image-match/python/py/Lib/site-packages/wandb/sdk/artifacts/artifact.py

"""Artifact class."""

from __future__ import annotations

import atexit
import contextlib
import json
import logging
import multiprocessing.dummy
import os
import re
import shutil
import stat
import tempfile
import time
from collections import deque
from collections.abc import Iterator, Sequence
from concurrent.futures import Executor, ThreadPoolExecutor, as_completed
from copy import copy
from dataclasses import asdict, replace
from datetime import timedelta
from itertools import filterfalse
from pathlib import Path, PurePosixPath
from typing import (  # noqa: UP035 (can't use `type` - shadows `Artifact.type`)
    IO,
    TYPE_CHECKING,
    Any,
    Callable,
    Final,
    Literal,
    Type,
)
from urllib.parse import quote, urljoin, urlparse

from pydantic import NonNegativeInt

import wandb
from wandb import data_types, env
from wandb._iterutils import one, unique_list
from wandb._pydantic import from_json
from wandb._strutils import nameof
from wandb.apis.normalize import normalize_exceptions
from wandb.apis.public import ArtifactCollection, ArtifactFiles, Run
from wandb.apis.public.utils import gql_compat
from wandb.data_types import WBValue
from wandb.errors import CommError
from wandb.errors.errors import UnsupportedError
from wandb.errors.term import termerror, termlog, termwarn
from wandb.proto import wandb_internal_pb2 as pb
from wandb.proto.wandb_telemetry_pb2 import Deprecated
from wandb.sdk import wandb_setup
from wandb.sdk.data_types._dtypes import Type as WBType
from wandb.sdk.data_types._dtypes import TypeRegistry
from wandb.sdk.lib import retry, telemetry
from wandb.sdk.lib.deprecation import warn_and_record_deprecation
from wandb.sdk.lib.filesystem import check_exists, system_preferred_path
from wandb.sdk.lib.hashutil import B64MD5, b64_to_hex_id, md5_file_b64
from wandb.sdk.lib.paths import FilePathStr, LogicalPath, StrPath, URIStr
from wandb.sdk.lib.runid import generate_fast_id, generate_id
from wandb.sdk.mailbox import MailboxHandle
from wandb.util import (
    alias_is_version_index,
    artifact_to_json,
    fsync_open,
    json_dumps_safer,
    uri_from_path,
    vendor_setup,
)

from ._factories import make_storage_policy
from ._gqlutils import org_info_from_entity, resolve_org_entity_name, server_supports
from ._validators import ensure_logged, ensure_not_finalized
from .artifact_download_logger import ArtifactDownloadLogger
from .artifact_instance_cache import (
    artifact_instance_cache,
    artifact_instance_cache_by_client_id,
)
from .artifact_manifest import ArtifactManifest
from .artifact_manifest_entry import ArtifactManifestEntry
from .artifact_manifests.artifact_manifest_v1 import ArtifactManifestV1
from .artifact_state import ArtifactState
from .artifact_ttl import ArtifactTTL
from .exceptions import (
    ArtifactNotLoggedError,
    TooFewItemsError,
    TooManyItemsError,
    WaitTimeoutError,
)
from .staging import get_staging_dir
from .storage_handlers.gcs_handler import _GCSIsADirectoryError
from .storage_policies._factories import make_http_session
from .storage_policies._multipart import should_multipart_download

reset_path = vendor_setup()

from wandb_gql import gql  # noqa: E402

reset_path()

if TYPE_CHECKING:
    from collections.abc import Iterable

    from wandb.apis.public import RetryingClient

    from ._generated import ArtifactFragment, ArtifactMembershipFragment
    from ._models.pagination import FileWithUrlConnection
    from ._validators import FullArtifactPath, LinkArtifactFields

logger = logging.getLogger(__name__)


_MB: Final[int] = 1024 * 1024
_FILE_EXECUTOR_WORKERS = 32
_MP_EXECUTOR_WORKERS = 32


class Artifact:
    """Flexible and lightweight building block for dataset and model versioning.

    Construct an empty W&B Artifact. Populate an artifacts contents with methods that
    begin with `add`. Once the artifact has all the desired files, you can call
    `run.log_artifact()` to log it.

    Args:
        name (str): A human-readable name for the artifact. Use the name to identify
            a specific artifact in the W&B App UI or programmatically. You can
            interactively reference an artifact with the `use_artifact` Public API.
            A name can contain letters, numbers, underscores, hyphens, and dots.
            The name must be unique across a project.
        type (str): The artifact's type. Use the type of an artifact to both organize
            and differentiate artifacts. You can use any string that contains letters,
            numbers, underscores, hyphens, and dots. Common types include `dataset` or
            `model`. Include `model` within your type string if you want to link the
            artifact to the W&B Model Registry.
            Note that some types reserved for internal use and cannot be set by users.
            Such types include `job` and types that start with `wandb-`.
        description (str | None) = None: A description of the artifact. For Model or
            Dataset Artifacts, add documentation for your standardized team model or
            dataset card. View an artifact's description programmatically with the
            `Artifact.description` attribute or programmatically with the W&B App UI.
            W&B renders the description as markdown in the W&B App.
        metadata (dict[str, Any] | None) = None: Additional information about an artifact.
            Specify metadata as a dictionary of key-value pairs. You can specify no more
            than 100 total keys.
        incremental: Use `Artifact.new_draft()` method instead to modify an
            existing artifact.
        use_as: Deprecated.

    Returns:
        An `Artifact` object.
    """

    _TMP_DIR = tempfile.TemporaryDirectory("wandb-artifacts")
    atexit.register(_TMP_DIR.cleanup)

    def __init__(
        self,
        name: str,
        type: str,
        description: str | None = None,
        metadata: dict[str, Any] | None = None,
        incremental: bool = False,
        use_as: str | None = None,
        storage_region: str | None = None,
    ) -> None:
        from wandb.sdk.artifacts._internal_artifact import InternalArtifact

        from ._validators import (
            validate_artifact_name,
            validate_artifact_type,
            validate_metadata,
        )

        if not re.match(r"^[a-zA-Z0-9_\-.]+$", name):
            raise ValueError(
                f"Artifact name may only contain alphanumeric characters, dashes, "
                f"underscores, and dots. Invalid name: {name!r}"
            )

        if incremental and not isinstance(self, InternalArtifact):
            termwarn("Using experimental arg `incremental`")

        # Internal.
        self._client: RetryingClient | None = None

        self._tmp_dir: tempfile.TemporaryDirectory | None = None
        self._added_objs: dict[int, tuple[WBValue, ArtifactManifestEntry]] = {}
        self._added_local_paths: dict[str, ArtifactManifestEntry] = {}
        self._save_handle: MailboxHandle[pb.Result] | None = None
        self._download_roots: set[str] = set()
        # Set by new_draft(), otherwise the latest artifact will be used as the base.
        self._base_id: str | None = None
        # Properties.
        self._id: str | None = None

        # Client IDs don't need cryptographic strength, so use a faster implementation.
        self._client_id: str = generate_fast_id(128)
        self._sequence_client_id: str = generate_fast_id(128)

        self._entity: str | None = None
        self._project: str | None = None
        self._name: str = validate_artifact_name(name)  # includes version after saving
        self._version: str | None = None
        self._source_entity: str | None = None
        self._source_project: str | None = None
        self._source_name: str = name  # includes version after saving
        self._source_version: str | None = None
        self._source_artifact: Artifact | None = None
        self._is_link: bool = False
        self._type: str = validate_artifact_type(type, name)
        self._description: str | None = description
        self._metadata: dict[str, Any] = validate_metadata(metadata)
        self._ttl_duration_seconds: int | None = None
        self._ttl_is_inherited: bool = True
        self._ttl_changed: bool = False
        self._aliases: list[str] = []
        self._saved_aliases: list[str] = []
        self._tags: list[str] = []
        self._saved_tags: list[str] = []
        self._distributed_id: str | None = None
        self._incremental: bool = incremental
        if use_as is not None:
            warn_and_record_deprecation(
                feature=Deprecated(artifact__init_use_as=True),
                message=(
                    "`use_as` argument is deprecated and does not affect the behaviour of `wandb.Artifact()`"
                ),
            )
        self._use_as: str | None = None
        self._state: ArtifactState = ArtifactState.PENDING

        # NOTE: These fields only reflect the last fetched response from the
        # server, if any. If the ArtifactManifest has already been fetched and/or
        # populated locally, it should take priority when determining these values.
        self._size: NonNegativeInt | None = None
        self._digest: str | None = None

        self._manifest: ArtifactManifest | None = ArtifactManifestV1(
            storage_policy=make_storage_policy(region=storage_region)
        )

        self._commit_hash: str | None = None
        self._file_count: int | None = None
        self._created_at: str | None = None
        self._updated_at: str | None = None
        self._final: bool = False
        self._history_step: int | None = None
        self._linked_artifacts: list[Artifact] = []

        self._fetch_file_urls_decorated: Callable[..., Any] | None = None

        # Cache.
        artifact_instance_cache_by_client_id[self._client_id] = self

    def __repr__(self) -> str:
        return f"<Artifact {self.id or self.name}>"

    @classmethod
    def _from_id(cls, artifact_id: str, client: RetryingClient) -> Artifact | None:
        from ._generated import ARTIFACT_BY_ID_GQL, ArtifactByID
        from ._validators import FullArtifactPath

        if cached_artifact := artifact_instance_cache.get(artifact_id):
            return cached_artifact

        gql_op = gql(ARTIFACT_BY_ID_GQL)

        data = client.execute(gql_op, variable_values={"id": artifact_id})
        result = ArtifactByID.model_validate(data)

        if (artifact := result.artifact) is None:
            return None

        src_collection = artifact.artifact_sequence
        src_project = src_collection.project

        entity_name = src_project.entity.name if src_project else ""
        project_name = src_project.name if src_project else ""

        name = f"{src_collection.name}:v{artifact.version_index}"

        path = FullArtifactPath(prefix=entity_name, project=project_name, name=name)
        return cls._from_attrs(path, artifact, client)

    @classmethod
    def _membership_from_name(
        cls, *, path: FullArtifactPath, client: RetryingClient
    ) -> Artifact:
        from ._generated import (
            ARTIFACT_MEMBERSHIP_BY_NAME_GQL,
            ArtifactMembershipByName,
        )

        if not server_supports(client, pb.PROJECT_ARTIFACT_COLLECTION_MEMBERSHIP):
            raise UnsupportedError(
                "Querying for the artifact collection membership is not supported "
                "by this version of wandb server. Consider updating to the latest version."
            )

        gql_op = gql(ARTIFACT_MEMBERSHIP_BY_NAME_GQL)
        gql_vars = {"entity": path.prefix, "project": path.project, "name": path.name}
        data = client.execute(gql_op, variable_values=gql_vars)
        result = ArtifactMembershipByName.model_validate(data)

        if not (project := result.project):
            msg = f"project {path.project!r} not found under entity {path.prefix!r}"
            raise ValueError(msg)

        if not (membership := project.artifact_collection_membership):
            entity_project = f"{path.prefix}/{path.project}"
            msg = f"artifact membership {path.name!r} not found in {entity_project!r}"
            raise ValueError(msg)

        return cls._from_membership(membership, target=path, client=client)

