--- apiVersion: v1 items: - apiVersion: v1 data: _example: |- ################################ # # # EXAMPLE CONFIGURATION # # # ################################ # This block is not actually functional configuration, # but serves to illustrate the available configuration # options and document them in a way that is accessible # to users that `kubectl edit` this config map. # # These sample configuration options may be copied out of # this example block and unindented to be in the data block # to actually change the configuration. # ====================================== EXPLAINERS CONFIGURATION ====================================== # Example explainers: |- { "art": { "image" : "kserve/art-explainer", "defaultImageVersion": "latest" } } # Art Explainer runtime configuration explainers: |- { # Art explainer runtime configuration "art": { # image contains the default Art explainer serving runtime image uri. "image" : "kserve/art-explainer", # defautltImageVersion contains the Art explainer serving runtime default image version. "defaultImageVersion": "latest" } } # ====================================== ISVC CONFIGURATION ====================================== # Example - setting custom annotation inferenceService: |- { "serviceAnnotationDisallowedList": [ "my.custom.annotation/1" ], "serviceLabelDisallowedList": [ "my.custom.label.1" ] } # Example - setting custom annotation inferenceService: |- { # ServiceAnnotationDisallowedList is a list of annotations that are not allowed to be propagated to Knative # revisions, which prevents the reconciliation loop to be triggered if the annotations is # configured here are used. # Default values are: # "autoscaling.knative.dev/min-scale", # "autoscaling.knative.dev/max-scale", # "internal.serving.kserve.io/storage-initializer-sourceuri", # "kubectl.kubernetes.io/last-applied-configuration", # "modelFormat" # Any new value will be appended to the list. "serviceAnnotationDisallowedList": [ "my.custom.annotation/1" ], # ServiceLabelDisallowedList is a list of labels that are not allowed to be propagated to Knative revisions # which prevents the reconciliation loop to be triggered if the labels is configured here are used. "serviceLabelDisallowedList": [ "my.custom.label.1" ] } # Example - setting custom resource inferenceService: |- { "resource": { "cpuLimit": "1", "memoryLimit": "2Gi", "cpuRequest": "1", "memoryRequest": "2Gi" } } # Example - setting custom resource inferenceService: |- { # resource contains the default resource configuration for the inference service. # you can override this configuration by specifying the resources in the inference service yaml. # If you want to unbound the resource (limits and requests), you can set the value to null or "" # or just remove the specific field from the config. "resource": { # cpuLimit is the limits.cpu to set for the inference service. "cpuLimit": "1", # memoryLimit is the limits.memory to set for the inference service. "memoryLimit": "2Gi", # cpuRequest is the requests.cpu to set for the inference service. "cpuRequest": "1", # memoryRequest is the requests.memory to set for the inference service. "memoryRequest": "2Gi" } } # ====================================== MultiNode CONFIGURATION ====================================== # Example multiNode: |- { "customGPUResourceTypeList": [ "custom.com/gpu" ] } # Example of multinode configuration multiNode: |- { # CustomGPUResourceTypeList is a list of custom GPU resource types intended to identify the GPU type of a resource, # not to restrict the user from using a specific GPU type. # The MultiNode runtime pod will dynamically add GPU resources based on the registered GPU types. "customGPUResourceTypeList": [ "custom.com/gpu" ] } # ====================================== OTelCollector CONFIGURATION ====================================== # Example opentelemetryCollector: |- { # scrapeInterval is the interval at which the OpenTelemetry Collector will scrape the metrics. "scrapeInterval": "5s", # metricScalerEndpoint is the endpoint from which the KEDA's ScaledObject will scrape the metrics. "metricScalerEndpoint": "keda-otel-scaler.keda.svc:4318", # metricReceiverEndpoint is the endpoint from which the OpenTelemetry Collector will scrape the metrics. "metricReceiverEndpoint": "keda-otel-scaler.keda.svc:4317" } # ====================================== AUTOSCALER CONFIGURATION ====================================== # Example autoscaler: |- { # scaleUpStabilizationWindowSeconds is the stabilization window in seconds for scale up. "scaleUpStabilizationWindowSeconds": "0", # scaleDownStabilizationWindowSeconds is the stabilization window in seconds for scale down. "scaleDownStabilizationWindowSeconds": "300" } # ====================================== STORAGE INITIALIZER CONFIGURATION ====================================== # Example storageInitializer: |- { "image" : "kserve/storage-initializer:latest", "memoryRequest": "100Mi", "memoryLimit": "1Gi", "cpuRequest": "100m", "cpuLimit": "1", "caBundleConfigMapName": "", "caBundleVolumeMountPath": "/etc/ssl/custom-certs", "enableModelcar": false, "enableOciModelSupport": false, "ociModelMode": "modelcar", "cpuModelcar": "10m", "memoryModelcar": "15Mi" } storageInitializer: |- { # image contains the default storage initializer image uri. "image" : "kserve/storage-initializer:latest", # memoryRequest is the requests.memory to set for the storage initializer init container. "memoryRequest": "100Mi", # memoryLimit is the limits.memory to set for the storage initializer init container. "memoryLimit": "1Gi", # cpuRequest is the requests.cpu to set for the storage initializer init container. "cpuRequest": "100m", # cpuLimit is the limits.cpu to set for the storage initializer init container. "cpuLimit": "1", # caBundleConfigMapName is the ConfigMap will be copied to a user namespace for the storage initializer init container. "caBundleConfigMapName": "", # caBundleVolumeMountPath is the mount point for the configmap set by caBundleConfigMapName for the storage initializer init container. "caBundleVolumeMountPath": "/etc/ssl/custom-certs", # enableModelcar enabled allows you to directly access an OCI container image by # using a source URL with an "oci://" schema. "enableModelcar": false, # enableOciModelSupport enables any OCI-backed model storage path (modelcar, native ImageVolume, or fetch). # This is the newer master switch; enableModelcar is kept as a backcompat alias for the "modelcar" mode. "enableOciModelSupport": false, # ociModelMode selects the materialization strategy when a storageUri uses oci:// without an explicit # suffix. Valid values: "modelcar" (default sidecar), "native" (K8s ImageVolume), "fetch" (init-container). "ociModelMode": "modelcar", # cpuModelcar is the cpu request and limit that is used for the passive modelcar container. It can be # set very low, but should be allowed by any Kubernetes LimitRange that might apply. "cpuModelcar": "10m", # cpuModelcar is the memory request and limit that is used for the passive modelcar container. It can be # set very low, but should be allowed by any Kubernetes LimitRange that might apply. "memoryModelcar": "15Mi", # uidModelcar is the UID under with which the modelcar process and the main container is running. # Some Kubernetes clusters might require this to be root (0). If not set the user id is left untouched (default) "uidModelcar": 10 } # ====================================== CREDENTIALS ====================================== # Example credentials: |- { "storageSpecSecretName": "storage-config", "storageSecretNameAnnotation": "serving.kserve.io/storageSecretName", "gcs": { "gcsCredentialFileName": "gcloud-application-credentials.json" }, "s3": { "s3AccessKeyIDName": "AWS_ACCESS_KEY_ID", "s3SecretAccessKeyName": "AWS_SECRET_ACCESS_KEY", "s3Endpoint": "", "s3UseHttps": "", "s3Region": "", "s3VerifySSL": "", "s3UseVirtualBucket": "", "s3UseAccelerate": "", "s3UseAnonymousCredential": "", "s3CABundleConfigMap": "", "s3CABundle": "" } } # This is a global configuration used for downloading models from the cloud storage. # You can override this configuration by specifying the annotations on service account or static secret. # https://kserve.github.io/website/master/modelserving/storage/s3/s3/ # For a quick reference about AWS ENV variables: # AWS Cli: https://docs.aws.amazon.com/cli/latest/userguide/cli-configure-envvars.html # Boto: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/configuration.html#using-environment-variables # # The `s3AccessKeyIDName` and `s3SecretAccessKeyName` fields are only used from this configmap when static credentials (IAM User Access Key Secret) # are used as the authentication method for AWS S3. # The rest of the fields are used in both authentication methods (IAM Role for Service Account & IAM User Access Key Secret) if a non-empty value is provided. credentials: |- { # storageSpecSecretName contains the secret name which has the credentials for downloading the model. # This option is used when specifying the storage spec on isvc yaml. "storageSpecSecretName": "storage-config", # The annotation can be specified on isvc yaml to allow overriding with the secret name reference from the annotation value. # When using storageUri the order of the precedence is: secret name reference annotation > secret name references from service account # When using storageSpec the order of the precedence is: secret name reference annotation > storageSpecSecretName in configmap # Configuration for google cloud storage "gcs": { # gcsCredentialFileName specifies the filename of the gcs credential "gcsCredentialFileName": "gcloud-application-credentials.json" }, # Configuration for aws s3 storage. This add the corresponding environmental variables to the storage initializer init container. # For more info on s3 storage see https://kserve.github.io/website/master/modelserving/storage/s3/s3/ "s3": { # s3AccessKeyIDName specifies the s3 access key id name "s3AccessKeyIDName": "AWS_ACCESS_KEY_ID", # s3SecretAccessKeyName specifies the s3 secret access key name "s3SecretAccessKeyName": "AWS_SECRET_ACCESS_KEY", # s3Endpoint specifies the s3 endpoint "s3Endpoint": "", # s3UseHttps controls whether to use secure https or unsecure http to download models. # Allowed values are 0 and 1. "s3UseHttps": "", # s3Region specifies the region of