zenml-nightly 0.68.1.dev20241112__py3-none-any.whl → 0.70.0.dev20241114__py3-none-any.whl

README.md

@@ -333,7 +333,7 @@ the Apache License Version 2.0.
     <a href="https://github.com/zenml-io/zenml-projects">Projects Showcase</a>
     <br />
     <br />
-    🎉 Version 0.68.1 is out. Check out the release notes
+    🎉 Version 0.70.0 is out. Check out the release notes
     <a href="https://github.com/zenml-io/zenml/releases">here</a>.
     <br />
     🖥️ Download our VS Code Extension <a href="https://marketplace.visualstudio.com/items?itemName=ZenML.zenml-vscode">here</a>.

RELEASE_NOTES.md

@@ -1,5 +1,82 @@
 <!-- markdown-link-check-disable -->
 
+# 0.70.0
+
+The **ZenML 0.70.0** release includes a significant number of database schema changes and migrations, which means upgrading to this version will require extra caution. As always, please make sure to make a copy of your production database before upgrading.
+
+## Key Changes
+
+* **Artifact Versioning Improvements**: The handling of artifact versions has been improved, including the API improvements like the ability to batch artifact version requests to improve the execution times and more types for the step input/output artifacts, including multiple versions of the same artifact (e.g. model checkpoints), to improve the UX using ZenML UI or while working directly with the API.
+* **Scalability Enhancements**: Various scalability improvements have been made, such as reducing unnecessary server requests and incrementing artifact versions server-side. These enhancements are expected to provide significant speed and scale improvements for ZenML users.
+* **Metadata management**: Now, all the metadata-creating functions are gathered under one method called `log_metadata`. It is possible to call this method with different inputs to log run metadata for artifact versions, model versions, steps, and runs.
+* **The oneof filtering**: This allows to filter entities using a new operator called `oneof`. You can use this with IDs (UUID type) or tags (or other string-typed attributes) like this `PipelineRunFilter(tag='oneof:["cats", "dogs"]')`.
+* **Documentation Improvements**: The ZenML documentation has been restructured and expanded, including the addition of new sections on [finetuning](https://docs.zenml.io/user-guide/llmops-guide/finetuning-llms) and [LLM/ML engineering](https://docs.zenml.io/user-guide/llmops-guide/evaluation) resources.
+* **Bug Fixes**: This release includes several bug fixes, including issues with in-process main module source loading, and more.
+
+## Caution: Make sure to back up your data before upgrading!
+While this release brings many valuable improvements, the database schema changes and migrations pose a potential risk to users. It is strongly recommended that users:
+
+* **Test the upgrade on a non-production environment**: Before upgrading a production system, test the upgrade process in a non-production environment to identify and address any issues.
+* **Back up your data**: Ensure that you have a reliable backup of your ZenML data before attempting the upgrade.
+
+## What's Changed
+* Optimizing the CI workflows by @bcdurak in https://github.com/zenml-io/zenml/pull/3145
+* Adding 0.68.0 to the migration tests by @bcdurak in https://github.com/zenml-io/zenml/pull/3144
+* Move step durations to body by @schustmi in https://github.com/zenml-io/zenml/pull/3046
+* Docs on ZenML setup by @strickvl in https://github.com/zenml-io/zenml/pull/3100
+* Remove wrongly set Model.was_created_in_this_run attribute by @schustmi in https://github.com/zenml-io/zenml/pull/3129
+* Allow specifying run tags in pipeline configuration by @schustmi in https://github.com/zenml-io/zenml/pull/3130
+* Fix materializer type compatibility check during loading by @schustmi in https://github.com/zenml-io/zenml/pull/3105
+* [docs] Add icons to headers in docs by @wjayesh in https://github.com/zenml-io/zenml/pull/3149
+* fix icons and remove redundant file by @wjayesh in https://github.com/zenml-io/zenml/pull/3150
+* Merge 0.68.1 release into develop by @schustmi in https://github.com/zenml-io/zenml/pull/3153
+* Allow filtering pipeline runs by stack component by @schustmi in https://github.com/zenml-io/zenml/pull/3142
