acryl-datahub-cloud 0.3.11.1rc8__py3-none-any.whl → 0.3.12rc1__py3-none-any.whl

acryl_datahub_cloud/_sdk_extras/assertions_client.py

@@ -1,23 +1,717 @@
 from __future__ import annotations
 
-from typing import TYPE_CHECKING, Union
+import logging
+from datetime import datetime, timezone
+from typing import TYPE_CHECKING, Any, Optional, Union
 
-from acryl_datahub_cloud._sdk_extras.assertion import Assertion
-from datahub.metadata.urns import AssertionUrn
+from acryl_datahub_cloud._sdk_extras.assertion import (
+    SmartFreshnessAssertion,
+)
+from acryl_datahub_cloud._sdk_extras.assertion_input import (
+    AssertionIncidentBehavior,
+    DetectionMechanismInputTypes,
+    ExclusionWindowInputTypes,
+    InferenceSensitivity,
+    _SmartFreshnessAssertionInput,
+)
+from acryl_datahub_cloud._sdk_extras.entities.assertion import Assertion, TagsInputType
+from acryl_datahub_cloud._sdk_extras.entities.monitor import Monitor
+from acryl_datahub_cloud._sdk_extras.errors import SDKUsageError
+from datahub.errors import ItemNotFoundError
+from datahub.metadata.urns import AssertionUrn, CorpUserUrn, DatasetUrn, MonitorUrn
 
 if TYPE_CHECKING:
     from datahub.sdk.main_client import DataHubClient
 
+logger = logging.getLogger(__name__)
+
+# TODO: Replace __datahub_system with the actual datahub system user https://linear.app/acryl-data/issue/OBS-1351/auditstamp-actor-hydration-pattern-for-sdk-calls
+DEFAULT_CREATED_BY = CorpUserUrn.from_string("urn:li:corpuser:__datahub_system")
+
 
 class AssertionsClient:
-    def __init__(self, client: DataHubClient):
+    def __init__(self, client: "DataHubClient"):
         self.client = client
+        _print_experimental_warning()
+
+    def upsert_smart_freshness_assertion(
+        self,
+        *,
+        dataset_urn: Union[str, DatasetUrn],
+        urn: Optional[Union[str, AssertionUrn]] = None,
+        display_name: Optional[str] = None,
+        detection_mechanism: DetectionMechanismInputTypes = None,
+        sensitivity: Optional[Union[str, InferenceSensitivity]] = None,
+        exclusion_windows: Optional[ExclusionWindowInputTypes] = None,
+        training_data_lookback_days: Optional[int] = None,
+        incident_behavior: Optional[
+            Union[AssertionIncidentBehavior, list[AssertionIncidentBehavior]]
+        ] = None,
+        tags: Optional[TagsInputType] = None,
+        updated_by: Optional[Union[str, CorpUserUrn]] = None,
+    ) -> SmartFreshnessAssertion:
+        """Upsert a smart freshness assertion.
+
+        Note: keyword arguments are required.
+
+        Upsert is a combination of create and update. If the assertion does not exist, it will be created.
+        If it does exist, it will be overwritten with the input values. If the input value is None,
+        the existing value will be overridden with a default value.
+
+        Args:
+            dataset_urn: The urn of the dataset to be monitored.
+            urn: The urn of the assertion. If not provided, a urn will be generated and the assertion
+                will be _created_ in the DataHub instance.
+            display_name: The display name of the assertion. If not provided, a random display name
+                will be generated.
