Operators, Sensors & Hooks

Complete reference for the Arize AX Airflow Provider. Every operator, sensor, and hook, grouped by domain.

This page is the full reference for the Arize AX Airflow Provider. It lists every operator, sensor, and hook in the package, grouped by the resource each one acts on. For setup, connection configuration, and patterns, start with the provider overview.

Import paths

Every class lives under the airflow.providers.arize_ax namespace:

# Operators: one module per resource domain
from airflow.providers.arize_ax.operators.datasets       import ...
from airflow.providers.arize_ax.operators.experiments    import ...
from airflow.providers.arize_ax.operators.projects       import ...
from airflow.providers.arize_ax.operators.spans          import ...
from airflow.providers.arize_ax.operators.ml             import ...
from airflow.providers.arize_ax.operators.evaluators     import ...
from airflow.providers.arize_ax.operators.tasks          import ...
from airflow.providers.arize_ax.operators.prompts        import ...
from airflow.providers.arize_ax.operators.spaces         import ...
from airflow.providers.arize_ax.operators.annotations    import ...
from airflow.providers.arize_ax.operators.api_keys       import ...
from airflow.providers.arize_ax.operators.ai_integrations import ...

# Sensors: a single module
from airflow.providers.arize_ax.sensors.arize_ax import ...

# Hook: used by every operator under the hood
from airflow.providers.arize_ax.hooks.arize_ax import ArizeAxHook

Common params. Every operator accepts arize_ax_conn_id (default arize_ax_default). Most accept space_id, which the provider resolves in this order: the operator argument, then the connection’s extra.default_space, then the Airflow Variable arize_ax_space_id. Create* operators accept if_exists="skip" for idempotent re-runs. Delete* and Get* operators accept ignore_if_missing=True. See design patterns.

Hook

ArizeAxHook is the only hook class. Every operator instantiates it in execute() to get a cached ArizeClient. You only use the hook directly when no operator covers your pattern.

from airflow.providers.arize_ax.hooks.arize_ax import ArizeAxHook

hook = ArizeAxHook(arize_ax_conn_id="arize_ax_default")
client = hook.get_conn()  # cached ArizeClient
datasets = hook.list_datasets(space_id="...", limit=50)

The hook exposes typed wrappers for every resource the operators cover (list_datasets, create_experiment, export_spans_to_df, trigger_task_run, log_spans, list_evaluators, and so on). It also handles pagination, retries transient Flight errors on span exports, and re-raises SDK 404s and backend errors as AirflowException with clear messages.

Datasets

Module: airflow.providers.arize_ax.operators.datasets

Operator	Purpose	Key params
ArizeAxListDatasetsOperator	List datasets in a space.	space_id, limit, cursor
ArizeAxCreateDatasetOperator	Create a dataset, optionally with starter examples.	space_id, name, examples or examples_path, if_exists
ArizeAxGetDatasetOperator	Fetch a dataset by ID.	dataset_id
ArizeAxDeleteDatasetOperator	Delete a dataset.	dataset_id, ignore_if_missing
ArizeAxListDatasetExamplesOperator	List examples in a dataset (or a specific version).	dataset_id, dataset_version_id, limit, all
ArizeAxAppendDatasetExamplesOperator	Append examples from memory, CSV, or JSON.	dataset_id, examples or examples_path
ArizeAxExportDatasetExamplesToFileOperator	Export all examples to JSON (paginated via SDK Flight).	dataset_id, path, dataset_version_id
ArizeAxEvalDatasetHealthOperator	Score a dataset on freshness, diversity, and coverage vs production.	dataset_id, project_id, start_time, end_time
ArizeAxSmartDatasetRefreshOperator	Evolve a dataset with diverse, high-value production examples.	dataset_id, project_id, start_time, end_time, diversity_strategy

