Passing metadata to the Validators

Why do some validators require metadata?

Some validators require additional metadata at runtime to perform their validation. For example, the Provenance_LLM validator at runtime requires the filepaths key in the metadata. This is because the Provenance_LLM checks for hallucinations in RAG workflows by comparing the generated output to the source documents. The filepaths key in the metadata is used to specify which documents to use for the comparison.

Passing metadata for a single validator

In order to pass metadata to a validator, you can pass the metadata dictionary to the guard.validate method. For example, below is an example for passing metadata to the Provenance_LLM validator.

First, we set up the validator and the guard.

# Import Guard and Validator
from guardrails.hub import ProvenanceLLM
from guardrails import Guard

# Import embedding model
from sentence_transformers import SentenceTransformer

# Initialize Validator
val = ProvenanceLLM(
    validation_method="sentence",
    llm_callable="gpt-3.5-turbo",
    top_k=3,
    max_tokens=2,
)

# Setup Guard
guard = Guard.for_string(validators=[val])

Below, we show how to pass the metadata to the guard.validate method.

# Setup text sources
sources = [
    "The sun is a star.",
    "The sun rises in the east and sets in the west."
]

# Load model for embedding function
model = SentenceTransformer('paraphrase-MiniLM-L6-v2')

# Create embed function
def embed_function(sources: list[str]) -> np.array:
    return model.encode(sources)

guard.validate(
    llm_output="The sun rises in the east.",
    metadata={
        "sources": sources,
        "embed_function": embed_function
    }
)

Passing metadata for multiple validators

If multiple validators require metadata, create a single metadata dictionary that contains the metadata keys for each validator. In the example below, both the Provenance_LLM and DetectPII validators require metadata.

from guardrails import Guard
from guardrails.hub import DetectPII, ProvenanceLLM

from sentence_transformers import SentenceTransformer


# Setup Guard with multiple validators
guard = Guard().use_many(
    ProvenanceLLM(validation_method="sentence"),
    DetectPII()
)

# Setup metadata for provenance validator
sources = [
    "The sun is a star.",
    "The sun rises in the east and sets in the west."
]
model = SentenceTransformer('paraphrase-MiniLM-L6-v2')

def embed_function(sources: list[str]) -> np.array:
    return model.encode(sources)

# Setup metadata for PII validator
pii_entities = ["EMAIL_ADDRESS", "PHONE_NUMBER"]

# Create a single metadata dictionary containing metadata keys for each validator
metadata = {
    'pii_entities': pii_entities,
    'sources': sources,
    'embed_function': embed_function
}

# Pass the metadata to the guard.validate method
guard.validate("some text", metadata=metadata)