Method●Since v2.0

invoke

Sanitize an LLM response using Model Armor.

invoke(
  self,
  input: T,
  config: Optional[RunnableConfig] = None,
  fail_open: Optional[bool] = None,
  **kwargs: Any = {}
) -> T

Parameters

Name	Type	Description
`input`*	`T`	The model response to sanitize. Can be `str`, `BaseMessage`, list of messages, or any object with string conversion.
`config`	`Optional[RunnableConfig]`	Default:`None` A config to use when invoking the `Runnable`.
`fail_open`	`Optional[bool]`	Default:`None` If `True`, allows unsafe responses to pass through.