Experiments

Module: airflow.providers.arize_ax.operators.experiments

Operator	Purpose	Key params
ArizeAxListExperimentsOperator	List experiments for a dataset.	dataset_id, limit, cursor
ArizeAxCreateExperimentOperator	Create experiment with pre-computed runs.	name, dataset_id, experiment_runs, task_fields, evaluator_columns, if_exists
ArizeAxGetExperimentOperator	Fetch an experiment by ID.	experiment_id
ArizeAxDeleteExperimentOperator	Delete an experiment.	experiment_id, ignore_if_missing
ArizeAxRunExperimentOperator	Run a task + evaluators end-to-end against a dataset.	name, dataset_id, task, evaluators, concurrency, dry_run
ArizeAxListExperimentRunsOperator	List runs for an experiment.	experiment_id, limit, cursor
ArizeAxExportExperimentRunsToFileOperator	Export all runs to a JSON file.	experiment_id, path
ArizeAxGetExperimentScoreOperator	Aggregate eval scores; gates on min_score.	experiment_id, metric_names, aggregation, min_score
ArizeAxCompareExperimentsOperator	Compare candidate vs baseline; gates on regression.	candidate_experiment_id, baseline_experiment_id, pass_threshold, fail_on_regression
ArizeAxDetectEvalDriftOperator	Detect metric drift between experiments; gates on drift.	experiment_id, baseline_id, drift_threshold, fail_on_drift
ArizeAxEvaluatorCalibrationOperator	Measure LLM-judge correlation vs human labels; gates on calibration.	experiment_id, human_label_column, calibration_threshold, fail_on_poor_calibration
ArizeAxBehavioralRegressionOperator	Compare behavioral distributions; gates on regression.	candidate_id, baseline_id, metrics, fail_on_regression
ArizeAxEvalBudgetAllocatorOperator	Distribute evaluation budget across projects.	projects, total_budget, allocation_strategy

Projects

Module: airflow.providers.arize_ax.operators.projects

Operator	Purpose	Key params
ArizeAxListProjectsOperator	List projects in a space.	space_id, limit, cursor
ArizeAxCreateProjectOperator	Create a project.	space_id, name, if_exists
ArizeAxGetProjectOperator	Fetch a project by ID.	project_id
ArizeAxDeleteProjectOperator	Delete a project.	project_id, ignore_if_missing

Spans

Module: airflow.providers.arize_ax.operators.spans

Operator	Purpose	Key params
ArizeAxListSpansOperator	List spans from a project (alpha).	project_id, space_id, filter, start_time, end_time
ArizeAxSpansLogOperator	Bulk-log spans from DataFrame, CSV, or JSON.	space_id, project_name, dataframe or dataframe_path, evals_dataframe
ArizeAxSpansUpdateEvaluationsOperator	Update span evaluations from a DataFrame.	space_id, project_name, dataframe or dataframe_path
ArizeAxSpansUpdateAnnotationsOperator	Update span annotations from a DataFrame.	space_id, project_name, dataframe or dataframe_path
ArizeAxSpansUpdateMetadataOperator	Update span metadata from a DataFrame.	space_id, project_name, dataframe or dataframe_path
ArizeAxSpansExportToDataframeOperator	Export filtered spans to a DataFrame.	space_id, project_name, start_time, end_time, where, columns
ArizeAxSpansExportToParquetOperator	Export filtered spans to a Parquet file.	space_id, project_name, start_time, end_time, path, where
ArizeAxExportSpansToFineTuningOperator	Export spans as OpenAI fine-tuning JSONL.	space_id, project_name, start_time, end_time, path
ArizeAxGetSpanMetricsOperator	Aggregate latency, cost, and token metrics.	space_id, project_name, start_time, end_time
ArizeAxCurateSpansToDatasetOperator	Curate filtered spans into a dataset.	space_id, project_name, dataset_id, where
ArizeAxExportAnnotatedSpansOperator	Export annotated spans to Parquet or JSON.	space_id, project_name, start_time, end_time, path, format
ArizeAxAdaptiveSamplingOperator	Pick priority spans for evaluation via uncertainty/novelty/anomaly.	space_id, project_name, sample_size, strategy

