1  """Guardrails for validating and filtering LLM outputs.
 2  
 3  Guardrails provide runtime checks to ensure LLM outputs meet specific criteria
 4  before they are used. They can block, filter, or transform outputs based on
 5  custom rules.
 6  
 7  Available guards:
 8  - `PIICheck`: Detect personally identifiable information
 9  - `IncludesWords`: Check if output contains specific words
10  - `WordsPresence`: Check for presence/absence of word lists
11  - `PythonFunction`: Custom validation using Python functions
12  
13  Use the `@guard` decorator to apply guards to LLM functions.
14  
15  Example:
16  ```python
17  from evidently.guardrails import guard, PIICheck
18  
19  @guard(PIICheck())
20  def generate_response(prompt: str) -> str:
21      # LLM generation code
22      pass
23  ```
24  """
25  
26  from .core import GuardException
27  from .decorators import guard
28  from .guards import IncludesWords
29  from .guards import PIICheck
30  from .guards import PythonFunction
31  from .guards import WordsPresence
32  
33  __all__ = ["guard", "PIICheck", "PythonFunction", "IncludesWords", "WordsPresence", "GuardException"]