Merge branch 'main' into pr/inferencepool-backendref

Author: isztldav

api/v1alpha1/ai_gateway_route.go

@@ -64,6 +64,16 @@ type AIGatewayRouteSpec struct {
 	// +optional
 	ParentRefs []gwapiv1.ParentReference `json:"parentRefs,omitempty"`
 
+	// Hostnames is a list of hostnames matched against the HTTP Host header to select an AIGatewayRoute
+	// used to process the request. This is equivalent to the Hostnames field in the Gateway API HTTPRouteSpec.
+	// When specified, the generated HTTPRoute will include these hostnames for hostname-based filtering.
+	// See https://gateway-api.sigs.k8s.io/reference/spec/#gateway.networking.k8s.io%2fv1.HTTPRouteSpec
+	// for the details of the Hostnames field in the Gateway API.
+	//
+	// +optional
+	// +kubebuilder:validation:MaxItems=16
+	Hostnames []gwapiv1.Hostname `json:"hostnames,omitempty"`
+
 	// Rules is the list of AIGatewayRouteRule that this AIGatewayRoute will match the traffic to.
 	// Each rule is a subset of the HTTPRoute in the Gateway API (https://gateway-api.sigs.k8s.io/api-types/httproute/).
 	//

api/v1alpha1/shared_types.go

@@ -22,6 +22,8 @@ type VersionedAPISchema struct {
 	//
 	// When the name is set to AzureOpenAI, this version maps to "API Version" in the
 	// Azure OpenAI API documentation (https://learn.microsoft.com/en-us/azure/ai-services/openai/reference#rest-api-versioning).
+	// This field is ignored for OpenAI, AWSBedrock, GCPVertexAI, and Anthropic.
+	// For OpenAI and Anthropic, use prefix to configure custom request paths.
 	//
 	// See https://aigateway.envoyproxy.io/docs/capabilities/llm-integrations/supported-providers for details.
 	// +optional
@@ -30,9 +32,11 @@ type VersionedAPISchema struct {
 	// Prefix is the prefix for the API.
 	//
 	// When the name is set to "OpenAI", "chat completions" API endpoint will be "${this_field}/chat/completions".
+	// When the name is set to "Anthropic", the "messages" API endpoint will be "${this_field}/messages".
 	// It can be with or without a leading slash ("/").
+	// This field is ignored for AWSAnthropic and GCPAnthropic.
 	//
-	// This is especially useful when routing to the backend that has an OpenAI compatible API but has a different
+	// This is especially useful when routing to a backend that has an OpenAI or Anthropic compatible API but has a different
 	// prefix. For example, Gemini OpenAI compatible API (https://ai.google.dev/gemini-api/docs/openai) uses
 	// "/v1beta/openai" prefix. Another example is that Cohere AI (https://docs.cohere.com/v2/docs/compatibility-api)
 	// uses "/compatibility/v1" prefix. On the other hand, DeepSeek (https://api-docs.deepseek.com/) doesn't

api/v1alpha1/zz_generated.deepcopy.go

@@ -64,6 +64,16 @@ type AIGatewayRouteSpec struct {
 	// +optional
 	ParentRefs []gwapiv1.ParentReference `json:"parentRefs,omitempty"`
 
+	// Hostnames is a list of hostnames matched against the HTTP Host header to select an AIGatewayRoute
+	// used to process the request. This is equivalent to the Hostnames field in the Gateway API HTTPRouteSpec.
+	// When specified, the generated HTTPRoute will include these hostnames for hostname-based filtering.
+	// See https://gateway-api.sigs.k8s.io/reference/spec/#gateway.networking.k8s.io%2fv1.HTTPRouteSpec
+	// for the details of the Hostnames field in the Gateway API.
+	//
+	// +optional
+	// +kubebuilder:validation:MaxItems=16
+	Hostnames []gwapiv1.Hostname `json:"hostnames,omitempty"`
+
 	// Rules is the list of AIGatewayRouteRule that this AIGatewayRoute will match the traffic to.
 	// Each rule is a subset of the HTTPRoute in the Gateway API (https://gateway-api.sigs.k8s.io/api-types/httproute/).
 	//

