lamindb 1.1.0__py3-none-any.whl → 1.2a2__py3-none-any.whl

lamindb/models/artifact.py

@@ -0,0 +1,2601 @@
+# ruff: noqa: TC004
+from __future__ import annotations
+
+import os
+import shutil
+from collections import defaultdict
+from pathlib import Path, PurePath, PurePosixPath
+from typing import TYPE_CHECKING, Any, Union, overload
+
+import fsspec
+import lamindb_setup as ln_setup
+import pandas as pd
+from anndata import AnnData
+from django.db import connections, models
+from django.db.models import CASCADE, PROTECT, Q
+from lamin_utils import colors, logger
+from lamindb_setup import settings as setup_settings
+from lamindb_setup._init_instance import register_storage_in_instance
+from lamindb_setup.core._settings_storage import init_storage
+from lamindb_setup.core.hashing import HASH_LENGTH, hash_dir, hash_file
+from lamindb_setup.core.types import UPathStr
+from lamindb_setup.core.upath import (
+    create_path,
+    extract_suffix_from_path,
+    get_stat_dir_cloud,
+    get_stat_file_cloud,
+)
+
+from lamindb.base import deprecated
+from lamindb.base.fields import (
+    BigIntegerField,
+    BooleanField,
+    CharField,
+    ForeignKey,
+)
+from lamindb.errors import FieldValidationError
+from lamindb.models.query_set import QuerySet
+
+from ..base.users import current_user_id
+from ..core.loaders import load_to_memory
+from ..core.storage import (
+    LocalPathClasses,
+    UPath,
+    delete_storage,
+    infer_suffix,
+    write_to_disk,
+)
+from ..core.storage._anndata_accessor import _anndata_n_observations
+from ..core.storage._pyarrow_dataset import PYARROW_SUFFIXES
+from ..core.storage._tiledbsoma import _soma_n_observations
+from ..core.storage.objects import is_package_installed
+from ..core.storage.paths import (
+    AUTO_KEY_PREFIX,
+    auto_storage_key_from_artifact,
+    auto_storage_key_from_artifact_uid,
+    check_path_is_child_of_root,
+    filepath_cache_key_from_artifact,
+    filepath_from_artifact,
+)
+from ..errors import IntegrityError, InvalidArgument, ValidationError
+from ..models._is_versioned import (
+    create_uid,
+    message_update_key_in_version_family,
+)
+from ._django import get_artifact_with_related
+from ._feature_manager import (
+    FeatureManager,
+    ParamManager,
+    ParamManagerArtifact,
+    add_label_feature_links,
+    get_label_links,
+)
+from ._is_versioned import IsVersioned
+from ._relations import (
+    dict_module_name_to_model_name,
+    dict_related_model_to_related_name,
+)
+from .core import Storage
+from .feature import Feature, FeatureValue
+from .has_parents import view_lineage
+from .record import (
+    BasicRecord,
+    LinkORM,
+    Record,
+    _get_record_kwargs,
+    record_repr,
+)
+from .run import ParamValue, Run, TracksRun, TracksUpdates, User
+from .schema import Schema
+from .ulabel import ULabel
+
+WARNING_RUN_TRANSFORM = "no run & transform got linked, call `ln.track()` & re-run"
+
+WARNING_NO_INPUT = "run input wasn't tracked, call `ln.track()` and re-run"
+
+try:
+    from ..core.storage._zarr import identify_zarr_type
+except ImportError:
+
+    def identify_zarr_type(storepath):  # type: ignore
+        raise ImportError("Please install zarr: pip install zarr<=2.18.4")
+
+
+if TYPE_CHECKING:
+    from collections.abc import Iterable
+
+    from mudata import MuData  # noqa: TC004
+    from pyarrow.dataset import Dataset as PyArrowDataset
+    from spatialdata import SpatialData  # noqa: TC004
+    from tiledbsoma import Collection as SOMACollection
+    from tiledbsoma import Experiment as SOMAExperiment
+    from tiledbsoma import Measurement as SOMAMeasurement
+
+    from lamindb.base.types import StrField
+    from lamindb.core.storage._backed_access import AnnDataAccessor, BackedAccessor
+
+    from ..base.types import (
+        ArtifactKind,
+    )
+    from ._label_manager import LabelManager
+    from .collection import Collection
+    from .project import Project, Reference
+    from .transform import Transform
+
+
+INCONSISTENT_STATE_MSG = (
+    "Trying to read a folder artifact from an outdated version, "
+    "this can result in an incosistent state.\n"
+    "Read from the latest version: artifact.versions.filter(is_latest=True).one()"
+)
+
+
+def process_pathlike(
+    filepath: UPath,
+    default_storage: Storage,
+    using_key: str | None,
+    skip_existence_check: bool = False,
+) -> tuple[Storage, bool]:
+    """Determines the appropriate storage for a given path and whether to use an existing storage key."""
+    if not skip_existence_check:
+        try:  # check if file exists
+            if not filepath.exists():
+                raise FileNotFoundError(filepath)
+        except PermissionError:
+            pass
+    if check_path_is_child_of_root(filepath, default_storage.root):
+        use_existing_storage_key = True
+        return default_storage, use_existing_storage_key
+    else:
+        # check whether the path is part of one of the existing
+        # already-registered storage locations
+        result = False
+        # within the hub, we don't want to perform check_path_in_existing_storage
+        if using_key is None:
+            result = check_path_in_existing_storage(filepath, using_key)
+        if isinstance(result, Storage):
+            use_existing_storage_key = True
+            return result, use_existing_storage_key
+        else:
+            # if the path is in the cloud, we have a good candidate
+            # for the storage root: the bucket
+            if not isinstance(filepath, LocalPathClasses):
+                # for a cloud path, new_root is always the bucket name
+                if filepath.protocol == "hf":
+                    hf_path = filepath.fs.resolve_path(filepath.as_posix())
+                    hf_path.path_in_repo = ""
+                    new_root = "hf://" + hf_path.unresolve()
+                else:
+                    if filepath.protocol == "s3":
+                        # check that endpoint_url didn't propagate here
+                        # as a part of the path string
+                        assert "?" not in filepath.path  # noqa: S101
+                    new_root = list(filepath.parents)[-1]
+                # do not register remote storage locations on hub if the current instance
+                # is not managed on the hub
+                storage_settings, _ = init_storage(
+                    new_root, prevent_register_hub=not setup_settings.instance.is_on_hub
+                )
+                storage_record = register_storage_in_instance(storage_settings)
+                use_existing_storage_key = True
+                return storage_record, use_existing_storage_key
+            # if the filepath is local
+            else:
+                use_existing_storage_key = False
+                # if the default storage is local we'll throw an error if the user
+                # doesn't provide a key
+                if default_storage.type == "local":
+                    return default_storage, use_existing_storage_key
+                # if the default storage is in the cloud (the file is going to
+                # be uploaded upon saving it), we treat the filepath as a cache
+                else:
+                    return default_storage, use_existing_storage_key
+
+
+def process_data(
+    provisional_uid: str,
+    data: UPathStr | pd.DataFrame | AnnData,
+    format: str | None,
+    key: str | None,
+    default_storage: Storage,
+    using_key: str | None,
+    skip_existence_check: bool = False,
+    is_replace: bool = False,
+) -> tuple[Any, Path | UPath, str, Storage, bool]:
+    """Serialize a data object that's provided as file or in memory.
+
+    if not overwritten, data gets stored in default storage
+    """
+    supported_data_types = [pd.DataFrame, AnnData]
+    if is_package_installed("mudata"):
+        from mudata import MuData
+
+        supported_data_types.append(MuData)
+    if is_package_installed("spatialdata"):
+        from spatialdata import SpatialData
+
+        supported_data_types.append(SpatialData)
+    supported_data_types = tuple(supported_data_types)  # type: ignore
+
+    if key is not None:
+        key_suffix = extract_suffix_from_path(PurePosixPath(key), arg_name="key")
+        # use suffix as the (adata) format if the format is not provided
+        if isinstance(data, AnnData) and format is None and len(key_suffix) > 0:
+            format = key_suffix[1:]
+    else:
+        key_suffix = None
+    if isinstance(data, (str, Path, UPath)):  # UPathStr, spelled out
+        access_token = (
+            default_storage._access_token
+            if hasattr(default_storage, "_access_token")
+            else None
+        )
+        path = create_path(data, access_token=access_token)
+        # we don't resolve http links because they can resolve into a different domain
+        # for example into a temporary url
+        if path.protocol not in {"http", "https"}:
+            path = path.resolve()
+        storage, use_existing_storage_key = process_pathlike(
+            path,
+            default_storage=default_storage,
+            using_key=using_key,
+            skip_existence_check=skip_existence_check,
+        )
+        suffix = extract_suffix_from_path(path)
+        memory_rep = None
+    elif isinstance(data, supported_data_types):
+        storage = default_storage
+        memory_rep = data
+        suffix = infer_suffix(data, format)
+    else:
+        raise NotImplementedError(
+            f"Do not know how to create a artifact object from {data}, pass a path instead!"
+        )
+    if key_suffix is not None and key_suffix != suffix and not is_replace:
+        # consciously omitting a trailing period
+        if isinstance(data, (str, Path, UPath)):
+            message = f"The suffix '{suffix}' of the provided path is inconsistent, it should be '{key_suffix}'"
+        else:
+            message = f"The suffix '{key_suffix}' of the provided key is inconsistent, it should be '{suffix}'"
+        raise InvalidArgument(message)
+    # in case we have an in-memory representation, we need to write it to disk
+    from lamindb import settings
+
+    if isinstance(data, supported_data_types):
+        path = settings.cache_dir / f"{provisional_uid}{suffix}"
+        write_to_disk(data, path)
+        use_existing_storage_key = False
+    return memory_rep, path, suffix, storage, use_existing_storage_key
+
+
+def get_stat_or_artifact(
+    path: UPath,
+    key: str | None = None,
+    check_hash: bool = True,
+    is_replace: bool = False,
+    instance: str | None = None,
+) -> Union[tuple[int, str | None, str | None, int | None, Artifact | None], Artifact]:
+    """Retrieves file statistics or an existing artifact based on the path, hash, and key."""
