[sttts]

Summary

v1beta2 model-only DGDRs currently fail under the default Helm values because dgdrAISimulateDefaultImage is empty, admission persists no spec.aisimulate.image, and validation rejects the request.

This change reuses the existing DGDR planner-image resolver for v1beta2. An empty operator option now derives a version-matched planner image from --operator-version; an explicitly configured dgdrAISimulateDefaultImage still wins.

With the release chart's appVersion: 1.6.0, this request:

apiVersion: nvidia.com/v1beta2
kind: DynamoGraphDeploymentRequest
metadata:
  name: qwen
spec:
  model:
    name: Qwen/Qwen3-0.6B

is admitted with:

spec:
  aisimulate:
    image: nvcr.io/nvidia/ai-dynamo/dynamo-planner:1.6.0

The chart already passes .Chart.AppVersion as --operator-version. Pre-release, nightly, or custom images can continue to set controllerManager.manager.dgdrAISimulateDefaultImage explicitly.

The production admission envtest no longer injects a fake AISimulate image. Its model-only case now proves that the operator version supplies and persists the default.

Stacked on ai-dynamo#16661.

Validation

• go test ./internal/webhook/defaulting
• KUBEBUILDER_ASSETS=<absolute envtest path> go test ./internal/webhook/validation -run 'TestDynamoGraphDeploymentRequestValidator_Validate/v1beta2_model-only_request_gets_schema_and_admission_defaults'
• go test ./cmd ./api/v1beta2
• make generate-api-docs
• pre-commit hooks run by the signed commit

[github-actions]

👋 Hi sttts! Thank you for contributing to devivasudevan/dynamo.

Just a reminder: The NVIDIA Test Github Validation CI runs an essential subset of the testing framework to quickly catch errors.Your PR reviewers may elect to test the changes comprehensively before approving your changes.

🚀