Implementation:Evidentlyai Evidently Legacy Runner

Knowledge Sources	Evidentlyai_Evidently
Domains	Data Monitoring, Pipeline Runner
Last Updated	2026-02-14 12:00 GMT

Overview

Provides a legacy runner infrastructure for loading reference and current datasets and executing Evidently data analysis pipelines with configurable options.

Description

The Legacy Runner module defines the RunnerOptions dataclass for capturing all parameters needed to run an Evidently analysis pipeline, including paths to reference and current data, sampling options, column mappings, analysis options, and output destination. It also provides the Runner base class whose _parse_data method uses DataLoader to load and optionally sample both reference and current datasets. The parse_options function converts raw dictionary-based configuration into strongly typed option objects (such as DataDriftOptions or QualityMetricsOptions) using a predefined options_mapping registry.

This module is part of the legacy pipeline execution path. The Runner class is intended to be subclassed by specific runner implementations that add their own run logic on top of the parsed data.

Usage

Use this module when running Evidently analyses via the legacy pipeline runner interface. Create a RunnerOptions instance with all necessary paths and options, instantiate a Runner (or subclass), and call _parse_data to load the datasets prior to running analysis. Use parse_options to convert raw configuration dictionaries (e.g., from YAML or JSON config files) into the appropriate option objects.

Code Reference

Source Location

• Repository: Evidentlyai_Evidently
• File: src/evidently/legacy/runner/runner.py

Signature

@dataclass
class RunnerOptions:
    reference_data_path: str
    reference_data_options: DataOptions
    reference_data_sampling: Optional[SamplingOptions]
    current_data_path: Optional[str]
    current_data_options: Optional[DataOptions]
    current_data_sampling: Optional[SamplingOptions]
    column_mapping: ColumnMapping
    options: List[object]
    output_path: str

def parse_options(raw_dict: Optional[Dict[str, Dict[str, object]]]) -> List[object]:
    ...

class Runner:
    def __init__(self, options: RunnerOptions):
        ...
    def _parse_data(self):
        ...

Import

from evidently.legacy.runner.runner import Runner, RunnerOptions, parse_options

I/O Contract

Inputs

Name	Type	Required	Description
reference_data_path	str	Yes	File path to the reference dataset
reference_data_options	DataOptions	Yes	Options controlling how the reference data file is loaded (format, separator, etc.)
reference_data_sampling	Optional[SamplingOptions]	No	Optional sampling configuration for the reference data
current_data_path	Optional[str]	No	File path to the current dataset (may be None for reference-only analysis)
current_data_options	Optional[DataOptions]	No	Options controlling how the current data file is loaded
current_data_sampling	Optional[SamplingOptions]	No	Optional sampling configuration for the current data
column_mapping	ColumnMapping	Yes	Mapping of column roles (target, prediction, features, etc.)
options	List[object]	Yes	List of analysis option objects (e.g., DataDriftOptions, QualityMetricsOptions)
output_path	str	Yes	File path for writing analysis output

Outputs

Name	Type	Description
_parse_data return	Tuple[DataFrame, Optional[DataFrame]]	A tuple of (reference_data, current_data) pandas DataFrames; current_data is None if no current data path was specified
parse_options return	List[object]	List of instantiated option objects parsed from the raw dictionary

Usage Examples

from evidently.legacy.runner.runner import Runner, RunnerOptions, parse_options
from evidently.legacy.runner.loader import DataOptions
from evidently.legacy.pipeline.column_mapping import ColumnMapping

# Parse options from a raw config dictionary
raw_opts = {
    "data_drift": {"confidence": 0.95, "nbinsx": 10}
}
options = parse_options(raw_opts)

# Create runner options
runner_options = RunnerOptions(
    reference_data_path="data/reference.csv",
    reference_data_options=DataOptions(),
    reference_data_sampling=None,
    current_data_path="data/current.csv",
    current_data_options=DataOptions(),
    current_data_sampling=None,
    column_mapping=ColumnMapping(),
    options=options,
    output_path="output/report.html",
)

# Instantiate runner and parse data
runner = Runner(runner_options)
reference_data, current_data = runner._parse_data()

Related Pages

• Environment:Evidentlyai_Evidently_Python_Core_Environment