ML

Module: airflow.providers.arize_ax.operators.ml

Operator	Purpose	Key params
ArizeAxMLLogBatchOperator	Batch-log ML predictions from DataFrame or CSV.	space_id, model_name, model_type, environment, dataframe, schema
ArizeAxMLLogStreamOperator	Stream a single ML prediction.	space_id, model_name, model_type, environment, log_params
ArizeAxMLExportToDataframeOperator	Export ML data to a DataFrame.	space_id, model_name, environment, start_time, end_time, where
ArizeAxMLExportToParquetOperator	Export ML data to a Parquet file.	space_id, model_name, environment, start_time, end_time, path

Evaluators

Module: airflow.providers.arize_ax.operators.evaluators (alpha)

Operator	Purpose	Key params
ArizeAxListEvaluatorsOperator	List LLM-template evaluators in Eval Hub.	space_id, limit, cursor, name_search
ArizeAxCreateEvaluatorOperator	Create an LLM-template evaluator.	space_id, name, evaluator_type, commit_message, template_config, if_exists
ArizeAxGetEvaluatorOperator	Fetch an evaluator by ID.	evaluator_id
ArizeAxUpdateEvaluatorOperator	Update evaluator metadata (name, description).	evaluator_id, name, description
ArizeAxDeleteEvaluatorOperator	Delete an evaluator.	evaluator_id, ignore_if_missing
ArizeAxAddEvaluatorVersionOperator	Add a new template version.	evaluator_id, commit_message, template_config
ArizeAxListEvaluatorVersionsOperator	List versions of an evaluator.	evaluator_id, limit, cursor
ArizeAxGetEvaluatorVersionOperator	Fetch a specific version.	evaluator_id, version_id

Tasks

Module: airflow.providers.arize_ax.operators.tasks (alpha)

Evaluation tasks attach evaluators to a project or dataset and execute on demand or on a schedule.

Operator	Purpose	Key params
ArizeAxListTasksOperator	List evaluation tasks (filter by space/project/dataset/type).	space_id, project_id, dataset_id, eval_task_type
ArizeAxGetTaskOperator	Fetch a task by ID.	task_id_param
ArizeAxCreateTaskOperator	Create a task that attaches evaluators to a project or dataset.	name, eval_task_type, evaluators, project_id or dataset_id, sampling_rate, is_continuous
ArizeAxListTaskRunsOperator	List runs for a task.	task_id_param, limit, cursor
ArizeAxGetTaskRunOperator	Fetch a task run by ID.	run_id
ArizeAxTriggerTaskRunOperator	Trigger an on-demand run; override_evaluations=True re-scores existing spans.	task_id_param, data_start_time, data_end_time, max_spans, override_evaluations
ArizeAxCancelTaskRunOperator	Cancel a pending or running task run.	run_id

Prompts

Module: airflow.providers.arize_ax.operators.prompts

Operator	Purpose	Key params
ArizeAxListPromptsOperator	List prompts in Prompt Hub.	space_id, limit, cursor
ArizeAxGetPromptOperator	Fetch a prompt by ID, version ID, or label.	prompt_id, version_id, version_label
ArizeAxCreatePromptOperator	Create a prompt with an initial version.	space_id, name, messages or messages_task_id, provider, model, if_exists
ArizeAxDeletePromptOperator	Delete a prompt.	prompt_id, ignore_if_missing
ArizeAxPromotePromptOperator	Apply a label to a prompt version (e.g. production).	prompt_id, version_id, label
ArizeAxComparePromptsOperator	Compare two prompt versions by experiment scores.	prompt_id_1, prompt_id_2, experiment_id

Spaces

Module: airflow.providers.arize_ax.operators.spaces

Operator	Purpose	Key params
ArizeAxListSpacesOperator	List spaces accessible to the authenticated user.	limit, cursor
ArizeAxGetSpaceOperator	Fetch a space by ID.	space_id
ArizeAxCreateSpaceOperator	Create a space in an organization.	name, organization_id, description
ArizeAxUpdateSpaceOperator	Update space name or description.	space_id, name, description
ArizeAxDeleteSpaceOperator	Delete a space (alpha).	space_id, ignore_if_missing

Annotation configs & queues

Module: airflow.providers.arize_ax.operators.annotations (alpha)

For configuring HITL annotation flows. Pair with ArizeAxAnnotationQueueSensor to gate downstream tasks until a minimum number of annotation configs exist in the space.