+            detection_mechanism: The detection mechanism to be used for the assertion. Information
+                schema is recommended. Valid values are:
+                - "information_schema" or DetectionMechanism.INFORMATION_SCHEMA
+                - "audit_log" or DetectionMechanism.AUDIT_LOG
+                - {
+                    "type": "last_modified_column",
+                    "column_name": "last_modified",
+                    "additional_filter": "last_modified > '2021-01-01'",
+                } or DetectionMechanism.LAST_MODIFIED_COLUMN(column_name='last_modified',
+                additional_filter='last_modified > 2021-01-01')
+                - "datahub_operation" or DetectionMechanism.DATAHUB_OPERATION
+            sensitivity: The sensitivity to be applied to the assertion. Valid values are:
+                - "low" or InferenceSensitivity.LOW
+                - "medium" or InferenceSensitivity.MEDIUM
+                - "high" or InferenceSensitivity.HIGH
+            exclusion_windows: The exclusion windows to be applied to the assertion, currently only
+                fixed range exclusion windows are supported. Valid values are:
+                - from datetime.datetime objects: {
+                    "start": "datetime(2025, 1, 1, 0, 0, 0)",
+                    "end": "datetime(2025, 1, 2, 0, 0, 0)",
+                }
+                - from string datetimes: {
+                    "start": "2025-01-01T00:00:00",
+                    "end": "2025-01-02T00:00:00",
+                }
+                - from FixedRangeExclusionWindow objects: FixedRangeExclusionWindow(
+                    start=datetime(2025, 1, 1, 0, 0, 0),
+                    end=datetime(2025, 1, 2, 0, 0, 0)
+                )
+            training_data_lookback_days: The training data lookback days to be applied to the
+                assertion as an integer.
+            incident_behavior: The incident behavior to be applied to the assertion. Valid values are:
+                - "raise_on_fail" or AssertionIncidentBehavior.RAISE_ON_FAIL
+                - "resolve_on_pass" or AssertionIncidentBehavior.RESOLVE_ON_PASS
+            tags: The tags to be applied to the assertion. Valid values are:
+                - a list of strings (strings will be converted to TagUrn objects)
+                - a list of TagUrn objects
+                - a list of TagAssociationClass objects
+            updated_by: Optional urn of the user who updated the assertion. The format is
+                "urn:li:corpuser:<username>", which you can find on the Users & Groups page.
+                The default is the datahub system user.
+                TODO: Retrieve the SDK user as the default instead of the datahub system user.
+
+        Returns:
+            SmartFreshnessAssertion: The created or updated assertion.
+        """
+        _print_experimental_warning()
+        now_utc = datetime.now(timezone.utc)
+
+        if updated_by is None:
+            logger.warning(
+                f"updated_by is not set, using {DEFAULT_CREATED_BY} as a placeholder"
+            )
+            updated_by = DEFAULT_CREATED_BY
 
-    def get_assertions(
-        self, urn: Union[str, list[str], AssertionUrn, list[AssertionUrn]]
-    ) -> list[Assertion]:
-        print(
-            "get_assertions is not implemented, this is a placeholder. Returning empty list."
+        # 1. If urn is not set, create a new assertion
+        if urn is None:
+            logger.info("URN is not set, creating a new assertion")
+            return self.create_smart_freshness_assertion(
+                dataset_urn=dataset_urn,
+                display_name=display_name,
+                detection_mechanism=detection_mechanism,
+                sensitivity=sensitivity,
+                exclusion_windows=exclusion_windows,
+                training_data_lookback_days=training_data_lookback_days,
+                incident_behavior=incident_behavior,
+                tags=tags,
+                created_by=updated_by,
+            )
+
+        # 2. If urn is set, first validate the input:
+        assertion_input = _SmartFreshnessAssertionInput(
+            urn=urn,
+            entity_client=self.client.entities,
+            dataset_urn=dataset_urn,
+            display_name=display_name,
+            detection_mechanism=detection_mechanism,
+            sensitivity=sensitivity,
+            exclusion_windows=exclusion_windows,
+            training_data_lookback_days=training_data_lookback_days,
+            incident_behavior=incident_behavior,
+            tags=tags,
+            created_by=updated_by,  # This will be overridden by the actual created_by
+            created_at=now_utc,  # This will be overridden by the actual created_at