    @classmethod
    def _from_name(
        cls,
        *,
        path: FullArtifactPath,
        client: RetryingClient,
        enable_tracking: bool = False,
    ) -> Artifact:
        from ._generated import ARTIFACT_BY_NAME_GQL, ArtifactByName

        if server_supports(client, pb.PROJECT_ARTIFACT_COLLECTION_MEMBERSHIP):
            return cls._membership_from_name(path=path, client=client)

        gql_vars = {
            "entity": path.prefix,
            "project": path.project,
            "name": path.name,
            "enableTracking": enable_tracking,
        }
        gql_op = gql(ARTIFACT_BY_NAME_GQL)
        data = client.execute(gql_op, variable_values=gql_vars)
        result = ArtifactByName.model_validate(data)

        if not (project := result.project):
            msg = f"project {path.project!r} not found in entity {path.prefix!r}"
            raise ValueError(msg)

        if not (artifact := project.artifact):
            entity_project = f"{path.prefix}/{path.project}"
            msg = f"artifact {path.name!r} not found in {entity_project!r}"
            raise ValueError(msg)

        return cls._from_attrs(path, artifact, client)

    @classmethod
    def _from_membership(
        cls,
        membership: ArtifactMembershipFragment,
        target: FullArtifactPath,
        client: RetryingClient,
    ) -> Artifact:
        from ._validators import is_artifact_registry_project

        if not (
            (collection := membership.artifact_collection)
            and (name := collection.name)
            and (proj := collection.project)
        ):
            raise ValueError("Missing artifact collection project in GraphQL response")

        if is_artifact_registry_project(proj.name) and (
            target.project == "model-registry"
        ):
            wandb.termwarn(
                "This model registry has been migrated and will be discontinued. "
                f"Your request was redirected to the corresponding artifact {name!r} in the new registry. "
                f"Please update your paths to point to the migrated registry directly, '{proj.name}/{name}'."
            )

        # Update the target path to use the actual project/entity names returned in the
        # response, in case they differ from the original target path
        # E.g. uppercase vs lowercase, migrated legacy model registry, etc.
        new_target = replace(target, prefix=proj.entity.name, project=proj.name)

        if not (artifact := membership.artifact):
            raise ValueError(f"Artifact {target.to_str()!r} not found in response")

        return cls._from_attrs(new_target, artifact, client, membership=membership)

    @classmethod
    def _from_attrs(
        cls,
        path: FullArtifactPath,
        src_art: ArtifactFragment,
        client: RetryingClient,
        *,
        # aliases/version_index are taken from the membership, if given
        membership: ArtifactMembershipFragment | None = None,
    ) -> Artifact:
        # Placeholder is required to skip validation.
        artifact = cls("placeholder", type="placeholder")
        artifact._client = client
        artifact._entity = path.prefix
        artifact._project = path.project
        artifact._name = path.name

        artifact._assign_attrs(src_art, membership=membership)

        artifact.finalize()

        # Cache.
        assert artifact.id is not None
        artifact_instance_cache[artifact.id] = artifact
        return artifact

    # TODO: Eventually factor out is_link. Have to currently use it since some forms of fetching the artifact
    # doesn't make it clear if the artifact is a link or not and have to manually set it.
    def _assign_attrs(
        self,
        src_art: ArtifactFragment,
        *,
        # aliases/version_index are taken from the membership, if given
        membership: ArtifactMembershipFragment | None = None,
        is_link: bool | None = None,
    ) -> None:
        """Update this Artifact's attributes using the server response."""
        from ._validators import validate_metadata, validate_ttl_duration_seconds

        self._id = src_art.id

        src_collection = src_art.artifact_sequence
        src_project = src_collection.project

        self._source_entity = src_project.entity.name if src_project else ""
        self._source_project = src_project.name if src_project else ""
        self._source_name = f"{src_collection.name}:v{src_art.version_index}"
        self._source_version = f"v{src_art.version_index}"

        self._entity = self._entity or self._source_entity
        self._project = self._project or self._source_project
        self._name = self._name or self._source_name

        # TODO: Refactor artifact query to fetch artifact via membership instead
        # and get the collection type
        if is_link is None:
            self._is_link = (
                self._entity != self._source_entity
                or self._project != self._source_project
                or self._name.split(":")[0] != self._source_name.split(":")[0]
            )
        else:
            self._is_link = is_link

        self._type = src_art.artifact_type.name
        self._description = src_art.description

        # The future of aliases is to move all alias fetches to the membership level
        # so we don't have to do the collection fetches below
        if membership:
            aliases = [a.alias for a in membership.aliases]
        elif src_art.aliases:
            entity = self._entity
            project = self._project
            collection = self._name.split(":")[0]
            aliases = [
                a.alias
                for a in src_art.aliases
                if (
                    (alias_coll := a.artifact_collection)
                    and (alias_proj := alias_coll.project)
                    and alias_proj.entity.name == entity
                    and alias_proj.name == project
                    and alias_coll.name == collection
                )
            ]
        else:
            aliases = []

        version_aliases = list(filter(alias_is_version_index, aliases))
        other_aliases = list(filterfalse(alias_is_version_index, aliases))

        try:
            version = one(
                version_aliases, too_short=TooFewItemsError, too_long=TooManyItemsError
            )
        except TooFewItemsError:
            # default to the membership version if passed to this method,
            # otherwise fallback to the source version
            if membership and (m_version_index := membership.version_index) is not None:
                version = f"v{m_version_index}"
            else:
                version = f"v{src_art.version_index}"
        except TooManyItemsError:
            msg = f"Expected at most one version alias, got {len(version_aliases)}: {version_aliases!r}"
            raise ValueError(msg) from None

        self._version = version
        self._name = self._name if (":" in self._name) else f"{self._name}:{version}"

        self._aliases = copy(other_aliases)
        self._saved_aliases = copy(other_aliases)

        self._tags = [tag.name for tag in src_art.tags]
        self._saved_tags = copy(self._tags)

        self._metadata = validate_metadata(src_art.metadata)

        self._ttl_duration_seconds = validate_ttl_duration_seconds(
            src_art.ttl_duration_seconds
        )
        self._ttl_is_inherited = src_art.ttl_is_inherited

        self._state = ArtifactState(src_art.state)
        self._size = src_art.size
        self._digest = src_art.digest

        self._manifest = None

        self._commit_hash = src_art.commit_hash
        self._file_count = src_art.file_count
        self._created_at = src_art.created_at
        self._updated_at = src_art.updated_at
        self._history_step = src_art.history_step

    @ensure_logged
    def new_draft(self) -> Artifact:
        """Create a new draft artifact with the same content as this committed artifact.

        Modifying an existing artifact creates a new artifact version known
        as an "incremental artifact". The artifact returned can be extended or
        modified and logged as a new version.

        Returns:
            An `Artifact` object.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
        """
        # Name, _entity and _project are set to the *source* name/entity/project:
        # if this artifact is saved it must be saved to the source sequence.
        artifact = Artifact(self.source_name.split(":")[0], self.type)
        artifact._entity = self._source_entity
        artifact._project = self._source_project
        artifact._source_entity = self._source_entity
        artifact._source_project = self._source_project

        # This artifact's parent is the one we are making a draft from.
        artifact._base_id = self.id

        # We can reuse the client, and copy over all the attributes that aren't
        # version-dependent and don't depend on having been logged.
        artifact._client = self._client
        artifact._description = self.description
        artifact._metadata = self.metadata
        artifact._manifest = ArtifactManifest.from_manifest_json(
            self.manifest.to_manifest_json()
        )
        return artifact

    # Properties (Python Class managed attributes).

    @property
    def id(self) -> str | None:
        """The artifact's ID."""
        if self.is_draft():
            return None
        assert self._id is not None
        return self._id

    @property
    @ensure_logged
    def entity(self) -> str:
        """The name of the entity that the artifact collection belongs to.

        If the artifact is a link, the entity will be the entity of the linked artifact.
        """
        assert self._entity is not None
        return self._entity

    @property
    @ensure_logged
    def project(self) -> str:
        """The name of the project that the artifact collection belongs to.

        If the artifact is a link, the project will be the project of the linked artifact.
        """
        assert self._project is not None
        return self._project

    @property
    def name(self) -> str:
        """The artifact name and version of the artifact.

        A string with the format `{collection}:{alias}`. If fetched before an artifact is
        logged/saved, the name won't contain the alias.
        If the artifact is a link, the name will be the name of the linked artifact.
        """
        return self._name

    @property
    def qualified_name(self) -> str:
        """The entity/project/name of the artifact.

        If the artifact is a link, the qualified name will be the qualified name of the
        linked artifact path.
        """
        return f"{self.entity}/{self.project}/{self.name}"

    @property
    @ensure_logged
    def version(self) -> str:
        """The artifact's version.

        A string with the format `v{number}`.
        If this is a link artifact, the version will be from the linked collection.
        """
        assert self._version is not None
        return self._version

    @property
    @ensure_logged
    def collection(self) -> ArtifactCollection:
        """The collection this artifact is retrieved from.

        A collection is an ordered group of artifact versions.
        If this artifact is retrieved from a collection that it is linked to,
        return that collection. Otherwise, return the collection
        that the artifact version originates from.

        The collection that an artifact originates from is known as
        the source sequence.
        """
        if (client := self._client) is None:
            raise RuntimeError("Client not initialized")
        base_name = self.name.split(":")[0]
        return ArtifactCollection(
            client, self.entity, self.project, base_name, self.type
        )

    @property
    @ensure_logged
    def source_entity(self) -> str:
        """The name of the entity of the source artifact."""
        assert self._source_entity is not None
        return self._source_entity

    @property
    @ensure_logged
    def source_project(self) -> str:
        """The name of the project of the source artifact."""
        assert self._source_project is not None
        return self._source_project

    @property
    def source_name(self) -> str:
        """The artifact name and version of the source artifact.

        A string with the format `{source_collection}:{alias}`. Before the artifact
        is saved, contains only the name since the version is not yet known.
        """
        return self._source_name

    @property
    def source_qualified_name(self) -> str:
        """The source_entity/source_project/source_name of the source artifact."""
        return f"{self.source_entity}/{self.source_project}/{self.source_name}"

    @property
    @ensure_logged
    def source_version(self) -> str:
        """The source artifact's version.

        A string with the format `v{number}`.
        """
        assert self._source_version is not None
        return self._source_version

    @property
    @ensure_logged
    def source_collection(self) -> ArtifactCollection:
        """The artifact's source collection.