+    n_files = None
+    from lamindb import settings
+
+    if settings.creation.artifact_skip_size_hash:
+        return None, None, None, n_files, None
+    stat = path.stat()  # one network request
+    if not isinstance(path, LocalPathClasses):
+        size, hash, hash_type = None, None, None
+        if stat is not None:
+            # convert UPathStatResult to fsspec info dict
+            stat = stat.as_info()
+            if (store_type := stat["type"]) == "file":
+                size, hash, hash_type = get_stat_file_cloud(stat)
+            elif store_type == "directory":
+                size, hash, hash_type, n_files = get_stat_dir_cloud(path)
+        if hash is None:
+            logger.warning(f"did not add hash for {path}")
+            return size, hash, hash_type, n_files, None
+    else:
+        if path.is_dir():
+            size, hash, hash_type, n_files = hash_dir(path)
+        else:
+            hash, hash_type = hash_file(path)
+            size = stat.st_size
+    if not check_hash:
+        return size, hash, hash_type, n_files, None
+    previous_artifact_version = None
+    if key is None or is_replace:
+        result = Artifact.objects.using(instance).filter(hash=hash).all()
+        artifact_with_same_hash_exists = len(result) > 0
+    else:
+        storage_id = settings.storage.id
+        result = (
+            Artifact.objects.using(instance)
+            .filter(Q(hash=hash) | Q(key=key, storage_id=storage_id))
+            .order_by("-created_at")
+            .all()
+        )
+        artifact_with_same_hash_exists = result.filter(hash=hash).count() > 0
+        if not artifact_with_same_hash_exists and len(result) > 0:
+            logger.important(
+                f"creating new artifact version for key='{key}' (storage: '{settings.storage.root_as_str}')"
+            )
+            previous_artifact_version = result[0]
+    if artifact_with_same_hash_exists:
+        message = "returning existing artifact with same hash"
+        if result[0]._branch_code == -1:
+            result[0].restore()
+            message = "restored artifact with same hash from trash"
+        logger.important(
+            f"{message}: {result[0]}; to track this artifact as an input, use: ln.Artifact.get()"
+        )
+        return result[0]
+    else:
+        return size, hash, hash_type, n_files, previous_artifact_version
+
+
+def check_path_in_existing_storage(
+    path: Path | UPath, using_key: str | None = None
+) -> Storage | bool:
+    for storage in Storage.objects.using(using_key).filter().all():
+        # if path is part of storage, return it
+        if check_path_is_child_of_root(path, root=storage.root):
+            return storage
+    return False
+
+
+def get_relative_path_to_directory(
+    path: PurePath | Path | UPath, directory: PurePath | Path | UPath
+) -> PurePath | Path:
+    if isinstance(directory, UPath) and not isinstance(directory, LocalPathClasses):
+        # UPath.relative_to() is not behaving as it should (2023-04-07)
+        # need to lstrip otherwise inconsistent behavior across trailing slashes
+        # see test_artifact.py: test_get_relative_path_to_directory
+        relpath = PurePath(
+            path.as_posix().replace(directory.as_posix(), "").lstrip("/")
+        )
+    elif isinstance(directory, Path):
+        relpath = path.resolve().relative_to(directory.resolve())  # type: ignore
+    elif isinstance(directory, PurePath):
+        relpath = path.relative_to(directory)
+    else:
+        raise TypeError("Directory not of type Path or UPath")
+    return relpath
+
+
+def get_artifact_kwargs_from_data(
+    *,
+    data: Path | UPath | str | pd.DataFrame | AnnData | MuData,
+    key: str | None,
+    run: Run | None,
+    format: str | None,
+    provisional_uid: str,
+    version: str | None,
+    default_storage: Storage,
+    using_key: str | None = None,
+    is_replace: bool = False,
+    skip_check_exists: bool = False,
+):
+    from lamindb import settings
+
+    run = get_run(run)
+    memory_rep, path, suffix, storage, use_existing_storage_key = process_data(
+        provisional_uid,
+        data,
+        format,
+        key,
+        default_storage,
+        using_key,
+        skip_check_exists,
+        is_replace=is_replace,
+    )
+    stat_or_artifact = get_stat_or_artifact(
+        path=path,
+        key=key,
+        instance=using_key,
+        is_replace=is_replace,
+    )
+    if isinstance(stat_or_artifact, Artifact):
+        existing_artifact = stat_or_artifact
+        if run is not None:
+            existing_artifact._populate_subsequent_runs(run)
+        return existing_artifact, None
+    else:
+        size, hash, hash_type, n_files, revises = stat_or_artifact
+
+    if revises is not None:  # update provisional_uid
+        provisional_uid, revises = create_uid(revises=revises, version=version)
+        if settings.cache_dir in path.parents:
+            path = path.rename(path.with_name(f"{provisional_uid}{suffix}"))
+
+    check_path_in_storage = False
+    if use_existing_storage_key:
+        inferred_key = get_relative_path_to_directory(
+            path=path, directory=UPath(storage.root)
+        ).as_posix()
+        if key is None:
+            key = inferred_key
+        else:
+            if not key == inferred_key:
+                raise InvalidArgument(
+                    f"The path '{data}' is already in registered storage"
+                    f" '{storage.root}' with key '{inferred_key}'\nYou passed"
+                    f" conflicting key '{key}': please move the file before"
+                    " registering it."
+                )
+        check_path_in_storage = True
+    else:
+        storage = default_storage
+
+    log_storage_hint(
+        check_path_in_storage=check_path_in_storage,
+        storage=storage,
+        key=key,
+        uid=provisional_uid,
+        suffix=suffix,
+        is_dir=n_files is not None,
+    )
+
+    # do we use a virtual or an actual storage key?
+    key_is_virtual = settings.creation._artifact_use_virtual_keys
+
+    # if the file is already in storage, independent of the default
+    # we use an actual storage key
+    if check_path_in_storage:
+        key_is_virtual = False
+
+    kwargs = {
+        "uid": provisional_uid,
+        "suffix": suffix,
+        "hash": hash,
+        "_hash_type": hash_type,
+        "key": key,
+        "size": size,
+        "storage_id": storage.id,
+        # passing both the id and the object
+        # to make them both available immediately
+        # after object creation
+        "n_files": n_files,
+        "_overwrite_versions": n_files is not None,  # True for folder, False for file
+        "n_observations": None,  # to implement
+        "run_id": run.id if run is not None else None,
+        "run": run,
+        "_key_is_virtual": key_is_virtual,
+        "revises": revises,
+    }
+    if not isinstance(path, LocalPathClasses):
+        local_filepath = None
+        cloud_filepath = path
+    else:
+        local_filepath = path
+        cloud_filepath = None
+    privates = {
+        "local_filepath": local_filepath,
+        "cloud_filepath": cloud_filepath,
+        "memory_rep": memory_rep,
+        "check_path_in_storage": check_path_in_storage,
+    }
+    return kwargs, privates
+
+
+def log_storage_hint(
+    *,
+    check_path_in_storage: bool,
+    storage: Storage | None,
+    key: str | None,
+    uid: str,
+    suffix: str,
+    is_dir: bool,
+) -> None:
+    hint = ""
+    if check_path_in_storage:
+        display_root = storage.root  # type: ignore
+        # check whether path is local
+        if fsspec.utils.get_protocol(storage.root) == "file":  # type: ignore
+            # if it's a local path, check whether it's in the current working directory
+            root_path = Path(storage.root)  # type: ignore
+            if check_path_is_child_of_root(root_path, Path.cwd()):
+                # only display the relative path, not the fully resolved path
+                display_root = root_path.relative_to(Path.cwd())  # type: ignore
+        hint += f"path in storage '{display_root}'"  # type: ignore
+    else:
+        hint += "path content will be copied to default storage upon `save()`"
+    if key is None:
+        storage_key = auto_storage_key_from_artifact_uid(uid, suffix, is_dir)
+        hint += f" with key `None` ('{storage_key}')"
+    else:
+        hint += f" with key '{key}'"
+    logger.hint(hint)
+
+
+def data_is_anndata(data: AnnData | UPathStr) -> bool:
+    if isinstance(data, AnnData):
+        return True
+    if isinstance(data, (str, Path, UPath)):
+        data_path = UPath(data)
+        if ".h5ad" in data_path.suffixes:  # ".h5ad.gz" is a valid suffix
+            return True
+        elif data_path.suffix == ".zarr":
+            # ".anndata.zarr" is a valid suffix (core.storage._valid_suffixes)
+            # TODO: the suffix based check should likely be moved to identify_zarr_type
+            if ".anndata" in data_path.suffixes:
+                return True
+            # check only for local, expensive for cloud
+            if fsspec.utils.get_protocol(data_path.as_posix()) == "file":
+                return identify_zarr_type(data_path) == "anndata"
+            else:
+                logger.warning("We do not check if cloud zarr is AnnData or not")
+                return False
+    return False
+
+
+def data_is_mudata(data: MuData | UPathStr) -> bool:
+    if is_package_installed("mudata"):
+        from mudata import MuData
+
+        if isinstance(data, MuData):
+            return True
+    if isinstance(data, (str, Path)):
+        return UPath(data).suffix == ".h5mu"
+    return False
+
+
+def data_is_spatialdata(data: SpatialData | UPathStr) -> bool:
+    if is_package_installed("spatialdata"):
+        from spatialdata import SpatialData
+
+        if isinstance(data, SpatialData):
+            return True
+        if isinstance(data, (str, Path)):
+            if UPath(data).suffix == ".zarr":
+                # TODO: inconsistent with anndata, where we run the storage
+                # check only for local, expensive for cloud
+                return identify_zarr_type(data, check=False) == "spatialdata"
+        return False
+
+
+def _check_otype_artifact(
+    data: UPathStr | pd.DataFrame | AnnData | MuData | SpatialData,
+    otype: str | None = None,
+) -> str:
+    if otype is None:
+        if isinstance(data, pd.DataFrame):
+            logger.warning("data is a DataFrame, please use .from_df()")
+            otype = "DataFrame"
+            return otype
+
+        data_is_path = isinstance(data, (str, Path))
+        if data_is_anndata(data):
+            if not data_is_path:
+                logger.warning("data is an AnnData, please use .from_anndata()")
+            otype = "AnnData"
+        elif data_is_mudata(data):
+            if not data_is_path:
+                logger.warning("data is a MuData, please use .from_mudata()")
+            otype = "MuData"
+        elif data_is_spatialdata(data):
+            if not data_is_path:
+                logger.warning("data is a SpatialData, please use .from_spatialdata()")
+            otype = "SpatialData"
+        elif not data_is_path:  # UPath is a subclass of Path
+            raise TypeError("data has to be a string, Path, UPath")
+    return otype
+
+
+def _populate_subsequent_runs_(record: Union[Artifact, Collection], run: Run):
+    if record.run is None:
+        record.run = run
+    elif record.run != run:
+        record._subsequent_runs.add(run)
+
+
+# also see current_run() in core._data
+def get_run(run: Run | None) -> Run | None:
+    from lamindb import settings
+
+    from .._tracked import get_current_tracked_run
+    from ..core._context import context
+
+    if run is None:
+        run = get_current_tracked_run()
+        if run is None:
+            run = context.run
+        if run is None and not settings.creation.artifact_silence_missing_run_warning:
+            # here we check that this is not a read-only connection
+            # normally for our connection strings the read-only role name has _read in it
+            # not absolutely safe but the worst case is that the warning is not shown
+            instance = setup_settings.instance
+            if instance.dialect != "postgresql" or "_read" not in instance.db:
+                logger.warning(WARNING_RUN_TRANSFORM)
+    # suppress run by passing False
+    elif not run:
+        run = None
+    return run
+
+
+def save_staged_feature_sets(self: Artifact) -> None:
+    if hasattr(self, "_staged_feature_sets"):
+        from lamindb.models._feature_manager import get_schema_by_slot_
+
+        existing_staged_feature_sets = get_schema_by_slot_(self)
+        saved_staged_feature_sets = {}
+        for key, schema in self._staged_feature_sets.items():
+            if isinstance(schema, Schema) and schema._state.adding:
+                schema.save()
+                saved_staged_feature_sets[key] = schema
+            if key in existing_staged_feature_sets:
+                # remove existing feature set on the same slot
+                self.feature_sets.remove(existing_staged_feature_sets[key])
+        if len(saved_staged_feature_sets) > 0:
+            s = "s" if len(saved_staged_feature_sets) > 1 else ""
+            display_schema_keys = ",".join(
+                f"'{key}'" for key in saved_staged_feature_sets.keys()
+            )
+            logger.save(
+                f"saved {len(saved_staged_feature_sets)} feature set{s} for slot{s}:"
+                f" {display_schema_keys}"
+            )
+
+
+def save_schema_links(self: Artifact) -> None:
+    from lamindb.models.save import bulk_create
+
