Guardrail

`Guardrail`

Bases: BaseModel

Represents a guardrail definition.

A guardrail is a security or content moderation mechanism that can be applied to pipelines, agents, and other entities. Guardrails contain filters for detecting specific content types and assignments that determine where they are applied.

Attributes:

Name	Type	Description
`id`	`str`	The unique identifier of the guardrail
`name`	`str`	The friendly display name of the guardrail
`description`	`str`	A description of what the guardrail does
`enabled`	`bool`	Whether the guardrail is currently enabled
`apply_datasource`	`bool`	Whether the guardrail should be applied to datasources
`apply_to_tools`	`bool`	Whether the guardrail should be applied to tool responses
`filters`	`List[GuardrailFilter]`	List of filters that define what the guardrail detects
`guardrail_assignments`	`List[GuardrailAssignment]`	List of entity assignments for the guardrail
`guardrail_targets`	`List[GuardrailTarget]`	List of target configurations for the guardrail

`GuardrailAssignment`

Bases: BaseModel

Represents a guardrail assignment definition.

Assignments define which entities (tenants, projects, agents, etc.) a guardrail is applied to. This allows for granular control over guardrail application across the organization hierarchy.

Attributes:

Name	Type	Description
`entity_type`	`str`	The type of entity (e.g., Tenant, Project, Agent, Model, Datasource)
`entity_id`	`str`	The unique identifier of the assigned entity
`project_id`	`Optional[str]`	Optional project identifier if the entity belongs to a project

`GuardrailFilter`

Bases: BaseModel

Represents a base filter definition within a guardrail.

Filters define the specific detection or processing rules that a guardrail applies to content. Each filter has a type indicating what kind of content it processes (e.g., PII, PCI, moderation).

Attributes:

Name	Type	Description
`enabled`	`bool`	Whether the filter is currently enabled
`guardrail_id`	`str`	The unique identifier of the parent guardrail
`tenant_id`	`str`	The tenant identifier this filter belongs to
`filter_type`	`str`	The type of filter (e.g., Pci, Pii, Phi, Moderation, PromptInjection)

`GuardrailTarget`

Bases: BaseModel

Guardrail target definition.

Targets specify how guardrails are applied to different entities, including whether they apply to input, output, or both, and whether they represent inclusions or exclusions.

Attributes:

Name	Type	Description
`entity_type`	`str`	The type of entity (e.g., Tenant, Project, Agent, Model)
`target_type`	`str`	Where the guardrail applies (None, Input, Output, Both)
`is_exception`	`bool`	Whether assignments represent exclusions rather than inclusions. When true, the guardrail applies to all entities EXCEPT those in the assignment list. When false (default), the guardrail applies ONLY to entities in the assignment list.