Add RetryDelay to ErrorHandling in WorkflowStep CRD

[JAORMX]

Summary

The VirtualMCPServer CRD's ErrorHandling struct is missing the RetryDelay field that is supported by the implementation.

Current State

Implementation (pkg/vmcp/config/config.go:399-409):

type StepErrorHandling struct {
    // Action: "abort", "continue", "retry"
    Action string `json:"action"`

    // RetryCount is the number of retry attempts (for retry action).
    RetryCount int `json:"retry_count,omitempty"`

    // RetryDelay is the initial delay between retries.
    RetryDelay Duration `json:"retry_delay,omitempty"`
}

CRD (cmd/thv-operator/api/v1alpha1/virtualmcpserver_types.go:274-285):

type ErrorHandling struct {
    // Action defines the action to take on error
    // +kubebuilder:validation:Enum=abort;continue;retry
    // +kubebuilder:default=abort
    Action string `json:"action,omitempty"`

    // MaxRetries is the maximum number of retries
    // Only used when Action is "retry"
    MaxRetries int `json:"maxRetries,omitempty"`
    
    // NO RetryDelay field
}

Impact

Users cannot configure the delay between retry attempts for workflow steps. This is documented in the proposal (docs/proposals/THV-2106-virtual-mcp-server.md:900-903):

on_error:
  action: abort  # abort | continue | retry
  retry_count: 3
  retry_delay: 5s

Proposed Solution

Add to ErrorHandling:

// RetryDelay is the delay between retry attempts
// Only used when Action is "retry"
// +kubebuilder:default="1s"
// +kubebuilder:validation:Pattern=`^([0-9]+(\.[0-9]+)?(ms|s|m))+$`
// +optional
RetryDelay string `json:"retryDelay,omitempty"`

Update the converter to parse and pass through the retry delay.

Related Files

• pkg/vmcp/config/config.go (implementation)
• cmd/thv-operator/api/v1alpha1/virtualmcpserver_types.go (CRD)
• cmd/thv-operator/api/v1alpha1/virtualmcpcompositetooldefinition_types.go (CRD)
• cmd/thv-operator/pkg/vmcpconfig/converter.go (converter)