+* Allow artifact response as step input by @schustmi in https://github.com/zenml-io/zenml/pull/3134
+* Filter component by user name by @schustmi in https://github.com/zenml-io/zenml/pull/3126
+* [docs] Restructure how-to section to make it more readable by @wjayesh in https://github.com/zenml-io/zenml/pull/3147
+* ZenML Pro web login implementation by @stefannica in https://github.com/zenml-io/zenml/pull/3141
+* Scalability improvements: Reduce misc/hydration server requests by @schustmi in https://github.com/zenml-io/zenml/pull/3093
+* Fix in-process main module source loading by @schustmi in https://github.com/zenml-io/zenml/pull/3119
+* Catch assertion in GH library by @schustmi in https://github.com/zenml-io/zenml/pull/3160
+* Enable cache precomputation for run templates by @schustmi in https://github.com/zenml-io/zenml/pull/3156
+* Add LLM and ML engineering books to README by @htahir1 in https://github.com/zenml-io/zenml/pull/3159
+* Add helper method to quickly create run template from pipeline by @schustmi in https://github.com/zenml-io/zenml/pull/3155
+* Add CLI command to export stack requirements by @schustmi in https://github.com/zenml-io/zenml/pull/3158
+* Scalability improvements: Increment artifact version server side by @schustmi in https://github.com/zenml-io/zenml/pull/3095
+* Update OpenAI integration by @safoinme in https://github.com/zenml-io/zenml/pull/3163
+* Remove deprecated torch version constraint by @safoinme in https://github.com/zenml-io/zenml/pull/3166
+* vLLM model deployer by @dudeperf3ct in https://github.com/zenml-io/zenml/pull/3032
+* Don't initialize client during flavor sync by @schustmi in https://github.com/zenml-io/zenml/pull/3168
+* Cleanup materializer temporary directories after step execution by @schustmi in https://github.com/zenml-io/zenml/pull/3162
+* Fix langchain in API docs by @avishniakov in https://github.com/zenml-io/zenml/pull/3171
+* Finetuning guide by @strickvl in https://github.com/zenml-io/zenml/pull/3157
+* Fix mypy issue vllm evidently by @safoinme in https://github.com/zenml-io/zenml/pull/3169
+* Add artifact version batch request by @schustmi in https://github.com/zenml-io/zenml/pull/3164
+* Add missing section links by @strickvl in https://github.com/zenml-io/zenml/pull/3172
+* Fix uvloop mypy by @avishniakov in https://github.com/zenml-io/zenml/pull/3174
+* Multiple output versions for a step outputs by @avishniakov in https://github.com/zenml-io/zenml/pull/3072
+* Simplify Metadata handling by @AlexejPenner in https://github.com/zenml-io/zenml/pull/3096
+* assign value to component_name in preset stack registration by @hirekk in https://github.com/zenml-io/zenml/pull/3178
+* Updating the template versions with `zenml login` by @bcdurak in https://github.com/zenml-io/zenml/pull/3177
+* Better input artifacts typing by @avishniakov in https://github.com/zenml-io/zenml/pull/3099
+* Refactor environment setup and caching by @safoinme in https://github.com/zenml-io/zenml/pull/3077
+* Fix spelling errors by @safoinme in https://github.com/zenml-io/zenml/pull/3181
+* Prevent some race conditions by @schustmi in https://github.com/zenml-io/zenml/pull/3167
+* Update stack deployments with latest features by @stefannica in https://github.com/zenml-io/zenml/pull/3183
+* Terraform best practices by @htahir1 in https://github.com/zenml-io/zenml/pull/3131
+* Fix sagemaker pipeline URLs by @stefannica in https://github.com/zenml-io/zenml/pull/3176
+* Fix lightning orchestrator for multi-step pipelines by @wjayesh in https://github.com/zenml-io/zenml/pull/3170
+* Port bugfixes from #2497 by @avishniakov in https://github.com/zenml-io/zenml/pull/3179
+* Removing the `enable_cache` from the config files  by @bcdurak in https://github.com/zenml-io/zenml/pull/3184
+* Don't pass tags to step config by @schustmi in https://github.com/zenml-io/zenml/pull/3186
+* New `log_metadata` function, new `oneof` filtering, additional `run_metadata` filtering by @bcdurak in https://github.com/zenml-io/zenml/pull/3182
+
+## New Contributors
+* @hirekk made their first contribution in https://github.com/zenml-io/zenml/pull/3178
+
+**Full Changelog**: https://github.com/zenml-io/zenml/compare/0.68.1...0.70.0
+
+
 # 0.68.1
 