        The source collection is the collection that the artifact was logged from.
        """
        if (client := self._client) is None:
            raise RuntimeError("Client not initialized")
        base_name = self.source_name.split(":")[0]
        return ArtifactCollection(
            client, self.source_entity, self.source_project, base_name, self.type
        )

    @property
    def is_link(self) -> bool:
        """Boolean flag indicating if the artifact is a link artifact.

        True: The artifact is a link artifact to a source artifact.
        False: The artifact is a source artifact.
        """
        return self._is_link

    @property
    @ensure_logged
    def linked_artifacts(self) -> list[Artifact]:
        """Returns a list of all the linked artifacts of a source artifact.

        If this artifact is a link artifact (`artifact.is_link == True`),
        it will return an empty list.

        Limited to 500 results.
        """
        if not self.is_link:
            self._linked_artifacts = self._fetch_linked_artifacts()
        return self._linked_artifacts

    @property
    @ensure_logged
    def source_artifact(self) -> Artifact:
        """Returns the source artifact, which is the original logged artifact.

        If this artifact is a source artifact (`artifact.is_link == False`),
        it will return itself.
        """
        from ._validators import FullArtifactPath

        if not self.is_link:
            return self
        if self._source_artifact is None:
            if (client := self._client) is None:
                raise ValueError("Client is not initialized")

            try:
                path = FullArtifactPath(
                    prefix=self.source_entity,
                    project=self.source_project,
                    name=self.source_name,
                )
                self._source_artifact = self._from_name(path=path, client=client)
            except Exception as e:
                raise ValueError(
                    f"Unable to fetch source artifact for linked artifact {self.name}"
                ) from e
        return self._source_artifact

    @property
    def type(self) -> str:
        """The artifact's type. Common types include `dataset` or `model`."""
        return self._type

    @property
    @ensure_logged
    def url(self) -> str:
        """
        Constructs the URL of the artifact.

        Returns:
            str: The URL of the artifact.
        """
        from ._validators import is_artifact_registry_project

        try:
            base_url = self._client.app_url  # type: ignore[union-attr]
        except AttributeError:
            return ""

        if not self.is_link:
            return self._construct_standard_url(base_url)
        if is_artifact_registry_project(self.project):
            return self._construct_registry_url(base_url)
        if self._type == "model" or self.project == "model-registry":
            return self._construct_model_registry_url(base_url)
        return self._construct_standard_url(base_url)

    def _construct_standard_url(self, base_url: str) -> str:
        if not all(
            [
                base_url,
                self.entity,
                self.project,
                self._type,
                self.collection.name,
                self._version,
            ]
        ):
            return ""
        return urljoin(
            base_url,
            f"{self.entity}/{self.project}/artifacts/{quote(self._type)}/{quote(self.collection.name)}/{self._version}",
        )

    def _construct_registry_url(self, base_url: str) -> str:
        from ._validators import remove_registry_prefix

        if not all(
            [
                base_url,
                self.entity,
                self.project,
                self.collection.name,
                self._version,
            ]
        ):
            return ""

        try:
            org_name = org_info_from_entity(self._client, self.entity).organization.name  # type: ignore[union-attr]
        except (AttributeError, ValueError):
            return ""

        selection_path = quote(
            f"{self.entity}/{self.project}/{self.collection.name}", safe=""
        )
        return urljoin(
            base_url,
            f"orgs/{org_name}/registry/{remove_registry_prefix(self.project)}?selectionPath={selection_path}&view=membership&version={self.version}",
        )

    def _construct_model_registry_url(self, base_url: str) -> str:
        if not all(
            [
                base_url,
                self.entity,
                self.project,
                self.collection.name,
                self._version,
            ]
        ):
            return ""
        selection_path = quote(
            f"{self.entity}/{self.project}/{self.collection.name}", safe=""
        )
        return urljoin(
            base_url,
            f"{self.entity}/registry/model?selectionPath={selection_path}&view=membership&version={self._version}",
        )

    @property
    def description(self) -> str | None:
        """A description of the artifact."""
        return self._description

    @description.setter
    def description(self, description: str | None) -> None:
        """Set the description of the artifact.

        For model or dataset Artifacts, add documentation for your
        standardized team model or dataset card. In the W&B UI the
        description is rendered as markdown.

        Editing the description will apply the changes to the source artifact
        and all linked artifacts associated with it.

        Args:
            description: Free text that offers a description of the artifact.
        """
        if self.is_link:
            wandb.termwarn(
                "Editing the description of this linked artifact will edit the description for the source artifact and it's linked artifacts as well."
            )
        self._description = description

    @property
    def metadata(self) -> dict:
        """User-defined artifact metadata.

        Structured data associated with the artifact.
        """
        return self._metadata

    @metadata.setter
    def metadata(self, metadata: dict) -> None:
        """User-defined artifact metadata.

        Metadata set this way will eventually be queryable and plottable in the UI; e.g.
        the class distribution of a dataset.

        Note: There is currently a limit of 100 total keys.
        Editing the metadata will apply the changes to the source artifact
        and all linked artifacts associated with it.

        Args:
            metadata: Structured data associated with the artifact.
        """
        from ._validators import validate_metadata

        if self.is_link:
            wandb.termwarn(
                "Editing the metadata of this linked artifact will edit the metadata for the source artifact and it's linked artifacts as well."
            )
        self._metadata = validate_metadata(metadata)

    @property
    def ttl(self) -> timedelta | None:
        """The time-to-live (TTL) policy of an artifact.

        Artifacts are deleted shortly after a TTL policy's duration passes.
        If set to `None`, the artifact deactivates TTL policies and will be not
        scheduled for deletion, even if there is a team default TTL.
        An artifact inherits a TTL policy from
        the team default if the team administrator defines a default
        TTL and there is no custom policy set on an artifact.

        Raises:
            ArtifactNotLoggedError: Unable to fetch inherited TTL if the
            artifact has not been logged or saved.
        """
        if self._ttl_is_inherited and (self.is_draft() or self._ttl_changed):
            raise ArtifactNotLoggedError(f"{nameof(type(self))}.ttl", self)
        if self._ttl_duration_seconds is None:
            return None
        return timedelta(seconds=self._ttl_duration_seconds)

    @ttl.setter
    def ttl(self, ttl: timedelta | ArtifactTTL | None) -> None:
        """The time-to-live (TTL) policy of an artifact.

        Artifacts are deleted shortly after a TTL policy's duration passes.
        If set to `None`, the artifact has no TTL policy set and it is not
        scheduled for deletion. An artifact inherits a TTL policy from
        the team default if the team administrator defines a default
        TTL and there is no custom policy set on an artifact.

        Args:
            ttl: The duration as a positive `datetime.timedelta` that represents
                how long the artifact will remain active from its creation.

        """
        if self.type == "wandb-history":
            raise ValueError("Cannot set artifact TTL for type wandb-history")

        if self.is_link:
            raise ValueError(
                "Cannot set TTL for link artifact. "
                "Unlink the artifact first then set the TTL for the source artifact"
            )

        self._ttl_changed = True
        if isinstance(ttl, ArtifactTTL):
            if ttl == ArtifactTTL.INHERIT:
                self._ttl_is_inherited = True
            else:
                raise ValueError(f"Unhandled ArtifactTTL enum {ttl}")
        else:
            self._ttl_is_inherited = False
            if ttl is None:
                self._ttl_duration_seconds = None
            else:
                if ttl.total_seconds() <= 0:
                    raise ValueError(
                        f"Artifact TTL Duration has to be positive. ttl: {ttl.total_seconds()}"
                    )
                self._ttl_duration_seconds = int(ttl.total_seconds())

    @property
    @ensure_logged
    def aliases(self) -> list[str]:
        """List of one or more semantically-friendly references or

        identifying "nicknames" assigned to an artifact version.

        Aliases are mutable references that you can programmatically reference.
        Change an artifact's alias with the W&B App UI or programmatically.
        See [Create new artifact versions](https://docs.wandb.ai/models/artifacts/create-a-new-artifact-version)
        for more information.
        """
        return self._aliases

    @aliases.setter
    @ensure_logged
    def aliases(self, aliases: list[str]) -> None:
        """Set the aliases associated with this artifact."""
        from ._validators import validate_aliases

        self._aliases = validate_aliases(aliases)

    @property
    @ensure_logged
    def tags(self) -> list[str]:
        """List of one or more tags assigned to this artifact version."""
        return self._tags

    @tags.setter
    @ensure_logged
    def tags(self, tags: list[str]) -> None:
        """Set the tags associated with this artifact.

        Editing tags will apply the changes to the source artifact
        and all linked artifacts associated with it.
        """
        from ._validators import validate_tags

        if self.is_link:
            wandb.termwarn(
                "Editing tags will apply the changes to the source artifact and all linked artifacts associated with it."
            )
        self._tags = validate_tags(tags)

    @property
    def distributed_id(self) -> str | None:
        """The distributed ID of the artifact.

        <!-- lazydoc-ignore: internal -->
        """
        return self._distributed_id

    @distributed_id.setter
    def distributed_id(self, distributed_id: str | None) -> None:
        self._distributed_id = distributed_id

    @property
    def incremental(self) -> bool:
        """Boolean flag indicating if the artifact is an incremental artifact.

        <!-- lazydoc-ignore: internal -->
        """
        return self._incremental

    @property
    def use_as(self) -> str | None:
        """Deprecated."""
        warn_and_record_deprecation(
            feature=Deprecated(artifact__use_as=True),
            message=("The use_as property of Artifact is deprecated."),
        )
        return self._use_as

    @property
    def state(self) -> str:
        """The status of the artifact. One of: "PENDING", "COMMITTED", or "DELETED"."""
        return self._state.value

    @property
    def manifest(self) -> ArtifactManifest:
        """The artifact's manifest.

        The manifest lists all of its contents, and can't be changed once the artifact
        has been logged.
        """
        if self._manifest is None:
            self._manifest = self._fetch_manifest()
        return self._manifest

    def _fetch_manifest(self) -> ArtifactManifest:
        """Fetch, parse, and load the full ArtifactManifest."""
        from ._generated import FETCH_ARTIFACT_MANIFEST_GQL, FetchArtifactManifest

        if (client := self._client) is None:
            raise RuntimeError("Client not initialized for artifact queries")

        # From the GraphQL API, get the (expiring) directUrl for downloading the manifest.
        gql_op = gql(FETCH_ARTIFACT_MANIFEST_GQL)
        gql_vars = {"id": self.id}
        data = client.execute(gql_op, variable_values=gql_vars)
        result = FetchArtifactManifest.model_validate(data)

        # Now fetch the actual manifest contents from the directUrl.
        if (artifact := result.artifact) and (manifest := artifact.current_manifest):
            # Create a short lived session instead of using requests.get()
            # because make_http_session() adds http headers from env vars.
            # Artifact manifest json is also downloaded from object storage
            # using presigned urls like artifact files, which requires adding
            # extra http headers when user specifies them in env vars.
            #
            # FIXME: For successive/repeated calls to `manifest`, figure out
            # how to reuse a single `requests.Session` within the constraints
            # of the current API. Creating a new session for _each_ fetch is
            # wasteful and introduces noticeable perf overhead when e.g.
            # downloading many artifacts sequentially or concurrently. The
            # storage policy's session is also not reused across different
            # artifacts.
            with make_http_session() as session:
                response = session.get(manifest.file.direct_url)
            return ArtifactManifest.from_manifest_json(from_json(response.content))

        raise ValueError("Failed to fetch artifact manifest")

    @property
    def digest(self) -> str:
        """The logical digest of the artifact.

