Model Usage TrackingΒΆ
Acumos AI is a platform and open source framework that makes it easy to build, share, and deploy AI apps (models).
With Boreas release, Acumos platform comes with AI Model usage tracking feature (for models that are deployed in a Kubernetes cluster).
Model usage tracking feature leverages on Elastic Stack which is part of the Acumos platform for collecting, processing, searching/indexing usage data and visualizing reports.
As the diagram illustrates, model access/calls are recorded by nginx sidecar to a persistent volume (log file), which the Filebeat daemon instance reads and pushes to Acumos platform Elastic Stack.
In this article, we are referring to Soup-to-Nuts Example that sets up face-privacy-filter - single and composite model solutions. You can onboard those models (or any other models) as illustrated in the Soup-to-Nuts example doc.
Following image illustrates AI model method calls and respective usage count.