+    if hasattr(self, "_staged_feature_sets"):
+        links = []
+        for slot, schema in self._staged_feature_sets.items():
+            kwargs = {
+                "artifact_id": self.id,
+                "schema_id": schema.id,
+                "slot": slot,
+            }
+            links.append(Artifact.feature_sets.through(**kwargs))
+        bulk_create(links, ignore_conflicts=True)
+
+
+# can restore later if needed
+# def format_provenance(self, fk_data, print_types):
+#     type_str = lambda attr: (
+#         f": {get_related_model(self.__class__, attr).__name__}" if print_types else ""
+#     )
+
+#     return "".join(
+#         [
+#             f"    .{field_name}{type_str(field_name)} = {format_field_value(value.get('name'))}\n"
+#             for field_name, value in fk_data.items()
+#             if value.get("name")
+#         ]
+#     )
+
+# can restore later if needed
+# def format_input_of_runs(self, print_types):
+#     if self.id is not None and self.input_of_runs.exists():
+#         values = [format_field_value(i.started_at) for i in self.input_of_runs.all()]
+#         type_str = ": Run" if print_types else ""  # type: ignore
+#         return f"    .input_of_runs{type_str} = {', '.join(values)}\n"
+#     return ""
+
+
+def _describe_postgres(self):  # for Artifact & Collection
+    from ._describe import describe_general
+    from ._feature_manager import describe_features
+
+    model_name = self.__class__.__name__
+    msg = f"{colors.green(model_name)}{record_repr(self, include_foreign_keys=False).lstrip(model_name)}\n"
+    if self._state.db is not None and self._state.db != "default":
+        msg += f"  {colors.italic('Database instance')}\n"
+        msg += f"    slug: {self._state.db}\n"
+
+    if model_name == "Artifact":
+        result = get_artifact_with_related(
+            self,
+            include_feature_link=True,
+            include_fk=True,
+            include_m2m=True,
+            include_schema=True,
+        )
+    else:
+        result = get_artifact_with_related(self, include_fk=True, include_m2m=True)
+    related_data = result.get("related_data", {})
+    # TODO: fk_data = related_data.get("fk", {})
+
+    tree = describe_general(self)
+    if model_name == "Artifact":
+        return describe_features(
+            self,
+            tree=tree,
+            related_data=related_data,
+            with_labels=True,
+            print_params=hasattr(self, "kind") and self.kind == "model",
+        )
+    else:
+        return tree
+
+
+def _describe_sqlite(self, print_types: bool = False):  # for artifact & collection
+    from ._describe import describe_general
+    from ._feature_manager import describe_features
+    from .collection import Collection
+
+    model_name = self.__class__.__name__
+    msg = f"{colors.green(model_name)}{record_repr(self, include_foreign_keys=False).lstrip(model_name)}\n"
+    if self._state.db is not None and self._state.db != "default":
+        msg += f"  {colors.italic('Database instance')}\n"
+        msg += f"    slug: {self._state.db}\n"
+
+    fields = self._meta.fields
+    direct_fields = []
+    foreign_key_fields = []
+    for f in fields:
+        if f.is_relation:
+            foreign_key_fields.append(f.name)
+        else:
+            direct_fields.append(f.name)
+    if not self._state.adding:
+        # prefetch foreign key relationships
+        self = (
+            self.__class__.objects.using(self._state.db)
+            .select_related(*foreign_key_fields)
+            .get(id=self.id)
+        )
+        # prefetch m-2-m relationships
+        many_to_many_fields = []
+        if isinstance(self, (Collection, Artifact)):
+            many_to_many_fields.append("input_of_runs")
+        if isinstance(self, Artifact):
+            many_to_many_fields.append("feature_sets")
+        self = (
+            self.__class__.objects.using(self._state.db)
+            .prefetch_related(*many_to_many_fields)
+            .get(id=self.id)
+        )
+    tree = describe_general(self)
+    if model_name == "Artifact":
+        return describe_features(
+            self,
+            tree=tree,
+            with_labels=True,
+            print_params=hasattr(self, "kind") and self.kind == "kind",
+        )
+    else:
+        return tree
+
+
+def describe_artifact_collection(self):  # for artifact & collection
+    from ._describe import print_rich_tree
+
+    if not self._state.adding and connections[self._state.db].vendor == "postgresql":
+        tree = _describe_postgres(self)
+    else:
+        tree = _describe_sqlite(self)
+
+    print_rich_tree(tree)
+
+
+def validate_feature(feature: Feature, records: list[Record]) -> None:
+    """Validate feature record, adjust feature.dtype based on labels records."""
+    if not isinstance(feature, Feature):
+        raise TypeError("feature has to be of type Feature")
+    if feature._state.adding:
+        registries = {record.__class__.__get_name_with_module__() for record in records}
+        registries_str = "|".join(registries)
+        msg = f"ln.Feature(name='{feature.name}', type='cat[{registries_str}]').save()"
+        raise ValidationError(f"Feature not validated. If it looks correct: {msg}")
+
+
+def get_labels(
+    self,
+    feature: Feature,
+    mute: bool = False,
+    flat_names: bool = False,
+) -> QuerySet | dict[str, QuerySet] | list:
+    """{}"""  # noqa: D415
+    if not isinstance(feature, Feature):
+        raise TypeError("feature has to be of type Feature")
+    if feature.dtype is None or not feature.dtype.startswith("cat["):
+        raise ValueError("feature does not have linked labels")
+    registries_to_check = feature.dtype.replace("cat[", "").rstrip("]").split("|")
+    if len(registries_to_check) > 1 and not mute:
+        logger.warning("labels come from multiple registries!")
+    # return an empty query set if self.id is still None
+    if self.id is None:
+        return QuerySet(self.__class__)
+    qs_by_registry = {}
+    for registry in registries_to_check:
+        # currently need to distinguish between ULabel and non-ULabel, because
+        # we only have the feature information for Label
+        if registry == "ULabel":
+            links_to_labels = get_label_links(self, registry, feature)
+            label_ids = [link.ulabel_id for link in links_to_labels]
+            qs_by_registry[registry] = ULabel.objects.using(self._state.db).filter(
+                id__in=label_ids
+            )
+        elif registry in self.features._accessor_by_registry:
+            qs_by_registry[registry] = getattr(
+                self, self.features._accessor_by_registry[registry]
+            ).all()
+    if flat_names:
+        # returns a flat list of names
+        from .record import get_name_field
+
+        values = []
+        for v in qs_by_registry.values():
+            values += v.list(get_name_field(v))
+        return values
+    if len(registries_to_check) == 1 and registry in qs_by_registry:
+        return qs_by_registry[registry]
+    else:
+        return qs_by_registry
+
+
+def add_labels(
+    self,
+    records: Record | list[Record] | QuerySet | Iterable,
+    feature: Feature | None = None,
+    *,
+    field: StrField | None = None,
+    feature_ref_is_name: bool | None = None,
+    label_ref_is_name: bool | None = None,
+    from_curator: bool = False,
+) -> None:
+    """{}"""  # noqa: D415
+    if self._state.adding:
+        raise ValueError("Please save the artifact/collection before adding a label!")
+
+    if isinstance(records, (QuerySet, QuerySet.__base__)):  # need to have both
+        records = records.list()
+    if isinstance(records, (str, Record)):
+        records = [records]
+    if not isinstance(records, list):  # avoids warning for pd Series
+        records = list(records)
+    # create records from values
+    if len(records) == 0:
+        return None
+    if isinstance(records[0], str):  # type: ignore
+        records_validated = []
+        # feature is needed if we want to create records from values
+        if feature is None:
+            raise ValueError(
+                "Please pass a feature, e.g., via: label = ln.ULabel(name='my_label',"
+                " feature=ln.Feature(name='my_feature'))"
+            )
+        if feature.dtype.startswith("cat["):
+            orm_dict = dict_module_name_to_model_name(Artifact)
+            for reg in feature.dtype.replace("cat[", "").rstrip("]").split("|"):
+                registry = orm_dict.get(reg)
+                records_validated += registry.from_values(records, field=field)
+
+        # feature doesn't have registries and therefore can't create records from values
+        # ask users to pass records
+        if len(records_validated) == 0:
+            raise ValueError(
+                "Please pass a record (a `Record` object), not a string, e.g., via:"
+                " label"
+                f" = ln.ULabel(name='{records[0]}')"  # type: ignore
+            )
+        records = records_validated
+
+    for record in records:
+        if record._state.adding:
+            raise ValidationError(
+                f"{record} not validated. If it looks correct: record.save()"
+            )
+
+    if feature is None:
+        d = dict_related_model_to_related_name(self.__class__)
+        # strategy: group records by registry to reduce number of transactions
+        records_by_related_name: dict = {}
+        for record in records:
+            related_name = d.get(record.__class__.__get_name_with_module__())
+            if related_name is None:
+                raise ValueError(f"Can't add labels to {record.__class__} record!")
+            if related_name not in records_by_related_name:
+                records_by_related_name[related_name] = []
+            records_by_related_name[related_name].append(record)
+        for related_name, records in records_by_related_name.items():
+            getattr(self, related_name).add(*records)
+    else:
+        validate_feature(feature, records)  # type:ignore
+        records_by_registry = defaultdict(list)
+        feature_sets = self.feature_sets.filter(itype="Feature").all()
+        internal_features = set()  # type: ignore
+        if len(feature_sets) > 0:
+            for schema in feature_sets:
+                internal_features = internal_features.union(
+                    set(schema.members.values_list("name", flat=True))
+                )  # type: ignore
+        for record in records:
+            records_by_registry[record.__class__.__get_name_with_module__()].append(
+                record
+            )
+        for registry_name, records in records_by_registry.items():
+            if not from_curator and feature.name in internal_features:
+                raise ValidationError(
+                    "Cannot manually annotate internal feature with label. Please use ln.Curator"
+                )
+            if registry_name not in feature.dtype:
+                if not feature.dtype.startswith("cat"):
+                    raise ValidationError(
+                        f"Feature {feature.name} needs dtype='cat' for label annotation, currently has dtype='{feature.dtype}'"
+                    )
+                if feature.dtype == "cat":
+                    feature.dtype = f"cat[{registry_name}]"  # type: ignore
+                    feature.save()
+                elif registry_name not in feature.dtype:
+                    new_dtype = feature.dtype.rstrip("]") + f"|{registry_name}]"
+                    raise ValidationError(
+                        f"Label type {registry_name} is not valid for Feature(name='{feature.name}', dtype='{feature.dtype}'), consider updating to dtype='{new_dtype}'"
+                    )
+
+            if registry_name not in self.features._accessor_by_registry:
+                logger.warning(f"skipping {registry_name}")
+                continue
+            if len(records) == 0:
+                continue
+            features_labels = {
+                registry_name: [(feature, label_record) for label_record in records]
+            }
+            add_label_feature_links(
+                self.features,
+                features_labels,
+                feature_ref_is_name=feature_ref_is_name,
+                label_ref_is_name=label_ref_is_name,
+            )
+
+
+class Artifact(Record, IsVersioned, TracksRun, TracksUpdates):
+    # Note that this docstring has to be consistent with Curator.save_artifact()
+    """Datasets & models stored as files, folders, or arrays.
+
+    Artifacts manage data in local or remote storage.
+
+    Some artifacts are array-like, e.g., when stored as `.parquet`, `.h5ad`,
+    `.zarr`, or `.tiledb`.
+
+    Args:
+        data: `UPathStr` A path to a local or remote folder or file.
+        kind: `Literal["dataset", "model"] | None = None` Distinguish models from datasets from other files & folders.
+        key: `str | None = None` A path-like key to reference artifact in default storage, e.g., `"myfolder/myfile.fcs"`. Artifacts with the same key form a version family.
+        description: `str | None = None` A description.
+        revises: `Artifact | None = None` Previous version of the artifact. Is an alternative way to passing `key` to trigger a new version.
+        run: `Run | None = None` The run that creates the artifact.
+
+    .. dropdown:: Typical storage formats & their API accessors
+
+        Arrays:
+
+        - Table: `.csv`, `.tsv`, `.parquet`, `.ipc` ⟷ `DataFrame`, `pyarrow.Table`
+        - Annotated matrix: `.h5ad`, `.h5mu`, `.zrad` ⟷ `AnnData`, `MuData`
+        - Generic array: HDF5 group, zarr group, TileDB store ⟷ HDF5, zarr, TileDB loaders
+
+        Non-arrays:
+
+        - Image: `.jpg`, `.png` ⟷ `np.ndarray`, ...
