guardrails

Creates, updates, deletes, gets or lists a guardrails resource.

Overview

Name	guardrails
Type	Resource
Id	google.ces.guardrails

Fields

The following fields are returned by SELECT queries:

get

Name	Datatype	Description
name	string	Identifier. The unique identifier of the guardrail. Format: projects/{project}/locations/{location}/apps/{app}/guardrails/{guardrail}
action	object	Optional. Action to take when the guardrail is triggered. (id: TriggerAction)
codeCallback	object	Optional. Guardrail that potentially blocks the conversation based on the result of the callback execution. (id: GuardrailCodeCallback)
contentFilter	object	Optional. Guardrail that bans certain content from being used in the conversation. (id: GuardrailContentFilter)
createTime	string (google-datetime)	Output only. Timestamp when the guardrail was created.
description	string	Optional. Description of the guardrail.
displayName	string	Required. Display name of the guardrail.
enabled	boolean	Optional. Whether the guardrail is enabled.
etag	string	Etag used to ensure the object hasn't changed during a read-modify-write operation. If the etag is empty, the update will overwrite any concurrent changes.
llmPolicy	object	Optional. Guardrail that blocks the conversation if the LLM response is considered violating the policy based on the LLM classification. (id: GuardrailLlmPolicy)
llmPromptSecurity	object	Optional. Guardrail that blocks the conversation if the prompt is considered unsafe based on the LLM classification. (id: GuardrailLlmPromptSecurity)
modelSafety	object	Optional. Guardrail that blocks the conversation if the LLM response is considered unsafe based on the model safety settings. (id: GuardrailModelSafety)
updateTime	string (google-datetime)	Output only. Timestamp when the guardrail was last updated.

list

Name	Datatype	Description
name	string	Identifier. The unique identifier of the guardrail. Format: projects/{project}/locations/{location}/apps/{app}/guardrails/{guardrail}
action	object	Optional. Action to take when the guardrail is triggered. (id: TriggerAction)
codeCallback	object	Optional. Guardrail that potentially blocks the conversation based on the result of the callback execution. (id: GuardrailCodeCallback)
contentFilter	object	Optional. Guardrail that bans certain content from being used in the conversation. (id: GuardrailContentFilter)
createTime	string (google-datetime)	Output only. Timestamp when the guardrail was created.
description	string	Optional. Description of the guardrail.
displayName	string	Required. Display name of the guardrail.
enabled	boolean	Optional. Whether the guardrail is enabled.
etag	string	Etag used to ensure the object hasn't changed during a read-modify-write operation. If the etag is empty, the update will overwrite any concurrent changes.
llmPolicy	object	Optional. Guardrail that blocks the conversation if the LLM response is considered violating the policy based on the LLM classification. (id: GuardrailLlmPolicy)
llmPromptSecurity	object	Optional. Guardrail that blocks the conversation if the prompt is considered unsafe based on the LLM classification. (id: GuardrailLlmPromptSecurity)
modelSafety	object	Optional. Guardrail that blocks the conversation if the LLM response is considered unsafe based on the model safety settings. (id: GuardrailModelSafety)
updateTime	string (google-datetime)	Output only. Timestamp when the guardrail was last updated.

Methods

The following methods are available for this resource:

Name	Accessible by	Required Params	Optional Params	Description
get	select	projectsId, locationsId, appsId, guardrailsId		Gets details of the specified guardrail.
list	select	projectsId, locationsId, appsId	pageSize, pageToken, filter, orderBy	Lists guardrails in the given app.
create	insert	projectsId, locationsId, appsId	guardrailId	Creates a new guardrail in the given app.
patch	update	projectsId, locationsId, appsId, guardrailsId	updateMask	Updates the specified guardrail.
delete	delete	projectsId, locationsId, appsId, guardrailsId	force, etag	Deletes the specified guardrail.

Parameters

Parameters can be passed in the WHERE clause of a query. Check the Methods section to see which parameters are required or optional for each operation.

