Skip to main content

databricks.labs.dqx.llm.llm_core

LLMModelConfigurator Objects​

class LLMModelConfigurator()

Configures DSPy language models.

__init__​

def __init__(model_config: LLMModelConfig)

Initialize model configurator.

Arguments:

  • model_config - Configuration for the LLM model.

configure​

def configure() -> None

Configure the DSPy language model with the provided settings.

DspySchemaGuesserSignature Objects​

class DspySchemaGuesserSignature(dspy.Signature)

Guess a table schema based on business description.

DspySchemaGuesser Objects​

class DspySchemaGuesser(dspy.Module)

Guess table schema from business description.

forward​

def forward(
        business_description: str) -> dspy.primitives.prediction.Prediction

Guess schema based on business description.

Arguments:

  • business_description - Natural language description of the dataset.

Returns:

Prediction containing guessed schema and assumptions.

DspyRuleSignature Objects​

class DspyRuleSignature(dspy.Signature)

Generate data quality rules with improved output format.

DspyRuleGeneration Objects​

class DspyRuleGeneration(dspy.Module)

Generate data quality rules.

Now focused solely on rule generation, with schema inference delegated.

forward​

def forward(schema_info: str, business_description: str,
            available_functions: str) -> dspy.primitives.prediction.Prediction

Generate data quality rules.

Arguments:

  • schema_info - JSON string containing table schema.
  • business_description - Natural language description of requirements.
  • available_functions - JSON string of available check functions.

Returns:

Prediction containing quality_rules and reasoning.

DspyRuleGenerationWithSchemaInference Objects​

class DspyRuleGenerationWithSchemaInference(dspy.Module)

Combines schema inference and rule generation.

Follows Dependency Inversion Principle by depending on abstractions (protocols).

forward​

def forward(schema_info: str, business_description: str,
            available_functions: str) -> dspy.primitives.prediction.Prediction

Generate rules with optional schema inference.

Arguments:

  • schema_info - JSON string of schema (can be empty to trigger inference).
  • business_description - Natural language requirements.
  • available_functions - JSON string of available functions.

Returns:

Prediction with quality_rules, reasoning, and optional schema info.

LLMRuleCompiler Objects​

class LLMRuleCompiler()

__init__​

def __init__(model_config: LLMModelConfig,
             custom_check_functions: dict[str, Callable] | None = None,
             rule_validator: RuleValidator | None = None,
             optimizer: BootstrapFewShotOptimizer | None = None)

Arguments:

  • model_config - Configuration for the LLM model.
  • custom_check_functions - Optional custom check functions.
  • rule_validator - Optional rule validator instance.
  • optimizer - Optional optimizer instance.

model​

@cached_property
def model() -> dspy.Module

Get the optimized DSPy model.

Returns:

Optimized DSPy module for generating data quality rules.