        The digest is the checksum of the artifact's contents. If an artifact has the
        same digest as the current `latest` version, then `log_artifact` is a no-op.
        """
        # Use the last fetched value of `Artifact.digest` ONLY if present AND the manifest
        # has not been fetched and/or populated locally.
        # Otherwise, use the manifest directly to recalculate the digest, as its contents
        # may have been locally modified.
        return (
            self._digest
            if (self._manifest is None) and (self._digest is not None)
            else self.manifest.digest()
        )

    @property
    def size(self) -> int:
        """The total size of the artifact in bytes.

        Includes any references tracked by this artifact.
        """
        # Use the last fetched value of `Artifact.size` ONLY if present AND the manifest
        # has not been fetched and/or populated locally.
        # Otherwise, use the manifest directly to recalculate the size, as its contents
        # may have been locally modified.
        #
        # NOTE on choice of GQL field: `Artifact.size` counts references, while
        # `Artifact.storageBytes` does not.
        return (
            self._size
            if (self._manifest is None) and (self._size is not None)
            else self.manifest.size()
        )

    @property
    @ensure_logged
    def commit_hash(self) -> str:
        """The hash returned when this artifact was committed."""
        assert self._commit_hash is not None
        return self._commit_hash

    @property
    @ensure_logged
    def file_count(self) -> int:
        """The number of files (including references)."""
        assert self._file_count is not None
        return self._file_count

    @property
    @ensure_logged
    def created_at(self) -> str:
        """Timestamp when the artifact was created."""
        assert self._created_at is not None
        return self._created_at

    @property
    @ensure_logged
    def updated_at(self) -> str:
        """The time when the artifact was last updated."""
        assert self._created_at is not None
        return self._updated_at or self._created_at

    @property
    @ensure_logged
    def history_step(self) -> int | None:
        """The nearest step which logged history metrics for this artifact's source run.

        Examples:
        ```python
        run = artifact.logged_by()
        if run and (artifact.history_step is not None):
            history = run.sample_history(
                min_step=artifact.history_step,
                max_step=artifact.history_step + 1,
                keys=["my_metric"],
            )
        ```
        """
        if self._history_step is None:
            return None
        return max(0, self._history_step - 1)

    # State management.

    def finalize(self) -> None:
        """Finalize the artifact version.

        You cannot modify an artifact version once it is finalized because the artifact
        is logged as a specific artifact version. Create a new artifact version
        to log more data to an artifact. An artifact is automatically finalized
        when you log the artifact with `log_artifact`.
        """
        self._final = True

    def is_draft(self) -> bool:
        """Check if artifact is not saved.

        Returns:
            Boolean. `False` if artifact is saved. `True` if artifact is not saved.
        """
        return self._state is ArtifactState.PENDING

    def _is_draft_save_started(self) -> bool:
        return self._save_handle is not None

    def save(
        self,
        project: str | None = None,
        settings: wandb.Settings | None = None,
    ) -> None:
        """Persist any changes made to the artifact.

        If currently in a run, that run will log this artifact. If not currently in a
        run, a run of type "auto" is created to track this artifact.

        Args:
            project: A project to use for the artifact in the case that a run is not
                already in context.
            settings: A settings object to use when initializing an automatic run. Most
                commonly used in testing harness.
        """
        if self._state is not ArtifactState.PENDING:
            return self._update()

        if self._incremental:
            with telemetry.context() as tel:
                tel.feature.artifact_incremental = True

        if run := wandb_setup.singleton().most_recent_active_run:
            # TODO: Deprecate and encourage explicit log_artifact().
            run.log_artifact(self)
        else:
            if settings is None:
                settings = wandb.Settings(silent="true")
            with wandb.init(  # type: ignore
                entity=self._source_entity,
                project=project or self._source_project,
                job_type="auto",
                settings=settings,
            ) as run:
                # redoing this here because in this branch we know we didn't
                # have the run at the beginning of the method
                if self._incremental:
                    with telemetry.context(run=run) as tel:
                        tel.feature.artifact_incremental = True
                run.log_artifact(self)

    def _set_save_handle(
        self,
        save_handle: MailboxHandle[pb.Result],
        client: RetryingClient,
    ) -> None:
        self._save_handle = save_handle
        self._client = client

    def wait(self, timeout: int | None = None) -> Artifact:
        """If needed, wait for this artifact to finish logging.

        Args:
            timeout: The time, in seconds, to wait.

        Returns:
            An `Artifact` object.
        """
        if self.is_draft():
            if self._save_handle is None:
                raise ArtifactNotLoggedError(nameof(self.wait), self)

            try:
                result = self._save_handle.wait_or(timeout=timeout)
            except TimeoutError as e:
                raise WaitTimeoutError(
                    "Artifact upload wait timed out, failed to fetch Artifact response"
                ) from e

            response = result.response.log_artifact_response
            if response.error_message:
                raise ValueError(response.error_message)
            self._populate_after_save(response.artifact_id)
        return self

    def _populate_after_save(self, artifact_id: str) -> None:
        from ._generated import ARTIFACT_BY_ID_GQL, ArtifactByID

        if (client := self._client) is None:
            raise RuntimeError("Client not initialized for artifact queries")

        gql_op = gql(ARTIFACT_BY_ID_GQL)
        data = client.execute(gql_op, variable_values={"id": artifact_id})
        result = ArtifactByID.model_validate(data)

        if not (artifact := result.artifact):
            raise ValueError(f"Unable to fetch artifact with id: {artifact_id!r}")

        # _populate_after_save is only called on source artifacts, not linked artifacts
        # We have to manually set is_link because we aren't fetching the collection
        # the artifact. That requires greater refactoring for commitArtifact to return
        # the artifact collection type.
        self._assign_attrs(artifact, is_link=False)

    @normalize_exceptions
    def _update(self) -> None:
        """Persists artifact changes to the wandb backend."""
        from ._generated import UPDATE_ARTIFACT_GQL, UpdateArtifact, UpdateArtifactInput
        from ._validators import FullArtifactPath, validate_tags

        if (client := self._client) is None:
            raise RuntimeError("Client not initialized for artifact mutations")

        entity, project, collection = self.entity, self.project, self.name.split(":")[0]

        old_aliases, new_aliases = set(self._saved_aliases), set(self.aliases)
        target = FullArtifactPath(prefix=entity, project=project, name=collection)
        self._add_aliases(new_aliases - old_aliases, target=target)
        self._delete_aliases(old_aliases - new_aliases, target=target)
        self._saved_aliases = copy(self.aliases)

        old_tags, new_tags = set(self._saved_tags), set(self.tags)

        gql_op = gql(UPDATE_ARTIFACT_GQL)
        gql_input = UpdateArtifactInput(
            artifact_id=self.id,
            description=self.description,
            metadata=json_dumps_safer(self.metadata),
            ttl_duration_seconds=self._ttl_duration_seconds_to_gql(),
            tags_to_add=[{"tagName": t} for t in validate_tags(new_tags - old_tags)],
            tags_to_delete=[{"tagName": t} for t in validate_tags(old_tags - new_tags)],
        )
        gql_vars = {"input": gql_input.model_dump()}
        data = client.execute(gql_op, variable_values=gql_vars)

        result = UpdateArtifact.model_validate(data).result
        if not (result and (artifact := result.artifact)):
            raise ValueError("Unable to parse updateArtifact response")
        self._assign_attrs(artifact)

        self._ttl_changed = False  # Reset after updating artifact

    def _add_aliases(self, alias_names: set[str], target: FullArtifactPath) -> None:
        from ._generated import ADD_ALIASES_GQL, AddAliasesInput

        if (client := self._client) is None:
            raise RuntimeError("Client not initialized for artifact mutations")

        # If there aren't any aliases to add, we can skip the GraphQL call.
        if alias_names:
            target_props = {
                "entityName": target.prefix,
                "projectName": target.project,
                "artifactCollectionName": target.name,
            }
            alias_inputs = [{**target_props, "alias": name} for name in alias_names]
            gql_op = gql(ADD_ALIASES_GQL)
            gql_input = AddAliasesInput(artifact_id=self.id, aliases=alias_inputs)
            gql_vars = {"input": gql_input.model_dump()}
            try:
                client.execute(gql_op, variable_values=gql_vars)
            except CommError as e:
                msg = (
                    "You do not have permission to add"
                    f" {'at least one of the following aliases' if len(alias_names) > 1 else 'the following alias'}"
                    f" to this artifact: {alias_names!r}"
                )
                raise CommError(msg) from e

    def _delete_aliases(self, alias_names: set[str], target: FullArtifactPath) -> None:
        from ._generated import DELETE_ALIASES_GQL, DeleteAliasesInput

        if (client := self._client) is None:
            raise RuntimeError("Client not initialized for artifact mutations")

        # If there aren't any aliases to delete, we can skip the GraphQL call.
        if alias_names:
            target_props = {
                "entityName": target.prefix,
                "projectName": target.project,
                "artifactCollectionName": target.name,
            }
            alias_inputs = [{**target_props, "alias": name} for name in alias_names]
            gql_op = gql(DELETE_ALIASES_GQL)
            gql_input = DeleteAliasesInput(artifact_id=self.id, aliases=alias_inputs)
            gql_vars = {"input": gql_input.model_dump()}
            try:
                client.execute(gql_op, variable_values=gql_vars)
            except CommError as e:
                msg = (
                    f"You do not have permission to delete"
                    f" {'at least one of the following aliases' if len(alias_names) > 1 else 'the following alias'}"
                    f" from this artifact: {alias_names!r}"
                )
                raise CommError(msg) from e

    # Adding, removing, getting entries.

    def __getitem__(self, name: str) -> WBValue | None:
        """Get the WBValue object located at the artifact relative `name`.

        Args:
            name: The artifact relative name to get.

        Returns:
            W&B object that can be logged with `run.log()` and visualized in the W&B UI.

        Raises:
            ArtifactNotLoggedError: If the artifact isn't logged or the run is offline.
        """
        return self.get(name)

    def __setitem__(self, name: str, item: WBValue) -> ArtifactManifestEntry:
        """Add `item` to the artifact at path `name`.

        Args:
            name: The path within the artifact to add the object.
            item: The object to add.

