nemoguardrails.library.llama_guard.actions

Module Contents

Functions

Name	Description
`_get_llama_guard_model`	-
`_llama_guard_outcome`	-
`llama_guard_check_input`	Checks user messages using the configured Llama Guard model
`llama_guard_check_output`	Check the bot response using the configured Llama Guard model
`parse_llama_guard_response`	Parses the response from the Llama Guard LLM and returns a tuple of:

Data

log

API

nemoguardrails.library.llama_guard.actions._get_llama_guard_model(
    llms: collections.abc.Mapping[str, nemoguardrails.types.LLMModel],
    model_name: str
) -> nemoguardrails.types.LLMModel

nemoguardrails.library.llama_guard.actions._llama_guard_outcome(
    allowed: bool,
    policy_violations: typing.Optional[typing.List[str]]
) -> nemoguardrails.actions.rail_outcome.RailOutcome

nemoguardrails.library.llama_guard.actions.llama_guard_check_input(
    llm_task_manager: nemoguardrails.llm.taskmanager.LLMTaskManager,
    llms: collections.abc.Mapping[str, nemoguardrails.types.LLMModel],
    model_name: str,
    context: typing.Optional[dict] = None,
    user_message: typing.Optional[str] = None,
    kwargs = {}
) -> nemoguardrails.actions.rail_outcome.RailOutcome

async

Checks user messages using the configured Llama Guard model and the configured prompt containing the safety guidelines.

nemoguardrails.library.llama_guard.actions.llama_guard_check_output(
    llm_task_manager: nemoguardrails.llm.taskmanager.LLMTaskManager,
    llms: collections.abc.Mapping[str, nemoguardrails.types.LLMModel],
    model_name: str,
    context: typing.Optional[dict] = None,
    user_message: typing.Optional[str] = None,
    bot_message: typing.Optional[str] = None
) -> nemoguardrails.actions.rail_outcome.RailOutcome

async

Check the bot response using the configured Llama Guard model and the configured prompt containing the safety guidelines.

nemoguardrails.library.llama_guard.actions.parse_llama_guard_response(
    response: str
) -> typing.Tuple[bool, typing.Optional[typing.List[str]]]

Parses the response from the Llama Guard LLM and returns a tuple of:

Whether the response is safe or not.
If not safe, a list of the violated policies.

nemoguardrails.library.llama_guard.actions.log = logging.getLogger(__name__)