What Is AI Execution Observability?

AI Execution Observability is a read-only troubleshooting feature that helps authorized users investigate AI requests and tool executions in HallianAI.

It provides organization-wide visibility into executions that:

• Succeeded
• Failed
• Timed out
• Were cancelled
• Did not complete
• Succeeded with warnings from child executions

Observability can help administrators, super users, and support teams determine what happened when an AI request, model interaction, or tool call does not behave as expected.

<aside> <img src="notion://custom_emoji/f2cca387-5803-81f0-a7b9-00035b6fc5fc/20bca387-5803-803b-80b8-007aadf2e11a" alt="notion://custom_emoji/f2cca387-5803-81f0-a7b9-00035b6fc5fc/20bca387-5803-803b-80b8-007aadf2e11a" width="40px" />

Important: Observability is a diagnostic and investigation feature. It cannot retry, replay, edit, or delete executions.

</aside>

Open Observability

1. Open HallianAI.
2. Open Settings.
3. Select Observability.

The Observability section is available only to users with the ObservabilityViewer permission. It provides read-only visibility across the organization.

If you do not see Observability in Settings, contact your HallianAI administrator.