oci-service-operator

generativeai.oracle.com/v1beta1

Back to API Reference

APIVersion: generativeai.oracle.com/v1beta1

This content is generated from the checked-in CRD schemas in config/crd/bases/. If a description is missing or incorrect, fix the source comments or generator inputs and rerun make generate manifests; do not hand-edit config/crd/bases/*.yaml.

Packages

Package Support Latest release Resources
Generative AI preview v2.1.0-alpha DedicatedAiCluster, Endpoint, Model

Resources

Kind Scope Sample Packages
DedicatedAiCluster Namespaced Sample Generative AI (v2.1.0-alpha)
Endpoint Namespaced Sample Generative AI (v2.1.0-alpha)
Model Namespaced Sample Generative AI (v2.1.0-alpha)

DedicatedAiCluster

DedicatedAiCluster is the Schema for the dedicatedaiclusters API.

Spec

DedicatedAiClusterSpec defines the desired state of DedicatedAiCluster.

Field Description Type Required Default Enum
compartmentId The compartment OCID to create the dedicated AI cluster in. string Yes - -
definedTags Defined tags for this resource. Each key is predefined and scoped to a namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Operations": {"CostCenter": "42"}} map[string, map[string, string]] No - -
description An optional description of the dedicated AI cluster. string No - -
displayName A user-friendly name. Does not have to be unique, and it’s changeable. string No - -
freeformTags Free-form tags for this resource. Each tag is a simple key-value pair with no predefined name, type, or namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Department": "Finance"} map[string, string] No - -
type The dedicated AI cluster type indicating whether this is a fine-tuning/training processor or hosting/inference processor. Allowed values are: - HOSTING - FINE_TUNING string Yes - -
unitCount The number of dedicated units in this AI cluster. integer Yes - -
unitShape The shape of dedicated unit in this AI cluster. The underlying hardware configuration is hidden from customers. Allowed values are: - LARGE_COHERE - LARGE_COHERE_V2 - SMALL_COHERE - SMALL_COHERE_V2 - SMALL_COHERE_4 - EMBED_COHERE - LLAMA2_70 - LARGE_GENERIC - LARGE_COHERE_V2_2 - LARGE_GENERIC_4 - SMALL_GENERIC_V2 - LARGE_GENERIC_2 - LARGE_GENERIC_V3 - LARGE_COHERE_V3 - RERANK_COHERE - SMALL_GENERIC_V1 - MEDIUM_GENERIC_V1 - LARGE_GENERIC_V1 - A10_X1 - A10_X2 - A10_X4 - A100_40G_X1 - A100_40G_X2 - A100_40G_X4 - A100_40G_X8 - A100_80G_X1 - A100_80G_X2 - A100_80G_X4 - A100_80G_X8 - H100_X1 - H100_X2 - H100_X4 - H100_X8 - H200_X1 - H200_X2 - H200_X4 - H200_X8 The following shapes can only be used to deploy imported models: - A10_X1, A10_X2, A10_X4 - A100_40G_X1, A100_40G_X2, A100_40G_X4, A100_40G_X8 - A100_80G_X1, A100_80G_X2, A100_80G_X4, A100_80G_X8 - H100_X1, H100_X2, H100_X4, H100_X8 - H200_X1, H200_X2, H200_X4, H200_X8 - OAI_A10_X2 - OAI_H100_X1 - OAI_H100_X2 - OAI_H200_X1 - OAI_A100_80G_X1 - OAI_A100_80G_X2 - OAI_A100_40G_X1 - OAI_A100_40G_X4 string Yes - -

Status

DedicatedAiClusterStatus defines the observed state of DedicatedAiCluster.

