gcloud alpha model-armor templates create

gcloud alpha model-armor templates create - create Model Armor Template

gcloud alpha model-armor templates create (TEMPLATE : --location=LOCATION) (--malicious-uri-filter-settings-enforcement=MALICIOUS_URI_FILTER_SETTINGS_ENFORCEMENT --rai-settings-filters=[confidenceLevel=CONFIDENCELEVEL],[filterType=FILTERTYPE] --basic-config-filter-enforcement=BASIC_CONFIG_FILTER_ENFORCEMENT     | --advanced-config-deidentify-template=ADVANCED_CONFIG_DEIDENTIFY_TEMPLATE --advanced-config-inspect-template=ADVANCED_CONFIG_INSPECT_TEMPLATE --pi-and-jailbreak-filter-settings-confidence-level=PI_AND_JAILBREAK_FILTER_SETTINGS_CONFIDENCE_LEVEL --pi-and-jailbreak-filter-settings-enforcement=PI_AND_JAILBREAK_FILTER_SETTINGS_ENFORCEMENT) [--labels=[LABELS,…]] [--request-id=REQUEST_ID] [--template-metadata-custom-llm-response-safety-error-code=TEMPLATE_METADATA_CUSTOM_LLM_RESPONSE_SAFETY_ERROR_CODE --template-metadata-custom-llm-response-safety-error-message=TEMPLATE_METADATA_CUSTOM_LLM_RESPONSE_SAFETY_ERROR_MESSAGE --template-metadata-custom-prompt-safety-error-code=TEMPLATE_METADATA_CUSTOM_PROMPT_SAFETY_ERROR_CODE --template-metadata-custom-prompt-safety-error-message=TEMPLATE_METADATA_CUSTOM_PROMPT_SAFETY_ERROR_MESSAGE --template-metadata-ignore-partial-invocation-failures --template-metadata-log-operations --template-metadata-log-sanitize-operations] [GCLOUD_WIDE_FLAG …]

Template resource - Identifier. name of resource The arguments in this group can be used to specify the attributes of this resource. (NOTE) Some attributes are not given arguments in this group but can be set in other ways.

To set the project attribute:

• provide the argument template on the command line with a fully specified name;
• provide the argument --project on the command line;
• set the property core/project.

This must be specified.

TEMPLATE

ID of the template or fully qualified identifier for the template.

To set the template attribute:

• provide the argument template on the command line.

This positional argument must be specified if any of the other arguments in this group are specified.

--location=LOCATION

The location id of the template resource.

To set the location attribute:

• provide the argument template on the command line with a fully specified name;
• provide the argument --location on the command line.

Filters configuration.

At least one of these must be specified:

Malicious URI filter settings.

--malicious-uri-filter-settings-enforcement=MALICIOUS_URI_FILTER_SETTINGS_ENFORCEMENT

Tells whether the Malicious URI filter is enabled or disabled. MALICIOUS_URI_FILTER_SETTINGS_ENFORCEMENT must be one of:

disabled

Disabled

enabled

Enabled

Responsible AI Filter settings.

--rai-settings-filters=[confidenceLevel=CONFIDENCELEVEL],[filterType=FILTERTYPE]

Required, List of Responsible AI filters enabled for template.

confidenceLevel

Confidence level for this RAI filter. During data sanitization, if data is classified under this filter with a confidence level equal to or greater than the specified level, a positive match is reported. If the confidence level is unspecified (i.e., 0), the system will use a reasonable default level based on the filter_type.

filterType

Type of responsible AI filter.

Shorthand Example:

--rai-settings-filters=confidenceLevel=string,filterType=string --rai-settings-filters=confidenceLevel=string,filterType=string

JSON Example:

--rai-settings-filters='[{"confidenceLevel": "string", "filterType": "string"}]'

File Example:

--rai-settings-filters=path_to_file.(yaml|json)

Sensitive Data Protection settings.

Arguments for the sdp configuration.

At most one of these can be specified:

Sensitive Data Protection basic configuration.

--basic-config-filter-enforcement=BASIC_CONFIG_FILTER_ENFORCEMENT

Tells whether the Sensitive Data Protection basic config is enabled or disabled. BASIC_CONFIG_FILTER_ENFORCEMENT must be one of:

disabled

Disabled

enabled

Enabled

Sensitive Data Protection Advanced configuration.

--advanced-config-deidentify-template=ADVANCED_CONFIG_DEIDENTIFY_TEMPLATE

Optional Sensitive Data Protection Deidentify template resource name.

