Skip to main content

Documentation Index

Fetch the complete documentation index at: https://koreai.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

Notifications display policy violations and system anomalies detected by the Agent Management Platform’s governance system. When an active policy evaluates an agent’s behavior and detects a violation, the platform creates a notification that includes the severity level, the triggering criteria, and the session that caused it. Use the Notifications page to monitor governance activity across your projects, prioritize high-severity issues, and investigate violations as they occur.
You are notified of new notifications via email or the in-app bell icon.

Why Notifications Matter

BenefitDescription
TriageHelps teams prioritize the most severe or time-sensitive issues.
Faster ResponseReduces the time to detect and resolve model or agent misbehavior.
AccountabilityRecords who viewed, acknowledged, or resolved each incident.

Key Capabilities

CapabilityDescription
Severity FilteringIsolate Critical or High issues to focus on the most urgent violations.
Visual IndicatorsColor-coded severity badges and read/unread markers for quick scanning.
Detailed Drill-DownOne-click access to the full execution trace for root-cause analysis.
Multi-Channel NotificationsSupports in-app and email notifications.

Core Terminology

TermDefinition
IncidentA specific occurrence of a policy violation.
Metric PillA badge that shows the value or score that caused the violation—for example, Toxicity: 0.95.
StateThe notification’s lifecycle status: Unread → Read.

Notification Statistics

The Notifications page includes summary metrics that help you quickly assess governance posture.
MetricDescription
Total NotificationsTotal notifications generated in the last 24 hours (Read + Unread).
CriticalNumber of critical-severity violations in the last 24 hours.
UnreadNotifications awaiting review.
Top ViolationThe most frequently triggered policy in the last 24 hours.
Policy list
Review Critical and High-severity items first to ensure urgent issues are resolved promptly.

Managing Notifications

Notifications require explicit acknowledgment. A notification stays in your inbox until you mark it as Read, ensuring each violation is reviewed before it is cleared.

Acknowledge Notifications

After reviewing a notification, mark it as Read to keep your inbox current. You can update items individually or select multiple notifications to acknowledge them at once.

Prioritize and Filter

Use sorting and filters to focus on the notifications that matter most.
FilterDescription
SeverityShow only Critical, High, Medium, or Low violations.
StatusFilter by Read, Unread, or Resolved.
PolicyNarrow results to a specific policy.
TimeframeLimit results to a defined date range.
ProjectAvailable in workspace-level views only.
In a workspace-level view, the Project filter is available to display violations from specific projects. In a project-level view, the filter is applied automatically for the current project and the project selector does not appear.

Viewing Notification Details

Use this procedure to open and review the details for any notification. Prerequisite: You must have access to the project where the notification was generated.
  1. Open the Notifications page.
  2. Identify a notification to review. Critical alerts are visually highlighted for quick scanning.
  3. Select the notification or choose View Details.
  4. Review the details in the notification panel, including the triggered policy, the session involved, and the condition that caused the violation.
  5. Take action:
    • Mark the alert as Read, or
    • Select the Session ID in the Context tab to open the session in Observability → Runs to view the full execution trace for deeper investigation.

Notification Details Panel

When you open a notification, a detail panel appears on the right with three tabs.
TabPurposeWhat to Review
SummaryConfirm the violation at a glance.Which rule was violated, which policy generated the alert, and when the event occurred.
ContextVerify the exact conditions that caused the violation.The session and event that triggered the violation, the associated policy, and a direct link to investigate the session in Runs.
Status & AuditTrack follow-up actions and support compliance review.When the notification was generated, when it was marked as Read, and any additional status updates.
Notifications detail

Investigating Policy Violations

When deeper analysis is needed, use the Context tab to open the related session in Observability → Runs, where you can review the complete execution trace—including LLM calls, tool usage, and timing data.
  1. In the Context tab of the selected notification, select the Session ID to view the full execution trace.
  2. The system opens the corresponding session in Observability → Runs.
  3. Review the spans and events that contributed to the violation.
  4. Determine whether the behavior was expected, caused by an edge case, or the result of incorrect configuration.
  5. Make any required updates, such as adjusting thresholds or modifying agent logic.
Notifications context