 Fixes an issue with some partially cached pipelines running on remote orchestrators.

zenml/VERSION

@@ -1 +1 @@
-0.68.1.dev20241112
+0.70.0.dev20241114

zenml/__init__.py

@@ -48,6 +48,7 @@ from zenml.model.model import Model
 from zenml.pipelines import get_pipeline_context, pipeline
 from zenml.steps import step, get_step_context
 from zenml.steps.utils import log_step_metadata
+from zenml.utils.metadata_utils import log_metadata
 from zenml.entrypoints import entrypoint
 
 __all__ = [
@@ -56,6 +57,7 @@ __all__ = [
     "get_pipeline_context",
     "get_step_context",
     "load_artifact",
+    "log_metadata",
     "log_artifact_metadata",
     "log_model_metadata",
     "log_step_metadata",

zenml/artifacts/utils.py

@@ -408,7 +408,7 @@ def log_artifact_metadata(
             not provided, when being called inside a step that produces an
             artifact named `artifact_name`, the metadata will be associated to
             the corresponding newly created artifact. Or, if not provided when
-            being called outside of a step, or in a step that does not produce
+            being called outside a step, or in a step that does not produce
             any artifact named `artifact_name`, the metadata will be associated
             to the latest version of that artifact.
 
@@ -417,6 +417,10 @@ def log_artifact_metadata(
             called inside a step with a single output, or, if neither an
             artifact nor an output with the given name exists.
     """
+    logger.warning(
+        "The `log_artifact_metadata` function is deprecated and will soon be "
+        "removed. Please use `log_metadata` instead."
+    )
     try:
         step_context = get_step_context()
         in_step_outputs = (artifact_name in step_context._outputs) or (

zenml/client.py

@@ -3796,6 +3796,7 @@ class Client(metaclass=ClientMetaClass):
         templatable: Optional[bool] = None,
         tag: Optional[str] = None,
         user: Optional[Union[UUID, str]] = None,
+        run_metadata: Optional[Dict[str, str]] = None,
         pipeline: Optional[Union[UUID, str]] = None,
         code_repository: Optional[Union[UUID, str]] = None,
         model: Optional[Union[UUID, str]] = None,
@@ -3835,6 +3836,7 @@ class Client(metaclass=ClientMetaClass):
             templatable: If the runs should be templatable or not.
             tag: Tag to filter by.
             user: The name/ID of the user to filter by.
+            run_metadata: The run_metadata of the run to filter by.
             pipeline: The name/ID of the pipeline to filter by.
             code_repository: Filter by code repository name/ID.
             model: Filter by model name/ID.
@@ -3874,6 +3876,7 @@ class Client(metaclass=ClientMetaClass):
             tag=tag,
             unlisted=unlisted,
             user=user,
+            run_metadata=run_metadata,
             pipeline=pipeline,
             code_repository=code_repository,
             stack=stack,
@@ -4194,7 +4197,7 @@ class Client(metaclass=ClientMetaClass):
                 ),
             )
         except RuntimeError:
-            pass  # Cannot link to step run if called outside of a step
+            pass  # Cannot link to step run if called outside a step
         return artifact
 
     def list_artifact_versions(
@@ -4222,6 +4225,7 @@ class Client(metaclass=ClientMetaClass):
         user: Optional[Union[UUID, str]] = None,
         model: Optional[Union[UUID, str]] = None,
         pipeline_run: Optional[Union[UUID, str]] = None,
+        run_metadata: Optional[Dict[str, str]] = None,
         tag: Optional[str] = None,
         hydrate: bool = False,
     ) -> Page[ArtifactVersionResponse]:
@@ -4253,6 +4257,7 @@ class Client(metaclass=ClientMetaClass):
             user: Filter by user name or ID.
             model: Filter by model name or ID.
             pipeline_run: Filter by pipeline run name or ID.
+            run_metadata: Filter by run metadata.
             hydrate: Flag deciding whether to hydrate the output model(s)
                 by including metadata fields in the response.
 

zenml/config/server_config.py

@@ -19,13 +19,14 @@ from secrets import token_hex
 from typing import Any, Dict, List, Optional, Union
 from uuid import UUID
 