+        - Fastq: `.fastq` ⟷ /
+        - VCF: `.vcf` ⟷ /
+        - QC: `.html` ⟷ /
+
+        You'll find these values in the `suffix` & `accessor` fields.
+
+        LaminDB makes some default choices (e.g., serialize a `DataFrame` as a `.parquet` file).
+
+    See Also:
+        :class:`~lamindb.Storage`
+            Storage locations for artifacts.
+        :class:`~lamindb.Collection`
+            Collections of artifacts.
+        :meth:`~lamindb.Artifact.from_df`
+            Create an artifact from a `DataFrame`.
+        :meth:`~lamindb.Artifact.from_anndata`
+            Create an artifact from an `AnnData`.
+
+    Examples:
+
+        Create an artifact by passing `key`:
+
+        >>> artifact = ln.Artifact("./my_file.parquet", key="example_datasets/my_file.parquet").save()
+        >>> artifact = ln.Artifact("./my_folder", key="project1/my_folder").save()
+
+        Calling `.save()` uploads the file to the default storage location of your lamindb instance.
+        (If it's a local instance, the "upload" is a mere copy operation.)
+
+        If your artifact is already in the cloud, lamindb auto-populates the `key` field based on the S3 key and there is no upload:
+
+        >>> artifact = ln.Artifact("s3://my_bucket/my_folder/my_file.csv").save()
+
+        You can make a new version of the artifact with `key = "example_datasets/my_file.parquet"`
+
+        >>> artifact_v2 = ln.Artifact("./my_file.parquet", key="example_datasets/my_file.parquet").save()
+        >>> artifact_v2.versions.df()  # see all versions
+
+        .. dropdown:: Why does the API look this way?
+
+            It's inspired by APIs building on AWS S3.
+
+            Both boto3 and quilt select a bucket (a storage location in LaminDB) and define a target path through a `key` argument.
+
+            In `boto3 <https://boto3.amazonaws.com/v1/documentation/api/latest/reference/services/s3/bucket/upload_file.html>`__::
+
+                # signature: S3.Bucket.upload_file(filepath, key)
+                import boto3
+                s3 = boto3.resource('s3')
+                bucket = s3.Bucket('mybucket')
+                bucket.upload_file('/tmp/hello.txt', 'hello.txt')
+
+            In `quilt3 <https://docs.quiltdata.com/api-reference/bucket>`__::
+
+                # signature: quilt3.Bucket.put_file(key, filepath)
+                import quilt3
+                bucket = quilt3.Bucket('mybucket')
+                bucket.put_file('hello.txt', '/tmp/hello.txt')
+
+        Sometimes you want to avoid mapping the artifact into a file hierarchy, and you can then _just_ populate `description` instead:
+
+        >>> artifact = ln.Artifact("s3://my_bucket/my_folder", description="My folder").save()
+        >>> artifact = ln.Artifact("./my_local_folder", description="My local folder").save()
+
+        Because you can then not use `key`-based versioning you have to pass `revises` to make a new artifact version:
+
+        >>> artifact_v2 = ln.Artifact("./my_file.parquet", revises=old_artifact).save()
+
+        If an artifact with the exact same hash already exists, `Artifact()` returns the existing artifact. In concurrent workloads where
+        the same artifact is created multiple times, `Artifact()` doesn't yet return the existing artifact but creates a new one; `.save()` however
+        detects the duplication and will return the existing artifact.
+
+    """
+
+    class Meta(Record.Meta, IsVersioned.Meta, TracksRun.Meta, TracksUpdates.Meta):
+        abstract = False
+
+    _len_full_uid: int = 20
+    _len_stem_uid: int = 16
+
+    params: ParamManager = ParamManagerArtifact  # type: ignore
+    """Param manager.
+
+    Example::
+
+        artifact.params.add_values({
+            "hidden_size": 32,
+            "bottleneck_size": 16,
+            "batch_size": 32,
+            "preprocess_params": {
+                "normalization_type": "cool",
+                "subset_highlyvariable": True,
+            },
+        })
+    """
+
+    features: FeatureManager = FeatureManager  # type: ignore
+    """Feature manager.
+
+    Features denote dataset dimensions, i.e., the variables that measure labels & numbers.
+
+    Annotate with features & values::
+
+       artifact.features.add_values({
+            "species": organism,  # here, organism is an Organism record
+            "scientist": ['Barbara McClintock', 'Edgar Anderson'],
+            "temperature": 27.6,
+            "study": "Candidate marker study"
+       })
+
+    Query for features & values::
+
+        ln.Artifact.features.filter(scientist="Barbara McClintock")
+
+    Features may or may not be part of the artifact content in storage. For
+    instance, the :class:`~lamindb.Curator` flow validates the columns of a
+    `DataFrame`-like artifact and annotates it with features corresponding to
+    these columns. `artifact.features.add_values`, by contrast, does not
+    validate the content of the artifact.
+    """
+
+    @property
+    def labels(self) -> LabelManager:
+        """Label manager.
+
+        To annotate with labels, you typically use the registry-specific accessors,
+        for instance :attr:`~lamindb.Artifact.ulabels`::
+
+            candidate_marker_study = ln.ULabel(name="Candidate marker study").save()
+            artifact.ulabels.add(candidate_marker_study)
+
+        Similarly, you query based on these accessors::
+
+            ln.Artifact.filter(ulabels__name="Candidate marker study").all()
+
+        Unlike the registry-specific accessors, the `.labels` accessor provides
+        a way of associating labels with features::
+
+            study = ln.Feature(name="study", dtype="cat").save()
+            artifact.labels.add(candidate_marker_study, feature=study)
+
+        Note that the above is equivalent to::
+
+            artifact.features.add_values({"study": candidate_marker_study})
+        """
+        from ._label_manager import LabelManager
+
+        return LabelManager(self)
+
+    id: int = models.AutoField(primary_key=True)
+    """Internal id, valid only in one DB instance."""
+    uid: str = CharField(
+        editable=False, unique=True, db_index=True, max_length=_len_full_uid
+    )
+    """A universal random id."""
+    key: str | None = CharField(db_index=True, null=True)
+    """A (virtual) relative file path within the artifact's storage location.
+
+    Setting a `key` is useful to automatically group artifacts into a version family.
+
+    LaminDB defaults to a virtual file path to make renaming of data in object storage easy.
+
+    If you register existing files in a storage location, the `key` equals the
+    actual filepath on the underyling filesytem or object store.
+    """
+    description: str | None = CharField(db_index=True, null=True)
+    """A description."""
+    storage: Storage = ForeignKey(
+        Storage, PROTECT, related_name="artifacts", editable=False
+    )
+    """Storage location, e.g. an S3 or GCP bucket or a local directory."""
+    suffix: str = CharField(max_length=30, db_index=True, editable=False)
+    # Initially, we thought about having this be nullable to indicate folders
+    # But, for instance, .zarr is stored in a folder that ends with a .zarr suffix
+    """Path suffix or empty string if no canonical suffix exists.
+
+    This is either a file suffix (`".csv"`, `".h5ad"`, etc.) or the empty string "".
+    """
+    kind: ArtifactKind | None = CharField(
+        max_length=20,
+        db_index=True,
+        null=True,
+    )
+    """:class:`~lamindb.base.types.ArtifactKind` (default `None`)."""
+    otype: str | None = CharField(
+        max_length=64, db_index=True, null=True, editable=False
+    )
+    """Default Python object type, e.g., DataFrame, AnnData."""
+    size: int | None = BigIntegerField(
+        null=True, db_index=True, default=None, editable=False
+    )
+    """Size in bytes.
+
+    Examples: 1KB is 1e3 bytes, 1MB is 1e6, 1GB is 1e9, 1TB is 1e12 etc.
+    """
+    hash: str | None = CharField(
+        max_length=HASH_LENGTH, db_index=True, null=True, unique=True, editable=False
+    )
+    """Hash or pseudo-hash of artifact content.
+
+    Useful to ascertain integrity and avoid duplication.
+    """
+    n_files: int | None = BigIntegerField(
+        null=True, db_index=True, default=None, editable=False
+    )
+    """Number of files for folder-like artifacts, `None` for file-like artifacts.
+
+    Note that some arrays are also stored as folders, e.g., `.zarr` or `.tiledbsoma`.
+
+    .. versionchanged:: 1.0
+        Renamed from `n_objects` to `n_files`.
+    """
+    n_observations: int | None = BigIntegerField(
+        null=True, db_index=True, default=None, editable=False
+    )
+    """Number of observations.
+
+    Typically, this denotes the first array dimension.
+    """
+    _hash_type: str | None = CharField(
+        max_length=30, db_index=True, null=True, editable=False
+    )
+    """Type of hash."""
+    ulabels: ULabel = models.ManyToManyField(
+        ULabel, through="ArtifactULabel", related_name="artifacts"
+    )
+    """The ulabels measured in the artifact (:class:`~lamindb.ULabel`)."""
+    run: Run | None = ForeignKey(
+        Run,
+        PROTECT,
+        related_name="output_artifacts",
+        null=True,
+        default=None,
+        editable=False,
+    )
+    """Run that created the artifact."""
+    input_of_runs: Run = models.ManyToManyField(Run, related_name="input_artifacts")
+    """Runs that use this artifact as an input."""
+    _subsequent_runs: Run = models.ManyToManyField(
+        "Run",
+        related_name="_recreated_artifacts",
+        db_table="lamindb_artifact__previous_runs",  # legacy name, change in lamindb v2
+    )
+    """Runs that re-created the record after initial creation."""
+    collections: Collection
+    """The collections that this artifact is part of."""
+    schema: Schema | None = ForeignKey(
+        Schema,
+        PROTECT,
+        null=True,
+        default=None,
+        related_name="validated_artifacts",
+    )
+    """The schema that validated this artifact in a :class:`~lamindb.curators.Curator`."""
+    feature_sets: Schema = models.ManyToManyField(
+        Schema, related_name="artifacts", through="ArtifactSchema"
+    )
+    """The feature sets measured by the artifact."""
+    _feature_values: FeatureValue = models.ManyToManyField(
+        FeatureValue, through="ArtifactFeatureValue", related_name="artifacts"
+    )
+    """Non-categorical feature values for annotation."""
+    _param_values: ParamValue = models.ManyToManyField(
+        ParamValue, through="ArtifactParamValue", related_name="artifacts"
+    )
+    """Parameter values."""
+    _key_is_virtual: bool = BooleanField()
+    """Indicates whether `key` is virtual or part of an actual file path."""
+    # be mindful that below, passing related_name="+" leads to errors
+    _actions: Artifact = models.ManyToManyField(
+        "self", symmetrical=False, related_name="_action_targets"
+    )
+    """Actions to attach for the UI."""
+    created_by: User = ForeignKey(
+        "lamindb.User",
+        PROTECT,
+        default=current_user_id,
+        related_name="created_artifacts",
+        editable=False,
+    )
+    """Creator of record."""
+    _overwrite_versions: bool = BooleanField(default=None)
+    """Indicates whether to store or overwrite versions.
+
+    It defaults to False for file-like artifacts and to True for folder-like artifacts.
+    """
+    projects: Project
+    """Linked projects."""
+    references: Reference
+    """Linked references."""
+
+    @overload
+    def __init__(
+        self,
+        # we're not choosing the name "path" for this arg because
+        # it'd be confusing with `artifact.path`, which is not the same
+        # so "data" conveys better that this is input data that's ingested
+        # and will be moved to a target path at `artifact.path`
+        # also internally, we sometimes pass "data objects" like a DataFrame
+        # here; and we might refactor this but we might also keep that internal
+        # usage
+        data: UPathStr,
+        kind: ArtifactKind | None = None,
+        key: str | None = None,
+        description: str | None = None,
+        revises: Artifact | None = None,
+        run: Run | None = None,
+    ): ...
