Overview
Playground
The validator playground is available to authenticated users. Please log in to use it.
log inThis validator checks that a value is similar to a list of previously known correct values.
For example, let’s say you’re extracting structured data from a PDF document, and extract some value. If you have an existing golden dataset of previous values, then this validator will ensure that the extracted value is not too different from known good values.
This validator works on numerical and string types in the following manner:
- For numbers, this validator checks that the extracted value is within
kstandard deviations of the validator. - For strings, this validator embeds the extracted value, generates embeddings for all reference values, and checks that the average semantic similarity is more than some threshold.
- Dependencies:
- guardrails-ai>=0.4.0
$ guardrails hub install hub://guardrails/similar_to_previous_values
In this example, we apply the validator to a string output generated by an LLM.
# Import Guard and Validator
from guardrails.hub import SimilarToPreviousValues
from guardrails import Guard
import numpy as np
import os
from typing import List, Union
try:
import cohere
except ImportError:
raise ImportError(
"This example requires the `cohere` package. "
"Install it with `pip install cohere`, and try again."
)
# Create a cohere client
cohere_key = os.environ["COHERE_API_KEY"]
cohere_client = cohere.Client(api_key=cohere_key)
def embed_function(text: Union[str, List[str]]) -> np.ndarray:
"""Embed the text using cohere's small model."""
# If text is a string, wrap it in a list
if isinstance(text, str):
text = [text]
response = cohere_client.embed(
model="embed-english-light-v2.0",
texts=text,
)
embeddings_list = response.embeddings
return np.array(embeddings_list)
# Use the Guard with the validator
guard = Guard().use(
SimilarToPreviousValues,
threshold=0.6, # Increase the threshold to make the validator stricter
on_fail="exception",
)
# Test passing response
guard.validate(
"""
You are so amazing!
""",
metadata={
"prev_values": ["You are amazing", "You are awesome.", "You are great!"],
"embed_function": embed_function,
},
)
try:
# Test failing response
guard.validate(
"""
Why don't you go to hell?
""",
metadata={
"prev_values": ["You are amazing", "You are awesome.", "You are great!"],
"embed_function": embed_function,
},
)
except Exception as e:
print(e)
Output:
Validation failed for field with errors: The value
- Why don't you go to hell?
is not semantically similar to the previous values. Avg. similarity: 0.24 < Threshold: 0.6.
__init__(self, standard_deviations=3, threshold=0.3, on_fail="noop")
Initializes a new instance of the Validator class.
Parameters
standard_deviations(int): Max number of standard deviations that the extracted value should be within. Required for numbers. Defaults to 3.threshold(float): Average similarity threshold below which the validator will fail. Required for strings. Defaults to 0.8.on_fail(str, Callable): The policy to enact when a validator fails. Ifstr, must be one ofreask,fix,filter,refrain,noop,exceptionorfix_reask. Otherwise, must be a function that is called when the validator fails.
__call__(self, value, metadata={}) -> ValidationResult
Validates the given value using the rules defined in this validator, relying on the metadata provided to customize the validation process. This method is automatically invoked by guard.parse(...), ensuring the validation logic is applied to the input data.
Note:
- This method should not be called directly by the user. Instead, invoke
guard.parse(...)where this method will be called internally for each associated Validator. - When invoking
guard.parse(...), ensure to pass the appropriatemetadatadictionary that includes keys and values required by this validator. Ifguardis associated with multiple validators, combine all necessary metadata into a single dictionary.
Parameters
-
value(Any): The input value to validate. -
metadata(dict): A dictionary containing metadata required for validation. Keys and values must match the expectations of this validator.Key Type Description Default prev_valslist List of previous values to pass to the validator N/A embed_functionCallable Function to embed the input text sentence-transformer's paraphrase-MiniLM-L6-v2