-from pydantic import BaseModel, ConfigDict, Field, model_validator
+from pydantic import BaseModel, ConfigDict, Field, PositiveInt, model_validator
 
 from zenml.constants import (
     DEFAULT_ZENML_JWT_TOKEN_ALGORITHM,
     DEFAULT_ZENML_JWT_TOKEN_LEEWAY,
     DEFAULT_ZENML_SERVER_DEVICE_AUTH_POLLING,
     DEFAULT_ZENML_SERVER_DEVICE_AUTH_TIMEOUT,
+    DEFAULT_ZENML_SERVER_GENERIC_API_TOKEN_LIFETIME,
     DEFAULT_ZENML_SERVER_LOGIN_RATE_LIMIT_DAY,
     DEFAULT_ZENML_SERVER_LOGIN_RATE_LIMIT_MINUTE,
     DEFAULT_ZENML_SERVER_MAX_DEVICE_AUTH_ATTEMPTS,
@@ -119,6 +120,8 @@ class ServerConfiguration(BaseModel):
             time of the JWT tokens issued to clients after they have
             authenticated with the ZenML server using an OAuth 2.0 device
             that has been marked as trusted.
+        generic_api_token_lifetime: The lifetime in seconds that generic
+            short-lived API tokens issued for automation purposes are valid.
         external_login_url: The login URL of an external authenticator service
             to use with the `EXTERNAL` authentication scheme.
         external_user_info_url: The user info URL of an external authenticator
@@ -230,6 +233,12 @@ class ServerConfiguration(BaseModel):
             deployment.
         max_request_body_size_in_bytes: The maximum size of the request body in
             bytes. If not specified, the default value of 256 Kb will be used.
+        memcache_max_capacity: The maximum number of entries that the memory
+            cache can hold. If not specified, the default value of 1000 will be
+            used.
+        memcache_default_expiry: The default expiry time in seconds for cache
+            entries. If not specified, the default value of 30 seconds will be
+            used.
     """
 
     deployment_type: ServerDeploymentType = ServerDeploymentType.OTHER
@@ -257,6 +266,10 @@ class ServerConfiguration(BaseModel):
     device_expiration_minutes: Optional[int] = None
     trusted_device_expiration_minutes: Optional[int] = None
 
+    generic_api_token_lifetime: PositiveInt = (
+        DEFAULT_ZENML_SERVER_GENERIC_API_TOKEN_LIFETIME
+    )
+
     external_login_url: Optional[str] = None
     external_user_info_url: Optional[str] = None
     external_cookie_name: Optional[str] = None
@@ -321,6 +334,9 @@ class ServerConfiguration(BaseModel):
         DEFAULT_ZENML_SERVER_MAX_REQUEST_BODY_SIZE_IN_BYTES
     )
 
+    memcache_max_capacity: int = 1000
+    memcache_default_expiry: int = 30
+
     _deployment_id: Optional[UUID] = None
 
     @model_validator(mode="before")

zenml/constants.py

@@ -167,9 +167,6 @@ ENV_ZENML_SERVER = "ZENML_SERVER"
 ENV_ZENML_ENFORCE_TYPE_ANNOTATIONS = "ZENML_ENFORCE_TYPE_ANNOTATIONS"
 ENV_ZENML_ENABLE_IMPLICIT_AUTH_METHODS = "ZENML_ENABLE_IMPLICIT_AUTH_METHODS"
 ENV_ZENML_DISABLE_STEP_LOGS_STORAGE = "ZENML_DISABLE_STEP_LOGS_STORAGE"
-ENV_ZENML_PIPELINE_API_TOKEN_EXPIRES_MINUTES = (
-    "ZENML_PIPELINE_API_TOKEN_EXPIRES_MINUTES"
-)
 ENV_ZENML_IGNORE_FAILURE_HOOK = "ZENML_IGNORE_FAILURE_HOOK"
 ENV_ZENML_CUSTOM_SOURCE_ROOT = "ZENML_CUSTOM_SOURCE_ROOT"
 ENV_ZENML_WHEEL_PACKAGE_NAME = "ZENML_WHEEL_PACKAGE_NAME"
@@ -270,6 +267,7 @@ ZENML_API_KEY_PREFIX = "ZENKEY_"
 DEFAULT_ZENML_SERVER_PIPELINE_RUN_AUTH_WINDOW = 60 * 48  # 48 hours
 DEFAULT_ZENML_SERVER_LOGIN_RATE_LIMIT_MINUTE = 5
 DEFAULT_ZENML_SERVER_LOGIN_RATE_LIMIT_DAY = 1000
+DEFAULT_ZENML_SERVER_GENERIC_API_TOKEN_LIFETIME = 60 * 60  # 1 hour
 