the bucket. "s3Region": "", # s3VerifySSL controls whether to verify the tls/ssl certificate. "s3VerifySSL": "", # s3UseVirtualBucket configures whether it is a virtual bucket or not. "s3UseVirtualBucket": "", # s3UseAccelerate configures whether to use transfer acceleration. "s3UseAccelerate": "", # s3UseAnonymousCredential configures whether to use anonymous credentials to download the model or not. "s3UseAnonymousCredential": "", # s3CABundleConfigMap specifies the mounted CA bundle config map name. "s3CABundleConfigMap": "", # s3CABundle specifies the full path (mount path + file name) for the mounted config map data when used with a configured CA bundle config map. # s3CABundle specifies the path to a certificate bundle to use for HTTPS certificate validation when used absent of a configured CA bundle config map. "s3CABundle": "" } } # ====================================== INGRESS CONFIGURATION ====================================== # Example ingress: |- { "enableGatewayApi": false, "kserveIngressGateway": "kserve/kserve-ingress-gateway", "ingressGateway" : "knative-serving/knative-ingress-gateway", "localGateway" : "knative-serving/knative-local-gateway", "localGatewayService" : "knative-local-gateway.istio-system.svc.cluster.local", "ingressDomain" : "example.com", "additionalIngressDomains": ["additional-example.com", "additional-example-1.com"], "ingressClassName" : "istio", "domainTemplate": "{{ .Name }}-{{ .Namespace }}.{{ .IngressDomain }}", "urlScheme": "http", "disableIstioVirtualHost": false, "disableIngressCreation": false, "disableHTTPRouteTimeout": false } ingress: |- { # enableGatewayApi specifies whether to use Gateway API instead of Ingress to serve external traffic. "enableGatewayApi": false, # KServe implements [Gateway API](https://gateway-api.sigs.k8s.io/) to serve external traffic. # By default, KServe configures a default gateway to serve external traffic. # But, KServe can be configured to use a custom gateway by modifying this configuration. # The gateway should be specified in format / # NOTE: This configuration only applicable for raw deployment. "kserveIngressGateway": "kserve/kserve-ingress-gateway", # ingressGateway specifies the ingress gateway to serve external traffic. # The gateway should be specified in format / # NOTE: This configuration only applicable for serverless deployment with Istio configured as network layer. "ingressGateway" : "knative-serving/knative-ingress-gateway", # knativeLocalGatewayService specifies the hostname of the Knative's local gateway service. # The default KServe configurations are re-using the Istio local gateways for Knative. In this case, this # knativeLocalGatewayService field can be left unset. When unset, the value of "localGatewayService" will be used. # However, sometimes it may be better to have local gateways specifically for KServe (e.g. when enabling strict mTLS in Istio). # Under such setups where KServe is needed to have its own local gateways, the values of the "localGateway" and # "localGatewayService" should point to the KServe local gateways. Then, this knativeLocalGatewayService field # should point to the Knative's local gateway service. # NOTE: This configuration only applicable for serverless deployment with Istio configured as network layer. "knativeLocalGatewayService": "", # localGateway specifies the gateway which handles the network traffic within the cluster. # NOTE: This configuration only applicable for serverless deployment with Istio configured as network layer. "localGateway" : "knative-serving/knative-local-gateway", # localGatewayService specifies the hostname of the local gateway service. # NOTE: This configuration only applicable for serverless deployment with Istio configured as network layer. "localGatewayService" : "knative-local-gateway.istio-system.svc.cluster.local", # ingressDomain specifies the domain name which is used for creating the url. # If ingressDomain is empty then example.com is used as default domain. # NOTE: This configuration only applicable for raw deployment. "ingressDomain" : "example.com", # additionalIngressDomains specifies the additional domain names which are used for creating the url. "additionalIngressDomains": ["additional-example.com", "additional-example-1.com"] # ingressClassName specifies the ingress controller to use for ingress traffic. # This is optional and if omitted the default ingress in the cluster is used. # https://kubernetes.io/docs/concepts/services-networking/ingress/#default-ingress-class # NOTE: This configuration only applicable for raw deployment. "ingressClassName" : "istio", # domainTemplate specifies the template for generating domain/url for each inference service by combining variable from: # Name of the inference service ( {{ .Name}} ) # Namespace of the inference service ( {{ .Namespace }} ) # Annotation of the inference service ( {{ .Annotations.key }} ) # Label of the inference service ( {{ .Labels.key }} ) # IngressDomain ( {{ .IngressDomain }} ) # If domain template is empty the default template {{ .Name }}-{{ .Namespace }}.{{ .IngressDomain }} is used. # NOTE: This configuration only applicable for raw deployment. "domainTemplate": "{{ .Name }}-{{ .Namespace }}.