+
+    @overload
+    def __init__(
+        self,
+        *db_args,
+    ): ...
+
+    def __init__(
+        self,
+        *args,
+        **kwargs,
+    ):
+        self.features = FeatureManager(self)  # type: ignore
+        self.params = ParamManager(self)  # type: ignore
+        # Below checks for the Django-internal call in from_db()
+        # it'd be better if we could avoid this, but not being able to create a Artifact
+        # from data with the default constructor renders the central class of the API
+        # essentially useless
+        # The danger below is not that a user might pass as many args (12 of it), but rather
+        # that at some point the Django API might change; on the other hand, this
+        # condition of for calling the constructor based on kwargs should always
+        # stay robust
+        if len(args) == len(self._meta.concrete_fields):
+            super().__init__(*args, **kwargs)
+            return None
+        # now we proceed with the user-facing constructor
+        if len(args) > 1:
+            raise ValueError("Only one non-keyword arg allowed: data")
+        data: str | Path = kwargs.pop("data") if len(args) == 0 else args[0]
+        kind: str = kwargs.pop("kind", None)
+        key: str | None = kwargs.pop("key", None)
+        run: Run | None = kwargs.pop("run", None)
+        description: str | None = kwargs.pop("description", None)
+        revises: Artifact | None = kwargs.pop("revises", None)
+        version: str | None = kwargs.pop("version", None)
+        if "visibility" in kwargs:  # backward compat
+            _branch_code = kwargs.pop("visibility")
+        elif "_branch_code" in kwargs:
+            _branch_code = kwargs.pop("_branch_code")
+        else:
+            _branch_code = 1
+        format = kwargs.pop("format", None)
+        _is_internal_call = kwargs.pop("_is_internal_call", False)
+        skip_check_exists = kwargs.pop("skip_check_exists", False)
+        if "default_storage" in kwargs:
+            default_storage = kwargs.pop("default_storage")
+        else:
+            if setup_settings.instance.keep_artifacts_local:
+                default_storage = setup_settings.instance.storage_local.record
+            else:
+                default_storage = setup_settings.instance.storage.record
+        using_key = kwargs.pop("using_key", None)
+        otype = kwargs.pop("otype") if "otype" in kwargs else None
+        otype = _check_otype_artifact(data=data, otype=otype)
+        if "type" in kwargs:
+            logger.warning("`type` will be removed soon, please use `kind`")
+            kind = kwargs.pop("type")
+        if not len(kwargs) == 0:
+            valid_keywords = ", ".join([val[0] for val in _get_record_kwargs(Artifact)])
+            raise FieldValidationError(
+                f"Only {valid_keywords} can be passed, you passed: {kwargs}"
+            )
+        if revises is not None and key is not None and revises.key != key:
+            note = message_update_key_in_version_family(
+                suid=revises.stem_uid,
+                existing_key=revises.key,
+                new_key=key,
+                registry="Artifact",
+            )
+            raise ValueError(
+                f"`key` is {key}, but `revises.key` is '{revises.key}'\n\n Either do *not* pass `key`.\n\n{note}"
+            )
+        if revises is not None:
+            if not isinstance(revises, Artifact):
+                raise TypeError("`revises` has to be of type `Artifact`")
+            if description is None:
+                description = revises.description
+        if key is not None and AUTO_KEY_PREFIX in key:
+            raise ValueError(
+                f"Do not pass key that contains a managed storage path in `{AUTO_KEY_PREFIX}`"
+            )
+        # below is for internal calls that require defining the storage location
+        # ahead of constructing the Artifact
+        if isinstance(data, (str, Path)) and AUTO_KEY_PREFIX in str(data):
+            if _is_internal_call:
+                is_automanaged_path = True
+                user_provided_key = key
+                key = None
+            else:
+                raise ValueError(
+                    f"Do not pass path inside the `{AUTO_KEY_PREFIX}` directory."
+                )
+        else:
+            is_automanaged_path = False
+        provisional_uid, revises = create_uid(revises=revises, version=version)
+        kwargs_or_artifact, privates = get_artifact_kwargs_from_data(
+            data=data,
+            key=key,
+            run=run,
+            format=format,
+            provisional_uid=provisional_uid,
+            version=version,
+            default_storage=default_storage,
+            using_key=using_key,
+            skip_check_exists=skip_check_exists,
+        )
+
+        # an object with the same hash already exists
+        if isinstance(kwargs_or_artifact, Artifact):
+            from .record import init_self_from_db, update_attributes
+
+            init_self_from_db(self, kwargs_or_artifact)
+            # adding "key" here is dangerous because key might be auto-populated
+            attr_to_update = {"description": description}
+            if kwargs_or_artifact._key_is_virtual and kwargs_or_artifact.key is None:
+                attr_to_update["key"] = key
+            elif self.key != key and key is not None:
+                logger.warning(
+                    f"key {self.key} on existing artifact differs from passed key {key}"
+                )
+            update_attributes(self, attr_to_update)
+            return None
+        else:
+            kwargs = kwargs_or_artifact
+
+        if revises is None:
+            revises = kwargs_or_artifact.pop("revises")
+
+        if data is not None:
+            self._local_filepath = privates["local_filepath"]
+            self._cloud_filepath = privates["cloud_filepath"]
+            self._memory_rep = privates["memory_rep"]
+            self._to_store = not privates["check_path_in_storage"]
+
+        if is_automanaged_path and _is_internal_call:
+            kwargs["_key_is_virtual"] = True
+            assert AUTO_KEY_PREFIX in kwargs["key"]  # noqa: S101
+            uid = (
+                kwargs["key"].replace(AUTO_KEY_PREFIX, "").replace(kwargs["suffix"], "")
+            )
+            kwargs["key"] = user_provided_key
+            if revises is not None:
+                assert uid.startswith(revises.stem_uid)  # noqa: S101
+            if len(uid) == 16:
+                if revises is None:
+                    uid += "0000"
+                else:
+                    uid, revises = create_uid(revises=revises, version=version)
+            kwargs["uid"] = uid
+
+        # only set key now so that we don't do a look-up on it in case revises is passed
+        if revises is not None:
+            kwargs["key"] = revises.key
+
+        kwargs["kind"] = kind
+        kwargs["version"] = version
+        kwargs["description"] = description
+        kwargs["_branch_code"] = _branch_code
+        kwargs["otype"] = otype
+        kwargs["revises"] = revises
+        # this check needs to come down here because key might be populated from an
+        # existing file path during get_artifact_kwargs_from_data()
+        if (
+            kwargs["key"] is None
+            and kwargs["description"] is None
+            and kwargs["run"] is None
+        ):
+            raise ValueError("Pass one of key, run or description as a parameter")
+
+        super().__init__(**kwargs)
+
+    @property
+    @deprecated("kind")
+    def type(self) -> str:
+        return self.kind
+
+    @property
+    @deprecated("otype")
+    def _accessor(self) -> str:
+        return self.otype
+
+    @property
+    def transform(self) -> Transform | None:
+        """Transform whose run created the artifact."""
+        return self.run.transform if self.run is not None else None
+
+    @property
+    @deprecated("n_files")
+    def n_objects(self) -> int:
+        return self.n_files
+
+    # add the below because this is what people will have in their code
+    # if they implement the recommended migration strategy
+    # - FeatureSet -> Schema
+    # - featureset -> schema
+    # - feature_set -> schema
+    # @property
+    # def schemas(self) -> QuerySet[Schema]:
+    #     """Schemas linked to artifact via many-to-many relationship.
+
+    #     Is now mediating the private `.feature_sets` relationship during
+    #     a transition period to better schema management.
+
+    #     .. versionchanged: 1.0
+    #        Was previously called `.feature_sets`.
+
+    #     """
+    #     return self.feature_sets
+
+    @property
+    def path(self) -> Path:
+        """Path.
+
+        File in cloud storage, here AWS S3:
+
+        >>> artifact = ln.Artifact("s3://my-bucket/my-file.csv").save()
+        >>> artifact.path
+        S3QueryPath('s3://my-bucket/my-file.csv')
+
+        File in local storage:
+
+        >>> ln.Artifact("./myfile.csv", key="myfile.csv").save()
+        >>> artifact.path
+        PosixPath('/home/runner/work/lamindb/lamindb/docs/guide/mydata/myfile.csv')
+        """
+        from lamindb import settings
+
+        filepath, _ = filepath_from_artifact(self, using_key=settings._using_key)
+        return filepath
+
+    @property
+    def _cache_path(self) -> UPath:
+        from lamindb import settings
+
+        filepath, cache_key = filepath_cache_key_from_artifact(
+            self, using_key=settings._using_key
+        )
+        if isinstance(filepath, LocalPathClasses):
+            return filepath
+        return setup_settings.paths.cloud_to_local_no_update(
+            filepath, cache_key=cache_key
+        )
+
+    @classmethod
+    def from_df(
+        cls,
+        df: pd.DataFrame,
+        *,
+        key: str | None = None,
+        description: str | None = None,
+        run: Run | None = None,
+        revises: Artifact | None = None,
+        **kwargs,
+    ) -> Artifact:
+        """Create from `DataFrame`, validate & link features.
+
+        Args:
+            df: A `DataFrame` object.
+            key: A relative path within default storage,
+                e.g., `"myfolder/myfile.parquet"`.
+            description: A description.
+            revises: An old version of the artifact.
+            run: The run that creates the artifact.
+
+        See Also:
+            :meth:`~lamindb.Collection`
+                Track collections.
+            :class:`~lamindb.Feature`
+                Track features.
+
+        Examples:
+            >>> df = ln.core.datasets.df_iris_in_meter_batch1()
+            >>> df.head()
+              sepal_length sepal_width petal_length petal_width iris_organism_code
+            0        0.051       0.035        0.014       0.002                 0
+            1        0.049       0.030        0.014       0.002                 0
+            2        0.047       0.032        0.013       0.002                 0
+            3        0.046       0.031        0.015       0.002                 0
+            4        0.050       0.036        0.014       0.002                 0
+            >>> artifact = ln.Artifact.from_df(df, description="Iris flower collection batch1")
+            >>> artifact.save()
+        """
+        artifact = Artifact(  # type: ignore
+            data=df,
+            key=key,
+            run=run,
+            description=description,
+            revises=revises,
+            otype="DataFrame",
+            kind="dataset",
+            **kwargs,
+        )
+        artifact.n_observations = len(df)
+        return artifact
+
+    @classmethod
+    def from_anndata(
+        cls,
+        adata: Union[AnnData, UPathStr],
+        *,
+        key: str | None = None,
+        description: str | None = None,
+        run: Run | None = None,
+        revises: Artifact | None = None,
+        **kwargs,
+    ) -> Artifact:
+        """Create from ``AnnData``, validate & link features.
+
+        Args:
+            adata: An `AnnData` object or a path of AnnData-like.
+            key: A relative path within default storage,
+                e.g., `"myfolder/myfile.h5ad"`.
+            description: A description.
+            revises: An old version of the artifact.
+            run: The run that creates the artifact.
+
+        See Also:
+
+            :meth:`~lamindb.Collection`
+                Track collections.
+            :class:`~lamindb.Feature`
+                Track features.