Field Description Type Required Default Enum
capacity DedicatedAiClusterCapacity defines nested fields for DedicatedAiCluster.Capacity. object No - -
compartmentId The compartment OCID to create the dedicated AI cluster in. string No - -
definedTags Defined tags for this resource. Each key is predefined and scoped to a namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Operations": {"CostCenter": "42"}} map[string, map[string, string]] No - -
description An optional description of the dedicated AI cluster. string No - -
displayName A user-friendly name. Does not have to be unique, and it’s changeable. string No - -
freeformTags Free-form tags for this resource. Each tag is a simple key-value pair with no predefined name, type, or namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Department": "Finance"} map[string, string] No - -
id The OCID (https://docs.oracle.com/iaas/Content/General/Concepts/identifiers.htm) of the dedicated AI cluster. string No - -
lifecycleDetails A message describing the current state with detail that can provide actionable information. string No - -
lifecycleState The current state of the dedicated AI cluster. string No - -
status - object Yes - -
systemTags System tags for this resource. Each key is predefined and scoped to a namespace. Example: {"orcl-cloud": {"free-tier-retained": "true"}} map[string, map[string, string]] No - -
timeCreated The date and time the dedicated AI cluster was created, in the format defined by RFC 3339 string No - -
timeUpdated The date and time the dedicated AI cluster was updated, in the format defined by RFC 3339 string No - -
type The dedicated AI cluster type indicating whether this is a fine-tuning/training processor or hosting/inference processor. string No - -
unitCount The number of dedicated units in this AI cluster. integer No - -
unitShape The shape of dedicated unit in this AI cluster. The underlying hardware configuration is hidden from customers. string No - -

Status.capacity

Back to DedicatedAiCluster status

DedicatedAiClusterCapacity defines nested fields for DedicatedAiCluster.Capacity.

Field Description Type Required Default Enum
capacityType - string No - -
jsonData - string No - -
totalEndpointCapacity The total number of endpoints that can be hosted on this dedicated AI cluster. integer No - -
usedEndpointCapacity The number of endpoints hosted on this dedicated AI cluster. integer No - -

Status.status

Back to DedicatedAiCluster status

Field Description Type Required Default Enum
async Async is the canonical controller-owned async contract. Resource-local legacy work-request fields may remain as compatibility mirrors while follow-on migrations land, but new async state should project here first. object No - -
conditions - list[object] No - -
createdAt - string (date-time) No - -
deletedAt - string (date-time) No - -
message - string No - -
ocid - string No - -
opcRequestId OpcRequestID is the latest non-empty OCI request ID from a mutating OCI response or surfaced OCI service error that materially contributed to the current shared status projection. Headerless follow-up observations keep the last non-empty value intact. string No - -
reason - string No - -
requestedAt - string (date-time) No - -
updatedAt - string (date-time) No - -

Status.status.async

Back to DedicatedAiCluster status

Async is the canonical controller-owned async contract. Resource-local legacy work-request fields may remain as compatibility mirrors while follow-on migrations land, but new async state should project here first.

Field Description Type Required Default Enum
current - object No - -

Status.status.async.current

Back to DedicatedAiCluster status

Field Description Type Required Default Enum
message - string No - -
normalizedClass - string Yes - attention, canceled, failed, pending, succeeded, unknown
percentComplete - number No - -
phase - string Yes - create, delete, update
rawOperationType - string No - -
rawStatus - string No - -
source - string Yes - lifecycle, none, workrequest
updatedAt - string (date-time) Yes - -
workRequestId - string No - -

Status.status.conditions[]

Back to DedicatedAiCluster status

Field Description Type Required Default Enum
lastTransitionTime - string (date-time) No - -
message - string No - -
reason - string No - -
status - string Yes - -
type - string Yes - -

Endpoint

Endpoint is the Schema for the endpoints API.

Spec

EndpointSpec defines the desired state of Endpoint.

