Monitor an automated long-running task

Learn how to monitor long-running task executions and review job details in Autonomous agents.

Prerequisites

You need an active Reltio agent with autonomous agents enabled.
You must have role-based access control (RBAC) permissions.

Use Task Management to monitor the jobs created by your long-running tasks. Each execution runs as a separate job with its own status, runtime details, execution metrics, outcome, and conversation.

To monitor and review an automated long-running task

Log in to autonomous agents.
Select your profile icon, and then select Task Management.

The Task Management page opens.

Select the task you want to review.

The task details panel opens. The following table describes the tabs available for the selected task and the information displayed in each tab.


Tab Name	Description	Image
Details	The Details tab contains the following sections: Task configuration The Task configuration section displays the following details: Task Name Task Status Agent Type Created at User Prompt The User Prompt section displays the prompt used to define the task scope. Schedule & Next Run The Schedule & Next Run section displays scheduling information based on the task type: For a one-time task, the section indicates that no additional runs are scheduled. For a recurring task, the section displays the next run date and time, followed by the frequency; for example, Daily at 07:00 AM.
Jobs	The Jobs tab displays the following details: History of all jobs Job ID Run date and time Current status For recurring tasks, each scheduled run appears as a separate job.

To monitor the details of a job:

Select the Jobs tab, and then select the job you want to review.

The Job Details page opens.

The following table describes the sections available on the Job Details page.


Section	Description
Job configuration details	Displays the configuration and execution details of the job.
Runtime Metrics	Displays the following metrics for the job: Total records Processed Failed Needs review Pending
User Prompt	Displays the prompt used to define the scope of the job.
Sources attached	Displays the files attached to the job.
Job Result	Displays the individual records processed by the job and their outcomes.

In the Job Result section, you can review the records processed by the job.

You can perform either of the following actions:

Select the Input Record link for a record to view the data that the job received. The Input record dialog box opens. The following table describes the tabs available in the dialog box.

Tab Description Image
Table Displays the input record as a table with the record fields and their corresponding values.

JSON Displays the input record in JSON format.

Tab	Description	Image
Table	Displays the input record as a table with the record fields and their corresponding values.
JSON	Displays the input record in JSON format.

Select a record to view the outcome for that record.

The Record Details page opens. The following table describes the elements on the Record Details page.


Element	Description
Input record	The record ID for the selected input record.
Status	The current status of the record.
Confidence	The confidence level assigned to the record based on the agent's assessment. The value can be High, Medium, Low, or NA.
Reason	Provides the rationale for the action taken by the agent on the record and explains the factors that contributed to the resulting outcome.
Details	The record attributes, field values, and metadata evaluated during processing.
Conversation	Displays the conversation associated with the selected record, including the messages exchanged while it was processed.

Select Back to return to the Job Details page.

To download a report file:

(Optional) Select a Status to narrow down your list of job records. For example, select Failed to list only the failed records.
To download a report file for the job, in the Job Result section, select Download.

The Download record dialog box opens.
In the Select Format field, select the format for the report file.

The following table describes the formats available for the report file.

Format Description
CSV A comma-separated values file with one row for each record and one column for each field.
JSONL A JSON Lines file with one JSON object for each record.
(Optional) To include the conversation for each record in the report file, turn on Include Conversation. Including conversation data may increase the file size and preparation time.
Select Download.
Reltio prepares the report file and downloads it to your device. To access all the report files generated for the job, select View Downloads.

Format	Description
CSV	A comma-separated values file with one row for each record and one column for each field.
JSONL	A JSON Lines file with one JSON object for each record.

To review the conversation associated with a job:

To open the conversation associated with the selected job, on the Job Details page, select Go to Conversation.
Review the conversation for details about the job execution.
The conversation shows additional details about the plan, execution flow, tool activity, and final outcomes for that job.
Inspect the job results and any items that require follow-up.
Continue follow-up in the same conversation if needed.
Select View Details to return to the Job Details page.

Result

You can monitor scheduled, in-progress, and completed jobs from Task Management.

Email notifications

Reltio sends email notifications to the user who created the job at key stages of an automated long-running task. Emails are sent at the following stages of a long-running task:

Started: The task is running. The email includes a link to track progress in autonomous agents.
Stopped: The task stopped before all records were processed. The email includes a summary of the partial results.
Completed: The task completed successfully. The email includes a summary of the results.
Failed: A problem prevented the task from completing. The email includes a summary of the results.

In the email, select View Job to open the job in autonomous agents, where you can review execution details, results, and the conversation associated with the job.

Unify and manage your data

Monitor an automated long-running task

Footer