 DEFAULT_ZENML_SERVER_SECURE_HEADERS_HSTS = (
     "max-age=63072000; includeSubdomains"
@@ -410,10 +408,6 @@ MODEL_METADATA_YAML_FILE_NAME = "model_metadata.yaml"
 
 # orchestrator constants
 ORCHESTRATOR_DOCKER_IMAGE_KEY = "orchestrator"
-PIPELINE_API_TOKEN_EXPIRES_MINUTES = handle_int_env_var(
-    ENV_ZENML_PIPELINE_API_TOKEN_EXPIRES_MINUTES,
-    default=60 * 24,  # 24 hours
-)
 
 # Secret constants
 SECRET_VALUES = "values"

zenml/enums.py

@@ -245,6 +245,13 @@ class OAuthDeviceStatus(StrEnum):
     LOCKED = "locked"
 
 
+class APITokenType(StrEnum):
+    """The API token type."""
+
+    GENERIC = "generic"
+    WORKLOAD = "workload"
+
+
 class GenericFilterOps(StrEnum):
     """Ops for all filters for string values on list methods."""
 
@@ -253,6 +260,7 @@ class GenericFilterOps(StrEnum):
     CONTAINS = "contains"
     STARTSWITH = "startswith"
     ENDSWITH = "endswith"
+    ONEOF = "oneof"
     GTE = "gte"
     GT = "gt"
     LTE = "lte"

zenml/model/utils.py

@@ -56,6 +56,11 @@ def log_model_metadata(
         ValueError: If no model name/version is provided and the function is not
             called inside a step with configured `model` in decorator.
     """
+    logger.warning(
+        "The `log_model_metadata` function is deprecated and will soon be "
+        "removed. Please use `log_metadata` instead."
+    )
+
     if model_name and model_version:
         from zenml import Model
 

zenml/models/v2/base/filter.py

@@ -13,6 +13,7 @@
 #  permissions and limitations under the License.
 """Base filter model definitions."""
 
+import json
 from abc import ABC, abstractmethod
 from datetime import datetime
 from typing import (
@@ -36,7 +37,7 @@ from pydantic import (
     field_validator,
     model_validator,
 )
-from sqlalchemy import asc, desc
+from sqlalchemy import Float, and_, asc, cast, desc
 from sqlmodel import SQLModel
 
 from zenml.constants import (
@@ -63,6 +64,11 @@ logger = get_logger(__name__)
 
 AnyQuery = TypeVar("AnyQuery", bound=Any)
 
+ONEOF_ERROR = (
+    "When you are using the 'oneof:' filtering make sure that the "
+    "provided value is a json formatted list."
+)
+
 
 class Filter(BaseModel, ABC):
     """Filter for all fields.
@@ -171,8 +177,28 @@ class StrFilter(Filter):
         GenericFilterOps.STARTSWITH,
         GenericFilterOps.CONTAINS,
         GenericFilterOps.ENDSWITH,
+        GenericFilterOps.ONEOF,
+        GenericFilterOps.GT,
+        GenericFilterOps.GTE,
+        GenericFilterOps.LT,
+        GenericFilterOps.LTE,
     ]
 
+    @model_validator(mode="after")
+    def check_value_if_operation_oneof(self) -> "StrFilter":
+        """Validator to check if value is a list if oneof operation is used.
+
+        Raises:
+            ValueError: If the value is not a list
+
+        Returns:
+            self
+        """
+        if self.operation == GenericFilterOps.ONEOF:
+            if not isinstance(self.value, list):
+                raise ValueError(ONEOF_ERROR)
+        return self
+
     def generate_query_conditions_from_column(self, column: Any) -> Any:
         """Generate query conditions for a string column.
 