{{ .IngressDomain }}", # urlScheme specifies the url scheme to use for inference service and inference graph. # If urlScheme is empty then by default http is used. "urlScheme": "http", # disableIstioVirtualHost controls whether to use istio as network layer. # By default istio is used as the network layer. When DisableIstioVirtualHost is true, KServe does not # create the top level virtual service thus Istio is no longer required for serverless mode. # By setting this field to true, user can use other networking layers supported by knative. # For more info https://github.com/kserve/kserve/pull/2380, https://kserve.github.io/website/master/admin/serverless/kourier_networking/. # NOTE: This configuration is only applicable to serverless deployment. "disableIstioVirtualHost": false, # disableIngressCreation controls whether to disable ingress creation for raw deployment mode. "disableIngressCreation": false, # disableHTTPRouteTimeout controls whether to omit the timeout field from HTTPRoute rules. # Set to true for Gateway controllers (e.g. GKE Gateway) that do not support the optional timeouts field. "disableHTTPRouteTimeout": false, # pathTemplate specifies the template for generating path based url for each inference service. # The following variables can be used in the template for generating url. # Name of the inference service ( {{ .Name}} ) # Namespace of the inference service ( {{ .Namespace }} ) # For more info https://github.com/kserve/kserve/issues/2257. # NOTE: This configuration only applicable to serverless deployment. "pathTemplate": "/serving/{{ .Namespace }}/{{ .Name }}" } # ====================================== LOGGER CONFIGURATION ====================================== # Example logger: |- { "image" : "kserve/agent:latest", "memoryRequest": "100Mi", "memoryLimit": "1Gi", "cpuRequest": "100m", "cpuLimit": "1", "defaultUrl": "http://default-broker" } logger: |- { # image contains the default logger image uri. "image" : "kserve/agent:latest", # memoryRequest is the requests.memory to set for the logger container. "memoryRequest": "100Mi", # memoryLimit is the limits.memory to set for the logger container. "memoryLimit": "1Gi", # cpuRequest is the requests.cpu to set for the logger container. "cpuRequest": "100m", # cpuLimit is the limits.cpu to set for the logger container. "cpuLimit": "1", # defaultUrl specifies the default logger url. If logger is not specified in the resource this url is used. "defaultUrl": "http://default-broker" } # ====================================== BATCHER CONFIGURATION ====================================== # Example batcher: |- { "image" : "kserve/agent:latest", "memoryRequest": "1Gi", "memoryLimit": "1Gi", "cpuRequest": "1", "cpuLimit": "1", "maxBatchSize": "32", "maxLatency": "5000" } batcher: |- { # image contains the default batcher image uri. "image" : "kserve/agent:latest", # memoryRequest is the requests.memory to set for the batcher container. "memoryRequest": "1Gi", # memoryLimit is the limits.memory to set for the batcher container. "memoryLimit": "1Gi", # cpuRequest is the requests.cpu to set for the batcher container. "cpuRequest": "1", # cpuLimit is the limits.cpu to set for the batcher container. "cpuLimit": "1" # maxBatchSize is the default maximum batch size for batcher. "maxBatchSize": "32", # maxLatency is the default maximum latency in milliseconds for batcher to wait and collect the batch. "maxLatency": "5000" } # ====================================== AGENT CONFIGURATION ====================================== # Example agent: |- { "image" : "kserve/agent:latest", "memoryRequest": "100Mi", "memoryLimit": "1Gi", "cpuRequest": "100m", "cpuLimit": "1" } agent: |- { # image contains the default agent image uri. "image" : "kserve/agent:latest", # memoryRequest is the requests.memory to set for the agent container. "memoryRequest": "100Mi", # memoryLimit is the limits.memory to set for the agent container. "memoryLimit": "1Gi", # cpuRequest is the requests.cpu to set for the agent container. "cpuRequest": "100m", # cpuLimit is the limits.cpu to set for the agent container. "cpuLimit": "1" } # ====================================== ROUTER CONFIGURATION ====================================== # Example router: |- { "image" : "kserve/router:latest", "memoryRequest": "100Mi", "memoryLimit": "1Gi", "cpuRequest": "100m", "cpuLimit": "1", "headers": { "propagate": [] }, "imagePullPolicy": "IfNotPresent", "imagePullSecrets": ["docker-secret"] } # router is the implementation of inference graph. router: |- { # image contains the default router image uri. "image" : "kserve/router:latest", # memoryRequest is the requests.memory to set for the router container. "memoryRequest": "100Mi", # memoryLimit is the limits.memory to set for the router container. "memoryLimit": "1Gi", # cpuRequest is the requests.cpu to set for the router container. "cpuRequest": "100m", # cpuLimit is the limits.cpu to set for the router container. "cpuLimit": "1", # Propagate the specified headers to all the steps specified in an InferenceGraph. # You can either specify the exact header names or use [Golang supported regex