        Returns:
            The added manifest entry

        Raises:
            ArtifactFinalizedError: You cannot make changes to the current
            artifact version because it is finalized. Log a new artifact
            version instead.
        """
        return self.add(item, name)

    @contextlib.contextmanager
    @ensure_not_finalized
    def new_file(
        self, name: str, mode: str = "x", encoding: str | None = None
    ) -> Iterator[IO]:
        """Open a new temporary file and add it to the artifact.

        Args:
            name: The name of the new file to add to the artifact.
            mode: The file access mode to use to open the new file.
            encoding: The encoding used to open the new file.

        Returns:
            A new file object that can be written to. Upon closing, the file
            is automatically added to the artifact.

        Raises:
            ArtifactFinalizedError: You cannot make changes to the current
            artifact version because it is finalized. Log a new artifact
            version instead.
        """
        overwrite: bool = "x" not in mode

        if self._tmp_dir is None:
            self._tmp_dir = tempfile.TemporaryDirectory()
        path = os.path.join(self._tmp_dir.name, name.lstrip("/"))

        Path(path).parent.mkdir(parents=True, exist_ok=True)
        try:
            with fsync_open(path, mode, encoding) as f:
                yield f
        except FileExistsError:
            raise ValueError(f"File with name {name!r} already exists at {path!r}")
        except UnicodeEncodeError as e:
            termerror(
                f"Failed to open the provided file ({nameof(type(e))}: {e}). Please "
                f"provide the proper encoding."
            )
            raise

        self.add_file(
            path, name=name, policy="immutable", skip_cache=True, overwrite=overwrite
        )

    @ensure_not_finalized
    def add_file(
        self,
        local_path: str,
        name: str | None = None,
        is_tmp: bool | None = False,
        skip_cache: bool | None = False,
        policy: Literal["mutable", "immutable"] | None = "mutable",
        overwrite: bool = False,
    ) -> ArtifactManifestEntry:
        """Add a local file to the artifact.

        Args:
            local_path: The path to the file being added.
            name: The path within the artifact to use for the file being added.
                Defaults to the basename of the file.
            is_tmp: If true, then the file is renamed deterministically to avoid
                collisions.
            skip_cache: If `True`, do not copy files to the cache
                after uploading.
            policy: By default, set to "mutable". If set to "mutable",
                create a temporary copy of the file to prevent corruption
                during upload. If set to "immutable", disable
                protection and rely on the user not to delete or change the
                file.
            overwrite: If `True`, overwrite the file if it already exists.

        Returns:
            The added manifest entry.

        Raises:
            ArtifactFinalizedError: You cannot make changes to the current
                artifact version because it is finalized. Log a new artifact
                version instead.
            ValueError: Policy must be "mutable" or "immutable"
        """
        if not os.path.isfile(local_path):
            raise ValueError(f"Path is not a file: {local_path!r}")

        name = LogicalPath(name or os.path.basename(local_path))
        digest = md5_file_b64(local_path)

        if is_tmp:
            file_path, file_name = os.path.split(name)
            file_name_parts = file_name.split(".")
            file_name_parts[0] = b64_to_hex_id(digest)[:20]
            name = os.path.join(file_path, ".".join(file_name_parts))

        return self._add_local_file(
            name,
            local_path,
            digest=digest,
            skip_cache=skip_cache,
            policy=policy,
            overwrite=overwrite,
        )

    @ensure_not_finalized
    def add_dir(
        self,
        local_path: str,
        name: str | None = None,
        skip_cache: bool | None = False,
        policy: Literal["mutable", "immutable"] | None = "mutable",
        merge: bool = False,
    ) -> None:
        """Add a local directory to the artifact.

        Args:
            local_path: The path of the local directory.
            name: The subdirectory name within an artifact. The name you
                specify appears in the W&B App UI nested by artifact's `type`.
                Defaults to the root of the artifact.
            skip_cache: If set to `True`, W&B will not copy/move files to
                the cache while uploading
            policy: By default, "mutable".
                - mutable: Create a temporary copy of the file to prevent
                    corruption during upload.
                - immutable: Disable protection, rely on the user not to delete
                    or change the file.
            merge: If `False` (default), throws ValueError if a file was already added
                in a previous add_dir call and its content has changed. If `True`,
                overwrites existing files with changed content. Always adds new files
                and never removes files. To replace an entire directory, pass a name
                when adding the directory using `add_dir(local_path, name=my_prefix)`
                and call `remove(my_prefix)` to remove the directory, then add it again.

        Raises:
            ArtifactFinalizedError: You cannot make changes to the current
                artifact version because it is finalized. Log a new artifact
                version instead.
            ValueError: Policy must be "mutable" or "immutable"
        """
        if not os.path.isdir(local_path):
            raise ValueError(f"Path is not a directory: {local_path!r}")

        termlog(
            f"Adding directory to artifact ({Path('.', local_path)})... ",
            newline=False,
        )
        start_time = time.monotonic()

        paths: deque[tuple[str, str]] = deque()
        logical_root = name or ""  # shared prefix, if any, for logical paths
        for dirpath, _, filenames in os.walk(local_path, followlinks=True):
            for fname in filenames:
                physical_path = os.path.join(dirpath, fname)
                logical_path = os.path.relpath(physical_path, start=local_path)
                logical_path = os.path.join(logical_root, logical_path)
                paths.append((logical_path, physical_path))

        def add_manifest_file(logical_pth: str, physical_pth: str) -> None:
            self._add_local_file(
                name=logical_pth,
                path=physical_pth,
                skip_cache=skip_cache,
                policy=policy,
                overwrite=merge,
            )

        num_threads = 8
        pool = multiprocessing.dummy.Pool(num_threads)
        pool.starmap(add_manifest_file, paths)
        pool.close()
        pool.join()

        termlog("Done. %.1fs" % (time.monotonic() - start_time), prefix=False)

    @ensure_not_finalized
    def add_reference(
        self,
        uri: ArtifactManifestEntry | str,
        name: StrPath | None = None,
        checksum: bool = True,
        max_objects: int | None = None,
    ) -> Sequence[ArtifactManifestEntry]:
        """Add a reference denoted by a URI to the artifact.

        Unlike files or directories that you add to an artifact, references are not
        uploaded to W&B. For more information,
        see [Track external files](https://docs.wandb.ai/models/artifacts/track-external-files).

        By default, the following schemes are supported:

        - http(s): The size and digest of the file will be inferred by the
          `Content-Length` and the `ETag` response headers returned by the server.
        - s3: The checksum and size are pulled from the object metadata.
          If bucket versioning is enabled, then the version ID is also tracked.
        - gs: The checksum and size are pulled from the object metadata. If bucket
          versioning is enabled, then the version ID is also tracked.
        - https, domain matching `*.blob.core.windows.net`
        - Azure: The checksum and size are be pulled from the blob metadata.
          If storage account versioning is enabled, then the version ID is
          also tracked.
        - file: The checksum and size are pulled from the file system. This scheme
          is useful if you have an NFS share or other externally mounted volume
          containing files you wish to track but not necessarily upload.

        For any other scheme, the digest is just a hash of the URI and the size is left
        blank.

        Args:
            uri: The URI path of the reference to add. The URI path can be an object
                returned from `Artifact.get_entry` to store a reference to another
                artifact's entry.
            name: The path within the artifact to place the contents of this reference.
            checksum: Whether or not to checksum the resource(s) located at the
                reference URI. Checksumming is strongly recommended as it enables
                automatic integrity validation. Disabling checksumming will speed up
                artifact creation but reference directories will not iterated through so
                the objects in the directory will not be saved to the artifact.
                We recommend setting `checksum=False` when adding reference objects,
                in which case a new version will only be created if the reference URI
                changes.
            max_objects: The maximum number of objects to consider when adding a
                reference that points to directory or bucket store prefix.
                By default, the maximum number of objects allowed for Amazon S3,
                GCS, Azure, and local files is 10,000,000. Other URI schemas
                do not have a maximum.

        Returns:
            The added manifest entries.

        Raises:
            ArtifactFinalizedError: You cannot make changes to the current
            artifact version because it is finalized. Log a new artifact
            version instead.
        """
        if name is not None:
            name = LogicalPath(name)

        # This is a bit of a hack, we want to check if the uri is a of the type
        # ArtifactManifestEntry. If so, then recover the reference URL.
        if isinstance(uri, ArtifactManifestEntry):
            uri_str = uri.ref_url()
        elif isinstance(uri, str):
            uri_str = uri
        url = urlparse(str(uri_str))
        if not url.scheme:
            raise ValueError(
                "References must be URIs. To reference a local file, use file://"
            )

        manifest_entries = self.manifest.storage_policy.store_reference(
            self,
            URIStr(uri_str),
            name=name,
            checksum=checksum,
            max_objects=max_objects,
        )
        for entry in manifest_entries:
            self.manifest.add_entry(entry)

        return manifest_entries

    @ensure_not_finalized
    def add(
        self, obj: WBValue, name: StrPath, overwrite: bool = False
    ) -> ArtifactManifestEntry:
        """Add wandb.WBValue `obj` to the artifact.

        Args:
            obj: The object to add. Currently support one of Bokeh, JoinedTable,
                PartitionedTable, Table, Classes, ImageMask, BoundingBoxes2D,
                Audio, Image, Video, Html, Object3D
            name: The path within the artifact to add the object.
            overwrite: If True, overwrite existing objects with the same file
                path if applicable.

        Returns:
            The added manifest entry

        Raises:
            ArtifactFinalizedError: You cannot make changes to the current
            artifact version because it is finalized. Log a new artifact
            version instead.
        """
        name = LogicalPath(name)

        # This is a "hack" to automatically rename tables added to
        # the wandb /media/tables directory to their sha-based name.
        # TODO: figure out a more appropriate convention.
        is_tmp_name = name.startswith("media/tables")

        # Validate that the object is one of the correct wandb.Media types
        # TODO: move this to checking subclass of wandb.Media once all are
        # generally supported
        allowed_types = (
            data_types.Bokeh,
            data_types.JoinedTable,
            data_types.PartitionedTable,
            data_types.Table,
            data_types.Classes,
            data_types.ImageMask,
            data_types.BoundingBoxes2D,
            data_types.Audio,
            data_types.Image,
            data_types.Video,
            data_types.Html,
            data_types.Object3D,
            data_types.Molecule,
            data_types._SavedModel,
        )
        if not isinstance(obj, allowed_types):
            raise TypeError(
                f"Found object of type {obj.__class__}, expected one of:"
                f" {allowed_types}"
            )

        obj_id = id(obj)
        if obj_id in self._added_objs:
            return self._added_objs[obj_id][1]