@@ -181,6 +207,9 @@ class StrFilter(Filter):
 
         Returns:
             A list of query conditions.
+
+        Raises:
+            ValueError: the comparison of the column to a numeric value fails.
         """
         if self.operation == GenericFilterOps.CONTAINS:
             return column.like(f"%{self.value}%")
@@ -190,6 +219,40 @@ class StrFilter(Filter):
             return column.endswith(f"{self.value}")
         if self.operation == GenericFilterOps.NOT_EQUALS:
             return column != self.value
+        if self.operation == GenericFilterOps.ONEOF:
+            return column.in_(self.value)
+        if self.operation in {
+            GenericFilterOps.GT,
+            GenericFilterOps.LT,
+            GenericFilterOps.GTE,
+            GenericFilterOps.LTE,
+        }:
+            try:
+                numeric_column = cast(column, Float)
+
+                assert self.value is not None
+
+                if self.operation == GenericFilterOps.GT:
+                    return and_(
+                        numeric_column, numeric_column > float(self.value)
+                    )
+                if self.operation == GenericFilterOps.LT:
+                    return and_(
+                        numeric_column, numeric_column < float(self.value)
+                    )
+                if self.operation == GenericFilterOps.GTE:
+                    return and_(
+                        numeric_column, numeric_column >= float(self.value)
+                    )
+                if self.operation == GenericFilterOps.LTE:
+                    return and_(
+                        numeric_column, numeric_column <= float(self.value)
+                    )
+            except Exception as e:
+                raise ValueError(
+                    f"Failed to compare the column '{column}' to the "
+                    f"value '{self.value}' (must be numeric): {e}"
+                )
 
         return column == self.value
 
@@ -211,6 +274,9 @@ class UUIDFilter(StrFilter):
         if isinstance(value, str):
             return value.replace("-", "")
 
+        if isinstance(value, list):
+            return [str(v).replace("-", "") for v in value]
+
         return value
 
     def generate_query_conditions_from_column(self, column: Any) -> Any:
@@ -588,6 +654,10 @@ class BaseFilter(BaseModel):
 
         Returns:
             A tuple of the filter value and the operator.
+
+        Raises:
+            ValueError: when we try to use the `oneof` operator with the wrong
+                value.
         """
         operator = GenericFilterOps.EQUALS  # Default operator
         if isinstance(value, str):
@@ -598,6 +668,15 @@ class BaseFilter(BaseModel):
             ):
                 value = split_value[1]
                 operator = GenericFilterOps(split_value[0])
+
+            if operator == operator.ONEOF:
+                try:
+                    value = json.loads(value)
+                    if not isinstance(value, list):
+                        raise ValueError
+                except ValueError:
+                    raise ValueError(ONEOF_ERROR)
+
         return value, operator
 
     def generate_name_or_id_query_conditions(
@@ -648,8 +727,8 @@ class BaseFilter(BaseModel):
 
         return or_(*conditions)
 
+    @staticmethod
     def generate_custom_query_conditions_for_column(
-        self,
         value: Any,
         table: Type[SQLModel],
         column: str,
@@ -833,16 +912,17 @@ class FilterGenerator:
 
         # Create str filters
         if self.is_str_field(column):
-            return StrFilter(
-                operation=GenericFilterOps(operator),
+            return self._define_str_filter(
+                operator=GenericFilterOps(operator),
                 column=column,
                 value=value,
             )
 
         # Handle unsupported datatypes
         logger.warning(
-            f"The Datatype {self._model_class.model_fields[column].annotation} might "
-            "not be supported for filtering. Defaulting to a string filter."
+            f"The Datatype {self._model_class.model_fields[column].annotation} "
+            "might not be supported for filtering. Defaulting to a string "
+            "filter."
         )
         return StrFilter(
             operation=GenericFilterOps(operator),
@@ -1032,8 +1112,9 @@ class FilterGenerator:
                     "Invalid value passed as UUID query parameter."
                 ) from e
 
-        # Cast the value to string for further comparisons.
-        value = str(value)
+        # For equality checks, ensure that the value is a valid UUID.
+        if operator == GenericFilterOps.ONEOF and not isinstance(value, list):
+            raise ValueError(ONEOF_ERROR)
 
         # Generate the filter.
         uuid_filter = UUIDFilter(
@@ -1043,6 +1124,38 @@ class FilterGenerator:
         )
         return uuid_filter
 