Name	Datatype	Description
appsId	string
guardrailsId	string
locationsId	string
projectsId	string
etag	string
filter	string
force	boolean
guardrailId	string
orderBy	string
pageSize	integer (int32)
pageToken	string
updateMask	string (google-fieldmask)

SELECT examples

get

Gets details of the specified guardrail.

SELECT
name,
action,
codeCallback,
contentFilter,
createTime,
description,
displayName,
enabled,
etag,
llmPolicy,
llmPromptSecurity,
modelSafety,
updateTime
FROM google.ces.guardrails
WHERE projectsId = '{{ projectsId }}' -- required
AND locationsId = '{{ locationsId }}' -- required
AND appsId = '{{ appsId }}' -- required
AND guardrailsId = '{{ guardrailsId }}' -- required
;

list

Lists guardrails in the given app.

SELECT
name,
action,
codeCallback,
contentFilter,
createTime,
description,
displayName,
enabled,
etag,
llmPolicy,
llmPromptSecurity,
modelSafety,
updateTime
FROM google.ces.guardrails
WHERE projectsId = '{{ projectsId }}' -- required
AND locationsId = '{{ locationsId }}' -- required
AND appsId = '{{ appsId }}' -- required
AND pageSize = '{{ pageSize }}'
AND pageToken = '{{ pageToken }}'
AND filter = '{{ filter }}'
AND orderBy = '{{ orderBy }}'
;

INSERT examples

create

Creates a new guardrail in the given app.

INSERT INTO google.ces.guardrails (
data__description,
data__codeCallback,
data__etag,
data__llmPolicy,
data__enabled,
data__llmPromptSecurity,
data__name,
data__displayName,
data__contentFilter,
data__modelSafety,
data__action,
projectsId,
locationsId,
appsId,
guardrailId
)
SELECT 
'{{ description }}',
'{{ codeCallback }}',
'{{ etag }}',
'{{ llmPolicy }}',
{{ enabled }},
'{{ llmPromptSecurity }}',
'{{ name }}',
'{{ displayName }}',
'{{ contentFilter }}',
'{{ modelSafety }}',
'{{ action }}',
'{{ projectsId }}',
'{{ locationsId }}',
'{{ appsId }}',
'{{ guardrailId }}'
RETURNING
name,
action,
codeCallback,
contentFilter,
createTime,
description,
displayName,
enabled,
etag,
llmPolicy,
llmPromptSecurity,
modelSafety,
updateTime
;