Field Description Type Required Default Enum
compartmentId The compartment OCID to create the endpoint in. string Yes - -
contentModerationConfig EndpointContentModerationConfig defines nested fields for Endpoint.ContentModerationConfig. object No - -
dedicatedAiClusterId The OCID of the dedicated AI cluster on which a model will be deployed to. string Yes - -
definedTags Defined tags for this resource. Each key is predefined and scoped to a namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Operations": {"CostCenter": "42"}} map[string, map[string, string]] No - -
description An optional description of the endpoint. string No - -
displayName A user-friendly name. Does not have to be unique, and it’s changeable. string No - -
freeformTags Free-form tags for this resource. Each tag is a simple key-value pair with no predefined name, type, or namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Department": "Finance"} map[string, string] No - -
generativeAiPrivateEndpointId The OCID of the Generative AI private endpoint to which this endpoint is attached to. string No - -
modelId The OCID of the model that’s used to create this endpoint. string Yes - -
piiDetectionConfig EndpointPiiDetectionConfig defines nested fields for Endpoint.PiiDetectionConfig. object No - -
promptInjectionConfig EndpointPromptInjectionConfig defines nested fields for Endpoint.PromptInjectionConfig. object No - -

Spec.contentModerationConfig

Back to Endpoint spec

EndpointContentModerationConfig defines nested fields for Endpoint.ContentModerationConfig.

Field Description Type Required Default Enum
isEnabled Whether to enable the content moderation feature. boolean Yes - -
mode Enum for the modes of operation for inference protection. string No - -
modelId The OCID of the model used for the feature. string No - -

Spec.piiDetectionConfig

Back to Endpoint spec

EndpointPiiDetectionConfig defines nested fields for Endpoint.PiiDetectionConfig.

Field Description Type Required Default Enum
isEnabled Whether to enable the PII detection feature. boolean Yes - -
mode The mode of operation for PII detection. string No - -
modelId The OCID of the model used for the feature. string No - -

Spec.promptInjectionConfig

Back to Endpoint spec

EndpointPromptInjectionConfig defines nested fields for Endpoint.PromptInjectionConfig.

Field Description Type Required Default Enum
isEnabled Whether to enable the prompt injection feature. boolean Yes - -
mode The mode of operation for prompt injection detection. string No - -
modelId The OCID of the model used for the feature. string No - -

Status

EndpointStatus defines the observed state of Endpoint.

Field Description Type Required Default Enum
compartmentId The compartment OCID to create the endpoint in. string No - -
contentModerationConfig EndpointContentModerationConfig defines nested fields for Endpoint.ContentModerationConfig. object No - -
dedicatedAiClusterId The OCID of the dedicated AI cluster on which the model will be deployed to. string No - -
definedTags Defined tags for this resource. Each key is predefined and scoped to a namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Operations": {"CostCenter": "42"}} map[string, map[string, string]] No - -
description An optional description of the endpoint. string No - -
displayName A user-friendly name. Does not have to be unique, and it’s changeable. string No - -
freeformTags Free-form tags for this resource. Each tag is a simple key-value pair with no predefined name, type, or namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Department": "Finance"} map[string, string] No - -
generativeAiPrivateEndpointId The OCID of the Generative AI private endpoint to which this endpoint is attached to. string No - -
id An OCID that uniquely identifies this endpoint resource. string No - -
lifecycleDetails A message describing the current state of the endpoint in more detail that can provide actionable information. string No - -
lifecycleState The current state of the endpoint. string No - -
modelId The OCID of the model that’s used to create this endpoint. string No - -
piiDetectionConfig EndpointPiiDetectionConfig defines nested fields for Endpoint.PiiDetectionConfig. object No - -
promptInjectionConfig EndpointPromptInjectionConfig defines nested fields for Endpoint.PromptInjectionConfig. object No - -
status - object Yes - -
systemTags System tags for this resource. Each key is predefined and scoped to a namespace. Example: {"orcl-cloud": {"free-tier-retained": "true"}} map[string, map[string, string]] No - -
timeCreated The date and time that the endpoint was created in the format of an RFC3339 datetime string. string No - -
timeUpdated The date and time that the endpoint was updated in the format of an RFC3339 datetime string. string No - -

Status.contentModerationConfig

Back to Endpoint status

EndpointContentModerationConfig defines nested fields for Endpoint.ContentModerationConfig.