patterns] # (https://pkg.go.dev/regexp/syntax@go1.21.3#hdr-Syntax) to propagate multiple headers. "headers": { "propagate": [ "Authorization", "Test-Header-*", "*Trace-Id*" ] } # imagePullPolicy specifies when the router image should be pulled from registry. "imagePullPolicy": "IfNotPresent", # # imagePullSecrets specifies the list of secrets to be used for pulling the router image from registry. # https://kubernetes.io/docs/tasks/configure-pod-container/pull-image-private-registry/ "imagePullSecrets": ["docker-secret"] } # ====================================== DEPLOYMENT CONFIGURATION ====================================== # Example deploy: |- { "defaultDeploymentMode": "Serverless", "deploymentRolloutStrategy": { "defaultRollout": { "maxSurge": "1", "maxUnavailable": "1" } } } deploy: |- { # defaultDeploymentMode specifies the default deployment mode of the kserve. The supported values are # Standard and Knative. Users can override the deployment mode at service level # by adding the annotation serving.kserve.io/deploymentMode. # "defaultDeploymentMode": "Standard", # deploymentRolloutStrategy specifies the default rollout strategy for the Standard deployment mode # "deploymentRolloutStrategy": { # defaultRollout specifies the default rollout configuration using Kubernetes deployment strategy # "defaultRollout": { # maxSurge specifies the maximum number of pods that can be created above the desired replica count # Can be an absolute number (ex: 5) or a percentage of desired pods (ex: 10%) # "maxSurge": "1", # maxUnavailable specifies the maximum number of pods that can be unavailable during the update # Can be an absolute number (ex: 5) or a percentage of desired pods (ex: 10%) # "maxUnavailable": "1" # } # } } # ====================================== SERVICE CONFIGURATION ====================================== # Example service: |- { "serviceClusterIPNone": false } service: |- { # ServiceClusterIPNone is a boolean flag to indicate if the service should have a clusterIP set to None. # If the DeploymentMode is Raw, the default value for ServiceClusterIPNone if not set is false # "serviceClusterIPNone": false } # ====================================== METRICS CONFIGURATION ====================================== # Example metricsAggregator: |- { "enableMetricAggregation": "false", "enablePrometheusScraping" : "false" } # For more info see https://github.com/kserve/kserve/blob/master/qpext/README.md metricsAggregator: |- { # enableMetricAggregation configures metric aggregation annotation. This adds the annotation serving.kserve.io/enable-metric-aggregation to every # service with the specified boolean value. If true enables metric aggregation in queue-proxy by setting env vars in the queue proxy container # to configure scraping ports. "enableMetricAggregation": "false", # enablePrometheusScraping configures metric aggregation annotation. This adds the annotation serving.kserve.io/enable-metric-aggregation to every # service with the specified boolean value. If true, prometheus annotations are added to the pod. If serving.kserve.io/enable-metric-aggregation is false, # the prometheus port is set with the default prometheus scraping port 9090, otherwise the prometheus port annotation is set with the metric aggregation port. "enablePrometheusScraping" : "false" } # ====================================== LOCALMODEL CONFIGURATION ====================================== # Example localModel: |- { "enabled": false, # jobNamespace specifies the namespace where the download job will be created. "jobNamespace": "kserve-localmodel-jobs", # defaultJobImage specifies the default image used for the download job. "defaultJobImage" : "kserve/storage-initializer:latest", # Kubernetes modifies the filesystem group ID on the attached volume. "fsGroup": 1000, # TTL for the download job after it is finished. "jobTTLSecondsAfterFinished": 3600, # The frequency at which the local model agent reconciles the local models # This is to detect if models are missing from local disk "reconcilationFrequencyInSecs": 60, # This is to disable localmodel pv and pvc management for namespaces without isvcs "disableVolumeManagement": false } agent: |- { "cpuLimit": "1", "cpuRequest": "100m", "image": "quay.io/opendatahub/kserve-agent@sha256:81de9c7b44afa2675e48ef6d97fb5e4ff2cc473e1368ec2f514df37916d89654", "memoryLimit": "1Gi", "memoryRequest": "100Mi" } autoscaler: |- { "scaleUpStabilizationWindowSeconds": "0", "scaleDownStabilizationWindowSeconds": "300" } autoscaling-wva-controller-config: |- { "prometheus": { "url": "https://thanos-querier.openshift-monitoring.svc.cluster.local:9091", "authModes": "bearer", "triggerAuthName": "ai-inference-keda-thanos", "triggerAuthKind": "ClusterTriggerAuthentication" } } batcher: |- { "cpuLimit": "1", "cpuRequest": "1", "image": "quay.io/opendatahub/kserve-agent@sha256:81de9c7b44afa2675e48ef6d97fb5e4ff2cc473e1368ec2f514df37916d89654", "memoryLimit": "1Gi", "memoryRequest": "1Gi" } credentials: |- { "storageSpecSecretName": "storage-config", "storageSecretNameAnnotation": "serving.kserve.io/storageSecretName", "gcs": { "gcsCredentialFileName": "gcloud-application-credentials.json" }, "s3": { "s3AccessKeyIDName": "AWS_ACCESS_KEY_ID", "s3SecretAccessKeyName": "AWS_SECRET_ACCESS_KEY", "s3Endpoint": "", "s3UseHttps": "", "s3Region": "", "s3VerifySSL": "", "s3UseVirtualBucket": "", "s3UseAccelerate": "", "s3UseAnonymousCredential": "", "s3CABundleConfigMap": "odh-kserve-custom-ca-bundle", "s3CABundle": "/etc/ssl/custom-certs/cabundle.crt" } } deploy: |- { "defaultDeploymentMode": "RawDeployment" } explainers: '{}' inferenceService: |- { "serviceAnnotationDisallowedList": [ "autoscaling.knative.dev/min-scale", "autoscaling.knative.dev/max-scale", "internal.serving.kserve.io/storage-initializer-sourceuri", "kubectl.kubernetes.io/last-applied-configuration", "security.opendatahub.io/enable-auth", "networking.knative.dev/visibility", "haproxy.router.openshift.io/timeout", "opendatahub.io/hardware-profile-name", "opendatahub.io/hardware-profile-namespace" ] } ingress: "{\n \"enableGatewayApi\": false,\n \"kserveIngressGateway\": \"openshift-ingress/openshift-ai-inference\",\n \ \"enableLLMInferenceServiceTLS\": true,\n \"ingressGateway\" : \"knative-serving/knative-ingress-gateway\",\n \ \"knativeLocalGatewayService\" : \"knative-local-gateway.istio-system.svc.cluster.local\",\n \ \"ingressService\" : \"istio-ingressgateway.istio-system.svc.cluster.local\",\n \ \"localGateway\" : \"istio-system/kserve-local-gateway\",\n \"localGatewayService\" : \"kserve-local-gateway.istio-system.svc.cluster.local\",\n \"ingressDomain\" \ : \"example.com\",\n \"ingressClassName\" : \"openshift-default\",\n \"domainTemplate\": \"{{ .Name }}-{{ .Namespace }}.{{ .IngressDomain }}\",\n \"urlScheme\": \"http\",\n \ \"disableIstioVirtualHost\": false, \n \"disableIngressCreation\": true\n}" localModel: |- { "enabled": false, "jobNamespace": "opendatahub", "defaultJobImage" : "REPLACE_IMAGE", "fsGroup": 1000, "localModelAgentImage": "REPLACE_IMAGE", "localModelAgentCpuRequest": "100m", "localModelAgentMemoryRequest": "200Mi", "localModelAgentCpuLimit": "100m", "localModelAgentMemoryLimit": "300Mi" } logger: |- { "cpuLimit": "1", "cpuRequest": "100m", "defaultUrl": "http://default-broker", "image": "quay.io/opendatahub/kserve-agent@sha256:81de9c7b44afa2675e48ef6d97fb5e4ff2cc473e1368ec2f514df37916d89654", "memoryLimit": "1Gi", "memoryRequest": "100Mi" } metricsAggregator: |- { "enableMetricAggregation": "false", "enablePrometheusScraping" : "false" } oauthProxy: |- { "cpuLimit": "200m", "cpuRequest": "100m", "image": "quay.io/opendatahub/odh-kube-auth-proxy@sha256:dcb09fbabd8811f0956ef612a0c9ddd5236804b9bd6548a0647d2b531c9d01b3", "memoryLimit": "128Mi", "memoryRequest": "64Mi" } openshiftConfig: |- { "modelcachePermissionFixImage": "REPLACE_IMAGE", "ovmsVersioningImage": "registry.redhat.io/ubi9/ubi-micro@sha256:38e934147827349f2b8b11ac9c38d7be23cb8e29f128b1c46e5c7ae54a2d23cd" } opentelemetryCollector: |- { "scrapeInterval": "5s", "metricReceiverEndpoint": "keda-otel-scaler.keda.svc:4317", "metricScalerEndpoint": "keda-otel-scaler.keda.svc:4318", "resource": { "cpuLimit": "1", "memoryLimit": "2Gi", "cpuRequest": "200m", "memoryRequest": "512Mi" } } router: |- { "cpuLimit": "1", "cpuRequest": "100m", "headers": { "propagate": [ "Authorization" ] }, "image": "quay.io/opendatahub/kserve-router@sha256:310f41f1c7eb1a56fda87465241323f4eb2e286e2805ba71b8015ffcf6fc5d8d", "memoryLimit": "1Gi", "memoryRequest": "100Mi" } security: |- { "autoMountServiceAccountToken": false } service: |- { "serviceClusterIPNone": false } storageInitializer: |- { "cpuLimit": "1", "cpuModelcar": "10m", "cpuRequest": "100m", "enableModelcar": true, "image": "quay.io/opendatahub/kserve-storage-initializer@sha256:45d91dc105b49144984afdbe9de3160da1c989d254fd58cacbd7a1fbe0d83b56", "memoryLimit": "24Gi", "memoryModelcar": "15Mi", "memoryRequest": "100Mi" } kind: ConfigMap metadata: creationTimestamp: "2026-07-28T12:41:20Z" managedFields: - apiVersion: v1 fieldsType: FieldsV1 fieldsV1: f:data: f:_example: {} f:agent: {} f:autoscaler: {} f:autoscaling-wva-controller-config: {} f:batcher: {} f:credentials: {} f:deploy: {} f:explainers: {} f:inferenceService: {} f:ingress: {} f:localModel: {} f:logger: {} f:metricsAggregator: {} f:oauthProxy: {} f:openshiftConfig: {} f:opentelemetryCollector: {} f:router: {} f:security: {} f:service: {} f:storageInitializer: {} manager: kubectl operation: Apply time: "2026-07-28T12:41:20Z" name: inferenceservice-config namespace: kserve resourceVersion: "12550" uid: d2261bf5-dead-4dd3-af9a-2602db0741e8 - apiVersion: v1 data: kserve-agent: quay.io/opendatahub/kserve-agent@sha256:81de9c7b44afa2675e48ef6d97fb5e4ff2cc473e1368ec2f514df37916d89654 kserve-controller: quay.io/opendatahub/kserve-controller@sha256:f5de9f19d844ff46d59774656aa5bcb57e119d6d3dfc6c865ff6d1196d3de102 kserve-llm-d: registry.redhat.io/rhaiis/vllm-cuda-rhel9@sha256:fc68d623d1bfc36c8cb2fe4a71f19c8578cfb420ce8ce07b20a02c1ee0be0cf3 kserve-llm-d-amd-rocm: registry.redhat.io/rhaiis/vllm-rocm-rhel9@sha256:d9a48add238cc095fa43eeee17c8c4d104de60c4dc623e0bc7f8c4b53b2b2e97 kserve-llm-d-amd-rocm-fast-1: registry.redhat.io/rhaiis/vllm-rocm-rhel9@sha256:d9a48add238cc095fa43eeee17c8c4d104de60c4dc623e0bc7f8c4b53b2b2e97 kserve-llm-d-amd-rocm-fast-1-upstream-version: 