+    @staticmethod
+    def _define_str_filter(
+        column: str, value: Any, operator: GenericFilterOps
+    ) -> StrFilter:
+        """Define a str filter for a given column.
+
+        Args:
+            column: The column to filter on.
+            value: The UUID value by which to filter.
+            operator: The operator to use for filtering.
+
+        Returns:
+            A Filter object.
+
+        Raises:
+            ValueError: If the value is not a proper value.
+        """
+        # For equality checks, ensure that the value is a valid UUID.
+        if operator == GenericFilterOps.ONEOF and not isinstance(value, list):
+            raise ValueError(
+                "If you are using `oneof:` as a filtering op, the value needs "
+                "to be a json formatted list string."
+            )
+
+        # Generate the filter.
+        str_filter = StrFilter(
+            operation=GenericFilterOps(operator),
+            column=column,
+            value=value,
+        )
+        return str_filter
+
     @staticmethod
     def _define_bool_filter(
         column: str, value: Any, operator: GenericFilterOps

zenml/models/v2/core/artifact_version.py

@@ -474,6 +474,7 @@ class ArtifactVersionFilter(WorkspaceScopedTaggableFilter):
         "user",
         "model",
         "pipeline_run",
+        "run_metadata",
     ]
     artifact_id: Optional[Union[UUID, str]] = Field(
         default=None,
@@ -545,6 +546,10 @@ class ArtifactVersionFilter(WorkspaceScopedTaggableFilter):
         description="Name/ID of a pipeline run that is associated with this "
         "artifact version.",
     )
+    run_metadata: Optional[Dict[str, str]] = Field(
+        default=None,
+        description="The run_metadata to filter the artifact versions by.",
+    )
 
     model_config = ConfigDict(protected_namespaces=())
 
@@ -564,6 +569,7 @@ class ArtifactVersionFilter(WorkspaceScopedTaggableFilter):
             ModelSchema,
             ModelVersionArtifactSchema,
             PipelineRunSchema,
+            RunMetadataSchema,
             StepRunInputArtifactSchema,
             StepRunOutputArtifactSchema,
             StepRunSchema,
@@ -645,6 +651,23 @@ class ArtifactVersionFilter(WorkspaceScopedTaggableFilter):
             )
             custom_filters.append(pipeline_run_filter)
 
+        if self.run_metadata is not None:
+            from zenml.enums import MetadataResourceTypes
+
+            for key, value in self.run_metadata.items():
+                additional_filter = and_(
+                    RunMetadataSchema.resource_id == ArtifactVersionSchema.id,
+                    RunMetadataSchema.resource_type
+                    == MetadataResourceTypes.ARTIFACT_VERSION,
+                    RunMetadataSchema.key == key,
+                    self.generate_custom_query_conditions_for_column(
+                        value=value,
+                        table=RunMetadataSchema,
+                        column="value",
+                    ),
+                )
+                custom_filters.append(additional_filter)
+
         return custom_filters
 
 

zenml/models/v2/core/model_version.py

@@ -590,6 +590,7 @@ class ModelVersionFilter(WorkspaceScopedTaggableFilter):
     FILTER_EXCLUDE_FIELDS: ClassVar[List[str]] = [
         *WorkspaceScopedTaggableFilter.FILTER_EXCLUDE_FIELDS,
         "user",
+        "run_metadata",
     ]
 
     name: Optional[str] = Field(
@@ -619,6 +620,10 @@ class ModelVersionFilter(WorkspaceScopedTaggableFilter):
         default=None,
         description="Name/ID of the user that created the model version.",
     )
+    run_metadata: Optional[Dict[str, str]] = Field(
+        default=None,
+        description="The run_metadata to filter the model versions by.",
+    )
 
     _model_id: UUID = PrivateAttr(None)
 
@@ -651,6 +656,7 @@ class ModelVersionFilter(WorkspaceScopedTaggableFilter):
 
         from zenml.zen_stores.schemas import (
             ModelVersionSchema,
+            RunMetadataSchema,
             UserSchema,
         )
 
@@ -665,6 +671,23 @@ class ModelVersionFilter(WorkspaceScopedTaggableFilter):
             )
             custom_filters.append(user_filter)
 
+        if self.run_metadata is not None:
+            from zenml.enums import MetadataResourceTypes
+
+            for key, value in self.run_metadata.items():
+                additional_filter = and_(
+                    RunMetadataSchema.resource_id == ModelVersionSchema.id,
+                    RunMetadataSchema.resource_type
+                    == MetadataResourceTypes.MODEL_VERSION,
+                    RunMetadataSchema.key == key,
+                    self.generate_custom_query_conditions_for_column(
+                        value=value,
+                        table=RunMetadataSchema,
+                        column="value",
+                    ),
+                )
+                custom_filters.append(additional_filter)
+
         return custom_filters
 
     def apply_filter(