Field Description Type Required Default Enum
isEnabled Whether to enable the content moderation feature. boolean Yes - -
mode Enum for the modes of operation for inference protection. string No - -
modelId The OCID of the model used for the feature. string No - -

Status.piiDetectionConfig

Back to Endpoint status

EndpointPiiDetectionConfig defines nested fields for Endpoint.PiiDetectionConfig.

Field Description Type Required Default Enum
isEnabled Whether to enable the PII detection feature. boolean Yes - -
mode The mode of operation for PII detection. string No - -
modelId The OCID of the model used for the feature. string No - -

Status.promptInjectionConfig

Back to Endpoint status

EndpointPromptInjectionConfig defines nested fields for Endpoint.PromptInjectionConfig.

Field Description Type Required Default Enum
isEnabled Whether to enable the prompt injection feature. boolean Yes - -
mode The mode of operation for prompt injection detection. string No - -
modelId The OCID of the model used for the feature. string No - -

Status.status

Back to Endpoint status

Field Description Type Required Default Enum
async Async is the canonical controller-owned async contract. Resource-local legacy work-request fields may remain as compatibility mirrors while follow-on migrations land, but new async state should project here first. object No - -
conditions - list[object] No - -
createdAt - string (date-time) No - -
deletedAt - string (date-time) No - -
message - string No - -
ocid - string No - -
opcRequestId OpcRequestID is the latest non-empty OCI request ID from a mutating OCI response or surfaced OCI service error that materially contributed to the current shared status projection. Headerless follow-up observations keep the last non-empty value intact. string No - -
reason - string No - -
requestedAt - string (date-time) No - -
updatedAt - string (date-time) No - -

Status.status.async

Back to Endpoint status

Async is the canonical controller-owned async contract. Resource-local legacy work-request fields may remain as compatibility mirrors while follow-on migrations land, but new async state should project here first.

Field Description Type Required Default Enum
current - object No - -

Status.status.async.current

Back to Endpoint status

Field Description Type Required Default Enum
message - string No - -
normalizedClass - string Yes - attention, canceled, failed, pending, succeeded, unknown
percentComplete - number No - -
phase - string Yes - create, delete, update
rawOperationType - string No - -
rawStatus - string No - -
source - string Yes - lifecycle, none, workrequest
updatedAt - string (date-time) Yes - -
workRequestId - string No - -

Status.status.conditions[]

Back to Endpoint status

Field Description Type Required Default Enum
lastTransitionTime - string (date-time) No - -
message - string No - -
reason - string No - -
status - string Yes - -
type - string Yes - -

Model

Model is the Schema for the models API.

Spec

ModelSpec defines the desired state of Model.

Field Description Type Required Default Enum
baseModelId The OCID of the base model that’s used for fine-tuning. string Yes - -
compartmentId The compartment OCID for fine-tuned models. For pretrained models, this value is null. string Yes - -
definedTags Defined tags for this resource. Each key is predefined and scoped to a namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Operations": {"CostCenter": "42"}} map[string, map[string, string]] No - -
description An optional description of the model. string No - -
displayName A user-friendly name. string No - -
fineTuneDetails ModelFineTuneDetails defines nested fields for Model.FineTuneDetails. object Yes - -
freeformTags Free-form tags for this resource. Each tag is a simple key-value pair with no predefined name, type, or namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Department": "Finance"} map[string, string] No - -
vendor The provider of the model. string No - -
version The version of the model. string No - -

Spec.fineTuneDetails

Back to Model spec

ModelFineTuneDetails defines nested fields for Model.FineTuneDetails.

Field Description Type Required Default Enum
dedicatedAiClusterId The OCID of the dedicated AI cluster this fine-tuning runs on. string Yes - -
trainingConfig ModelFineTuneDetailsTrainingConfig defines nested fields for Model.FineTuneDetails.TrainingConfig. object No - -
trainingDataset ModelFineTuneDetailsTrainingDataset defines nested fields for Model.FineTuneDetails.TrainingDataset. object Yes - -

Spec.fineTuneDetails.trainingConfig

Back to Model spec

ModelFineTuneDetailsTrainingConfig defines nested fields for Model.FineTuneDetails.TrainingConfig.