0.11.0+rhai5 kserve-llm-d-amd-rocm-fast-2: registry.redhat.io/rhaiis/vllm-rocm-rhel9@sha256:d9a48add238cc095fa43eeee17c8c4d104de60c4dc623e0bc7f8c4b53b2b2e97 kserve-llm-d-amd-rocm-fast-2-upstream-version: 0.11.0+rhai5 kserve-llm-d-amd-rocm-upstream-version: 0.11.0+rhai5 kserve-llm-d-ibm-spyre: registry.redhat.io/rhaiis/vllm-spyre-rhel9@sha256:80ae3e435a5be2c1f117f36599103ab05357917dd6e37f0df6613cb3ac2c13ea kserve-llm-d-ibm-spyre-fast-1: registry.redhat.io/rhaiis/vllm-spyre-rhel9@sha256:80ae3e435a5be2c1f117f36599103ab05357917dd6e37f0df6613cb3ac2c13ea kserve-llm-d-ibm-spyre-fast-1-upstream-version: 0.10.2 kserve-llm-d-ibm-spyre-fast-2: registry.redhat.io/rhaiis/vllm-spyre-rhel9@sha256:80ae3e435a5be2c1f117f36599103ab05357917dd6e37f0df6613cb3ac2c13ea kserve-llm-d-ibm-spyre-fast-2-upstream-version: 0.10.2 kserve-llm-d-ibm-spyre-upstream-version: 0.10.2 kserve-llm-d-inference-scheduler: quay.io/opendatahub/odh-llm-d-router-endpoint-picker:v0.9.0 kserve-llm-d-intel-gaudi: registry.redhat.io/rhaii-early-access/vllm-gaudi-rhel9:3.4.0-ea.2 kserve-llm-d-intel-gaudi-fast-1: registry.redhat.io/rhaii-early-access/vllm-gaudi-rhel9:3.4.0-ea.2 kserve-llm-d-intel-gaudi-fast-1-upstream-version: 0.16.0 kserve-llm-d-intel-gaudi-fast-2: registry.redhat.io/rhaii-early-access/vllm-gaudi-rhel9:3.4.0-ea.2 kserve-llm-d-intel-gaudi-fast-2-upstream-version: 0.16.0 kserve-llm-d-intel-gaudi-upstream-version: 0.16.0 kserve-llm-d-latency-predictor-prediction: quay.io/opendatahub/odh-latency-predictor-prediction:odh-stable kserve-llm-d-latency-predictor-training: quay.io/opendatahub/odh-latency-predictor-training:odh-stable kserve-llm-d-nvidia-cuda: registry.redhat.io/rhaiis/vllm-cuda-rhel9@sha256:fc68d623d1bfc36c8cb2fe4a71f19c8578cfb420ce8ce07b20a02c1ee0be0cf3 kserve-llm-d-nvidia-cuda-fast-1: registry.redhat.io/rhaiis/vllm-cuda-rhel9@sha256:fc68d623d1bfc36c8cb2fe4a71f19c8578cfb420ce8ce07b20a02c1ee0be0cf3 kserve-llm-d-nvidia-cuda-fast-1-upstream-version: 0.11.0+rhai5 kserve-llm-d-nvidia-cuda-fast-2: registry.redhat.io/rhaiis/vllm-cuda-rhel9@sha256:fc68d623d1bfc36c8cb2fe4a71f19c8578cfb420ce8ce07b20a02c1ee0be0cf3 kserve-llm-d-nvidia-cuda-fast-2-upstream-version: 0.11.0+rhai5 kserve-llm-d-nvidia-cuda-upstream-version: 0.11.0+rhai5 kserve-llm-d-routing-sidecar: quay.io/opendatahub/odh-llm-d-router-disagg-sidecar:v0.9.0 kserve-llm-d-uds-tokenizer: quay.io/opendatahub/llm-d-kv-cache:v0.8.0 kserve-localmodel-controller: quay.io/opendatahub/odh-kserve-localmodel-controller:odh-master kserve-localmodelnode-agent: quay.io/opendatahub/odh-kserve-localmodelnode-agent:odh-master kserve-router: quay.io/opendatahub/kserve-router@sha256:310f41f1c7eb1a56fda87465241323f4eb2e286e2805ba71b8015ffcf6fc5d8d kserve-storage-initializer: quay.io/opendatahub/kserve-storage-initializer@sha256:45d91dc105b49144984afdbe9de3160da1c989d254fd58cacbd7a1fbe0d83b56 kube-rbac-proxy: quay.io/opendatahub/odh-kube-auth-proxy@sha256:dcb09fbabd8811f0956ef612a0c9ddd5236804b9bd6548a0647d2b531c9d01b3 llmisvc-controller: quay.io/opendatahub/odh-kserve-llmisvc-controller@sha256:16bf1af46c7c514036004731466ff3316d17d8a1e200348a212b6308af4c48f8 ovms-versioning-ubi-micro: registry.redhat.io/ubi9/ubi-micro@sha256:38e934147827349f2b8b11ac9c38d7be23cb8e29f128b1c46e5c7ae54a2d23cd kind: ConfigMap metadata: creationTimestamp: "2026-07-28T12:41:20Z" managedFields: - apiVersion: v1 fieldsType: FieldsV1 fieldsV1: f:data: f:kserve-agent: {} f:kserve-controller: {} f:kserve-llm-d: {} f:kserve-llm-d-amd-rocm: {} f:kserve-llm-d-amd-rocm-fast-1: {} f:kserve-llm-d-amd-rocm-fast-1-upstream-version: {} f:kserve-llm-d-amd-rocm-fast-2: {} f:kserve-llm-d-amd-rocm-fast-2-upstream-version: {} f:kserve-llm-d-amd-rocm-upstream-version: {} f:kserve-llm-d-ibm-spyre: {} f:kserve-llm-d-ibm-spyre-fast-1: {} f:kserve-llm-d-ibm-spyre-fast-1-upstream-version: {} f:kserve-llm-d-ibm-spyre-fast-2: {} f:kserve-llm-d-ibm-spyre-fast-2-upstream-version: {} f:kserve-llm-d-ibm-spyre-upstream-version: {} f:kserve-llm-d-inference-scheduler: {} f:kserve-llm-d-intel-gaudi: {} f:kserve-llm-d-intel-gaudi-fast-1: {} f:kserve-llm-d-intel-gaudi-fast-1-upstream-version: {} f:kserve-llm-d-intel-gaudi-fast-2: {} f:kserve-llm-d-intel-gaudi-fast-2-upstream-version: {} f:kserve-llm-d-intel-gaudi-upstream-version: {} f:kserve-llm-d-latency-predictor-prediction: {} f:kserve-llm-d-latency-predictor-training: {} f:kserve-llm-d-nvidia-cuda: {} f:kserve-llm-d-nvidia-cuda-fast-1: {} f:kserve-llm-d-nvidia-cuda-fast-1-upstream-version: {} f:kserve-llm-d-nvidia-cuda-fast-2: {} f:kserve-llm-d-nvidia-cuda-fast-2-upstream-version: {} f:kserve-llm-d-nvidia-cuda-upstream-version: {} f:kserve-llm-d-routing-sidecar: {} f:kserve-llm-d-uds-tokenizer: {} f:kserve-localmodel-controller: {} f:kserve-localmodelnode-agent: {} f:kserve-router: {} f:kserve-storage-initializer: {} f:kube-rbac-proxy: {} f:llmisvc-controller: {} f:ovms-versioning-ubi-micro: {} manager: kubectl operation: Apply time: "2026-07-28T12:41:20Z" name: kserve-parameters namespace: kserve resourceVersion: "12551" uid: 6c0ec6f6-527d-4eaf-95de-0217fb47fc4e - apiVersion: v1 data: ca.crt: | -----BEGIN CERTIFICATE----- MIIDPDCCAiSgAwIBAgIIDkiKpa/ZZ5owDQYJKoZIhvcNAQELBQAwJjESMBAGA1UE