+
+        Examples:
+            >>> import bionty as bt
+            >>> bt.settings.organism = "human"
+            >>> adata = ln.core.datasets.anndata_with_obs()
+            >>> artifact = ln.Artifact.from_anndata(adata, description="mini anndata with obs")
+            >>> artifact.save()
+        """
+        if not data_is_anndata(adata):
+            raise ValueError(
+                "data has to be an AnnData object or a path to AnnData-like"
+            )
+        _anndata_n_observations(adata)
+        artifact = Artifact(  # type: ignore
+            data=adata,
+            key=key,
+            run=run,
+            description=description,
+            revises=revises,
+            otype="AnnData",
+            kind="dataset",
+            **kwargs,
+        )
+        # this is done instead of _anndata_n_observations(adata)
+        # because we need a proper path through create_path for cloud paths
+        # for additional upath options etc that create_path adds
+        obj_for_obs: AnnData | UPath
+        if hasattr(artifact, "_memory_rep") and artifact._memory_rep is not None:
+            obj_for_obs = artifact._memory_rep
+        else:
+            # returns ._local_filepath for local files
+            # and the proper path through create_path for cloud paths
+            obj_for_obs = artifact.path
+        artifact.n_observations = _anndata_n_observations(obj_for_obs)
+        return artifact
+
+    @classmethod
+    def from_mudata(
+        cls,
+        mdata: Union[MuData, UPathStr],
+        *,
+        key: str | None = None,
+        description: str | None = None,
+        run: Run | None = None,
+        revises: Artifact | None = None,
+        **kwargs,
+    ) -> Artifact:
+        """Create from ``MuData``, validate & link features.
+
+        Args:
+            mdata: A `MuData` object.
+            key: A relative path within default storage,
+                e.g., `"myfolder/myfile.h5mu"`.
+            description: A description.
+            revises: An old version of the artifact.
+            run: The run that creates the artifact.
+
+        See Also:
+            :meth:`~lamindb.Collection`
+                Track collections.
+            :class:`~lamindb.Feature`
+                Track features.
+
+        Examples:
+            >>> import bionty as bt
+            >>> bt.settings.organism = "human"
+            >>> mdata = ln.core.datasets.mudata_papalexi21_subset()
+            >>> artifact = ln.Artifact.from_mudata(mdata, description="a mudata object")
+            >>> artifact.save()
+        """
+        if not data_is_mudata(mdata):
+            raise ValueError("data has to be a MuData object or a path to MuData-like")
+        artifact = Artifact(  # type: ignore
+            data=mdata,
+            key=key,
+            run=run,
+            description=description,
+            revises=revises,
+            otype="MuData",
+            kind="dataset",
+            **kwargs,
+        )
+        if not isinstance(mdata, UPathStr):
+            artifact.n_observations = mdata.n_obs
+        return artifact
+
+    @classmethod
+    def from_spatialdata(
+        cls,
+        sdata: Union[SpatialData, UPathStr],
+        *,
+        key: str | None = None,
+        description: str | None = None,
+        run: Run | None = None,
+        revises: Artifact | None = None,
+        **kwargs,
+    ) -> Artifact:
+        """Create from ``SpatialData``, validate & link features.
+
+        Args:
+            mdata: A `SpatialData` object.
+            key: A relative path within default storage,
+                e.g., `"myfolder/myfile.zarr"`.
+            description: A description.
+            revises: An old version of the artifact.
+            run: The run that creates the artifact.
+
+        See Also:
+            :meth:`~lamindb.Collection`
+                Track collections.
+            :class:`~lamindb.Feature`
+                Track features.
+
+        Examples:
+            >>> artifact = ln.Artifact.from_spatialdata(sdata, key="my_dataset.zarr")
+        """
+        if not data_is_spatialdata(sdata):
+            raise ValueError(
+                "data has to be a SpatialData object or a path to SpatialData-like"
+            )
+        artifact = Artifact(  # type: ignore
+            data=sdata,
+            key=key,
+            run=run,
+            description=description,
+            revises=revises,
+            otype="SpatialData",
+            kind="dataset",
+            **kwargs,
+        )
+        # ill-defined https://scverse.zulipchat.com/#narrow/channel/315824-spatial/topic/How.20to.20calculate.20the.20number.20of.20observations.3F
+        # artifact.n_observations = ...
+        return artifact
+
+    @classmethod
+    def from_tiledbsoma(
+        cls,
+        path: UPathStr,
+        *,
+        key: str | None = None,
+        description: str | None = None,
+        run: Run | None = None,
+        revises: Artifact | None = None,
+        **kwargs,
+    ) -> Artifact:
+        """Create from a tiledbsoma store.
+
+        Args:
+            path: A tiledbsoma store with .tiledbsoma suffix.
+            key: A relative path within default storage,
+                e.g., `"myfolder/mystore.tiledbsoma"`.
+            description: A description.
+            revises: An old version of the artifact.
+            run: The run that creates the artifact.
+
+        Examples:
+            >>> artifact = ln.Artifact.from_tiledbsoma("s3://mybucket/store.tiledbsoma", description="a tiledbsoma store")
+            >>> artifact.save()
+        """
+        if UPath(path).suffix != ".tiledbsoma":
+            raise ValueError(
+                "A tiledbsoma store should have .tiledbsoma suffix to be registered."
+            )
+        artifact = Artifact(  # type: ignore
+            data=path,
+            key=key,
+            run=run,
+            description=description,
+            revises=revises,
+            otype="tiledbsoma",
+            kind="dataset",
+            **kwargs,
+        )
+        artifact.n_observations = _soma_n_observations(artifact.path)
+        return artifact
+
+    @classmethod
+    def from_dir(
+        cls,
+        path: UPathStr,
+        *,
+        key: str | None = None,
+        run: Run | None = None,
+    ) -> list[Artifact]:
+        """Create a list of artifact objects from a directory.
+
+        Hint:
+            If you have a high number of files (several 100k) and don't want to
+            track them individually, create a single :class:`~lamindb.Artifact` via
+            ``Artifact(path)`` for them. See, e.g., :doc:`docs:rxrx`.
+
+        Args:
+            path: Source path of folder.
+            key: Key for storage destination. If `None` and
+                directory is in a registered location, the inferred `key` will
+                reflect the relative position. If `None` and directory is outside
+                of a registered storage location, the inferred key defaults to `path.name`.
+            run: A `Run` object.
+
+        Examples:
+            >>> dir_path = ln.core.datasets.generate_cell_ranger_files("sample_001", ln.settings.storage)
+            >>> artifacts = ln.Artifact.from_dir(dir_path)
+            >>> ln.save(artifacts)
+        """
+        from lamindb import settings
+
+        folderpath: UPath = create_path(path)  # returns Path for local
+        default_storage = settings.storage.record
+        using_key = settings._using_key
+        storage, use_existing_storage = process_pathlike(
+            folderpath, default_storage, using_key
+        )
+        folder_key_path: PurePath | Path
+        if key is None:
+            if not use_existing_storage:
+                logger.warning(
+                    "folder is outside existing storage location, will copy files from"
+                    f" {path} to {storage.root}/{folderpath.name}"
+                )
+                folder_key_path = Path(folderpath.name)
+            else:
+                # maintain the hierachy within an existing storage location
+                folder_key_path = get_relative_path_to_directory(
+                    folderpath, UPath(storage.root)
+                )
+        else:
+            folder_key_path = Path(key)
+
+        folder_key = folder_key_path.as_posix()
+        # silence fine-grained logging
+        verbosity = settings.verbosity
+        verbosity_int = settings._verbosity_int
+        if verbosity_int >= 1:
+            settings.verbosity = "warning"
+        artifacts_dict = {}
+        for filepath in folderpath.rglob("*"):
+            if filepath.is_file():
+                relative_path = get_relative_path_to_directory(filepath, folderpath)
+                artifact_key = folder_key + "/" + relative_path.as_posix()
+                # if creating from rglob, we don't need to check for existence
+                artifact = Artifact(
+                    filepath, run=run, key=artifact_key, skip_check_exists=True
+                )
+                artifacts_dict[artifact.uid] = artifact
+        settings.verbosity = verbosity
+
+        # run sanity check on hashes
+        hashes = [
+            artifact.hash
+            for artifact in artifacts_dict.values()
+            if artifact.hash is not None
+        ]
+        uids = artifacts_dict.keys()
+        n_unique_hashes = len(set(hashes))
+        if n_unique_hashes == len(hashes):
+            artifacts = list(artifacts_dict.values())
+        else:
+            # consider exact duplicates (same id, same hash)
+            # below can't happen anymore because artifacts is a dict now
+            # if len(set(uids)) == len(set(hashes)):
+            #     logger.warning("dropping duplicate records in list of artifact records")
+            #     artifacts = list(set(uids))
+            # consider false duplicates (different id, same hash)
+            if not len(set(uids)) == n_unique_hashes:
+                seen_hashes = set()
+                non_unique_artifacts = {
+                    hash: artifact
+                    for hash, artifact in artifacts_dict.items()
+                    if artifact.hash in seen_hashes or seen_hashes.add(artifact.hash)  # type: ignore
+                }
+                display_non_unique = "\n    ".join(
+                    f"{artifact}" for artifact in non_unique_artifacts
+                )
+                logger.warning(
+                    "there are multiple artifact uids with the same hashes, dropping"
+                    f" {len(non_unique_artifacts)} duplicates out of"
+                    f" {len(artifacts_dict)} artifacts:\n    {display_non_unique}"
+                )
+                artifacts = [
+                    artifact
+                    for artifact in artifacts_dict.values()
+                    if artifact not in non_unique_artifacts.values()
+                ]
+        logger.success(
+            f"created {len(artifacts)} artifacts from directory using storage"
+            f" {storage.root} and key = {folder_key}/"
+        )
+        return artifacts
+
+    def replace(
+        self,
+        data: Union[UPathStr, pd.DataFrame, AnnData, MuData],
+        run: Run | None = None,
+        format: str | None = None,
+    ) -> None:
+        """Replace artifact content.
+
+        Args:
+            data: A file path.
+            run: The run that created the artifact gets
+                auto-linked if ``ln.track()`` was called.
+
+        Examples:
+            Say we made a change to the content of an artifact, e.g., edited the image
+            `paradisi05_laminopathic_nuclei.jpg`.
+
+            This is how we replace the old file in storage with the new file:
+
+            >>> artifact.replace("paradisi05_laminopathic_nuclei.jpg")
+            >>> artifact.save()
+
+            Note that this neither changes the storage key nor the filename.
+
+            However, it will update the suffix if it changes.
+        """
+        from lamindb import settings
+
+        default_storage = settings.storage.record
+        kwargs, privates = get_artifact_kwargs_from_data(
+            provisional_uid=self.uid,
+            data=data,
+            key=self.key,
+            run=run,
+            format=format,
+            default_storage=default_storage,
+            version=None,
+            is_replace=True,
+        )
+
+        # this artifact already exists
+        if privates is None:
+            return kwargs
+
+        check_path_in_storage = privates["check_path_in_storage"]
+        if check_path_in_storage:
+            err_msg = (
+                "Can only replace with a local path not in any Storage. "
+                f"This data is in {Storage.objects.get(id=kwargs['storage_id'])}."
+            )
+            raise ValueError(err_msg)
+
+        _overwrite_versions = kwargs["_overwrite_versions"]
+        if self._overwrite_versions != _overwrite_versions:
+            err_msg = "It is not allowed to replace "
+            err_msg += "a folder" if self._overwrite_versions else "a file"
+            err_msg += " with " + ("a folder." if _overwrite_versions else "a file.")
+            raise ValueError(err_msg)
+
+        if self.key is not None and not self._key_is_virtual:
+            key_path = PurePosixPath(self.key)
+            new_filename = f"{key_path.stem}{kwargs['suffix']}"
+            # the following will only be true if the suffix changes!