Field Description Type Required Default Enum
earlyStoppingPatience Stop training if the loss metric does not improve beyond ‘early_stopping_threshold’ for this many times of evaluation. integer No - -
earlyStoppingThreshold How much the loss must improve to prevent early stopping. number No - -
jsonData - string No - -
learningRate The initial learning rate to be used during training number No - -
logModelMetricsIntervalInSteps Determines how frequently to log model metrics. Every step is logged for the first 20 steps and then follows this parameter for log frequency. Set to 0 to disable logging the model metrics. integer No - -
loraAlpha This parameter represents the scaling factor for the weight matrices in LoRA. integer No - -
loraDropout This parameter indicates the dropout probability for LoRA layers. number No - -
loraR This parameter represents the LoRA rank of the update matrices. integer No - -
numOfLastLayers The number of last layers to be fine-tuned. integer No - -
totalTrainingEpochs The maximum number of training epochs to run for. integer No - -
trainingBatchSize The batch size used during training. integer No - -
trainingConfigType - string No - -

Spec.fineTuneDetails.trainingDataset

Back to Model spec

ModelFineTuneDetailsTrainingDataset defines nested fields for Model.FineTuneDetails.TrainingDataset.

Field Description Type Required Default Enum
bucketName The Object Storage bucket name. string No - -
datasetType - string No - -
jsonData - string No - -
namespaceName The Object Storage namespace. string No - -
objectName The Object Storage object name. string No - -

Status

ModelStatus defines the observed state of Model.

Field Description Type Required Default Enum
baseModelId The OCID of the base model that’s used for fine-tuning. For pretrained models, the value is null. string No - -
capabilities Describes what this model can be used for. list[string] No - -
compartmentId The compartment OCID for fine-tuned models. For pretrained models, this value is null. string No - -
definedTags Defined tags for this resource. Each key is predefined and scoped to a namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Operations": {"CostCenter": "42"}} map[string, map[string, string]] No - -
description An optional description of the model. string No - -
displayName A user-friendly name. string No - -
fineTuneDetails ModelFineTuneDetails defines nested fields for Model.FineTuneDetails. object No - -
freeformTags Free-form tags for this resource. Each tag is a simple key-value pair with no predefined name, type, or namespace. For more information, see Resource Tags (https://docs.oracle.com/iaas/Content/General/Concepts/resourcetags.htm). Example: {"Department": "Finance"} map[string, string] No - -
id An ID that uniquely identifies a pretrained or fine-tuned model. string No - -
isLongTermSupported Whether a model is supported long-term. Only applicable to base models. boolean No - -
lifecycleDetails A message describing the current state of the model in more detail that can provide actionable information. string No - -
lifecycleState The lifecycle state of the model. string No - -
modelMetrics ModelMetrics defines nested fields for Model.ModelMetrics. object No - -
status - object Yes - -
systemTags System tags for this resource. Each key is predefined and scoped to a namespace. Example: {"orcl-cloud": {"free-tier-retained": "true"}} map[string, map[string, string]] No - -
timeCreated The date and time that the model was created in the format of an RFC3339 datetime string. string No - -
timeDedicatedRetired The timestamp indicating when the custom model and its associated foundation model will be fully retired. string No - -
timeDeprecated Corresponds to the time when the custom model and its associated foundation model will be deprecated. string No - -
timeOnDemandRetired The timestamp indicating when the base model will no longer be available for on-demand usage. string No - -
timeUpdated The date and time that the model was updated in the format of an RFC3339 datetime string. string No - -
type The model type indicating whether this is a pretrained/base model or a custom/fine-tuned model. string No - -
vendor The provider of the base model. string No - -
version The version of the model. string No - -

Status.fineTuneDetails

Back to Model status

ModelFineTuneDetails defines nested fields for Model.FineTuneDetails.

