- artschema.catalog: catalog with metadata, verify specs
- artschema/LoadArtifactSpec: source run property, export options
- artschema: bugfix in pkg reqs check with markers
- artschema: bugfix in pkg requirements check
- artschema/specs: add
place
method, extend behavior ofoptional
flag for loaded run artifacts
- mlflow/api/base: fix dir permissions after artifact place by copy
- mlflow/api/mv: return api copy after
place_artifact
- BaseMlflowApi: use file links when placing artifact dirs
- artschema.specs: let run artifact be optional
- mlflow.api.common.transfer: fix scheme translation
- mlflow.api.mv: fix list_artifacts
- scheme translation for artifact urls
- pull artifacts in offline mode (if at least metadata is cached)
- enum json encoding
- kedro.NodeFunc: let users implement either call or async acall
- providers/mlflow: use http client settings in healthcheck, stringify tags/params before max length check, fix parsing local urls
- kedro.hooks.mlflow-artifacts: allow pipeline inptus/outputs to be a mapping
- artschema.tags: forbid range constraint on required pkg version if there's build metadata (e.g.: 1.2.3+hash)
- artschema.tags: account for recursive extras when checking requirements of artschema
- artschema.spec: fix entity api property in LoadArtifactSpec
- artschema.tags: fix detection of installed extras and pkg spec verification
- artschema.spec: expose entity api in LoadArtifactSpec
- utils.dicts: preserve empty dict leaf vals in deep merge (used in kedro conf merge)
- kedro.session: use only root package name for determining version
- mlflow.api.transfer: apply prog bar flag, allow extra args, copy links by default
- kedro.hooks.artifacts: fix file permissions after input setup
- artschema: skip saving dumper conf if empty, adjust spec parser for pydantic v2
- utils.pydantic: coerce int to str
- typing_utils: fix typevar and typing alias coercion to type (used in artschema type param inference)
- kedro: fix registering extra scopes as config resolvers, fix artschema dataset config parsing in pydantic v2
- adopt pydantic v2 (#11)
- artschema.framework: skip warning when there's no dumper conf to be saved
- kedro.pipeline_factory: accept mapping instead of config loader
- mlflow.traits: allow passing parent run ID to run manager
- kedro.hooks.mlflow_tracker: fix param
prepend_dataset
for mlflow metrics
- kedro.hooks.mlflow_tracker: allow disabling dataset name prefix in metrics
- kedro.config_loader: let source default to base, auto-register new scopes as resolvers
- artschema: hint to correct format of fq class name
- artschema: allow passing str path, auto convert to pathlib
- kedro hooks, datasets, config loader and session patch
- BaseMlflowApi: protect cached files, allow duplication by default if using caching from local repo
- artschema, lineage
- package metadata
- base mlflow api and plugins