        # If the object is coming from another artifact, save it as a reference
        ref_path = obj._get_artifact_entry_ref_url()
        if ref_path is not None:
            return self.add_reference(ref_path, type(obj).with_suffix(name))[0]

        val = obj.to_json(self)
        name = obj.with_suffix(name)
        entry = self.manifest.get_entry_by_path(name)
        if (not overwrite) and (entry is not None):
            return entry

        if is_tmp_name:
            file_path = os.path.join(self._TMP_DIR.name, str(id(self)), name)
            folder_path, _ = os.path.split(file_path)
            os.makedirs(folder_path, exist_ok=True)
            with open(file_path, "w", encoding="utf-8") as tmp_f:
                json.dump(val, tmp_f, sort_keys=True)
        else:
            filemode = "w" if overwrite else "x"
            with self.new_file(name, mode=filemode, encoding="utf-8") as f:
                json.dump(val, f, sort_keys=True)
                file_path = f.name

        # Note, we add the file from our temp directory.
        # It will be added again later on finalize, but succeed since
        # the checksum should match
        entry = self.add_file(file_path, name, is_tmp_name)
        # We store a reference to the obj so that its id doesn't get reused.
        self._added_objs[obj_id] = (obj, entry)
        if obj._artifact_target is None:
            obj._set_artifact_target(self, entry.path)

        if is_tmp_name:
            with contextlib.suppress(FileNotFoundError):
                os.remove(file_path)

        return entry

    def _add_local_file(
        self,
        name: StrPath,
        path: StrPath,
        digest: B64MD5 | None = None,
        skip_cache: bool | None = False,
        policy: Literal["mutable", "immutable"] | None = "mutable",
        overwrite: bool = False,
    ) -> ArtifactManifestEntry:
        policy = policy or "mutable"
        if policy not in ["mutable", "immutable"]:
            raise ValueError(
                f"Invalid policy {policy!r}. Policy may only be `mutable` or `immutable`."
            )
        upload_path = path
        if policy == "mutable":
            with tempfile.NamedTemporaryFile(dir=get_staging_dir(), delete=False) as f:
                staging_path = f.name
                shutil.copyfile(path, staging_path)
                # Set as read-only to prevent changes to the file during upload process
                os.chmod(staging_path, stat.S_IRUSR)
                upload_path = staging_path

        entry = ArtifactManifestEntry(
            path=name,
            digest=digest or md5_file_b64(upload_path),
            size=os.path.getsize(upload_path),
            local_path=upload_path,
            skip_cache=skip_cache,
        )
        self.manifest.add_entry(entry, overwrite=overwrite)
        self._added_local_paths[os.fspath(path)] = entry
        return entry

    @ensure_not_finalized
    def remove(self, item: StrPath | ArtifactManifestEntry) -> None:
        """Remove an item from the artifact.

        Args:
            item: The item to remove. Can be a specific manifest entry
                or the name of an artifact-relative path. If the item
                matches a directory all items in that directory will be removed.

        Raises:
            ArtifactFinalizedError: You cannot make changes to the current
                artifact version because it is finalized. Log a new artifact
                version instead.
            FileNotFoundError: If the item isn't found in the artifact.
        """
        if isinstance(item, ArtifactManifestEntry):
            self.manifest.remove_entry(item)
            return

        path = str(PurePosixPath(item))
        if entry := self.manifest.get_entry_by_path(path):
            return self.manifest.remove_entry(entry)

        entries = self.manifest.get_entries_in_directory(path)
        if not entries:
            raise FileNotFoundError(f"No such file or directory: {path}")
        for entry in entries:
            self.manifest.remove_entry(entry)

    def get_path(self, name: StrPath) -> ArtifactManifestEntry:
        """Deprecated. Use `get_entry(name)`."""
        warn_and_record_deprecation(
            feature=Deprecated(artifact__get_path=True),
            message="Artifact.get_path(name) is deprecated, use Artifact.get_entry(name) instead.",
        )
        return self.get_entry(name)

    @ensure_logged
    def get_entry(self, name: StrPath) -> ArtifactManifestEntry:
        """Get the entry with the given name.

        Args:
            name: The artifact relative name to get

        Returns:
            A `W&B` object.

        Raises:
            ArtifactNotLoggedError: if the artifact isn't logged or the run is offline.
            KeyError: if the artifact doesn't contain an entry with the given name.
        """
        name = LogicalPath(name)
        entry = self.manifest.entries.get(name) or self._get_obj_entry(name)[0]
        if entry is None:
            raise KeyError(f"Path not contained in artifact: {name}")
        entry._parent_artifact = self
        return entry

    @ensure_logged
    def get(self, name: str) -> WBValue | None:
        """Get the WBValue object located at the artifact relative `name`.

        Args:
            name: The artifact relative name to retrieve.

        Returns:
            W&B object that can be logged with `run.log()` and
            visualized in the W&B UI.

        Raises:
            ArtifactNotLoggedError: if the artifact isn't logged or the
                run is offline.
        """
        entry, wb_class = self._get_obj_entry(name)
        if entry is None or wb_class is None:
            return None

        # If the entry is a reference from another artifact, then get it directly from
        # that artifact.
        if referenced_id := entry._referenced_artifact_id():
            assert self._client is not None
            artifact = self._from_id(referenced_id, client=self._client)
            assert artifact is not None
            return artifact.get(uri_from_path(entry.ref))

        # Special case for wandb.Table. This is intended to be a short term
        # optimization. Since tables are likely to download many other assets in
        # artifact(s), we eagerly download the artifact using the parallelized
        # `artifact.download`. In the future, we should refactor the deserialization
        # pattern such that this special case is not needed.
        if wb_class == wandb.Table:
            self.download()

        # Get the ArtifactManifestEntry
        item = self.get_entry(entry.path)
        item_path = item.download()

        # Load the object from the JSON blob
        with open(item_path) as file:
            json_obj = json.load(file)

        result = wb_class.from_json(json_obj, self)
        result._set_artifact_source(self, name)
        return result

    def get_added_local_path_name(self, local_path: str) -> str | None:
        """Get the artifact relative name of a file added by a local filesystem path.

        Args:
            local_path: The local path to resolve into an artifact relative name.

        Returns:
            The artifact relative name.
        """
        if entry := self._added_local_paths.get(local_path):
            return entry.path
        return None

    def _get_obj_entry(
        self, name: str
    ) -> tuple[ArtifactManifestEntry, Type[WBValue]] | tuple[None, None]:  # noqa: UP006  # `type` shadows `Artifact.type`
        """Return an object entry by name, handling any type suffixes.

        When objects are added with `.add(obj, name)`, the name is typically changed to
        include the suffix of the object type when serializing to JSON. So we need to be
        able to resolve a name, without tasking the user with appending .THING.json.
        This method returns an entry if it exists by a suffixed name.

        Args:
            name: name used when adding
        """
        for wb_class in WBValue.type_mapping().values():
            wandb_file_name = wb_class.with_suffix(name)
            if entry := self.manifest.entries.get(wandb_file_name):
                return entry, wb_class
        return None, None

    # Downloading.

    @ensure_logged
    def download(
        self,
        root: StrPath | None = None,
        allow_missing_references: bool = False,
        skip_cache: bool | None = None,
        path_prefix: StrPath | None = None,
        multipart: bool | None = None,
    ) -> FilePathStr:
        """Download the contents of the artifact to the specified root directory.

        Existing files located within `root` are not modified. Explicitly delete `root`
        before you call `download` if you want the contents of `root` to exactly match
        the artifact.

        Args:
            root: The directory W&B stores the artifact's files.
            allow_missing_references: If set to `True`, any invalid reference paths
                will be ignored while downloading referenced files.
            skip_cache: If set to `True`, the artifact cache will be skipped when
                downloading and W&B will download each file into the default root or
                specified download directory.
            path_prefix: If specified, only files with a path that starts with the given
                prefix will be downloaded. Uses unix format (forward slashes).
            multipart: If set to `None` (default), the artifact will be downloaded
                in parallel using multipart download if individual file size is greater
                than 2GB. If set to `True` or `False`, the artifact will be downloaded in
                parallel or serially regardless of the file size.

        Returns:
            The path to the downloaded contents.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
        """
        root = self._add_download_root(root)

        # TODO: download artifacts using core when implemented
        # if is_require_core():
        #     return self._download_using_core(
        #         root=root,
        #         allow_missing_references=allow_missing_references,
        #         skip_cache=bool(skip_cache),
        #         path_prefix=path_prefix,
        #     )
        return self._download(
            root=root,
            allow_missing_references=allow_missing_references,
            skip_cache=skip_cache,
            path_prefix=path_prefix,
            multipart=multipart,
        )

    def _download_using_core(
        self,
        root: str,
        allow_missing_references: bool = False,
        skip_cache: bool = False,
        path_prefix: StrPath | None = None,
    ) -> FilePathStr:
        import pathlib

        from wandb.sdk.backend.backend import Backend

        # TODO: Create a special stream instead of relying on an existing run.
        if wandb.run is None:
            wl = wandb_setup.singleton()

            stream_id = generate_id()

            settings = wl.settings.to_proto()
            # TODO: remove this
            tmp_dir = pathlib.Path(tempfile.mkdtemp())

            settings.sync_dir.value = str(tmp_dir)
            settings.sync_file.value = str(tmp_dir / f"{stream_id}.wandb")
            settings.run_id.value = stream_id

            service = wl.ensure_service()
            service.inform_init(settings=settings, run_id=stream_id)

            backend = Backend(settings=wl.settings, service=service)
            backend.ensure_launched()

            assert backend.interface
            backend.interface._stream_id = stream_id  # type: ignore
        else:
            assert wandb.run._backend
            backend = wandb.run._backend

        assert backend.interface
        handle = backend.interface.deliver_download_artifact(
            self.id,  # type: ignore
            root,
            allow_missing_references,
            skip_cache,
            path_prefix,  # type: ignore
        )
        # TODO: Start the download process in the user process too, to handle reference downloads
        self._download(
            root=root,
            allow_missing_references=allow_missing_references,
            skip_cache=skip_cache,
            path_prefix=path_prefix,
        )
        result = handle.wait_or(timeout=None)

        response = result.response.download_artifact_response
        if response.error_message:
            raise ValueError(f"Error downloading artifact: {response.error_message}")

        return FilePathStr(root)

    def _download(
        self,
        root: str,
        allow_missing_references: bool = False,
        skip_cache: bool | None = None,
        path_prefix: StrPath | None = None,
        multipart: bool | None = None,
    ) -> FilePathStr:
        nfiles = len(self.manifest.entries)
        size_mb = self.size / _MB

        if log := (nfiles > 5000 or size_mb > 50):
            termlog(
                f"Downloading large artifact {self.name!r}, {size_mb:.2f}MB. {nfiles!r} files...",
            )
            start_time = time.monotonic()

        download_logger = ArtifactDownloadLogger(nfiles=nfiles)

        def _download_entry(
            entry: ArtifactManifestEntry, mp_executor: Executor
        ) -> None:
            multipart_executor = (
                mp_executor
                if should_multipart_download(entry.size, override=multipart)
                else None
            )
            try:
                entry.download(root, skip_cache=skip_cache, executor=multipart_executor)
            except FileNotFoundError as e:
                if allow_missing_references:
                    wandb.termwarn(str(e))
                    return
                raise
            except _GCSIsADirectoryError as e:
                logger.debug(str(e))
                return
            except IsADirectoryError:
                wandb.termwarn(
                    f"Unable to download file {entry.path!r} as there is a directory with the same path, skipping."
                )
                return
            except NotADirectoryError:
                wandb.termwarn(
                    f"Unable to download file {entry.path!r} as there is a file with the same path as a directory this file is expected to be in, skipping."
                )
                return
            download_logger.notify_downloaded()

        with (
            ThreadPoolExecutor(max_workers=_FILE_EXECUTOR_WORKERS) as file_executor,
            ThreadPoolExecutor(max_workers=_MP_EXECUTOR_WORKERS) as mp_executor,
        ):
            batch_size = env.get_artifact_fetch_file_url_batch_size()

            active_futures = set()
            cursor, has_more = None, True
            while has_more:
                files_page = self._fetch_file_urls(cursor=cursor, per_page=batch_size)

                has_more = files_page.page_info.has_next_page
                cursor = files_page.page_info.end_cursor

                # `File` nodes are formally nullable, so filter them out just in case.
                file_nodes = (e.node for e in files_page.edges if e.node)
                for node in file_nodes:
                    entry = self.get_entry(node.name)
                    # TODO: uncomment once artifact downloads are supported in core
                    # if require_core and entry.ref is None:
                    #     # Handled by core
                    #     continue
                    entry._download_url = node.direct_url
                    if (not path_prefix) or entry.path.startswith(str(path_prefix)):
                        active_futures.add(
                            file_executor.submit(
                                _download_entry, entry, mp_executor=mp_executor
                            )
                        )

                # Wait for download threads to catch up.
                #
                # Extra context and observations (tonyyli):
                # - Even though the ThreadPoolExecutor limits the number of
                #   concurrently-executed tasks, its internal task queue is unbounded.
                #   The code below seems intended to ensure that at most `batch_size`
                #   "backlogged" futures are held in memory at any given time.  This seems
                #   like a reasonable safeguard against unbounded memory consumption.
                #
                # - We should probably use a builtin bounded Queue or Semaphore instead.
                #   Consider this for a future change, or (depending on appetite for risk)
                #   managing this logic via asyncio instead, if viable.
                if len(active_futures) > batch_size:
                    for future in as_completed(active_futures):
                        future.result()  # check for errors
                        active_futures.remove(future)
                        if len(active_futures) <= batch_size:
                            break

            # Check for errors.
            for future in as_completed(active_futures):
                future.result()

        if log:
            # If you're wondering if we can display a `timedelta`, note that it
            # doesn't really support custom string format specifiers (compared to
            # e.g. `datetime` objs).  To truncate the number of decimal places for
            # the seconds part, we manually convert/format each part below.
            dt_secs = abs(time.monotonic() - start_time)
            hrs, mins = divmod(dt_secs, 3600)
            mins, secs = divmod(mins, 60)
            termlog(
                f"Done. {int(hrs):02d}:{int(mins):02d}:{secs:04.1f} ({size_mb / dt_secs:.1f}MB/s)",
                prefix=False,
            )
        return FilePathStr(root)

    def _build_fetch_file_urls_wrapper(self) -> Callable[..., Any]:
        import requests

        @retry.retriable(
            retry_timedelta=timedelta(minutes=3),
            retryable_exceptions=(requests.RequestException),
        )
        def _impl(cursor: str | None, per_page: int = 5000) -> FileWithUrlConnection:
            from ._generated import (
                GET_ARTIFACT_FILE_URLS_GQL,
                GET_ARTIFACT_MEMBERSHIP_FILE_URLS_GQL,
                GetArtifactFileUrls,
                GetArtifactMembershipFileUrls,
            )
            from ._models.pagination import FileWithUrlConnection

            if self._client is None:
                raise RuntimeError("Client not initialized")

            if server_supports(self._client, pb.ARTIFACT_COLLECTION_MEMBERSHIP_FILES):
                query = gql(GET_ARTIFACT_MEMBERSHIP_FILE_URLS_GQL)
                gql_vars = {
                    "entity": self.entity,
                    "project": self.project,
                    "collection": self.name.split(":")[0],
                    "alias": self.version,
                    "cursor": cursor,
                    "perPage": per_page,
                }
                data = self._client.execute(query, variable_values=gql_vars, timeout=60)
                result = GetArtifactMembershipFileUrls.model_validate(data)

                if not (
                    (project := result.project)
                    and (collection := project.artifact_collection)
                    and (membership := collection.artifact_membership)
                    and (files := membership.files)
                ):
                    raise ValueError(
                        f"Unable to fetch files for artifact: {self.name!r}"
                    )
                return FileWithUrlConnection.model_validate(files)
            else:
                query = gql(GET_ARTIFACT_FILE_URLS_GQL)
                gql_vars = {"id": self.id, "cursor": cursor, "perPage": per_page}
                data = self._client.execute(query, variable_values=gql_vars, timeout=60)
                result = GetArtifactFileUrls.model_validate(data)

                if not ((artifact := result.artifact) and (files := artifact.files)):
                    raise ValueError(
                        f"Unable to fetch files for artifact: {self.name!r}"
                    )
                return FileWithUrlConnection.model_validate(files)

        return _impl

    def _fetch_file_urls(
        self, cursor: str | None, per_page: int = 5000
    ) -> FileWithUrlConnection:
        if self._fetch_file_urls_decorated is None:
            self._fetch_file_urls_decorated = self._build_fetch_file_urls_wrapper()
        return self._fetch_file_urls_decorated(cursor, per_page)

    @ensure_logged
    def checkout(self, root: str | None = None) -> str:
        """Replace the specified root directory with the contents of the artifact.

        WARNING: This will delete all files in `root` that are not included in the
        artifact.

        Args:
            root: The directory to replace with this artifact's files.

        Returns:
           The path of the checked out contents.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
        """
        root = root or self._default_root(include_version=False)

        for dirpath, _, files in os.walk(root):
            for file in files:
                full_path = os.path.join(dirpath, file)
                artifact_path = os.path.relpath(full_path, start=root)
                try:
                    self.get_entry(artifact_path)
                except KeyError:
                    # File is not part of the artifact, remove it.
                    os.remove(full_path)

        return self.download(root=root)

    @ensure_logged
    def verify(self, root: str | None = None) -> None:
        """Verify that the contents of an artifact match the manifest.

        All files in the directory are checksummed and the checksums are then
        cross-referenced against the artifact's manifest. References are not verified.

        Args:
            root: The directory to verify. If None artifact will be downloaded to
                './artifacts/self.name/'.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
            ValueError: If the verification fails.
        """
        root = root or self._default_root()

        for dirpath, _, files in os.walk(root):
            for file in files:
                full_path = os.path.join(dirpath, file)
                artifact_path = os.path.relpath(full_path, start=root)
                try:
                    self.get_entry(artifact_path)
                except KeyError:
                    raise ValueError(
                        f"Found file {full_path} which is not a member of artifact {self.name}"
                    )

        ref_count = 0
        for entry in self.manifest.entries.values():
            if entry.ref is None:
                if md5_file_b64(os.path.join(root, entry.path)) != entry.digest:
                    raise ValueError(f"Digest mismatch for file: {entry.path}")
            else:
                ref_count += 1
        if ref_count > 0:
            termwarn(f"skipped verification of {ref_count} refs")

    @ensure_logged
    def file(self, root: str | None = None) -> StrPath:
        """Download a single file artifact to the directory you specify with `root`.

        Args:
            root: The root directory to store the file. Defaults to
                `./artifacts/self.name/`.

        Returns:
            The full path of the downloaded file.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
            ValueError: If the artifact contains more than one file.
        """
        if root is None:
            root = os.path.join(".", "artifacts", self.name)

        if len(self.manifest.entries) > 1:
            raise ValueError(
                "This artifact contains more than one file, call `.download()` to get "
                'all files or call .get_entry("filename").download()'
            )

        return self.get_entry(list(self.manifest.entries)[0]).download(root)

    @ensure_logged
    def files(
        self, names: list[str] | None = None, per_page: int = 50
    ) -> ArtifactFiles:
        """Iterate over all files stored in this artifact.

        Args:
            names: The filename paths relative to the root of the artifact you wish to
                list.
            per_page: The number of files to return per request.

        Returns:
            An iterator containing `File` objects.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
        """
        if (client := self._client) is None:
            raise RuntimeError("Client not initialized")
        return ArtifactFiles(client, self, names, per_page)

    def _default_root(self, include_version: bool = True) -> FilePathStr:
        name = self.source_name if include_version else self.source_name.split(":")[0]
        root = os.path.join(env.get_artifact_dir(), name)
        # In case we're on a system where the artifact dir has a name corresponding to
        # an unexpected filesystem, we'll check for alternate roots. If one exists we'll
        # use that, otherwise we'll fall back to the system-preferred path.
        return FilePathStr(check_exists(root) or system_preferred_path(root))

    def _add_download_root(self, dir_path: StrPath | None) -> FilePathStr:
        root = str(dir_path or self._default_root())
        self._download_roots.add(os.path.abspath(root))
        return root

    def _local_path_to_name(self, file_path: str) -> str | None:
        """Convert a local file path to a path entry in the artifact."""
        abs_file_path = os.path.abspath(file_path)
        abs_file_parts = abs_file_path.split(os.sep)
        for i in range(len(abs_file_parts) + 1):
            if os.path.join(os.sep, *abs_file_parts[:i]) in self._download_roots:
                return os.path.join(*abs_file_parts[i:])
        return None

    # Others.

    @ensure_logged
    def delete(self, delete_aliases: bool = False) -> None:
        """Delete an artifact and its files.

        If called on a linked artifact, only the link is deleted, and the
        source artifact is unaffected.

        Use `Artifact.unlink()` instead of `Artifact.delete()` to remove a
        link between a source artifact and a collection.

        Args:
            delete_aliases: If set to `True`, delete all aliases associated
                with the artifact. If `False`, raise an exception if
                the artifact has existing aliases. This parameter is ignored
                if the artifact is retrieved from a collection it is linked to.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
        """
        if self.is_link:
            wandb.termwarn(
                "Deleting a link artifact will only unlink the artifact from the source artifact and not delete the source artifact and the data of the source artifact."
            )
            self._unlink()
        else:
            self._delete(delete_aliases)

    @normalize_exceptions
    def _delete(self, delete_aliases: bool = False) -> None:
        from ._generated import DELETE_ARTIFACT_GQL, DeleteArtifactInput

        if self._client is None:
            raise RuntimeError("Client not initialized for artifact mutations")

        gql_op = gql(DELETE_ARTIFACT_GQL)
        gql_input = DeleteArtifactInput(
            artifact_id=self.id,
            delete_aliases=delete_aliases,
        )
        self._client.execute(gql_op, variable_values={"input": gql_input.model_dump()})

    @normalize_exceptions
    def link(self, target_path: str, aliases: Iterable[str] | None = None) -> Artifact:
        """Link this artifact to a collection.

