Skip to main content

Monitoring

1. Overview

The Monitoring screen gives you health control over every application running on the platform — Flows, Apps, and Add-ons — in a single place: executions, errors, resource consumption, AI token consumption, detailed tracing, and active alerts. Instead of hunting for information across several different systems, you can quickly see what's running, what's failing, and why.

2. Filters

In the side panel, filters narrow down the monitoring view:

FilterDescription
SquadSquad that owns the monitored resource.
EnvironmentMonitored environment (e.g. test).
TypeResource type: Flows, Apps, or Addons.
FlowSpecific Flow to monitor (when Type = Flows), or the All Flows (aggregate) aggregated view.
PeriodTime window of the data shown (e.g. last 15 min).

3. Aggregated Indicators

At the top of the screen, indicators summarize the overall state of what's selected in the filter:

IndicatorDescription
Total ExecutionsTotal executions in the period.
Error ExecutionsExecutions that ended in error.
OOM ContainersContainers that suffered an Out of Memory event in the period.
Restart DeploymentsNumber of deployment restarts in the period.
Running InstancesInstances currently running.

4. Overview Tab

Shows a summary of the selected resource (e.g. a specific Flow), with the same types of indicator at a finer granularity:

IndicatorDescription
ExecutionsExecutions in the period.
ErrorsExecutions with errors.
OOMsOut of Memory occurrences.
RestartsResource restarts.
Running InstancesInstances currently running.
CPU NowCurrent CPU usage (%).
Memory NowCurrent memory usage (GB).
DeploymentsNumber of deployments performed.

Right below, two charts complement the view:

  • Top Executions — ranking of the services/flows with the most executions in the period (configurable, e.g. Top 5).
  • Top Errored Executions — ranking of the services with the most failed executions in the period ("Services with the most failed executions in the selected period"), with the chart itself indicating when there are no errors in the period ("No errored executions in this period").

5. Metrics Tab

TODO: detail the charts and specific metrics on this tab (not yet captured on screen).

6. Logs Tab

Queries the execution logs of the monitored Flows/Apps, letting you investigate the detailed behavior of a specific execution without leaving the Monitoring screen.

TODO: detail the filters and log display format for this tab (not yet captured on screen).

7. Trace Tab

Shows the complete execution trace — the step-by-step path an execution took (for example, between a Flow's components, or between a Backend App's calls), letting you pinpoint exactly which step a problem or slowdown occurred at.

TODO: detail the trace visualization (not yet captured on screen).

8. Alerts Tab

Shows the error alerts currently occurring for that monitored instance/resource — a single place to track active problems without having to cross-reference logs manually.

TODO: detail alert types and rule configuration (not yet captured on screen).

9. Agent AI Tab

Shows the AI token consumption generated by the AI Agent connector within the monitored Flows — useful for correlating AI usage with cost (see also Usage) and with the behavior of executions that use AI.

IndicatorDescription
Input TokensTotal input (prompt) tokens consumed in the period.
Output TokensTotal output (response) tokens generated in the period.
Total TokensSum of input and output tokens in the period.
Avg Input TokensAverage input tokens per execution.
Avg Output TokensAverage output tokens per execution.
Avg Total TokensAverage total tokens per execution.

Time-series charts (Input Tokens/s, Output Tokens/s, Total Tokens/s) show the token consumption rate over the selected period.

Below that, percentile charts (Input Percentiles, Output Percentiles, Total Percentiles) show the per-execution token consumption distribution at P50, P75, and P99 — useful for identifying whether typical consumption is stable or whether some executions consume much more tokens than most.

The Models Breakdown (breakdown by AI model used) and Performance Metrics sections are still under development (Coming Soon).

10. FAQ

What causes an "OOM Container"? It happens when a container exceeds its allocated memory limit and is terminated by Kubernetes due to an Out of Memory condition.

Can alerts be configured to notify me automatically (e.g. email, Slack)? TODO: confirm whether the Alerts tab supports external notifications or is just a visualization panel.

Is the token consumption on the Agent AI tab the same value charged in Usage? TODO: confirm whether the numbers match 1:1 with the "Tokens cost" on the Usage screen.