+            updated_by=updated_by,
+            updated_at=now_utc,
+        )
+
+        # 3. Upsert the assertion and monitor entities:
+        assertion_entity, monitor_entity = (
+            assertion_input.to_assertion_and_monitor_entities()
         )
-        print(f"urn provided: {urn}")
-        return []
+        # If assertion upsert fails, we won't try to upsert the monitor
+        self.client.entities.upsert(assertion_entity)
+        # TODO: Wrap monitor upsert in a try-except and delete the assertion if monitor upsert fails (once delete is implemented https://linear.app/acryl-data/issue/OBS-1350/add-delete-method-to-entity-clientpy)
+        # try:
+        self.client.entities.upsert(monitor_entity)
+        # except Exception as e:
+        #     logger.error(f"Error upserting monitor: {e}")
+        #     self.client.entities.delete(assertion_entity)
+        #     raise e
+
+        return SmartFreshnessAssertion.from_entities(assertion_entity, monitor_entity)
+
+    def _upsert_and_merge_smart_freshness_assertion(
+        self,
+        *,
+        dataset_urn: Union[str, DatasetUrn],
+        urn: Optional[Union[str, AssertionUrn]] = None,
+        display_name: Optional[str] = None,
+        detection_mechanism: DetectionMechanismInputTypes = None,
+        sensitivity: Optional[Union[str, InferenceSensitivity]] = None,
+        exclusion_windows: Optional[ExclusionWindowInputTypes] = None,
+        training_data_lookback_days: Optional[int] = None,
+        incident_behavior: Optional[
+            Union[AssertionIncidentBehavior, list[AssertionIncidentBehavior]]
+        ] = None,
+        tags: Optional[TagsInputType] = None,
+        updated_by: Optional[Union[str, CorpUserUrn]] = None,
+    ) -> SmartFreshnessAssertion:
+        """Upsert and merge a smart freshness assertion.
+
+        Note: keyword arguments are required.
+
+        Upsert and merge is a combination of create and update. If the assertion does not exist,
+        it will be created. If it does exist, it will be updated. Existing assertion fields will
+        be updated if the input value is not None. If the input value is None, the existing value
+        will be preserved. If the input value can be un-set e.g. by passing an empty list or
+        empty string.
+
+        NOTE: This method is private and is not part of the public API. It will be used by the
+        yaml client to manage assertions.
+
+        Args:
+            dataset_urn: The urn of the dataset to be monitored.
+            urn: The urn of the assertion. If not provided, a urn will be generated and the
+                assertion will be _created_ in the DataHub instance.
+            display_name: The display name of the assertion. If not provided, a random display
+                name will be generated.
+            detection_mechanism: The detection mechanism to be used for the assertion. Information
+                schema is recommended. Valid values are:
+                - "information_schema" or DetectionMechanism.INFORMATION_SCHEMA
+                - "audit_log" or DetectionMechanism.AUDIT_LOG
+                - {
+                    "type": "last_modified_column",
+                    "column_name": "last_modified",
+                    "additional_filter": "last_modified > '2021-01-01'",
+                } or DetectionMechanism.LAST_MODIFIED_COLUMN(column_name='last_modified',
+                additional_filter='last_modified > 2021-01-01')
+                - "datahub_operation" or DetectionMechanism.DATAHUB_OPERATION
+            sensitivity: The sensitivity to be applied to the assertion. Valid values are:
+                - "low" or InferenceSensitivity.LOW
+                - "medium" or InferenceSensitivity.MEDIUM
+                - "high" or InferenceSensitivity.HIGH
+            exclusion_windows: The exclusion windows to be applied to the assertion, currently only
+                fixed range exclusion windows are supported. Valid values are:
+                - from datetime.datetime objects: {
+                    "start": "datetime(2025, 1, 1, 0, 0, 0)",
+                    "end": "datetime(2025, 1, 2, 0, 0, 0)",
+                }
+                - from string datetimes: {
+                    "start": "2025-01-01T00:00:00",
+                    "end": "2025-01-02T00:00:00",
+                }
+                - from FixedRangeExclusionWindow objects: FixedRangeExclusionWindow(
+                    start=datetime(2025, 1, 1, 0, 0, 0),
+                    end=datetime(2025, 1, 2, 0, 0, 0)
+                )
+            training_data_lookback_days: The training data lookback days to be applied to the
+                assertion as an integer.