        Args:
            target_path: The path of the collection. Path consists of the prefix
                "wandb-registry-" along with the registry name and the
                collection name `wandb-registry-{REGISTRY_NAME}/{COLLECTION_NAME}`.
            aliases: Add one or more aliases to the linked artifact. The
                "latest" alias is automatically applied to the most recent artifact
                you link.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.

        Returns:
            The linked artifact.
        """
        from wandb import Api
        from wandb.sdk.internal.internal_api import Api as InternalApi

        from ._generated import LINK_ARTIFACT_GQL, LinkArtifact, LinkArtifactInput
        from ._validators import ArtifactPath, FullArtifactPath, validate_aliases

        if self.is_link:
            wandb.termwarn(
                "Linking to a link artifact will result in directly linking to the source artifact of that link artifact."
            )

        # Save the artifact first if necessary
        if self.is_draft():
            if not self._is_draft_save_started():
                # Avoiding public `.source_project` property here,
                # as it requires the artifact is logged first.
                self.save(project=self._source_project)

            # Wait until the artifact is committed before trying to link it.
            self.wait()

        if (client := self._client) is None:
            raise RuntimeError("Client not initialized for artifact mutations")

        # FIXME: Find a way to avoid using InternalApi here, due to the perf overhead
        settings = InternalApi().settings()

        target = ArtifactPath.from_str(target_path).with_defaults(
            project=settings.get("project") or "uncategorized",
        )

        # Parse the entity (first part of the path) appropriately,
        # depending on whether we're linking to a registry
        if target.is_registry_path():
            # In a Registry linking, the entity is used to fetch the organization of the
            # artifact, therefore the source artifact's entity is passed to the backend
            org = target.prefix or settings.get("organization") or None
            target.prefix = resolve_org_entity_name(client, self.source_entity, org)
        else:
            target = target.with_defaults(prefix=self.source_entity)

        # Explicitly convert to FullArtifactPath to ensure all fields are present
        target = FullArtifactPath(**asdict(target))

        # Prepare the validated GQL input, send it
        alias_inputs = [
            {"artifactCollectionName": target.name, "alias": a}
            for a in validate_aliases(aliases or [])
        ]
        gql_input = LinkArtifactInput(
            artifact_id=self.id,
            artifact_portfolio_name=target.name,
            entity_name=target.prefix,
            project_name=target.project,
            aliases=alias_inputs,
        )
        gql_vars = {"input": gql_input.model_dump()}

        # Newer server versions can return `artifactMembership` directly in the response,
        # avoiding the need to re-fetch the linked artifact at the end.
        omit_variables = omit_fields = None
        if not server_supports(
            client, pb.ARTIFACT_MEMBERSHIP_IN_LINK_ARTIFACT_RESPONSE
        ):
            omit_variables = {"includeAliases"}
            omit_fields = {"artifactMembership"}

        gql_op = gql_compat(
            LINK_ARTIFACT_GQL, omit_variables=omit_variables, omit_fields=omit_fields
        )
        data = client.execute(gql_op, variable_values=gql_vars)
        result = LinkArtifact.model_validate(data).result

        # Newer server versions can return artifactMembership directly in the response
        if result and (membership := result.artifact_membership):
            return self._from_membership(membership, target=target, client=client)

        # Old behavior, which requires re-fetching the linked artifact to return it
        if not (result and (version_idx := result.version_index) is not None):
            raise ValueError("Unable to parse linked artifact version from response")

        link_name = f"{target.to_str()}:v{version_idx}"
        return Api(overrides={"entity": self.source_entity})._artifact(link_name)

    @ensure_logged
    def unlink(self) -> None:
        """Unlink this artifact if it is a linked member of an artifact collection.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
            ValueError: If the artifact is not linked to any collection.
        """
        # Fail early if this isn't a linked artifact to begin with
        if not self.is_link:
            raise ValueError(
                f"Artifact {self.qualified_name!r} is not a linked artifact and cannot be unlinked.  "
                f"To delete it, use {nameof(self.delete)!r} instead."
            )

        self._unlink()

    @normalize_exceptions
    def _unlink(self) -> None:
        from ._generated import UNLINK_ARTIFACT_GQL, UnlinkArtifactInput

        if self._client is None:
            raise RuntimeError("Client not initialized for artifact mutations")

        mutation = gql(UNLINK_ARTIFACT_GQL)
        gql_input = UnlinkArtifactInput(
            artifact_id=self.id,
            artifact_portfolio_id=self.collection.id,
        )
        gql_vars = {"input": gql_input.model_dump()}
        try:
            self._client.execute(mutation, variable_values=gql_vars)
        except CommError as e:
            raise CommError(
                f"You do not have permission to unlink the artifact {self.qualified_name!r}"
            ) from e

    @ensure_logged
    def used_by(self) -> list[Run]:
        """Get a list of the runs that have used this artifact and its linked artifacts.

        Returns:
            A list of `Run` objects.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
        """
        from ._generated import ARTIFACT_USED_BY_GQL, ArtifactUsedBy

        if (client := self._client) is None:
            raise RuntimeError("Client not initialized for artifact queries")

        query = gql(ARTIFACT_USED_BY_GQL)
        gql_vars = {"id": self.id}
        data = client.execute(query, variable_values=gql_vars)
        result = ArtifactUsedBy.model_validate(data)

        if (
            (artifact := result.artifact)
            and (used_by := artifact.used_by)
            and (edges := used_by.edges)
        ):
            run_nodes = (e.node for e in edges)
            return [
                Run(client, proj.entity.name, proj.name, run.name)
                for run in run_nodes
                if (proj := run.project)
            ]
        return []

    @ensure_logged
    def logged_by(self) -> Run | None:
        """Get the W&B run that originally logged the artifact.

        Returns:
            The name of the W&B run that originally logged the artifact.

        Raises:
            ArtifactNotLoggedError: If the artifact is not logged.
        """
        from ._generated import ARTIFACT_CREATED_BY_GQL, ArtifactCreatedBy

        if (client := self._client) is None:
            raise RuntimeError("Client not initialized for artifact queries")

        gql_op = gql(ARTIFACT_CREATED_BY_GQL)
        gql_vars = {"id": self.id}
        data = client.execute(gql_op, variable_values=gql_vars)
        result = ArtifactCreatedBy.model_validate(data)

        if (
            (artifact := result.artifact)
            and (creator := artifact.created_by)
            and (name := creator.name)
            and (project := creator.project)
        ):
            return Run(client, project.entity.name, project.name, name)
        return None

    @ensure_logged
    def json_encode(self) -> dict[str, Any]:
        """Returns the artifact encoded to the JSON format.

        Returns:
            A `dict` with `string` keys representing attributes of the artifact.
        """
        return artifact_to_json(self)

    @staticmethod
    def _expected_type(
        entity_name: str, project_name: str, name: str, client: RetryingClient
    ) -> str | None:
        """Returns the expected type for a given artifact name and project."""
        from ._generated import ARTIFACT_TYPE_GQL, ArtifactType

        name = name if (":" in name) else f"{name}:latest"

        gql_op = gql(ARTIFACT_TYPE_GQL)
        gql_vars = {"entity": entity_name, "project": project_name, "name": name}
        data = client.execute(gql_op, variable_values=gql_vars)
        result = ArtifactType.model_validate(data)
        if (project := result.project) and (artifact := project.artifact):
            return artifact.artifact_type.name
        return None

    def _ttl_duration_seconds_to_gql(self) -> int | None:
        # Set the artifact TTL to `ttl_duration_seconds` if the user provided a value.
        # Otherwise, use `ttl_status` to indicate backend values INHERIT (-1) or
        # DISABLED (-2) when the TTL is None.
        # When `ttl_change is None`, nothing changed and this is a no-op.
        INHERIT = -1  # noqa: N806
        DISABLED = -2  # noqa: N806

        if not self._ttl_changed:
            return None
        if self._ttl_is_inherited:
            return INHERIT
        return self._ttl_duration_seconds or DISABLED

    def _fetch_linked_artifacts(self) -> list[Artifact]:
        """Fetches all linked artifacts from the server."""
        from wandb._pydantic import gql_typename

        from ._generated import (
            FETCH_LINKED_ARTIFACTS_GQL,
            ArtifactPortfolioTypeFields,
            FetchLinkedArtifacts,
        )
        from ._validators import LinkArtifactFields

        if self.id is None:
            raise ValueError(
                "Unable to find any artifact memberships for artifact without an ID"
            )

        if (client := self._client) is None:
            raise ValueError("Client is not initialized")

        gql_op = gql_compat(FETCH_LINKED_ARTIFACTS_GQL)
        data = client.execute(gql_op, variable_values={"artifactID": self.id})
        result = FetchLinkedArtifacts.model_validate(data)

        if not (
            (artifact := result.artifact)
            and (memberships := artifact.artifact_memberships)
            and (membership_edges := memberships.edges)
        ):
            raise ValueError("Unable to find any artifact memberships for artifact")

        linked_artifacts: deque[Artifact] = deque()
        linked_nodes = (
            node
            for edge in membership_edges
            if (
                (node := edge.node)
                and (col := node.artifact_collection)
                and (col.typename__ == gql_typename(ArtifactPortfolioTypeFields))
            )
        )
        for node in linked_nodes:
            alias_names = unique_list(a.alias for a in node.aliases)
            version = f"v{node.version_index}"
            aliases = (
                [*alias_names, version]
                if version not in alias_names
                else [*alias_names]
            )

            if not (
                node
                and (col := node.artifact_collection)
                and (proj := col.project)
                and (proj.entity.name and proj.name)
            ):
                raise ValueError("Unable to fetch fields for linked artifact")

            link_fields = LinkArtifactFields(
                entity_name=proj.entity.name,
                project_name=proj.name,
                name=f"{col.name}:{version}",
                version=version,
                aliases=aliases,
            )
            link = self._create_linked_artifact_using_source_artifact(link_fields)
            linked_artifacts.append(link)
        return list(linked_artifacts)

    def _create_linked_artifact_using_source_artifact(
        self,
        link_fields: LinkArtifactFields,
    ) -> Artifact:
        """Copies the source artifact to a linked artifact."""
        linked_artifact = copy(self)
        linked_artifact._version = link_fields.version
        linked_artifact._aliases = link_fields.aliases
        linked_artifact._saved_aliases = copy(link_fields.aliases)
        linked_artifact._name = link_fields.name
        linked_artifact._entity = link_fields.entity_name
        linked_artifact._project = link_fields.project_name
        linked_artifact._is_link = link_fields.is_link
        linked_artifact._linked_artifacts = link_fields.linked_artifacts
        return linked_artifact


class _ArtifactVersionType(WBType):
    name = "artifactVersion"
    types = [Artifact]


TypeRegistry.add(_ArtifactVersionType)