+            if key_path.name != new_filename:
+                self._clear_storagekey = self.key
+                self.key = str(key_path.with_name(new_filename))
+                # update old key with the new one so that checks in record pass
+                self._old_key = self.key
+                logger.warning(
+                    f"replacing the file will replace key '{key_path}' with '{self.key}'"
+                    f" and delete '{key_path}' upon `save()`"
+                )
+        else:
+            old_storage = auto_storage_key_from_artifact(self)
+            is_dir = self.n_files is not None
+            new_storage = auto_storage_key_from_artifact_uid(
+                self.uid, kwargs["suffix"], is_dir
+            )
+            if old_storage != new_storage:
+                self._clear_storagekey = old_storage
+                if self.key is not None:
+                    new_key_path = PurePosixPath(self.key).with_suffix(kwargs["suffix"])
+                    self.key = str(new_key_path)
+                    # update old key with the new one so that checks in record pass
+                    self._old_key = self.key
+
+        self.suffix = kwargs["suffix"]
+        self.size = kwargs["size"]
+        self.hash = kwargs["hash"]
+        self._hash_type = kwargs["_hash_type"]
+        self.run_id = kwargs["run_id"]
+        self.run = kwargs["run"]
+        self.n_files = kwargs["n_files"]
+
+        self._local_filepath = privates["local_filepath"]
+        self._cloud_filepath = privates["cloud_filepath"]
+        self._memory_rep = privates["memory_rep"]
+        # no need to upload if new file is already in storage
+        self._to_store = not check_path_in_storage
+
+    def open(
+        self, mode: str = "r", is_run_input: bool | None = None, **kwargs
+    ) -> Union[
+        AnnDataAccessor,
+        BackedAccessor,
+        SOMACollection,
+        SOMAExperiment,
+        SOMAMeasurement,
+        PyArrowDataset,
+    ]:
+        """Return a cloud-backed data object.
+
+        Works for `AnnData` (`.h5ad` and `.zarr`), generic `hdf5` and `zarr`,
+        `tiledbsoma` objects (`.tiledbsoma`), `pyarrow` compatible formats.
+
+        Args:
+            mode: can only be `"w"` (write mode) for `tiledbsoma` stores,
+                otherwise should be always `"r"` (read-only mode).
+
+        Notes:
+            For more info, see tutorial: :doc:`/arrays`.
+
+        Examples:
+
+            Read AnnData in backed mode from cloud:
+
+            >>> artifact = ln.Artifact.get(key="lndb-storage/pbmc68k.h5ad")
+            >>> artifact.open()
+            AnnDataAccessor object with n_obs × n_vars = 70 × 765
+                constructed for the AnnData object pbmc68k.h5ad
+                ...
+        """
+        if self._overwrite_versions and not self.is_latest:
+            raise ValueError(INCONSISTENT_STATE_MSG)
+        # all hdf5 suffixes including gzipped
+        h5_suffixes = [".h5", ".hdf5", ".h5ad"]
+        h5_suffixes += [s + ".gz" for s in h5_suffixes]
+        # ignore empty suffix for now
+        suffixes = (
+            (
+                "",
+                ".zarr",
+                ".anndata.zarr",
+                ".tiledbsoma",
+            )
+            + tuple(h5_suffixes)
+            + PYARROW_SUFFIXES
+            + tuple(
+                s + ".gz" for s in PYARROW_SUFFIXES
+            )  # this doesn't work for externally gzipped files, REMOVE LATER
+        )
+        if self.suffix not in suffixes:
+            raise ValueError(
+                "Artifact should have a zarr, h5, tiledbsoma object"
+                " or a compatible `pyarrow.dataset.dataset` directory"
+                " as the underlying data, please use one of the following suffixes"
+                f" for the object name: {', '.join(suffixes[1:])}."
+                f" Or no suffix for a folder with {', '.join(PYARROW_SUFFIXES)} files"
+                " (no mixing allowed)."
+            )
+        if self.suffix != ".tiledbsoma" and self.key != "soma" and mode != "r":
+            raise ValueError(
+                "Only a tiledbsoma store can be openened with `mode!='r'`."
+            )
+
+        from lamindb import settings
+        from lamindb.core.storage._backed_access import (
+            _track_writes_factory,
+            backed_access,
+        )
+
+        using_key = settings._using_key
+        filepath, cache_key = filepath_cache_key_from_artifact(
+            self, using_key=using_key
+        )
+        is_tiledbsoma_w = (
+            filepath.name == "soma" or self.suffix == ".tiledbsoma"
+        ) and mode == "w"
+        # consider the case where an object is already locally cached
+        localpath = setup_settings.paths.cloud_to_local_no_update(
+            filepath, cache_key=cache_key
+        )
+        if is_tiledbsoma_w:
+            open_cache = False
+        else:
+            open_cache = not isinstance(
+                filepath, LocalPathClasses
+            ) and not filepath.synchronize(localpath, just_check=True)
+        if open_cache:
+            try:
+                access = backed_access(localpath, mode, using_key, **kwargs)
+            except Exception as e:
+                if isinstance(filepath, LocalPathClasses):
+                    raise e
+                logger.warning(
+                    f"The cache might be corrupted: {e}. Trying to open directly."
+                )
+                access = backed_access(filepath, mode, using_key, **kwargs)
+                # happens only if backed_access has been successful
+                # delete the corrupted cache
+                if localpath.is_dir():
+                    shutil.rmtree(localpath)
+                else:
+                    localpath.unlink(missing_ok=True)
+        else:
+            access = backed_access(filepath, mode, using_key, **kwargs)
+            if is_tiledbsoma_w:
+
+                def finalize():
+                    nonlocal self, filepath, localpath
+                    if not isinstance(filepath, LocalPathClasses):
+                        _, hash, _, _ = get_stat_dir_cloud(filepath)
+                    else:
+                        # this can be very slow
+                        _, hash, _, _ = hash_dir(filepath)
+                    if self.hash != hash:
+                        from .record import init_self_from_db
+
+                        new_version = Artifact(
+                            filepath, revises=self, _is_internal_call=True
+                        ).save()
+                        init_self_from_db(self, new_version)
+
+                        if localpath != filepath and localpath.exists():
+                            shutil.rmtree(localpath)
+
+                access = _track_writes_factory(access, finalize)
+        # only call if open is successfull
+        _track_run_input(self, is_run_input)
+        return access
+
+    def load(self, is_run_input: bool | None = None, **kwargs) -> Any:
+        """Cache and load into memory.
+
+        See all :mod:`~lamindb.core.loaders`.
+
+        Examples:
+
+            Load a `DataFrame`-like artifact:
+
+            >>> artifact.load().head()
+            sepal_length sepal_width petal_length petal_width iris_organism_code
+            0        0.051       0.035        0.014       0.002                 0
+            1        0.049       0.030        0.014       0.002                 0
+            2        0.047       0.032        0.013       0.002                 0
+            3        0.046       0.031        0.015       0.002                 0
+            4        0.050       0.036        0.014       0.002                 0
+
+            Load an `AnnData`-like artifact:
+
+            >>> artifact.load()
+            AnnData object with n_obs × n_vars = 70 × 765
+
+            Fall back to :meth:`~lamindb.Artifact.cache` if no in-memory representation is configured:
+
+            >>> artifact.load()
+            PosixPath('/home/runner/work/lamindb/lamindb/docs/guide/mydata/.lamindb/jb7BY5UJoQVGMUOKiLcn.jpg')
+        """
+        from lamindb import settings
+
+        if self._overwrite_versions and not self.is_latest:
+            raise ValueError(INCONSISTENT_STATE_MSG)
+
+        if hasattr(self, "_memory_rep") and self._memory_rep is not None:
+            access_memory = self._memory_rep
+        else:
+            filepath, cache_key = filepath_cache_key_from_artifact(
+                self, using_key=settings._using_key
+            )
+            cache_path = _synchronize_cleanup_on_error(filepath, cache_key=cache_key)
+            try:
+                # cache_path is local so doesn't trigger any sync in load_to_memory
+                access_memory = load_to_memory(cache_path, **kwargs)
+            except Exception as e:
+                # raise the exception if it comes from not having a correct loader
+                # or if the original path is local
+                if isinstance(e, NotImplementedError) or isinstance(
+                    filepath, LocalPathClasses
+                ):
+                    raise e
+                logger.warning(
+                    f"The cache might be corrupted: {e}. Retrying to synchronize."
+                )
+                # delete the existing cache
+                if cache_path.is_dir():
+                    shutil.rmtree(cache_path)
+                else:
+                    cache_path.unlink(missing_ok=True)
+                # download again and try to load into memory
+                cache_path = _synchronize_cleanup_on_error(
+                    filepath, cache_key=cache_key
+                )
+                access_memory = load_to_memory(cache_path, **kwargs)
+        # only call if load is successfull
+        _track_run_input(self, is_run_input)
+        return access_memory
+
+    def cache(self, is_run_input: bool | None = None) -> Path:
+        """Download cloud artifact to local cache.
+
+        Follows synching logic: only caches an artifact if it's outdated in the local cache.
+
+        Returns a path to a locally cached on-disk object (say a `.jpg` file).
+
+        Examples:
+
+            Sync file from cloud and return the local path of the cache:
+
+            >>> artifact.cache()
+            PosixPath('/home/runner/work/Caches/lamindb/lamindb-ci/lndb-storage/pbmc68k.h5ad')
+        """
+        from lamindb import settings
+
+        if self._overwrite_versions and not self.is_latest:
+            raise ValueError(INCONSISTENT_STATE_MSG)
+
+        filepath, cache_key = filepath_cache_key_from_artifact(
+            self, using_key=settings._using_key
+        )
+        cache_path = _synchronize_cleanup_on_error(filepath, cache_key=cache_key)
+        # only call if sync is successfull
+        _track_run_input(self, is_run_input)
+        return cache_path
+
+    def delete(
+        self,
+        permanent: bool | None = None,
+        storage: bool | None = None,
+        using_key: str | None = None,
+    ) -> None:
+        """Trash or permanently delete.
+
+        A first call to `.delete()` puts an artifact into the trash (sets `_branch_code` to `-1`).
+        A second call permanently deletes the artifact.
+        If it is a folder artifact with multiple versions, deleting a non-latest version
+        will not delete the underlying storage by default (if `storage=True` is not specified).
+        Deleting the latest version will delete all the versions for folder artifacts.
+
+        FAQ: :doc:`docs:faq/storage`
+
+        Args:
+            permanent: Permanently delete the artifact (skip trash).
+            storage: Indicate whether you want to delete the artifact in storage.
+
+        Examples:
+
+            For an `Artifact` object `artifact`, call:
+
+            >>> artifact = ln.Artifact.filter(key="some.csv").one()
+            >>> artifact.delete() # delete a single file artifact
+
+            >>> artifact = ln.Artifact.filter(key="some.tiledbsoma". is_latest=False).first()
+            >>> artiact.delete() # delete an old version, the data will not be deleted
+
+            >>> artifact = ln.Artifact.filter(key="some.tiledbsoma". is_latest=True).one()
+            >>> artiact.delete() # delete all versions, the data will be deleted or prompted for deletion.
+        """
+        # this first check means an invalid delete fails fast rather than cascading through
+        # database and storage permission errors
+        if os.getenv("LAMINDB_MULTI_INSTANCE") is None:
+            isettings = setup_settings.instance
+            if self.storage.instance_uid != isettings.uid and (
+                storage or storage is None
+            ):
+                raise IntegrityError(
+                    "Cannot simply delete artifacts outside of this instance's managed storage locations."
+                    "\n(1) If you only want to delete the metadata record in this instance, pass `storage=False`"
+                    f"\n(2) If you want to delete the artifact in storage, please load the managing lamindb instance (uid={self.storage.instance_uid})."