+            incident_behavior: The incident behavior to be applied to the assertion. Valid values are:
+                - "raise_on_fail" or AssertionIncidentBehavior.RAISE_ON_FAIL
+                - "resolve_on_pass" or AssertionIncidentBehavior.RESOLVE_ON_PASS
+            tags: The tags to be applied to the assertion. Valid values are:
+                - a list of strings (strings will be converted to TagUrn objects)
+                - a list of TagUrn objects
+                - a list of TagAssociationClass objects
+            updated_by: Optional urn of the user who updated the assertion. The format is
+                "urn:li:corpuser:<username>", which you can find on the Users & Groups page.
+                The default is the datahub system user.
+                TODO: Retrieve the SDK user as the default instead of the datahub system user.
+
+        Returns:
+            SmartFreshnessAssertion: The created or updated assertion.
+        """
+        _print_experimental_warning()
+        now_utc = datetime.now(timezone.utc)
+
+        if updated_by is None:
+            logger.warning(
+                f"updated_by is not set, using {DEFAULT_CREATED_BY} as a placeholder"
+            )
+            updated_by = DEFAULT_CREATED_BY
+
+        # 1. If urn is not set, create a new assertion
+        if urn is None:
+            logger.info("URN is not set, creating a new assertion")
+            return self.create_smart_freshness_assertion(
+                dataset_urn=dataset_urn,
+                display_name=display_name,
+                detection_mechanism=detection_mechanism,
+                sensitivity=sensitivity,
+                exclusion_windows=exclusion_windows,
+                training_data_lookback_days=training_data_lookback_days,
+                incident_behavior=incident_behavior,
+                tags=tags,
+                created_by=updated_by,
+            )
+
+        # 2. If urn is set, first validate the input:
+        assertion_input = _SmartFreshnessAssertionInput(
+            urn=urn,
+            entity_client=self.client.entities,
+            dataset_urn=dataset_urn,
+            display_name=display_name,
+            detection_mechanism=detection_mechanism,
+            sensitivity=sensitivity,
+            exclusion_windows=exclusion_windows,
+            training_data_lookback_days=training_data_lookback_days,
+            incident_behavior=incident_behavior,
+            tags=tags,
+            created_by=updated_by,  # This will be overridden by the actual created_by
+            created_at=now_utc,  # This will be overridden by the actual created_at
+            updated_by=updated_by,
+            updated_at=now_utc,
+        )
+
+        # 3. Merge the assertion input with the existing assertion and monitor entities or create a new assertion
+        # if the assertion does not exist:
+        merged_assertion_input_or_created_assertion = (
+            self._retrieve_and_merge_assertion_and_monitor(
+                assertion_input=assertion_input,
+                dataset_urn=dataset_urn,
+                urn=urn,
+                display_name=display_name,
+                detection_mechanism=detection_mechanism,
+                sensitivity=sensitivity,
+                exclusion_windows=exclusion_windows,
+                training_data_lookback_days=training_data_lookback_days,
+                incident_behavior=incident_behavior,
+                tags=tags,
+                updated_by=updated_by,
+                now_utc=now_utc,
+            )
+        )
+
+        # Return early if we created a new assertion in the merge:
+        if isinstance(
+            merged_assertion_input_or_created_assertion, SmartFreshnessAssertion
+        ):
+            return merged_assertion_input_or_created_assertion
+
+        # 4. Upsert the assertion and monitor entities:
+        assertion_entity, monitor_entity = (
+            merged_assertion_input_or_created_assertion.to_assertion_and_monitor_entities()
+        )
+        # If assertion upsert fails, we won't try to upsert the monitor
+        self.client.entities.upsert(assertion_entity)
+        # TODO: Wrap monitor upsert in a try-except and delete the assertion if monitor upsert fails (once delete is implemented https://linear.app/acryl-data/issue/OBS-1350/add-delete-method-to-entity-clientpy)
+        # try:
+        self.client.entities.upsert(monitor_entity)
+        # except Exception as e:
