API Reference

Main Validation Class

Configuration

Validation Report

class AisertReport(status: bool, rules: dict)

Final validation report containing overall status and detailed results.

Returned by Aisert.collect to provide comprehensive validation outcomes. Contains both high-level pass/fail status and granular per-validation details.

Parameters:

• status (bool) – True if all validations passed, False if any failed

• rules (dict) – Dictionary mapping execution order to validation results

The rules dictionary has the format:

{1: {'validator': 'ContainsValidator', 'status': True, 'reason': '...'}}

Example usage:

report = Aisert("Hello world").assert_contains(["Hello"]).collect()
if report.status:
    print("All validations passed!")
for order, result in report.rules.items():
    print(f"{order}: {result['validator']} - {result['status']}")

Added in version 0.1.0.

__init__(status: bool, rules: dict)

Create a new validation report.

Parameters:

• status (bool) – Overall validation status (True if all passed)

• rules (dict) – Dictionary of validation results keyed by execution order

Exceptions

exception AisertError

Base exception for all Aisert errors.

exception SchemaValidationError

Schema validation specific errors.

exception ContainsValidationError

Text contain validation errors

exception TokenValidationError

Token counting specific errors.

exception SemanticValidationError

Semantic validation specific errors.

Custom Validator Base Classes

Base classes for creating custom validators.

Token Validator Base

class TokenValidatorBase

Abstract base class for token counting implementations across different AI providers.

Provides a common interface for token counting while allowing provider-specific implementations (OpenAI, Anthropic, HuggingFace, Google). Each provider has different tokenization methods and APIs.

Subclasses must implement: - get_instance(): Factory method for singleton pattern - count(): Token counting logic for the specific provider

Example

class MyTokenValidator(TokenValidatorBase):

def count(self, text: str) -> int:

return len(text.split()) # Simple word count

__init__()

Initialize base token validator.

classmethod get_instance(**kwargs)

Factory method to get validator instance (typically singleton).

Should be implemented by subclasses to return cached instances for performance, as tokenizers can be expensive to initialize.

Parameters:

**kwargs – Provider-specific configuration parameters

Returns:

Instance of the token validator

Raises:

NotImplementedError – If subclass doesn’t implement this method

count(text: str) → int

Count tokens in the provided text using provider-specific logic.

Parameters:

text – Input text to count tokens for

Returns:

Number of tokens in the text

Raises:

NotImplementedError – If subclass doesn’t implement this method

Example

validator = OpenAITokenValidator.get_instance(token_model=”gpt-4”) count = validator.count(“Hello world”) # Returns token count

Semantic Validator Base

Validator Factories

Factories for registering and managing custom validators.

Token Validator Factory

class TokenValidatorFactory

Factory class for creating instances of token validators based on the model provider. This class provides a method to get an instance of a specific token validator based on the model provider and token model.

classmethod register_token_validator(model_provider: str, token_validator_class)

A class method ot register custom token validators :param model_provider: provider of the model :param token_validator_class: custom token validator class

static get_instance(model_provider: str, token_model: str = None, token_encoding: str = None, api_key: str = None, **kwargs)

Get an instance of TokenValidator with the specified provider and model. :param model_provider: Provider name :param token_model: Model name :param token_encoding: Encoding method :param api_key: API key if needed :return: An instance of TokenValidator.

Semantic Validator Factory