A proper form for creating inference services

[Mark Winter]

Currently when you click "New Model Server" it just lets you paste YAML.

It would be nicer if it presented a form to create an Inference Service without YAML knowledge (and also allow pasting YAML if desired)

I can work on this if the idea is approved.

Form would be something like this, most fields would be optional:

• Name Field [REQUIRED]

• Model Format [REQUIRED]

  • Make backend call to get available formats from ClusterServingRuntime and ServingRuntime

• Serving Runtime (auto set by model format but editable)

• StorageURI Field [REQUIRED]

• Resources

  • Requests
    • CPU
    • Memory
  • Limits
    • CPU
    • Memory
    • GPU

• Custom Ports

  • Port field

• Replicas

  • MinReplicas Field
  • MaxReplicas Field

• Enable Autoscaling Checkbox

  • HPA, KPA Selector
    • KPA CPU Utilisation Field
    • HPA Metric+Target Field
  • ContainerConcurrency

• Enable Transformer Checkbox

  • Container Image Field
  • Container Command Field
  • Resources
    • Requests
      • CPU
      • Memory
    • Limits
      • CPU
      • Memory
  • Custom ENV
  • Custom Ports
    • Add port field

• ImagePullSecrets Ref Field

[ConverJens]

We would also love to have this!