+        #     logger.error(f"Error upserting monitor: {e}")
+        #     self.client.entities.delete(assertion_entity)
+        #     raise e
+
+        return SmartFreshnessAssertion.from_entities(assertion_entity, monitor_entity)
+
+    def _retrieve_and_merge_assertion_and_monitor(
+        self,
+        assertion_input: _SmartFreshnessAssertionInput,
+        dataset_urn: Union[str, DatasetUrn],
+        urn: Union[str, AssertionUrn],
+        display_name: Optional[str],
+        detection_mechanism: DetectionMechanismInputTypes,
+        sensitivity: Optional[Union[str, InferenceSensitivity]],
+        exclusion_windows: Optional[ExclusionWindowInputTypes],
+        training_data_lookback_days: Optional[int],
+        incident_behavior: Optional[
+            Union[AssertionIncidentBehavior, list[AssertionIncidentBehavior]]
+        ],
+        tags: Optional[TagsInputType],
+        updated_by: Optional[Union[str, CorpUserUrn]],
+        now_utc: datetime,
+    ) -> Union[SmartFreshnessAssertion, _SmartFreshnessAssertionInput]:
+        # 1. Retrieve any existing assertion and monitor entities:
+        maybe_assertion_entity, monitor_urn, maybe_monitor_entity = (
+            self._retrieve_assertion_and_monitor(assertion_input)
+        )
+
+        # 2.1 If the assertion and monitor entities exist, create a SmartFreshnessAssertion object from them:
+        if maybe_assertion_entity and maybe_monitor_entity:
+            existing_assertion = SmartFreshnessAssertion.from_entities(
+                maybe_assertion_entity, maybe_monitor_entity
+            )
+        # 2.2 If the assertion exists but the monitor does not, create a placeholder monitor entity to be able to create the assertion:
+        elif maybe_assertion_entity and not maybe_monitor_entity:
+            existing_assertion = SmartFreshnessAssertion.from_entities(
+                maybe_assertion_entity,
+                Monitor(
+                    id=monitor_urn, info=("ASSERTION", "ACTIVE")
+                ),  # TODO: Set active based on enabled parameter once it is added
+            )
+        # 2.3 If the assertion does not exist, create a new assertion with a generated urn and return the assertion input:
+        elif not maybe_assertion_entity:
+            logger.info(
+                f"No existing assertion entity found for assertion urn {urn}, creating a new assertion with a generated urn"
+            )
+            return self.create_smart_freshness_assertion(
+                dataset_urn=dataset_urn,
+                display_name=display_name,
+                detection_mechanism=detection_mechanism,
+                sensitivity=sensitivity,
+                exclusion_windows=exclusion_windows,
+                training_data_lookback_days=training_data_lookback_days,
+                incident_behavior=incident_behavior,
+                tags=tags,
+                created_by=updated_by,
+            )
+
+        # 3. Check for any issues e.g. different dataset urns
+        if (
+            existing_assertion
+            and existing_assertion.dataset_urn != assertion_input.dataset_urn
+        ):
+            raise SDKUsageError(
+                f"Dataset URN mismatch, existing assertion: {existing_assertion.dataset_urn} != new assertion: {dataset_urn}"
+            )
+
+        # 4. Merge the existing assertion with the validated input:
+        merged_assertion_input = self._merge_input(
+            dataset_urn=dataset_urn,
+            urn=urn,
+            display_name=display_name,
+            detection_mechanism=detection_mechanism,
+            sensitivity=sensitivity,
+            exclusion_windows=exclusion_windows,
+            training_data_lookback_days=training_data_lookback_days,
+            incident_behavior=incident_behavior,
+            tags=tags,
+            now_utc=now_utc,
+            assertion_input=assertion_input,
+            maybe_assertion_entity=maybe_assertion_entity,
+            maybe_monitor_entity=maybe_monitor_entity,
+            existing_assertion=existing_assertion,
+        )
+
+        return merged_assertion_input
+
+    def _retrieve_assertion_and_monitor(
+        self, assertion_input: _SmartFreshnessAssertionInput
+    ) -> tuple[Optional[Assertion], MonitorUrn, Optional[Monitor]]:
+        """Retrieve the assertion and monitor entities from the DataHub instance.
+
+        Args:
+            assertion_input: The validated input to the function.
+
+        Returns:
+            The assertion and monitor entities.