If provided then DeidentifyContent action is performed during Sanitization using this template and inspect template. The De-identified data will be returned in SdpDeidentifyResult. Note that all info-types present in the deidentify template must be present in inspect template.

e.g. organizations/{organization}/deidentifyTemplates/{deidentify_template}, projects/{project}/deidentifyTemplates/{deidentify_template} organizations/{organization}/locations/{location}/deidentifyTemplates/{deidentify_template} projects/{project}/locations/{location}/deidentifyTemplates/{deidentify_template}

--advanced-config-inspect-template=ADVANCED_CONFIG_INSPECT_TEMPLATE

Sensitive Data Protection inspect template resource name

If only inspect template is provided (de-identify template not provided), then Sensitive Data Protection InspectContent action is performed during Sanitization. All Sensitive Data Protection findings identified during inspection will be returned as SdpFinding in SdpInsepctionResult e.g. organizations/{organization}/inspectTemplates/{inspect_template}, projects/{project}/inspectTemplates/{inspect_template} organizations/{organization}/locations/{location}/inspectTemplates/{inspect_template} projects/{project}/locations/{location}/inspectTemplates/{inspect_template}

Prompt injection and Jailbreak Filter settings.

--pi-and-jailbreak-filter-settings-confidence-level=PI_AND_JAILBREAK_FILTER_SETTINGS_CONFIDENCE_LEVEL

Confidence level for this filter. Confidence level is used to determine the threshold for the filter. If detection confidence is equal to or greater than the specified level, a positive match is reported. Confidence level will only be used if the filter is enabled. PI_AND_JAILBREAK_FILTER_SETTINGS_CONFIDENCE_LEVEL must be one of:

high

Low chance of false positives.

low-and-above

Highest chance of a false positive.

medium-and-above

Some chance of false positives.

--pi-and-jailbreak-filter-settings-enforcement=PI_AND_JAILBREAK_FILTER_SETTINGS_ENFORCEMENT

Tells whether Prompt injection and Jailbreak filter is enabled or disabled. PI_AND_JAILBREAK_FILTER_SETTINGS_ENFORCEMENT must be one of:

disabled

Enabled

enabled

Enabled

--labels=[LABELS,…]

Labels as key value pairs.

KEY

Keys must start with a lowercase character and contain only hyphens (-), underscores (_), lowercase characters, and numbers.

VALUE

Values must contain only hyphens (-), underscores (_), lowercase characters, and numbers.

Shorthand Example:

--labels=string=string

JSON Example:

--labels='{"string": "string"}'

File Example:

--labels=path_to_file.(yaml|json)

--request-id=REQUEST_ID

An optional request ID to identify requests. Specify a unique request ID so that if you must retry your request, the server will know to ignore the request if it has already been completed. The server stores the request ID for 60 minutes after the first request.

For example, consider a situation where you make an initial request and the request times out. If you make the request again with the same request ID, the server can check if original operation with the same request ID was received, and if so, will ignore the second request. This prevents clients from accidentally creating duplicate commitments.

The request ID must be a valid UUID with the exception that zero UUID is not supported (00000000-0000-0000-0000-000000000000).

Message describing TemplateMetadata

--template-metadata-custom-llm-response-safety-error-code=TEMPLATE_METADATA_CUSTOM_LLM_RESPONSE_SAFETY_ERROR_CODE

Indicates the custom error code set by the user to be returned to the end user if the LLM response trips Model Armor filters.

--template-metadata-custom-llm-response-safety-error-message=TEMPLATE_METADATA_CUSTOM_LLM_RESPONSE_SAFETY_ERROR_MESSAGE

Indicates the custom error message set by the user to be returned to the end user if the LLM response trips Model Armor filters.

--template-metadata-custom-prompt-safety-error-code=TEMPLATE_METADATA_CUSTOM_PROMPT_SAFETY_ERROR_CODE

Indicates the custom error code set by the user to be returned to the end user by the service extension if the prompt trips Model Armor filters.

--template-metadata-custom-prompt-safety-error-message=TEMPLATE_METADATA_CUSTOM_PROMPT_SAFETY_ERROR_MESSAGE

Indicates the custom error message set by the user to be returned to the end user if the prompt trips Model Armor filters.

--template-metadata-ignore-partial-invocation-failures

If true, partial detector failures should be ignored.

--template-metadata-log-operations

If true, log template crud operations.

--template-metadata-log-sanitize-operations

If true, log sanitize operations.

Run $ gcloud help for details.