Advanced
Observability
Advanced
Observability
End-to-end Observability
The Playground provides a user interface for monitoring all events within your cluster in real-time. This provides the ability to:
- View a chronological list of all events across your clusters and runs
- Filter events by cluster, run, or specific machines
- Review individual event details, including input, output, and execution time
Events can also be retreived via the API.
Observable Event Types
The following event types are available.
| Event Type | Description |
|---|---|
jobCreated | A new job has been created |
jobAcknowledged | A job has been acknowledged by a machine |
jobStatusRequest | A request for the current status of a job |
jobResulted | A job has produced a result |
jobStalled | A job did not produce a result within the expected time. This may be retried based on the function’s retry configuration. |
jobStalledTooManyTimes | A job stalled more times than the maximum allowed. |
jobRecovered | A previously stalled job has recovered via a retry. |
machineRegistered | A new machine has been registered with the control plane. |
machinePing | A heartbeat was received from a machine. |
machineStalled | A machine has become unresponsive |
modelInvocation | An LLM has been invoked |
agentTool | An agent has invoked a tool |
agentToolError | An error occurred when an agent tried to use a tool |
workflowFeedbackSubmitted | Feedback has been submitted for a run |
External Observability
Inferable integrates with Langfuse for tracing and evaluation. For more information, see the Langfuse integration docs.
Was this page helpful?