Once the pipeline is deployed and running, AI Asset Manager Agent gathers metric data about the pipeline and its environment and forwards them to AI Asset Manager. Such metric data can be useful to identify potential current or future problems with pipeline execution. For a detailed list of available metrics, refer to the Collected metrics page.
Besides the standard metrics listed, AI Asset Manager Agent also automatically collects custom metrics defined in pipelines generated by AI SDK >v2.4.0. These metric data are automatically published to a pre-defined ZMQ topic.
Note: These metric data can only be collected with AI Asset Manager Agent, but not with AI Model Monitor Agent, since the latter application only connects to Databus topics, not to ZMQ, and the connector of custom metric pipeline outputs generated by AI SDK >v2.4.0 cannot be reconfigured.
In order to browse collected metrics, click on the "Go To Metrics" button on the Target overview page.
Clicking the button opens a Prometheus user interface. Here you can list metrics, write and evaluate query expressions using the collected metric data.
| ① | Query expression field |
| ② | Browse metrics |
| ③ | Table view |
| ④ | Graph view |
The current value of the metric or expression is displayed in Table view.
Previously collected datapoints are plotted under the Graph view tab.
In case you wish to be notified if certain metrics reach a given threshold or if a defined expression evaluates to true, you can configure alerts under the Alert configuration menu.