Docs

Observability Platform

AI Application Insights

Insight tasks

An insight task can be used for batch analysis on historical sessions within a specified range based on the specified insight rule. For each session, the task generates a tag and an evaluation rationale. By creating an insight task, you can filter a large number of sessions to identify the data that requires attention. You can also view the overall distribution of results and locate issues based on the original sessions. The analysis results provide a basis for optimizing system prompts, tool calls, or product workflows. This topic describes how to manage insight tasks.

Limits

You have view permissions on the target workspace. For details, see Workspace-level IAM permission control.

The target workspace contains historical session data of AI applications. For details about how to integrate and query session data, see LLM session analysis.

Create an insight task

In the left navigation pane, select AI Application Insights > Insight task.

In the upper-right corner of the page, select the target workspace. An insight task can only be used to analyze the session data in the selected workspace.

Click "Create insight task" and then configure the parameters.

Parameters	Description
Task name	Enter a custom name for the insight task. Duplicate names are not allowed within the same workspace.
Associated insight rules	Select the insight rule used by the current task. For details, see Insight rules.
Analysis time range	Select the time range in which the sessions were generated. The system analyzes only the sessions within the specified time range. You can select "Last 1 day", "Last 7 days", "Last 14 days", "Last 30 days", or a custom time range.
Agent Service	Select the agent services that you want the current task to analyze. If you do not specify a service, the task analyzes all agent services within the workspace by default.
Data filtering	Click Add Filter to further narrow down the data to be analyzed within the analysis time range and the Agent service scope. You can configure fields such as session ID, user ID, session duration, and trace count. tip If you do not add any filter criteria, the system analyzes all eligible sessions within the analysis time range. If you add multiple filter criteria, the criteria are combined with a logical AND. That is, only the sessions that meet all the specified criteria are included in the analysis.

View an insight task

In the left navigation pane, select AI Application Insights > Insight task.

In the upper-right corner of the page, select the target workspace. The system displays all insight tasks in the workspace.

In the task list, click the name of the target task, or click Details in the Actions column to view the basic information about the task.

The task result overview page displays the number and percentage of each tag in the current task. Click View Rule Configuration to view the insight rule associated with the task. For details, see Insight rules.

Task result details
Associated insight rules
Execution details
Click the Task result details tab to view the analysis results of the task. This tab consists of a filter section and a result details section.
In the filter section, you can filter sessions based on the following criteria:
Status: Filter sessions by the analysis status.
Rule name: Filter sessions by the insight rule associated with the current task.
Filter by Session ID: Enter the complete session ID to locate the analysis result of a specific session.
The result details section displays the analysis records for each session:
Creation time: The time when the insight result was generated. This information helps you confirm whether the result comes from the target analysis period. It also helps you distinguish between results that are generated for the same session by different tasks or at different times.
Session ID: The ID of the session. Click the session ID to go to the LLM session analysis page, where you can view the complete conversation, trace, and related context. For details, see LLM session analysis.
Tag attribute: The classification result output by the rule. Use this result to determine the extent to which user needs are met in a complete session.
Evaluation rationale: The explanation that the large language model provides for the current tag. This explanation describes which parts of the conversation affected the satisfaction assessment.
Click the Associated insight rules tab to view information about the insight rule used by the current task, including the rule name, rule ID, version, output type, and description. On this tab, you can check the evaluation criteria on which the current analysis is based.

Click the Execution details tab to view the execution records of the task, including the execution round, the number of records, the start time, the end time, and the status.

Delete an insight task

warning

Deleted tasks cannot be recovered. Proceed with caution.

In the left navigation pane, select AI Application Insights > Insight task.

In the upper-right corner of the page, select the target workspace. The system displays all insight tasks in the workspace.

Click Delete in the Actions column of the target insight task to delete the task.

Last updated: 2026.09.21 10:36:58