Skip to main content

databricks.labs.dqx.anomaly.model_registry

Model registry utilities for row anomaly detection.

Persistence only: schema and AnomalyModelRegistry. Record types live in model_config.

AnomalyModelRegistry Objects​

class AnomalyModelRegistry()

Manage anomaly model metadata in a Delta table.

convert_decimals​

@staticmethod
def convert_decimals(obj: Any) -> Any

Recursively convert Decimal values to float for PyArrow compatibility.

This is a public utility method that can be used by tests and other code to handle Decimal to float conversions for PyArrow compatibility.

build_model_df​

@staticmethod
def build_model_df(spark: SparkSession,
                   record: AnomalyModelRecord) -> DataFrame

Convert a registry record into a DataFrame with nested structure.

save_model​

def save_model(record: AnomalyModelRecord, table: str) -> None

Archive previous active model with the same name and insert the new record.

get_active_model​

def get_active_model(table: str, model_name: str) -> AnomalyModelRecord | None

Fetch the active model for a given name.

get_segment_model​

def get_segment_model(
        table: str, base_model_name: str,
        segment_values: dict[str, str]) -> AnomalyModelRecord | None

Fetch model for specific segment combination.

get_all_segment_models​

def get_all_segment_models(table: str,
                           base_model_name: str) -> list[AnomalyModelRecord]

Fetch all segment models for a base name.