Spark NQL capabilities

The data model will change between the technical preview and general availability. Therefore, you might have to adjust certain NQL queries you created during the TP later, when moving to the released feature.

Spark uses the following NQL data model tables to query Spark-user interaction data:

Table	Description
agent.conversations	Store information about conversations between employees and the Spark agent.

agent.conversations table fields

The query below uses conversation_id field object from the conversations table of the agent NQL namespace.

agent.conversations 
| list conversation_id

agent.conversations table field	Description
conversation_id	Unique identifier of the conversation.
time	The date and time at which the last message in the conversation was sent or received.
first_message_time	The time at which the user initiated the conversation.
conversation_duration	The elapsed time between the first and last messages in the conversation.
number_of_turns	The number of turns in the conversation between the user and the agent.
outcome	The outcome of the conversation. Only set for completed conversations.
state	The current state of the conversation.¨

Spark KPI widget NQL examples

The examples below describe how the system updates the corresponding investigation queries with changes made from the Dashboards page.

Viewing an investigation of a KPI dashboard showing total conversations

agent.conversations during past 30d
| summarize active_users = user.count(), total_conversations = conversation_id.count(), active_conversations = conversation_id.countif((state == in_progress or state == null))

Drilling down to an investigation from a dashboard

Drill down… on a KPI dashboard widget showing the average resolution time of devices over the past 30 days

devices during past 30d
| include agent.conversations during past 30d
| where agent.conversations.outcome in [resolved, escalated]
| compute number_of_conversations_ = conversations.number_of_conversations.sum(), average_resolution_time = agent.conversations.conversation_duration.avg()
| where number_of_conversations_ != 0
| where average_resolution_time != 0
| list device.name, device.entity, device.hardware.model, device.hardware.type, device.operating_system.name, number_of_conversations_, average_resolution_time
| sort number_of_conversations_ desc

RELATED TOPICS

• Getting started with Spark