+                    f"\nThese are all managed storage locations of this instance:\n{Storage.filter(instance_uid=isettings.uid).df()}"
+                )
+        # by default, we only move artifacts into the trash (_branch_code = -1)
+        trash__branch_code = -1
+        if self._branch_code > trash__branch_code and not permanent:
+            if storage is not None:
+                logger.warning("moving artifact to trash, storage arg is ignored")
+            # move to trash
+            self._branch_code = trash__branch_code
+            self.save()
+            logger.important(
+                f"moved artifact to trash (_branch_code = {trash__branch_code})"
+            )
+            return
+
+        # if the artifact is already in the trash
+        # permanent delete skips the trash
+        if permanent is None:
+            # ask for confirmation of permanent delete
+            response = input(
+                "Artifact record is already in trash! Are you sure you want to permanently"
+                " delete it? (y/n) You can't undo this action."
+            )
+            delete_record = response == "y"
+        else:
+            assert permanent  # noqa: S101
+            delete_record = True
+
+        if delete_record:
+            # need to grab file path before deletion
+            try:
+                path, _ = filepath_from_artifact(self, using_key)
+            except OSError:
+                # we can still delete the record
+                logger.warning("Could not get path")
+                storage = False
+            # only delete in storage if DB delete is successful
+            # DB delete might error because of a foreign key constraint violated etc.
+            if self._overwrite_versions and self.is_latest:
+                # includes self
+                for version in self.versions.all():
+                    _delete_skip_storage(version)
+            else:
+                self._delete_skip_storage()
+            # by default do not delete storage if deleting only a previous version
+            # and the underlying store is mutable
+            if self._overwrite_versions and not self.is_latest:
+                delete_in_storage = False
+                if storage:
+                    logger.warning(
+                        "Storage argument is ignored; can't delete storage on an previous version"
+                    )
+            elif self.key is None or self._key_is_virtual:
+                # do not ask for confirmation also if storage is None
+                delete_in_storage = storage is None or storage
+            else:
+                # for artifacts with non-virtual semantic storage keys (key is not None)
+                # ask for extra-confirmation
+                if storage is None:
+                    response = input(
+                        f"Are you sure to want to delete {path}? (y/n)  You can't undo"
+                        " this action."
+                    )
+                    delete_in_storage = response == "y"
+                else:
+                    delete_in_storage = storage
+            if not delete_in_storage:
+                logger.important(f"a file/folder remains here: {path}")
+            # we don't yet have logic to bring back the deleted metadata record
+            # in case storage deletion fails - this is important for ACID down the road
+            if delete_in_storage:
+                delete_msg = delete_storage(path, raise_file_not_found_error=False)
+                if delete_msg != "did-not-delete":
+                    logger.success(f"deleted {colors.yellow(f'{path}')}")
+
+    def save(self, upload: bool | None = None, **kwargs) -> Artifact:
+        """Save to database & storage.
+
+        Args:
+            upload: Trigger upload to cloud storage in instances with hybrid storage mode.
+
+        Examples:
+            >>> artifact = ln.Artifact("./myfile.csv", description="myfile")
+            >>> artifact.save()
+        """
+        state_was_adding = self._state.adding
+        print_progress = kwargs.pop("print_progress", True)
+        store_kwargs = kwargs.pop(
+            "store_kwargs", {}
+        )  # kwargs for .upload_from in the end
+        access_token = kwargs.pop("access_token", None)
+        local_path = None
+        if upload and setup_settings.instance.keep_artifacts_local:
+            # switch local storage location to cloud
+            local_path = self.path
+            self.storage_id = setup_settings.instance.storage.id
+            self._local_filepath = local_path
+            # switch to virtual storage key upon upload
+            # the local filepath is already cached at that point
+            self._key_is_virtual = True
+            # ensure that the artifact is uploaded
+            self._to_store = True
+
+        self._save_skip_storage(**kwargs)
+
+        from .save import check_and_attempt_clearing, check_and_attempt_upload
+
+        using_key = None
+        if "using" in kwargs:
+            using_key = kwargs["using"]
+        exception_upload = check_and_attempt_upload(
+            self,
+            using_key,
+            access_token=access_token,
+            print_progress=print_progress,
+            **store_kwargs,
+        )
+        if exception_upload is not None:
+            # we do not want to raise file not found on cleanup if upload of a file failed
+            # often it is ACID in the filesystem itself
+            # for example, s3 won't have the failed file, so just skip the delete in this case
+            raise_file_not_found_error = False
+            self._delete_skip_storage()
+        else:
+            # this is the case when it is cleaned on .replace
+            raise_file_not_found_error = True
+        # this is triggered by an exception in check_and_attempt_upload or by replace.
+        exception_clear = check_and_attempt_clearing(
+            self,
+            raise_file_not_found_error=raise_file_not_found_error,
+            using_key=using_key,
+        )
+        if exception_upload is not None:
+            raise RuntimeError(exception_upload)
+        if exception_clear is not None:
+            raise RuntimeError(exception_clear)
+        # this is only for keep_artifacts_local
+        if local_path is not None and not state_was_adding:
+            # only move the local artifact to cache if it was not newly created
+            local_path_cache = ln_setup.settings.cache_dir / local_path.name
+            # don't use Path.rename here because of cross-device link error
+            # https://laminlabs.slack.com/archives/C04A0RMA0SC/p1710259102686969
+            shutil.move(
+                local_path,  # type: ignore
+                local_path_cache,
+            )
+            logger.important(f"moved local artifact to cache: {local_path_cache}")
+        return self
+
+    def restore(self) -> None:
+        """Restore from trash.
+
+        Examples:
+            >>> artifact.restore()
+        """
+        self._branch_code = 1
+        self.save()
+
+    def describe(self) -> None:
+        """Describe relations of record.
+
+        Examples:
+            >>> artifact.describe()
+        """
+        return describe_artifact_collection(self)
+
+    def _populate_subsequent_runs(self, run: Run) -> None:
+        _populate_subsequent_runs_(self, run)
+
+
+# can't really just call .cache in .load because of double tracking
+def _synchronize_cleanup_on_error(
+    filepath: UPath, cache_key: str | None = None
+) -> UPath:
+    try:
+        cache_path = setup_settings.paths.cloud_to_local(
+            filepath, cache_key=cache_key, print_progress=True
+        )
+    except Exception as e:
+        if not isinstance(filepath, LocalPathClasses):
+            cache_path = setup_settings.paths.cloud_to_local_no_update(
+                filepath, cache_key=cache_key
+            )
+            if cache_path.is_dir():
+                shutil.rmtree(cache_path)
+            else:
+                cache_path.unlink(missing_ok=True)
+        raise e
+    return cache_path
+
+
+def _delete_skip_storage(artifact, *args, **kwargs) -> None:
+    super(Artifact, artifact).delete(*args, **kwargs)
+
+
+def _save_skip_storage(artifact, **kwargs) -> None:
+    save_staged_feature_sets(artifact)
+    super(Artifact, artifact).save(**kwargs)
+    save_schema_links(artifact)
+
+
+class ArtifactFeatureValue(BasicRecord, LinkORM, TracksRun):
+    id: int = models.BigAutoField(primary_key=True)
+    artifact: Artifact = ForeignKey(Artifact, CASCADE, related_name="+")
+    # we follow the lower() case convention rather than snake case for link models
+    featurevalue = ForeignKey(FeatureValue, PROTECT, related_name="+")
+
+    class Meta:
+        unique_together = ("artifact", "featurevalue")
+
+
+class ArtifactParamValue(BasicRecord, LinkORM, TracksRun):
+    id: int = models.BigAutoField(primary_key=True)
+    artifact: Artifact = ForeignKey(Artifact, CASCADE, related_name="+")
+    # we follow the lower() case convention rather than snake case for link models
+    paramvalue: ParamValue = ForeignKey(ParamValue, PROTECT, related_name="+")
+
+    class Meta:
+        unique_together = ("artifact", "paramvalue")
+
+
+def _track_run_input(
+    data: Artifact
+    | Iterable[Artifact],  # can also be Collection | Iterable[Collection]
+    is_run_input: bool | Run | None = None,
+    run: Run | None = None,
+):
+    from lamindb import settings
+
+    from .._tracked import get_current_tracked_run
+    from ..core._context import context
+    from .collection import Collection
+
+    if isinstance(is_run_input, Run):
+        run = is_run_input
+        is_run_input = True
+    elif run is None:
+        run = get_current_tracked_run()
+        if run is None:
+            run = context.run
+    # consider that data is an iterable of Data
+    data_iter: Iterable[Artifact] | Iterable[Collection] = (
+        [data] if isinstance(data, (Artifact, Collection)) else data
+    )
+    track_run_input = False
+    input_data = []
+    if run is not None:
+        # avoid cycles: data can't be both input and output
+        def is_valid_input(data: Artifact | Collection):
+            is_valid = False
+            if data._state.db == "default":
+                # things are OK if the record is on the default db
+                is_valid = True
+            elif data._state.db is None:
+                # if a record is not yet saved, it can't be an input
+                # we silently ignore because what likely happens is that
+                # the user works with an object that's about to be saved
+                # in the current Python session
+                is_valid = False
+            else:
+                # record is on another db
+                # we have to save the record into the current db with
+                # the run being attached to a transfer transform
+                logger.important(
+                    f"completing transfer to track {data.__class__.__name__}('{data.uid[:8]}') as input"
+                )
+                data.save()
+                is_valid = True
+            return (
+                data.run_id != run.id
+                and not data._state.adding  # this seems duplicated with data._state.db is None
+                and is_valid
+            )
+
+        input_data = [data for data in data_iter if is_valid_input(data)]
+        input_data_ids = [data.id for data in input_data]
+    if input_data:
+        data_class_name = input_data[0].__class__.__name__.lower()
+    # let us first look at the case in which the user does not
+    # provide a boolean value for `is_run_input`
+    # hence, we need to determine whether we actually want to
+    # track a run or not
+    if is_run_input is None:
+        # we don't have a run record
+        if run is None:
+            if settings.track_run_inputs:
+                # here we check that this is not a read-only connection
+                # normally for our connection strings the read-only role name has _read in it
+                # not absolutely safe but the worst case is that the warning is not shown
+                instance = setup_settings.instance
+                if instance.dialect != "postgresql" or "_read" not in instance.db:
+                    logger.warning(WARNING_NO_INPUT)
+        # assume we have a run record
+        else:
+            # assume there is non-cyclic candidate input data
+            if input_data:
+                if settings.track_run_inputs:
+                    transform_note = ""
+                    if len(input_data) == 1:
+                        if input_data[0].transform is not None:
+                            transform_note = (
+                                ", adding parent transform"
+                                f" {input_data[0].transform.id}"
+                            )
+                    logger.info(
+                        f"adding {data_class_name} ids {input_data_ids} as inputs for run"
+                        f" {run.id}{transform_note}"
+                    )
+                    track_run_input = True
+                else:
+                    logger.hint(
+                        "track these data as a run input by passing `is_run_input=True`"
+                    )
+    else:
+        track_run_input = is_run_input
+    if track_run_input:
+        if run is None:
+            raise ValueError("No run context set. Call `ln.track()`.")
+        # avoid adding the same run twice
+        run.save()
+        if data_class_name == "artifact":
+            LinkORM = run.input_artifacts.through
+            links = [
+                LinkORM(run_id=run.id, artifact_id=data_id)
+                for data_id in input_data_ids
+            ]
+        else:
+            LinkORM = run.input_collections.through
+            links = [
+                LinkORM(run_id=run.id, collection_id=data_id)
+                for data_id in input_data_ids
+            ]
+        LinkORM.objects.bulk_create(links, ignore_conflicts=True)
+        # generalize below for more than one data batch
+        if len(input_data) == 1:
+            if input_data[0].transform is not None:
+                run.transform.predecessors.add(input_data[0].transform)
+
+
+# privates currently dealt with separately
+# mypy: ignore-errors
+Artifact._delete_skip_storage = _delete_skip_storage
+Artifact._save_skip_storage = _save_skip_storage
+Artifact.view_lineage = view_lineage