+        """
+        assert assertion_input.urn is not None, "URN is required"
+
+        # Get assertion entity
+        maybe_assertion_entity: Optional[Assertion] = None
+        try:
+            entity = self.client.entities.get(assertion_input.urn)
+            if entity is not None:
+                assert isinstance(entity, Assertion)
+                maybe_assertion_entity = entity
+        except ItemNotFoundError:
+            pass
+
+        # Get monitor entity
+        monitor_urn = Monitor._ensure_id(
+            id=(assertion_input.dataset_urn, assertion_input.urn)
+        )
+        maybe_monitor_entity: Optional[Monitor] = None
+        try:
+            entity = self.client.entities.get(monitor_urn)
+            if entity is not None:
+                assert isinstance(entity, Monitor)
+                maybe_monitor_entity = entity
+        except ItemNotFoundError:
+            pass
+
+        return maybe_assertion_entity, monitor_urn, maybe_monitor_entity
+
+    def _merge_input(
+        self,
+        dataset_urn: Union[str, DatasetUrn],
+        urn: Union[str, AssertionUrn],
+        display_name: Optional[str],
+        detection_mechanism: DetectionMechanismInputTypes,
+        sensitivity: Optional[Union[str, InferenceSensitivity]],
+        exclusion_windows: Optional[ExclusionWindowInputTypes],
+        training_data_lookback_days: Optional[int],
+        incident_behavior: Optional[
+            Union[AssertionIncidentBehavior, list[AssertionIncidentBehavior]]
+        ],
+        tags: Optional[TagsInputType],
+        now_utc: datetime,
+        assertion_input: _SmartFreshnessAssertionInput,
+        maybe_assertion_entity: Optional[Assertion],
+        maybe_monitor_entity: Optional[Monitor],
+        existing_assertion: SmartFreshnessAssertion,
+    ) -> _SmartFreshnessAssertionInput:
+        """Merge the input with the existing assertion and monitor entities.
+
+        Args:
+            dataset_urn: The urn of the dataset to be monitored.
+            urn: The urn of the assertion.
+            display_name: The display name of the assertion.
+            detection_mechanism: The detection mechanism to be used for the assertion.
+            sensitivity: The sensitivity to be applied to the assertion.
+            exclusion_windows: The exclusion windows to be applied to the assertion.
+            training_data_lookback_days: The training data lookback days to be applied to the assertion.
+            incident_behavior: The incident behavior to be applied to the assertion.
+            tags: The tags to be applied to the assertion.
+            now_utc: The current UTC time from when the function is called.
+            assertion_input: The validated input to the function.
+            maybe_assertion_entity: The existing assertion entity from the DataHub instance.
+            maybe_monitor_entity: The existing monitor entity from the DataHub instance.
+            existing_assertion: The existing assertion from the DataHub instance.
+
+        Returns:
+            The merged assertion input.
+        """
+        merged_assertion_input = _SmartFreshnessAssertionInput(
+            urn=urn,
+            entity_client=self.client.entities,
+            dataset_urn=dataset_urn,
+            display_name=_merge_field(
+                display_name,
+                "display_name",
+                assertion_input,
+                existing_assertion,
+                maybe_assertion_entity.description if maybe_assertion_entity else None,
+            ),
+            detection_mechanism=_merge_field(
+                detection_mechanism,
+                "detection_mechanism",
+                assertion_input,
+                existing_assertion,
+                SmartFreshnessAssertion._get_detection_mechanism(  # TODO: Consider moving this conversion to DetectionMechanism.parse(), it could avoid having to use Optional on the return type of SmartFreshnessAssertion.get_detection_mechanism()
+                    maybe_assertion_entity, maybe_monitor_entity, default=None
+                )
+                if maybe_assertion_entity and maybe_monitor_entity
+                else None,
+            ),
+            sensitivity=_merge_field(
+                sensitivity,
+                "sensitivity",
+                assertion_input,
+                existing_assertion,
+                maybe_monitor_entity.sensitivity if maybe_monitor_entity else None,
+            ),
+            exclusion_windows=_merge_field(
+                exclusion_windows,
+                "exclusion_windows",
+                assertion_input,
+                existing_assertion,
+                maybe_monitor_entity.exclusion_windows
+                if maybe_monitor_entity
+                else None,
+            ),
+            training_data_lookback_days=_merge_field(
+                training_data_lookback_days,
+                "training_data_lookback_days",
+                assertion_input,
+                existing_assertion,
+                maybe_monitor_entity.training_data_lookback_days
+                if maybe_monitor_entity
+                else None,
+            ),
+            incident_behavior=_merge_field(
+                incident_behavior,
+                "incident_behavior",
+                assertion_input,
+                existing_assertion,
+                SmartFreshnessAssertion._get_incident_behavior(maybe_assertion_entity)
+                if maybe_assertion_entity
+                else None,
+            ),
+            tags=_merge_field(
+                tags,
+                "tags",
+                assertion_input,
+                existing_assertion,
+                maybe_assertion_entity.tags if maybe_assertion_entity else None,
+            ),
+            created_by=existing_assertion.created_by
+            or DEFAULT_CREATED_BY,  # Override with the existing assertion's created_by or the default created_by if not set
+            created_at=existing_assertion.created_at
+            or now_utc,  # Override with the existing assertion's created_at or now if not set
+            updated_by=assertion_input.updated_by,  # Override with the input's updated_by
+            updated_at=assertion_input.updated_at,  # Override with the input's updated_at (now)
+        )
+
+        return merged_assertion_input
+
+    def create_smart_freshness_assertion(
+        self,
+        *,
+        dataset_urn: Union[str, DatasetUrn],
+        display_name: Optional[str] = None,
+        detection_mechanism: DetectionMechanismInputTypes = None,
+        sensitivity: Optional[Union[str, InferenceSensitivity]] = None,
+        exclusion_windows: Optional[ExclusionWindowInputTypes] = None,
+        training_data_lookback_days: Optional[int] = None,
+        incident_behavior: Optional[
+            Union[AssertionIncidentBehavior, list[AssertionIncidentBehavior]]
+        ] = None,
+        tags: Optional[TagsInputType] = None,
+        created_by: Optional[Union[str, CorpUserUrn]] = None,
+    ) -> SmartFreshnessAssertion:
+        """Create a smart freshness assertion.
+
+        Note: keyword arguments are required.
+
+        Args:
+            dataset_urn: The urn of the dataset to be monitored.
+            display_name: The display name of the assertion. If not provided, a random display
+                name will be generated.
+            detection_mechanism: The detection mechanism to be used for the assertion. Information
+                schema is recommended. Valid values are:
+                - "information_schema" or DetectionMechanism.INFORMATION_SCHEMA
+                - "audit_log" or DetectionMechanism.AUDIT_LOG
+                - {
+                    "type": "last_modified_column",
+                    "column_name": "last_modified",
+                    "additional_filter": "last_modified > '2021-01-01'",
+                } or DetectionMechanism.LAST_MODIFIED_COLUMN(column_name='last_modified',
+                additional_filter='last_modified > 2021-01-01')
+                - "datahub_operation" or DetectionMechanism.DATAHUB_OPERATION
+            sensitivity: The sensitivity to be applied to the assertion. Valid values are:
+                - "low" or InferenceSensitivity.LOW
+                - "medium" or InferenceSensitivity.MEDIUM
+                - "high" or InferenceSensitivity.HIGH
+            exclusion_windows: The exclusion windows to be applied to the assertion, currently only
+                fixed range exclusion windows are supported. Valid values are:
+                - from datetime.datetime objects: {
+                    "start": "datetime(2025, 1, 1, 0, 0, 0)",
+                    "end": "datetime(2025, 1, 2, 0, 0, 0)",
+                }
+                - from string datetimes: {
+                    "start": "2025-01-01T00:00:00",
+                    "end": "2025-01-02T00:00:00",
+                }
+                - from FixedRangeExclusionWindow objects: FixedRangeExclusionWindow(
+                    start=datetime(2025, 1, 1, 0, 0, 0),
+                    end=datetime(2025, 1, 2, 0, 0, 0)
+                )
+            training_data_lookback_days: The training data lookback days to be applied to the
+                assertion as an integer.
+            incident_behavior: The incident behavior to be applied to the assertion. Valid values are:
+                - "raise_on_fail" or AssertionIncidentBehavior.RAISE_ON_FAIL
+                - "resolve_on_pass" or AssertionIncidentBehavior.RESOLVE_ON_PASS
+            tags: The tags to be applied to the assertion. Valid values are:
+                - a list of strings (strings will be converted to TagUrn objects)
+                - a list of TagUrn objects
+                - a list of TagAssociationClass objects
+            created_by: Optional urn of the user who created the assertion. The format is
+                "urn:li:corpuser:<username>", which you can find on the Users & Groups page.
+                The default is the datahub system user.
+                TODO: Retrieve the SDK user as the default instead of the datahub system user.
+
+        Returns:
+            SmartFreshnessAssertion: The created assertion.
+        """
+        _print_experimental_warning()
+        now_utc = datetime.now(timezone.utc)
+        if created_by is None:
+            logger.warning(
+                f"Created by is not set, using {DEFAULT_CREATED_BY} as a placeholder"
+            )
+            created_by = DEFAULT_CREATED_BY
+        assertion_input = _SmartFreshnessAssertionInput(
+            urn=None,
+            entity_client=self.client.entities,
+            dataset_urn=dataset_urn,
+            display_name=display_name,
+            detection_mechanism=detection_mechanism,
+            sensitivity=sensitivity,
+            exclusion_windows=exclusion_windows,
+            training_data_lookback_days=training_data_lookback_days,
+            incident_behavior=incident_behavior,
+            tags=tags,
+            created_by=created_by,
+            created_at=now_utc,
+            updated_by=created_by,
+            updated_at=now_utc,
+        )
+        assertion_entity, monitor_entity = (
+            assertion_input.to_assertion_and_monitor_entities()
+        )
+        # If assertion creation fails, we won't try to create the monitor
+        self.client.entities.create(assertion_entity)
+        # TODO: Wrap monitor creation in a try-except and delete the assertion if monitor creation fails (once delete is implemented https://linear.app/acryl-data/issue/OBS-1350/add-delete-method-to-entity-clientpy)
+        # try:
+        self.client.entities.create(monitor_entity)
+        # except Exception as e:
+        #     logger.error(f"Error creating monitor: {e}")
+        #     self.client.entities.delete(assertion_entity)
+        #     raise e
+        return SmartFreshnessAssertion.from_entities(assertion_entity, monitor_entity)
+
+
+def _merge_field(
+    input_field_value: Any,
+    input_field_name: str,
+    validated_assertion_input: _SmartFreshnessAssertionInput,
+    validated_existing_assertion: SmartFreshnessAssertion,
+    existing_entity_value: Optional[Any] = None,  # TODO: Can we do better than Any?
+) -> Any:
+    """Merge the input field value with any existing entity value or default value.
+
+    The merge logic is as follows:
+    - If the input is None, use the existing value
+    - If the input is not None, use the input value
+    - If the input is an empty list or empty string, still use the input value (falsy values can be used to unset fields)
+    - If the input is a non-empty list or non-empty string, use the input value
+    - If the input is None and the existing value is None, use the default value from _AssertionInput
+
+    Args:
+        input_field_value: The value of the field in the input e.g. passed to the function.
+        input_field_name: The name of the field in the input.
+        validated_assertion_input: The *validated* input to the function.
+        validated_existing_assertion: The *validated* existing assertion from the DataHub instance.
+        existing_entity_value: The value of the field in the existing entity from the DataHub instance, directly retrieved from the entity.
+
+        Returns:
+            The merged value of the field.
+
+    """
+    if input_field_value is None:  # Input value default
+        if existing_entity_value is not None:  # Existing entity value set
+            return existing_entity_value
+        elif (
+            getattr(validated_existing_assertion, input_field_name) is None
+        ):  # Validated existing value not set
+            return getattr(validated_assertion_input, input_field_name)
+        else:  # Validated existing value set
+            return getattr(validated_existing_assertion, input_field_name)
+    else:  # Input value set
+        return input_field_value
+
+
+def _print_experimental_warning() -> None:
+    print(
+        "Warning: The assertions client is experimental and under heavy development. Expect breaking changes."
+    )