Operator	Purpose	Key params
ArizeAxListAnnotationConfigsOperator	List annotation configurations.	space_id, limit, cursor
ArizeAxCreateAnnotationConfigOperator	Create a config (freeform, category, or numeric).	space_id, name, config_type, minimum_score, maximum_score
ArizeAxDeleteAnnotationConfigOperator	Delete a config.	annotation_config, space_id
ArizeAxListAnnotationQueuesOperator	List queues.	space_id, limit, cursor
ArizeAxGetAnnotationQueueOperator	Fetch a queue by ID or name.	queue_id or queue_name, space_id
ArizeAxCreateAnnotationQueueOperator	Create a queue with annotators and configs.	space_id, name, description, annotators, configs
ArizeAxUpdateAnnotationQueueOperator	Update queue metadata, annotators, or configs.	queue_id, name, instructions, configs, annotators
ArizeAxDeleteAnnotationQueueOperator	Delete a queue.	queue_id, ignore_if_missing
ArizeAxListAnnotationQueueRecordsOperator	List pending records (max 500/page).	queue_id, limit, cursor
ArizeAxAddAnnotationQueueRecordsOperator	Add record sources to a queue (max 2 sources).	queue_id, sources, num_records
ArizeAxAnnotateQueueRecordOperator	Submit annotations on a record.	queue_id, record_id, annotations
ArizeAxAssignQueueRecordOperator	Assign a record to one or more annotators.	queue_id, record_id, annotators

API keys

Module: airflow.providers.arize_ax.operators.api_keys

Operator	Purpose	Key params
ArizeAxListAPIKeysOperator	List API keys (user or service).	limit, cursor
ArizeAxCreateAPIKeyOperator	Create an API key.	key_type, description, organization_id
ArizeAxDeleteAPIKeyOperator	Delete an API key.	api_key_id, ignore_if_missing
ArizeAxRefreshAPIKeyOperator	Rotate (refresh) an API key.	api_key_id

AI integrations

Module: airflow.providers.arize_ax.operators.ai_integrations (alpha)

Register external agent platforms (LangGraph, OpenAI Agents, Autogen) so Arize evaluators can run against them.

Operator	Purpose	Key params
ArizeAxListAIIntegrationsOperator	List AI integrations.	space_id, limit, cursor
ArizeAxGetAIIntegrationOperator	Fetch an integration by ID.	integration_id
ArizeAxCreateAIIntegrationOperator	Create an integration.	name, provider, space_id, api_key, base_url, model_names, if_exists
ArizeAxUpdateAIIntegrationOperator	Update an integration.	integration_id, name, provider, api_key
ArizeAxDeleteAIIntegrationOperator	Delete an integration.	integration_id, ignore_if_missing

Sensors

Module: airflow.providers.arize_ax.sensors.arize_ax

All sensors accept the standard Airflow poke_interval, timeout, mode, and soft_fail.

Sensor	Waits until…	Key params
ArizeAxExperimentCompleteSensor	Experiment reaches a terminal state.	experiment_id, min_runs
ArizeAxExperimentRunCountSensor	Experiment has at least N completed runs.	experiment_id, min_runs
ArizeAxEvaluationScoreSensor	Experiment metric mean ≥ threshold.	experiment_id, metric_name, min_score
ArizeAxDatasetReadySensor	Dataset has ≥ N examples.	dataset_id, min_examples, dataset_version_id
ArizeAxSpanCountSensor	Project has ≥ N spans in the window.	project_id, min_count, start_time, end_time, filter
ArizeAxSpanIngestionSensor	Span count stops growing rapidly (ingestion stable).	project_id, stable_window_seconds
ArizeAxAnnotationQueueSensor	At least N annotation configs exist in the space.	space_id, min_count
ArizeAxTaskRunSensor	Evaluation task run reaches a terminal state.	run_id

Resources

Provider overview

Install, connection setup, design patterns, and example DAG walkthroughs.

Example DAGs

Runnable DAGs covering CI/CD gating, prompt lifecycle, dataset curation, and more.

Python SDK v8

The underlying ArizeClient API the provider wraps.