ReferencePython SDK ReferenceNemoguardrailsLibraryGuardrails Ai

nemoguardrails.library.guardrails_ai.actions

View as Markdown

Dynamic validator loading for Guardrails AI integration.

Module Contents

Classes

NameDescription
Guard-

Functions

NameDescription
_get_guardGet or create a Guard instance for a validator.
_load_validator_classDynamically load a validator class.
guardrails_ai_validation_mappingMap Guardrails AI validation result to NeMo Guardrails format.
validate_guardrails_aiUnified action for all Guardrails AI validators.
validate_guardrails_ai_inputUnified action for all Guardrails AI validators.
validate_guardrails_ai_outputUnified action for all Guardrails AI validators.

Data

_guard_cache

_validator_class_cache

log

API

class nemoguardrails.library.guardrails_ai.actions.Guard()
 
nemoguardrails.library.guardrails_ai.actions.Guard.use(
    validator
)
nemoguardrails.library.guardrails_ai.actions.Guard.validate(
    text,
    metadata = None
)
nemoguardrails.library.guardrails_ai.actions._get_guard(
    validator_name: str,
    validator_params = {}
) -> nemoguardrails.library.guardrails_ai.actions.Guard

Get or create a Guard instance for a validator.

nemoguardrails.library.guardrails_ai.actions._load_validator_class(
    validator_name: str
) -> typing.Type

Dynamically load a validator class.

nemoguardrails.library.guardrails_ai.actions.guardrails_ai_validation_mapping(
    result: typing.Dict[str, typing.Any]
) -> bool

Map Guardrails AI validation result to NeMo Guardrails format.

nemoguardrails.library.guardrails_ai.actions.validate_guardrails_ai(
    validator_name: str,
    text: str,
    kwargs = {}
) -> typing.Dict[str, typing.Any]

Unified action for all Guardrails AI validators.

Parameters:

validator

Name of the validator to use (from VALIDATOR_REGISTRY)

text
str

Text to validate

Returns: Dict[str, Any]

Dict with validation_result

nemoguardrails.library.guardrails_ai.actions.validate_guardrails_ai_input(
    validator: str,
    config: nemoguardrails.rails.llm.config.RailsConfig,
    context: typing.Optional[dict] = None,
    text: typing.Optional[str] = None,
    kwargs = {}
) -> typing.Dict[str, typing.Any]

Unified action for all Guardrails AI validators.

Parameters:

validator
str

Name of the validator to use (from VALIDATOR_REGISTRY)

text
Optional[str]Defaults to None

Text to validate

context
Optional[dict]Defaults to None

Optional context dictionary

Returns: Dict[str, Any]

Dict with validation_result and valid (bool derived from validation_passed).

nemoguardrails.library.guardrails_ai.actions.validate_guardrails_ai_output(
    validator: str,
    context: typing.Optional[dict] = None,
    text: typing.Optional[str] = None,
    config: typing.Optional[nemoguardrails.rails.llm.config.RailsConfig] = None,
    kwargs = {}
) -> typing.Dict[str, typing.Any]

Unified action for all Guardrails AI validators.

Parameters:

validator
str

Name of the validator to use (from VALIDATOR_REGISTRY)

text
Optional[str]Defaults to None

Text to validate

context
Optional[dict]Defaults to None

Optional context dictionary

Returns: Dict[str, Any]

Dict with validation_result and valid (bool derived from validation_passed).

nemoguardrails.library.guardrails_ai.actions._guard_cache: Dict[tuple, Guard] = {}
 
nemoguardrails.library.guardrails_ai.actions._validator_class_cache: Dict[str, Type] = {}
 
nemoguardrails.library.guardrails_ai.actions.log = logging.getLogger(__name__)