api/v1beta1/ai_gateway_route.go

@@ -26,6 +26,8 @@ type VersionedAPISchema struct {
 	//
 	// When the name is set to AzureOpenAI, this version maps to "API Version" in the
 	// Azure OpenAI API documentation (https://learn.microsoft.com/en-us/azure/ai-services/openai/reference#rest-api-versioning).
+	// This field is ignored for OpenAI, AWSBedrock, GCPVertexAI, and Anthropic.
+	// For OpenAI and Anthropic, use prefix to configure custom request paths.
 	//
 	// See https://aigateway.envoyproxy.io/docs/capabilities/llm-integrations/supported-providers for details.
 	// +optional
@@ -34,9 +36,11 @@ type VersionedAPISchema struct {
 	// Prefix is the prefix for the API.
 	//
 	// When the name is set to "OpenAI", "chat completions" API endpoint will be "${this_field}/chat/completions".
+	// When the name is set to "Anthropic", the "messages" API endpoint will be "${this_field}/messages".
 	// It can be with or without a leading slash ("/").
+	// This field is ignored for AWSAnthropic and GCPAnthropic.
 	//
-	// This is especially useful when routing to the backend that has an OpenAI compatible API but has a different
+	// This is especially useful when routing to a backend that has an OpenAI or Anthropic compatible API but has a different
 	// prefix. For example, Gemini OpenAI compatible API (https://ai.google.dev/gemini-api/docs/openai) uses
 	// "/v1beta/openai" prefix. Another example is that Cohere AI (https://docs.cohere.com/v2/docs/compatibility-api)
 	// uses "/compatibility/v1" prefix. On the other hand, DeepSeek (https://api-docs.deepseek.com/) doesn't

api/v1beta1/shared_types.go

@@ -355,6 +355,8 @@ func (c *AIGatewayRouteController) newHTTPRoute(ctx context.Context, dst *gwapiv
 	dst.Annotations[httpRouteAnnotationForAIGatewayGeneratedIndication] = "true"
 
 	dst.Spec.ParentRefs = aiGatewayRoute.Spec.ParentRefs
+
+	dst.Spec.Hostnames = aiGatewayRoute.Spec.Hostnames
 	return nil
 }
 

api/v1beta1/zz_generated.deepcopy.go

@@ -555,6 +555,90 @@ func Test_newHTTPRoute_LabelAndAnnotationPropagation(t *testing.T) {
 	require.Equal(t, "ann-value-2", httpRoute.Annotations["custom-annotation-2"])
 }
 
+func Test_newHTTPRoute_Hostnames_NotSet(t *testing.T) {
+	c := requireNewFakeClientWithIndexes(t)
+
+	backend := &aigv1b1.AIServiceBackend{
+		ObjectMeta: metav1.ObjectMeta{Name: "test-backend", Namespace: "test-ns"},
+		Spec: aigv1b1.AIServiceBackendSpec{
+			BackendRef: gwapiv1.BackendObjectReference{Name: "some-backend", Namespace: ptr.To(gwapiv1.Namespace("test-ns"))},
+		},
+	}
+	require.NoError(t, c.Create(context.Background(), backend))
+
+	aiGatewayRoute := &aigv1b1.AIGatewayRoute{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-route",
+			Namespace: "test-ns",
+		},
+		Spec: aigv1b1.AIGatewayRouteSpec{
+			Rules: []aigv1b1.AIGatewayRouteRule{
+				{
+					BackendRefs: []aigv1b1.AIGatewayRouteRuleBackendRef{
+						{Name: "test-backend", Weight: ptr.To[int32](100)},
+					},
+				},
+			},
+		},
+	}
+
+	controller := &AIGatewayRouteController{client: c}
+	httpRoute := &gwapiv1.HTTPRoute{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-route",
+			Namespace: "test-ns",
+		},
+	}
+
+	err := controller.newHTTPRoute(context.Background(), httpRoute, aiGatewayRoute)
+	require.NoError(t, err)
+
+	require.Empty(t, httpRoute.Spec.Hostnames)
+}
+
+func Test_newHTTPRoute_Hostnames_Set(t *testing.T) {
+	c := requireNewFakeClientWithIndexes(t)
+
+	backend := &aigv1b1.AIServiceBackend{
+		ObjectMeta: metav1.ObjectMeta{Name: "test-backend", Namespace: "test-ns"},
+		Spec: aigv1b1.AIServiceBackendSpec{
+			BackendRef: gwapiv1.BackendObjectReference{Name: "some-backend", Namespace: ptr.To(gwapiv1.Namespace("test-ns"))},
+		},
+	}
+	require.NoError(t, c.Create(context.Background(), backend))
+
+	aiGatewayRoute := &aigv1b1.AIGatewayRoute{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-route",
+			Namespace: "test-ns",
+		},
+		Spec: aigv1b1.AIGatewayRouteSpec{
+			Hostnames: []gwapiv1.Hostname{"api.example.com", "*.example.net", "sub.example.com"},
+			Rules: []aigv1b1.AIGatewayRouteRule{
+				{
+					BackendRefs: []aigv1b1.AIGatewayRouteRuleBackendRef{
+						{Name: "test-backend", Weight: ptr.To[int32](100)},
+					},
+				},
+			},
+		},
+	}
+
+	controller := &AIGatewayRouteController{client: c, logger: logr.Discard()}
+	httpRoute := &gwapiv1.HTTPRoute{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-route",
+			Namespace: "test-ns",
+		},
+	}
+
+	err := controller.newHTTPRoute(context.Background(), httpRoute, aiGatewayRoute)
+	require.NoError(t, err)
+
+	expected := []gwapiv1.Hostname{"api.example.com", "*.example.net", "sub.example.com"}
+	require.Equal(t, expected, httpRoute.Spec.Hostnames)
+}
+
 func TestAIGatewayRouteController_syncGateways_NamespaceDetermination(t *testing.T) {
 	fakeClient := requireNewFakeClientWithIndexes(t)
 	eventCh := internaltesting.NewControllerEventChan[*gwapiv1.Gateway]()