Field Description Type Required Default Enum
dedicatedAiClusterId The OCID of the dedicated AI cluster this fine-tuning runs on. string Yes - -
trainingConfig ModelFineTuneDetailsTrainingConfig defines nested fields for Model.FineTuneDetails.TrainingConfig. object No - -
trainingDataset ModelFineTuneDetailsTrainingDataset defines nested fields for Model.FineTuneDetails.TrainingDataset. object Yes - -

Status.fineTuneDetails.trainingConfig

Back to Model status

ModelFineTuneDetailsTrainingConfig defines nested fields for Model.FineTuneDetails.TrainingConfig.

Field Description Type Required Default Enum
earlyStoppingPatience Stop training if the loss metric does not improve beyond ‘early_stopping_threshold’ for this many times of evaluation. integer No - -
earlyStoppingThreshold How much the loss must improve to prevent early stopping. number No - -
jsonData - string No - -
learningRate The initial learning rate to be used during training number No - -
logModelMetricsIntervalInSteps Determines how frequently to log model metrics. Every step is logged for the first 20 steps and then follows this parameter for log frequency. Set to 0 to disable logging the model metrics. integer No - -
loraAlpha This parameter represents the scaling factor for the weight matrices in LoRA. integer No - -
loraDropout This parameter indicates the dropout probability for LoRA layers. number No - -
loraR This parameter represents the LoRA rank of the update matrices. integer No - -
numOfLastLayers The number of last layers to be fine-tuned. integer No - -
totalTrainingEpochs The maximum number of training epochs to run for. integer No - -
trainingBatchSize The batch size used during training. integer No - -
trainingConfigType - string No - -

Status.fineTuneDetails.trainingDataset

Back to Model status

ModelFineTuneDetailsTrainingDataset defines nested fields for Model.FineTuneDetails.TrainingDataset.

Field Description Type Required Default Enum
bucketName The Object Storage bucket name. string No - -
datasetType - string No - -
jsonData - string No - -
namespaceName The Object Storage namespace. string No - -
objectName The Object Storage object name. string No - -

Status.modelMetrics

Back to Model status

ModelMetrics defines nested fields for Model.ModelMetrics.

Field Description Type Required Default Enum
finalAccuracy Fine-tuned model accuracy. number No - -
finalLoss Fine-tuned model loss. number No - -
jsonData - string No - -
modelMetricsType - string No - -

Status.status

Back to Model status

Field Description Type Required Default Enum
async Async is the canonical controller-owned async contract. Resource-local legacy work-request fields may remain as compatibility mirrors while follow-on migrations land, but new async state should project here first. object No - -
conditions - list[object] No - -
createdAt - string (date-time) No - -
deletedAt - string (date-time) No - -
message - string No - -
ocid - string No - -
opcRequestId OpcRequestID is the latest non-empty OCI request ID from a mutating OCI response or surfaced OCI service error that materially contributed to the current shared status projection. Headerless follow-up observations keep the last non-empty value intact. string No - -
reason - string No - -
requestedAt - string (date-time) No - -
updatedAt - string (date-time) No - -

Status.status.async

Back to Model status

Async is the canonical controller-owned async contract. Resource-local legacy work-request fields may remain as compatibility mirrors while follow-on migrations land, but new async state should project here first.

Field Description Type Required Default Enum
current - object No - -

Status.status.async.current

Back to Model status

Field Description Type Required Default Enum
message - string No - -
normalizedClass - string Yes - attention, canceled, failed, pending, succeeded, unknown
percentComplete - number No - -
phase - string Yes - create, delete, update
rawOperationType - string No - -
rawStatus - string No - -
source - string Yes - lifecycle, none, workrequest
updatedAt - string (date-time) Yes - -
workRequestId - string No - -

Status.status.conditions[]

Back to Model status

Field Description Type Required Default Enum
lastTransitionTime - string (date-time) No - -
message - string No - -
reason - string No - -
status - string Yes - -
type - string Yes - -
This site is open source. Improve this page.