CxMJb3BlbnNoaWZ0MRAwDgYDVQQDEwdyb290LWNhMB4XDTI2MDcyODEyMjYyNFoX DTM2MDcyNTEyMjYyNFowJjESMBAGA1UECxMJb3BlbnNoaWZ0MRAwDgYDVQQDEwdy b290LWNhMIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEA40rjXwjDUa99 hkLiNaXmWtjFcQ/9oCshGWhjrDcmM5isbgPu+O/zLMoBVdq5YvNdazROqUR9hrb2 2takL7GlNYAnHQgqTxM5pAnk3BGF7kqX4RfPeuU6MQPiQZIjaBXwroDriGEMCVsI I2Q/9MTMnzLNzX3cS8pYtWTGJ5sNFblLR9lDXTfwukjuL7E2xnJ0Kf5nVzU2BZMb ZQ+VuEDvfKbjCxprMiqM0dvE39uwZijPYoyItJV2V8hLK5PR1LAUuenslCvoUcU5 Xu3aOdPgXjTug5SajKUt4pUgO6nHDBYzDi+N8oCNqbmiJPNQzY/1roBArNRwqZ6o y1AWikhlYwIDAQABo24wbDAOBgNVHQ8BAf8EBAMCAqQwDwYDVR0TAQH/BAUwAwEB /zBJBgNVHQ4EQgRAgcJhauJNS2PZ0f1NPVbWd9M2LZo0AApfAm2QN+xaQ9rU56QW 6oiGvNrhNohVuoZmLH36b49Ld8cfMnWKuWgj8zANBgkqhkiG9w0BAQsFAAOCAQEA pwvp0/OaakM6K7rmN++x14J2Zadvt4yzaifT3qY6URqH6FqHvbOWxEae2q9DLbu4 7ZFCxNCpdvOcz65b07rmWr19McUG1zTPsYuU/TsygaJfNo8im+MCDcGPNkHT/1Q6 MspyMd+rs5Y/8Tz9qFe3QUmyuEyaQ8RLgznqHXom1IG7PT/Az4Ra8UyLD9rgLXG9 BX7J0rxpZ+0vuvfAkxL7AEEDZ8qIUxyCQR32dr4YXFWr3+Kevam+P4v6DE6HuWU7 SK+riuSOpgKTn00EGN+48cKBMyjONIY/cSVxtJZnX8sJ9D9K+43AIPqQq4VMCbXO cZyYvgEeV8FCjjVxZi2aVg== -----END CERTIFICATE----- -----BEGIN CERTIFICATE----- MIIEADCCAuigAwIBAgIIAyzgY/couWQwDQYJKoZIhvcNAQELBQAwJjESMBAGA1UE CxMJb3BlbnNoaWZ0MRAwDgYDVQQDEwdyb290LWNhMB4XDTI2MDcyODEyMjcwMVoX DTI3MDcyODEyMjcwMVowMDESMBAGA1UEChMJb3BlbnNoaWZ0MRowGAYDVQQDExFv cGVuc2hpZnQtaW5ncmVzczCCASIwDQYJKoZIhvcNAQEBBQADggEPADCCAQoCggEB AMFbdrmU7/Vsbm6ElAZH4yQ4x9IoNAuY1nFDD7+WMK4P+vnMRp7Jo9R9zDBerMDI SkQhcyUPz1cYGrOchee30pDzqRfc2thuVzYxnPVoS1d3Nn4hweBTywvdVCmu5/zE 0YtvhUuUnCa8067tOjVfcmdUcOQGwJVDD4zLzAszzdix92svvAYSAxois9uG6S8G IX+D58tPSfsxTQ9w0yqx3wkKcaCECnH2VUv/F4Jly9I+EDwMbgPjMfNC673UezHu ApFysev+moAW1PMAg6Q1wxXtHMM8g0v/zzEWv/BsPGoLoVingKzsbT+iHh1AtdqZ I8jaKiU/D61JIYJdhWxeaf8CAwEAAaOCASYwggEiMA4GA1UdDwEB/wQEAwIFoDAd BgNVHSUEFjAUBggrBgEFBQcDAgYIKwYBBQUHAwEwDAYDVR0TAQH/BAIwADBJBgNV HQ4EQgRAvZsOMddz6faZK9Y/UveBu1DGNwL/yyxAxw12ckHUP5Bjhx9wuKRtHN3C yPDX+KMfUqeB6hWjFx4u2CxZBlfmvDBLBgNVHSMERDBCgECBwmFq4k1LY9nR/U09 VtZ30zYtmjQACl8CbZA37FpD2tTnpBbqiIa82uE2iFW6hmYsffpvj0t3xx8ydYq5 aCPzMEsGA1UdEQREMEKCQCouYXBwcy41NjZhY2E5Yy04Y2FjLTRmMmQtYTA1Yi0w MWVjNzJkMjQzNDkucHJvZC5rb25mbHV4ZWFhcy5jb20wDQYJKoZIhvcNAQELBQAD ggEBAILifN6ui/3FzgrWHyeAmmfm14HEnIey6QFJNQxJr/HYwXYHhyrPbpZF1rFh cKmRwSA1+AG9NN070YTtt4DRZ+QWsVjtBs9sdp3tDvbpJW2hl0BJzPP3LIZV+Vhv lli0mkkOj1sPEtkxsDJ0ACAtnSCIaqjlMCmDdXiUDn+CdXmUOhGj0J9ls9b1LWCF WRUQNFtHwRjf4yfpHT5Ut60lrXwUMQH4404VmMpu433Rs+rxFss3TWslfY7+pucS qYhJyBL5vwOEjlJ2lb8qE4kbTbYz37kGAa3IKnoQQuGmiKV+KcE8hf9/mq03dtfB /QST1Rzm8aFtAMWXPvCpzFS+nlo= -----END CERTIFICATE----- kind: ConfigMap metadata: annotations: kubernetes.io/description: Contains a CA bundle that can be used to verify the kube-apiserver when using internal endpoints such as the internal service IP or kubernetes.default.svc. No other usage is guaranteed across distributions of Kubernetes clusters. creationTimestamp: "2026-07-28T12:39:31Z" managedFields: - apiVersion: v1 fieldsType: FieldsV1 fieldsV1: f:data: .: {} f:ca.crt: {} f:metadata: f:annotations: .: {} f:kubernetes.io/description: {} manager: kube-controller-manager operation: Update time: "2026-07-28T12:39:31Z" name: kube-root-ca.crt namespace: kserve resourceVersion: "11264" uid: 216bb900-729d-4a9b-945c-a84c01ab36dc - apiVersion: v1 data: service-ca.crt: | -----BEGIN CERTIFICATE----- MIIDUTCCAjmgAwIBAgIIGbYJ5MLo3iEwDQYJKoZIhvcNAQELBQAwNjE0MDIGA1UE Awwrb3BlbnNoaWZ0LXNlcnZpY2Utc2VydmluZy1zaWduZXJAMTc4NTI0MjE0NDAe Fw0yNjA3MjgxMjM1NDNaFw0yODA5MjUxMjM1NDRaMDYxNDAyBgNVBAMMK29wZW5z aGlmdC1zZXJ2aWNlLXNlcnZpbmctc2lnbmVyQDE3ODUyNDIxNDQwggEiMA0GCSqG SIb3DQEBAQUAA4IBDwAwggEKAoIBAQCtsI/N2TqvJTBp4yYuo+BRvcGbeP0m2JG9 GuDoMefFZrOIqQDotA9KG6ZBLt9+5tr+yqRRo6wS0J7YYx0afovV4CP94PxWtkA2 jfE60zLAzhbs6kGkBvnWnOSDo2a8vmLmAlmWvgZzgLxYMsCHrxzhVBakBxMYuz+z 5yAyEyiMA5Cn5tysSF0iNlkQbgY86AQ6xoEukHiCr7jh0lFWhsfZ9TqCgyGKYOcc FOJ9DYwgeupg+6QffeCeNnkVGtVJ9MMaYG5BzMPzQw7+k0C2GMYVqqRSAVnk6hOY OXZFeN/Ilv6t5mUMqzV1ynr7KT7JpwzdhGfORnI+IxkCj+Acj8c3AgMBAAGjYzBh MA4GA1UdDwEB/wQEAwICpDAPBgNVHRMBAf8EBTADAQH/MB0GA1UdDgQWBBTRJpYZ 8b4yuATo0Z9OMsHvALf+MjAfBgNVHSMEGDAWgBTRJpYZ8b4yuATo0Z9OMsHvALf+ MjANBgkqhkiG9w0BAQsFAAOCAQEAcH/LgS9qE/7d/U04NpTbGkvOf7U7ukzovWAO dikdLMV3u6c2xLVymBEHp9gNFttldx5zgGMq95+MRdj57LaZtjKBxfkIWV31EQTJ vBfY2Tn5iiF7zPOq2SU0+8iqEHnUm0ToqcjMpRK0LRbp0NFzIj3T1JvrfzOF55iv nqEeGBR/y1lG9DdD48mbfQMreTwoA3pet1PJoVEpvMldUVPrr2e6Q6k5dTUSqV2Z QyAg6MCEepdaMJWwxPFxSaxrfiSODZyW5XXdpmz8+gbF9r7viw9JYbpaUqXlzdTS tpnzJppU38RBYYzbiaBu50iIQw8WmXZ1Qz7Ypy2XfB09q3YJkg== -----END CERTIFICATE----- kind: ConfigMap metadata: annotations: service.beta.openshift.io/inject-cabundle: "true" creationTimestamp: "2026-07-28T12:39:31Z" managedFields: - apiVersion: v1 fieldsType: FieldsV1 fieldsV1: f:data: {} f:metadata: f:annotations: .: {} f:service.beta.openshift.io/inject-cabundle: {} manager: kube-controller-manager operation: Update time: "2026-07-28T12:39:31Z" - apiVersion: v1 fieldsType: FieldsV1 fieldsV1: f:data: f:service-ca.crt: {} manager: service-ca-operator operation: Update time: "2026-07-28T12:39:31Z" name: openshift-service-ca.crt namespace: kserve resourceVersion: "11269" uid: 8655d883-676f-45f1-aba8-00190c7bbc94 kind: ConfigMapList metadata: resourceVersion: "25338"