Manifest

# Description fields are for documentation purposes
- name: guardrails
props:
  - name: projectsId
    value: "{{ projectsId }}"
    description: Required parameter for the guardrails resource.
  - name: locationsId
    value: "{{ locationsId }}"
    description: Required parameter for the guardrails resource.
  - name: appsId
    value: "{{ appsId }}"
    description: Required parameter for the guardrails resource.
  - name: description
    value: "{{ description }}"
    description: |
      Optional. Description of the guardrail.
  - name: codeCallback
    description: |
      Optional. Guardrail that potentially blocks the conversation based on the result of the callback execution.
    value:
      beforeAgentCallback:
        pythonCode: "{{ pythonCode }}"
        disabled: {{ disabled }}
        description: "{{ description }}"
        proactiveExecutionEnabled: {{ proactiveExecutionEnabled }}
      afterModelCallback:
        pythonCode: "{{ pythonCode }}"
        disabled: {{ disabled }}
        description: "{{ description }}"
        proactiveExecutionEnabled: {{ proactiveExecutionEnabled }}
      beforeModelCallback:
        pythonCode: "{{ pythonCode }}"
        disabled: {{ disabled }}
        description: "{{ description }}"
        proactiveExecutionEnabled: {{ proactiveExecutionEnabled }}
      afterAgentCallback:
        pythonCode: "{{ pythonCode }}"
        disabled: {{ disabled }}
        description: "{{ description }}"
        proactiveExecutionEnabled: {{ proactiveExecutionEnabled }}
  - name: etag
    value: "{{ etag }}"
    description: |
      Etag used to ensure the object hasn't changed during a read-modify-write operation. If the etag is empty, the update will overwrite any concurrent changes.
  - name: llmPolicy
    description: |
      Optional. Guardrail that blocks the conversation if the LLM response is considered violating the policy based on the LLM classification.
    value:
      policyScope: "{{ policyScope }}"
      maxConversationMessages: {{ maxConversationMessages }}
      modelSettings:
        model: "{{ model }}"
        temperature: {{ temperature }}
      failOpen: {{ failOpen }}
      allowShortUtterance: {{ allowShortUtterance }}
      prompt: "{{ prompt }}"
  - name: enabled
    value: {{ enabled }}
    description: |
      Optional. Whether the guardrail is enabled.
  - name: llmPromptSecurity
    description: |
      Optional. Guardrail that blocks the conversation if the prompt is considered unsafe based on the LLM classification.
    value:
      defaultSettings:
        defaultPromptTemplate: "{{ defaultPromptTemplate }}"
      failOpen: {{ failOpen }}
      customPolicy:
        policyScope: "{{ policyScope }}"
        maxConversationMessages: {{ maxConversationMessages }}
        modelSettings:
          model: "{{ model }}"
          temperature: {{ temperature }}
        failOpen: {{ failOpen }}
        allowShortUtterance: {{ allowShortUtterance }}
        prompt: "{{ prompt }}"
  - name: name
    value: "{{ name }}"
    description: |
      Identifier. The unique identifier of the guardrail. Format: `projects/{project}/locations/{location}/apps/{app}/guardrails/{guardrail}`
  - name: displayName
    value: "{{ displayName }}"
    description: |
      Required. Display name of the guardrail.
  - name: contentFilter
    description: |
      Optional. Guardrail that bans certain content from being used in the conversation.
    value:
      matchType: "{{ matchType }}"
      disregardDiacritics: {{ disregardDiacritics }}
      bannedContentsInAgentResponse:
        - "{{ bannedContentsInAgentResponse }}"
      bannedContentsInUserInput:
        - "{{ bannedContentsInUserInput }}"
      bannedContents:
        - "{{ bannedContents }}"
  - name: modelSafety
    description: |
      Optional. Guardrail that blocks the conversation if the LLM response is considered unsafe based on the model safety settings.
    value:
      safetySettings:
        - category: "{{ category }}"
          threshold: "{{ threshold }}"
  - name: action
    description: |
      Optional. Action to take when the guardrail is triggered.
    value:
      generativeAnswer:
        prompt: "{{ prompt }}"
      respondImmediately:
        responses:
          - disabled: {{ disabled }}
            text: "{{ text }}"
      transferAgent:
        agent: "{{ agent }}"
  - name: guardrailId
    value: "{{ guardrailId }}"

UPDATE examples

patch

Updates the specified guardrail.

UPDATE google.ces.guardrails
SET 
data__description = '{{ description }}',
data__codeCallback = '{{ codeCallback }}',
data__etag = '{{ etag }}',
data__llmPolicy = '{{ llmPolicy }}',
data__enabled = {{ enabled }},
data__llmPromptSecurity = '{{ llmPromptSecurity }}',
data__name = '{{ name }}',
data__displayName = '{{ displayName }}',
data__contentFilter = '{{ contentFilter }}',
data__modelSafety = '{{ modelSafety }}',
data__action = '{{ action }}'
WHERE 
projectsId = '{{ projectsId }}' --required
AND locationsId = '{{ locationsId }}' --required
AND appsId = '{{ appsId }}' --required
AND guardrailsId = '{{ guardrailsId }}' --required
AND updateMask = '{{ updateMask}}'
RETURNING
name,
action,
codeCallback,
contentFilter,
createTime,
description,
displayName,
enabled,
etag,
llmPolicy,
llmPromptSecurity,
modelSafety,
updateTime;

DELETE examples

delete

Deletes the specified guardrail.

DELETE FROM google.ces.guardrails
WHERE projectsId = '{{ projectsId }}' --required
AND locationsId = '{{ locationsId }}' --required
AND appsId = '{{ appsId }}' --required
AND guardrailsId = '{{ guardrailsId }}' --required
AND force = '{{ force }}'
AND etag = '{{ etag }}'
;