Adds Initial InferencePool Support to GatewayParameters

Signed-off-by: Daneyon Hansen <[email protected]>
kgateway-dev · Feb 11, 2025 · 110b220 · 110b220
1 parent ee7fa7b
commit 110b220
Show file tree

Hide file tree

Showing 10 changed files with 391 additions and 11 deletions.
diff --git a/api/applyconfiguration/api/v1alpha1/endpointpickerextension.go b/api/applyconfiguration/api/v1alpha1/endpointpickerextension.go
diff --git a/api/applyconfiguration/api/v1alpha1/inferenceextension.go b/api/applyconfiguration/api/v1alpha1/inferenceextension.go
diff --git a/api/applyconfiguration/api/v1alpha1/inferencepoolobjref.go b/api/applyconfiguration/api/v1alpha1/inferencepoolobjref.go
diff --git a/api/applyconfiguration/api/v1alpha1/kubernetesproxyconfig.go b/api/applyconfiguration/api/v1alpha1/kubernetesproxyconfig.go
diff --git a/api/applyconfiguration/internal/internal.go b/api/applyconfiguration/internal/internal.go
diff --git a/api/applyconfiguration/utils.go b/api/applyconfiguration/utils.go
diff --git a/api/v1alpha1/gateway_parameters_types.go b/api/v1alpha1/gateway_parameters_types.go
@@ -106,6 +106,14 @@ type KubernetesProxyConfig struct {
 	// +kubebuilder:validation:Optional
 	AiExtension *AiExtension `json:"aiExtension,omitempty"`
 
+	// InferenceExtension defines the desired state of the Gateway API inference extension.
+	// For additional details, see: https://gateway-api-inference-extension.sigs.k8s.io/.
+	//
+	// InferenceExtension can only be specified when GatewayParameters is referenced by a Gateway.
+	//
+	// +kubebuilder:validation:Optional
+	InferenceExtension *InferenceExtension `json:"inferenceExtension,omitempty"`
+
 	// Used to unset the `runAsUser` values in security contexts.
 	FloatingUserId *bool `json:"floatingUserId,omitempty"`
 }
@@ -612,6 +620,52 @@ type AiExtension struct {
 	Stats *AiExtensionStats `json:"stats,omitempty"`
 }
 
+// InferenceExtension defines the desired state of the Gateway API inference extension.
+// For additional details, see: https://gateway-api-inference-extension.sigs.k8s.io/.
+type InferenceExtension struct {
+	// EndpointPickers defines a list of EndpointPicker extensions.
+	//
+	// +kubebuilder:validation:MaxItems=8
+	EndpointPickers []EndpointPickerExtension `json:"endpointPickers"`
+}
+
+// EndpointPickerExtension defines the desired state of an EndpointPicker extension.
+type EndpointPickerExtension struct {
+	// PoolRef is a reference to an InferencePool associated to the Endpoint Picker extension.
+	// The referenced InferencePool must exist in the same namespace as the Gateway that references
+	// this GatewayParameters.
+	//
+	// +kubebuilder:validation:Required
+	PoolRef InferencePoolObjRef `json:"poolRef"`
+}
+
+// InferencePoolObjRef is a reference to an InferencePool.
+type InferencePoolObjRef struct {
+	// Group is the group of the referent.
+	//
+	// +optional
+	// +kubebuilder:default="inference.networking.x-k8s.io"
+	// +kubebuilder:validation:MaxLength=253
+	// +kubebuilder:validation:Pattern=`^$|^[a-z0-9]([-a-z0-9]*[a-z0-9])?(\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$`
+	Group string `json:"group,omitempty"`
+
+	// Kind is kind of the referent.
+	//
+	// +optional
+	// +kubebuilder:default="InferencePool"
+	// +kubebuilder:validation:MinLength=1
+	// +kubebuilder:validation:MaxLength=63
+	// +kubebuilder:validation:Pattern=`^[a-zA-Z]([-a-zA-Z0-9]*[a-zA-Z0-9])?$`
+	Kind string `json:"kind,omitempty"`
+
+	// Name is the name of the referent.
+	//
+	// +kubebuilder:validation:MinLength=1
+	// +kubebuilder:validation:MaxLength=253
+	// +kubebuilder:validation:Required
+	Name string `json:"name"`
+}
+
 func (in *AiExtension) GetEnabled() *bool {
 	if in == nil {
 		return nil

diff --git a/api/v1alpha1/zz_generated.deepcopy.go b/api/v1alpha1/zz_generated.deepcopy.go
diff --git a/install/helm/kgateway/crds/gateway.kgateway.dev_gatewayparameters.yaml b/install/helm/kgateway/crds/gateway.kgateway.dev_gatewayparameters.yaml
@@ -406,6 +406,39 @@ spec:
                     type: object
                   floatingUserId:
                     type: boolean
+                  inferenceExtension:
+                    properties:
+                      endpointPickers:
+                        items:
+                          properties:
+                            poolRef:
+                              properties:
+                                group:
+                                  default: inference.networking.x-k8s.io
+                                  maxLength: 253
+                                  pattern: ^$|^[a-z0-9]([-a-z0-9]*[a-z0-9])?(\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$
+                                  type: string
+                                kind:
+                                  default: InferencePool
+                                  maxLength: 63
+                                  minLength: 1
+                                  pattern: ^[a-zA-Z]([-a-zA-Z0-9]*[a-zA-Z0-9])?$
+                                  type: string
+                                name:
+                                  maxLength: 253
+                                  minLength: 1
+                                  type: string
+                              required:
+                              - name
+                              type: object
+                          required:
+                          - poolRef
+                          type: object
+                        maxItems: 8
+                        type: array
+                    required:
+                    - endpointPickers
+                    type: object
                   istio:
                     properties:
                       customSidecars: