Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.1.0-alpha.2] - 2026-04-15

This minor release includes 1 commit.

Bug Fixes

  • Patch azureml artifact builder for MLflow 3.10+ compatibility (#12) by @gtauzin

Contributors

Thanks to all contributors for this release: - @gtauzin

[0.1.0-alpha.1] - 2026-04-01

This minor release includes 1 commit.

Features

  • kedro azureml run -j <job> command for running named jobs immediately on Azure ML. Supports --dry-run (preview), --wait-for-completion (CI blocking), and --on-job-scheduled (callback). by @gtauzin
  • kedro azureml schedule -j <job> command for creating or updating persistent Azure ML schedules. Requires each job to have a schedule configured. Supports --dry-run (preview). by @gtauzin
  • kedro azureml compile -j <job> for compiling named job pipelines to YAML. by @gtauzin
  • schedules and jobs config sections with cron and recurrence triggers, pipeline filtering (from_nodes, to_nodes, tags, etc.), per-job display name, compute, and experiment name. by @gtauzin
  • Named workspaces: workspace is now a dict of named workspace configs (with mandatory __default__). Jobs can reference a specific workspace via workspace: key. CLI --workspace/-w selects a workspace at run/schedule time. by @gtauzin
  • Full kedro-mlflow compatibility: unified experiment naming via mlflow.yml, MLflow run tagging hook, and env var injection into Azure ML component jobs. by @gtauzin
  • Support for Python 3.13. by @gtauzin
  • Support factory-resolved datasets in the runner. by @gtauzin

Refactoring

  • Config restructure: the azure: top-level key is replaced by three flat sections -- workspace, compute, execution. compute and workspace are flat dicts keyed by name (with mandatory __default__). experiment_name moves into per-job config. The temporary_storage and pipeline_data_passing config sections are removed. by @gtauzin
  • kedro azureml run is replaced by kedro azureml run -j <job> (immediate execution) and kedro azureml schedule -j <job> (persistent schedules). kedro azureml compile now requires -j <job>. --subscription-id replaced by --workspace. by @gtauzin
  • Blob storage removal: KedroAzureRunnerDataset, KedroAzureRunnerDistributedDataset, BlobStorageDataPassing, KedroAzureRunnerConfig, and runner_dataset.py module deleted. Pipeline data passing via AzureMLPipelineDataset is now the only mode. by @gtauzin
  • Removed kedro azureml run command and all its options (--display-name, --compute-name, --experiment-name, -p/--pipeline, --wait-for-completion, --on-job-scheduled). by @gtauzin
  • Removed deprecated docker config section; environment configuration now uses execution.environment. by @gtauzin
  • Removed deprecated SDK v1 dataset stubs (AzureMLPandasDataset, AzureMLFileDataset) and v1_datasets module. by @gtauzin
  • Migrated project following the stateful-y/python-package-copier template. by @gtauzin
  • kedro azureml init no longer accepts positional arguments or --aml-env. It generates conf/base/azureml.yml with placeholder values to be filled in manually. by @gtauzin

Documentation

  • Migrated documentation from Sphinx (RST) to MkDocs with Material theme. by @gtauzin
  • Rewrote all documentation pages based on diataxis approach: getting started, user guide, API reference, and contributing guide. by @gtauzin
  • Added NumPy-style docstrings to all public modules, classes, and functions (interrogate coverage at 100%). by @gtauzin

Contributors

Thanks to all contributors for this release: - @gtauzin

[kedro-azureml v1.0.0]

This project is a fork of kedro-azureml originally created by GetInData | Part of Xebia.