internal/controller/ai_gateway_route.go

@@ -164,7 +164,7 @@ func (c *GatewayController) Reconcile(ctx context.Context, req ctrl.Request) (ct
 func schemaToFilterAPI(schema aigv1b1.VersionedAPISchema) filterapi.VersionedAPISchema {
 	ret := filterapi.VersionedAPISchema{}
 	ret.Name = filterapi.APISchemaName(schema.Name)
-	if schema.Name == aigv1b1.APISchemaOpenAI {
+	if schema.Name == aigv1b1.APISchemaOpenAI || schema.Name == aigv1b1.APISchemaAnthropic {
 		ret.Prefix = cmp.Or(ptr.Deref(schema.Prefix, ""), "v1")
 	} else {
 		ret.Version = ptr.Deref(schema.Version, "")
@@ -367,6 +367,11 @@ func (c *GatewayController) reconcileFilterConfigSecret(
 		ec.GlobalLLMRequestCosts = append(ec.GlobalLLMRequestCosts, fc)
 	}
 
+	// Models contributed by routes with no Spec.Hostnames. We only promote these to
+	// ec.UnscopedModels (and merge them into ec.ModelsByHost) when at least one route
+	// IS hostname-scoped; otherwise the existing ec.Models list already covers them.
+	var unscopedModels []filterapi.Model
+
 	for i := range aiGatewayRoutes {
 		aiGatewayRoute := &aiGatewayRoutes[i]
 		if !aiGatewayRoute.GetDeletionTimestamp().IsZero() {
@@ -375,6 +380,7 @@ func (c *GatewayController) reconcileFilterConfigSecret(
 		}
 		hasEffectiveRoute = true
 		routeName := fmt.Sprintf("%s/%s", aiGatewayRoute.Namespace, aiGatewayRoute.Name)
+		hostnames := aiGatewayRoute.Spec.Hostnames
 		spec := aiGatewayRoute.Spec
 		routeBackendNamesSet := map[string]struct{}{}
 		routeBackendNames := []string{}
@@ -389,11 +395,25 @@ func (c *GatewayController) reconcileFilterConfigSecret(
 					if (h.Type != nil && *h.Type != gwapiv1.HeaderMatchExact) || string(h.Name) != internalapi.ModelNameHeaderKeyDefault {
 						continue
 					}
-					ec.Models = append(ec.Models, filterapi.Model{
+					model := filterapi.Model{
 						Name:      h.Value,
 						CreatedAt: ptr.Deref[metav1.Time](rule.ModelsCreatedAt, aiGatewayRoute.CreationTimestamp).UTC(),
 						OwnedBy:   ptr.Deref(rule.ModelsOwnedBy, defaultOwnedBy),
-					})
+					}
+					ec.Models = append(ec.Models, model)
+					if len(hostnames) > 0 {
+						if ec.ModelsByHost == nil {
+							ec.ModelsByHost = make(map[string][]filterapi.Model)
+						}
+						for _, hn := range hostnames {
+							ec.ModelsByHost[string(hn)] = append(ec.ModelsByHost[string(hn)], model)
+						}
+					} else {
+						// Routes without hostnames are "unscoped": they apply to every host.
+						// Tracked in unscopedModels for now; only promoted to ec.UnscopedModels
+						// after the loop if at least one scoped route is also present.
+						unscopedModels = append(unscopedModels, model)
+					}
 				}
 			}
 			for backendRefIndex := range rule.BackendRefs {
@@ -483,6 +503,18 @@ func (c *GatewayController) reconcileFilterConfigSecret(
 		}
 	}
 
+	// If at least one route is hostname-scoped, promote the unscoped models to ec.UnscopedModels
+	// so the runtime can fall back to them on unmatched hosts, and merge them into every per-host
+	// list so a host-matched request still sees the models from routes that didn't declare hostnames.
+	// When no route uses hostname scoping, ec.Models is the sole source of truth and we skip both
+	// steps to avoid serializing a redundant UnscopedModels duplicate of Models.
+	if len(ec.ModelsByHost) > 0 && len(unscopedModels) > 0 {
+		ec.UnscopedModels = unscopedModels
+		for hn := range ec.ModelsByHost {
+			ec.ModelsByHost[hn] = append(ec.ModelsByHost[hn], unscopedModels...)
+		}
+	}
+
 	// Configuration for MCP processor.
 	var effectiveMCPRoute bool
 	ec.MCPConfig, effectiveMCPRoute = mcpConfig(mcpRoutes)