message to a user-friendly name to identify and define what the step does. This improves the readability of the skill.
Add the ‘Message’ to be displayed. This is optional.
Add as many Cards as required. You could add buttons that trigger another skill or conversation using ‘+Options' or open a URL using '+URL’
Click on the tag icon to change the value. This is the action taken when the option is selected. For more information, refer to Building Conversation Flow | Show Message.
Now you can pass global attributes from message cards to send in web widget context. This is applicable only for web widget.
The Card Message is now ready.
...
Drag and drop the Set/Transformation attribute construct from the palette onto the canvas.
On the properties panel, you can now Set Attributes or Transform Attributes.
To Set Attribute:
Click on ‘+Set attribute’.
Now, on the Set Attribute control, configure the details.
Select the attribute to be set.
Select the scope of the attribute, i.e., Global, Local, or User Custom.
Select the Attribute from the dropdown.
Now add the Expression to indicate what type of value is to be assigned to the attribute. Add You could use a Value, JSON Path, indirect JSON, XPath, Indirect Xpath or Script to the set the value.
Add an expression that holds to assign value.
When using a script, add the script that returns a value to be set to the attribute. Use the following syntax to use existing attributes in the script:
contextAttributes.getAttribute("@{global.global1}"),
contextAttributes.getAttribute("@{local.local1}")
To Transform value:
Click on ‘+Transform’.
Now, on the Transform Attribute control, configure the details.
Select the attribute to be set after transforming a value.
Select the scope of the attribute, i.e., Global, Local, or User Custom.
Select the Attribute from the dropdown.
Now select the Transformation function.
Select the Function Mode.
Based on the Function mode, provide information to the input parameters. For more details on how to use the Transformation function, refer to Using Transformation Function.
You can add as many ‘Set and transform’ attributes as required. The functions are executed in the sequence in which these are added.
...
To use Set KPI in your skill follow the below steps:
Drag and drop "Set KPI" on the canvas.
...
On the properties panel, select the KPIs to be updated and the effort (in minutes) saved.
...
You may update the following KPIs:
Deflected Support Requests: The KPI indicates that the skill can deflect a support request. Specify the number of minutes that the skill can save as the user request is directly resolved and a support request is avoided
Resolved with Knowledge: The KPI is linked to ‘Deflected Support Requests’. Mark 'Resolved with Knowledge' if a Knowledge Article deflects the ticket.
Resolved with External Information: The KPI is linked to ‘Deflected Support Requests.’ Mark 'Resolved with External Information' if the ticket is deflected using the information generated from external sources like the Internet or generative AI.
Actionable Support Requests: Use the KPI to indicate that the skill creates a well-structured, actionable ticket. Specify the number of support minutes that can be saved by creating an actionable support request through Skill. The default value is set to 5 minutes.
Automated Resolutions: The KPI indicates that the skill resolves user issues using an automated service. Specify the number of support minutes that can be saved by the execution of the automated service, skill, or job by executing the Skill.
Ineffective Best Response: This is a Luma Knowledge KPI. The KPI indicates that the Knowledge Articles returned in the result set are incorrect or irrelevant to the user’s request.
Knowledge Gap: This is a Luma Knowledge KPI. The KPI indicates that Knowledge articles are unavailable for the user’s request and the phrase should be marked as a Gap in Luma Knowledge.
Knowledge Feedback: The Luma Knowledge KPI is used to record feedback on a Knowledge Article or a FAQ presented to the end-user. Select the type of article , and provide the Artifact ID and Feedback (Helpful/Not Helpful).
...
Skill Gap: This is a Luma VA Skill kpi. If the users response is negative during the conversation, using this kpi one can identify the cases where skill training or new skills are required.
| Info |
|---|
|
...
Drag and drop Capture Feedback on the canvas.
On the properties panel, select the type of feedback in the Feedback Prompt
For Text type:
Add the feedback question specific to the skill in the Message field.
For Quick Replies:
Add the response questions.
You may add a follow-up action or skill in Payload to automatically trigger an action based on the feedback response selected by the end user.
Click on the tag icon on the option and add the required action or skill in the ‘Value.’
Add a Post Feedback Message. The message is triggered once the response to the survey is received.
Script
The Script block empowers skill developers to add custom scripts to their skills, enabling handling complex scenarios such as performing calculations, data manipulations, or implementing other custom logic. It provides flexibility and capability of your skills by allowing you to inject JavaScript directly into your conversation flow. Using scripts, you can perform complex processing and set or clear context attributes.
To use a script in the conversation, follow the below steps:
Drag and drop the Script on the canvas.
On the properties panel, add the script to the script block.
Click on the maximize button on the screen to view the script pop-up.
| Info |
|---|
Use the following syntax to use existing attributes in the script To refer to the value of a context variables:
To set a context attribute:
To clear attributes:
|
End
The 'End' block signifies the conclusion of a flow within the skill. When the skill execution reaches the End block, it signals the end of the skill's execution, triggering the Global Survey based on the specified configuration. While the End block is not mandatory, reaching a control without a subsequent node also marks the end of the skill execution and automatically accounts for the Survey, following the global configuration. To disable the survey for a specific path in the skill, you can incorporate the End Block accordingly.
...
Utilize the ‘Search Skill' to find relevant services at any point during a conversation using search phrases. ‘Search Skill’ looks for the skills that match the provided phrase. The identified results can be accessed using the 'matchedSkill’ Context variables. For more information, refer to Use Variables in Skills.
Follow the below steps to add Search Skills to your skill:
...
Drag and drop Suggest Skills on the canvas
On the properties panel, select Configuring the skills to be suggested.
To configure a custom list of skills:
Select Custom.
Choose the desired Display Format, Cards, or Quick Replies.
Type in the Message you wish the user to view.
Under Select Skill, type in the skill name you want to display as a suggested skill.
A minimum of one skill is to be selected.
To reuse an existing configuration:
Select Suggest Skills.
Select the Suggest Skill Configuration to be displayed to the user. For more information, refer to Customize System Skills behavior.
Click on the pop-out icon to view the Suggest Skill configuration in a new tab.
...
Search Knowledge construct eliminates configuring integration with the Luma Knowledge tenant. Luma VA and Luma Knowledge tenants are automatically associated with each other. 'Search Knowledge looks for articles directly in the associated Luma Knowledge system. The Knowledge Context variables can access the Knowledge articles in the results. For more information, refer to Use Variables in Skills.
Follow the below steps to add Search Knowledge to your skill:
Drag and drop the Search Knowledge construct from the palette onto the canvas.
On the properties panel, select the search methods:
Metadata - To search for Knowledge Articles using a phrase. You could also supply additional context for the search by providing metadata,
Article ID - To search for a specific Knowledge article using a Knowledge Article ID.
Now, based on the selected search method, provide the search filters:
For the ‘Metadata’ Method, you may provide the following:
Search Query: Provide the search phrase to search for Knowledge. The user’s initial request (available in context variable @{conv.skillPhrase}) is assigned to the field by default. You could set the field to a static value or another attribute with a text value.
Domain: Domain is the first level of categorization applicable to the Knowledge Artifact. In Luma Knowledge, all Knowledge Articles are categorized under domains and usually refer to the department or functional division the article is related. For example, HR, IT, Network, etc. These domains must be available in the Knowledge graph of your tenant
Topics: Information on products, services, projects, or sub-teams related to the article. For example, Microsoft, Asset Management, etc.
Select ‘Exclude Previously Presented Articles’ to exclude articles already presented in the conversation to avoid duplicates and streamline the result presentation. This ensures duplicate articles are not displayed when the system tries to search for Knowledge within the same conversation.
...
to exclude articles already presented in the conversation to avoid duplicates and streamline the result presentation. This ensures duplicate articles are not displayed when the system tries to search for Knowledge within the same conversation.
Include All Matched Articles - select this option to Include matching articles in the search results along with the precise answer.
| Info |
|---|
|
...
Add Display Precise Answer to the Conversation flow.
On the properties panel, select the checkbox 'Select View Articles…' to provide an option to view the article along with the precise answer.
Provide the label for the article view option.
| Info |
|---|
Source and Parent Knowledge Article information is displayed with each Knowledge article in the results. Source articles will now open in the web browser if the corresponding article is an external document (has external redirection). |
Create Artifact
Use the Create Artifact step to generate a knowledge article based on the information extracted during an earlier conversation step, enhancing the accessibility of relevant information. The Create Artifact conversation step allows the skill developer to create a Knowledge Artifact in the associated Luma Knowledge Tenant. You may add 'Create Artifact' into your skill when you wish to save some knowledge to be available for later access. The conversation step adds an article in Luma Knowledge so that is available to users in later searches.
...
'Search Catalog' searches Service Catalogs within your Luma Knowledge Tenant. Add 'Search Catalog' into your skill to seamlessly search for catalogs during its execution, using a search query or a specific catalog ID. This eliminates the necessity of configuring integrations with the Luma Knowledge tenant, enabling Luma to access catalogs within the associated Luma Knowledge system directly. The Catalogs in results can be accessed using the Catalog Context variables. For more information, refer to Use Variables in Skills.
Follow the below steps to add Search Catalog to your skill:
...
Drag and drop the 'AI Task' conversation step on the skill.
On the properties panel, select Existing to execute the OOTB tasks or Adhoc to add a new request for Generative AI.
For an Existing Task:
Select the Task Category.
Select Sub Category.
Select the required Task.
Note: Only the prompt which have Visible in AI Prompt checked will be available here.
Now click on Test Response to verify if the selected task generates the information required in the skill.
The AI Task Response Validation screen is used toverify the AI task and configurations. Here, you can validate the response of the AI task by providing values to the input parameters of the selected task.
Task Category, Subcategory, and Task section to select the task. The section is populated based on the selection in the Properties panel. You may test and change the task as required.
The Task Description section provides information on the task being executed.
Map Task Input Parameters lists all the input parameters required for the task. You can assign a value or use an attribute to refer to the value available in the attribute.
Now click on Test to verify the response of the AI Task.
When an attribute is used while mapping an input parameter, Assign Test values for each attribute. These values are used to test the task. Add the test values and click Test.
The system executes the task using the test values.
For an Adhoc task, select 'Adhoc Task' in the properties panel.
Add the task details to be executed in the Task Description field. This is a free text field where you can describe the task to be executed using Generative AI.
Now, Click on Test Response.
On the AI Task Response Validation screen, validate the response of the AI task.
Task Description indicates the action to be performed in the step.
Click on Test and validate the response in Test Response. Refine the Task description to ensure it generates the information as required.
...