Uploaded image for project: 'Red Hat OpenShift AI Engineering'
  1. Red Hat OpenShift AI Engineering
  2. RHOAIENG-1102

Incomplete address for internal model serving endpoint

XMLWordPrintable

    • False
    • Hide

      None

      Show
      None
    • False
    • Moderate

      Is there an existing issue for this?

      •  I have searched the existing issues

      Current Behavior

      When the model-server is configured to be presented through an external Route, you get all the info to connect to it:

      When you create a model-server that is only available internally to the cluster, the popup giving the address of the endpoint is missing the full information. The /v2/models/model_name/infer is not there, making the URL unusable unless you know the format KServer is using.

      Also, url is redundant with grpcUrl, it's the same. We should only keep Rest URL and gRPC URL.

      Expected Behavior

      Internal address information should give the same full information on how to reach the model API internally.
      Optionally, offering the same copy/paste functionality as for the external config would be good.
      Optionally, a gRPC Route when creating an external access would be useful, as it's a preferred endpoint type for model APIs.

      Steps To Reproduce

      No response

      Workaround (if any)

      Read the KServe doc and figure out what the API endpoint path is...

            Unassigned Unassigned
            gmoutier@redhat.com Guillaume Moutier
            RHOAI Dashboard
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

              Created:
              Updated: