frogml 1.2.21__tar.gz → 1.2.23__tar.gz

{frogml-1.2.21 → frogml-1.2.23}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: frogml
-Version: 1.2.21
+Version: 1.2.23
 Summary: frogml contains the necessary objects and communication tools for using the JFrog ml Platform
 License: Apache-2.0
 Keywords: mlops,ml,deployment,serving,model

{frogml-1.2.21 → frogml-1.2.23}/frogml/__init__.py

@@ -1,5 +1,5 @@
 __author__ = "jfrog"
-__version__ = "1.2.21"
+__version__ = "1.2.23"
 
 from frogml.sdk.model.decorators.api import api_decorator as api
 from frogml.sdk.model_loggers.model_logger import load_model, log_model

frogml-1.2.23/frogml/_proto/qwak/model_group/model_group_repository_details_pb2.py

@@ -0,0 +1,43 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# source: frogml._proto.qwak.model_group/model_group_repository_details.proto
+# Protobuf Python Version: 4.25.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n5qwak/model_group/model_group_repository_details.proto\x12\x1cqwak.model_groups.management\"r\n\x11RepositoryDetails\x12\x45\n\x0frepository_type\x18\x01 \x01(\x0b\x32,.qwak.model_groups.management.RepositoryType\x12\x16\n\x0erepository_key\x18\x02 \x01(\t\"\x80\x03\n\x17RemoteRepositoryDetails\x12K\n\x11\x64ocker_repository\x18\x01 \x01(\x0b\x32..qwak.model_groups.management.DockerRepositoryH\x00\x12V\n\x17hugging_face_repository\x18\x02 \x01(\x0b\x32\x33.qwak.model_groups.management.HuggingFaceRepositoryH\x00\x12\x45\n\x0enpm_repository\x18\x04 \x01(\x0b\x32+.qwak.model_groups.management.NpmRepositoryH\x00\x12G\n\x0fpypi_repository\x18\x05 \x01(\x0b\x32,.qwak.model_groups.management.PypiRepositoryH\x00\x12\x1d\n\x15repository_remote_url\x18\x03 \x01(\tB\x11\n\x0frepository_type\"\x12\n\x10\x44ockerRepository\"\x13\n\x11\x44\x61tasetRepository\"\x14\n\x12\x41rtifactRepository\"\x17\n\x15HuggingFaceRepository\"\x0f\n\rNpmRepository\"\x10\n\x0ePypiRepository\"\xed\x03\n\x0eRepositoryType\x12K\n\x11\x64ocker_repository\x18\x01 \x01(\x0b\x32..qwak.model_groups.management.DockerRepositoryH\x00\x12M\n\x12\x64\x61taset_repository\x18\x02 \x01(\x0b\x32/.qwak.model_groups.management.DatasetRepositoryH\x00\x12O\n\x13\x61rtifact_repository\x18\x03 \x01(\x0b\x32\x30.qwak.model_groups.management.ArtifactRepositoryH\x00\x12V\n\x17hugging_face_repository\x18\x04 \x01(\x0b\x32\x33.qwak.model_groups.management.HuggingFaceRepositoryH\x00\x12\x45\n\x0enpm_repository\x18\x05 \x01(\x0b\x32+.qwak.model_groups.management.NpmRepositoryH\x00\x12G\n\x0fpypi_repository\x18\x06 \x01(\x0b\x32,.qwak.model_groups.management.PypiRepositoryH\x00\x42\x06\n\x04typeBN\n&com.qwak.ai.management.model.group.apiB\"ModelGroupRepositoriesDetailsProtoP\x01\x62\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'qwak.model_group.model_group_repository_details_pb2', _globals)
+if _descriptor._USE_C_DESCRIPTORS == False:
+  _globals['DESCRIPTOR']._options = None
+  _globals['DESCRIPTOR']._serialized_options = b'\n&com.qwak.ai.management.model.group.apiB\"ModelGroupRepositoriesDetailsProtoP\001'
+  _globals['_REPOSITORYDETAILS']._serialized_start=87
+  _globals['_REPOSITORYDETAILS']._serialized_end=201
+  _globals['_REMOTEREPOSITORYDETAILS']._serialized_start=204
+  _globals['_REMOTEREPOSITORYDETAILS']._serialized_end=588
+  _globals['_DOCKERREPOSITORY']._serialized_start=590
+  _globals['_DOCKERREPOSITORY']._serialized_end=608
+  _globals['_DATASETREPOSITORY']._serialized_start=610
+  _globals['_DATASETREPOSITORY']._serialized_end=629
+  _globals['_ARTIFACTREPOSITORY']._serialized_start=631
+  _globals['_ARTIFACTREPOSITORY']._serialized_end=651
+  _globals['_HUGGINGFACEREPOSITORY']._serialized_start=653
+  _globals['_HUGGINGFACEREPOSITORY']._serialized_end=676
+  _globals['_NPMREPOSITORY']._serialized_start=678
+  _globals['_NPMREPOSITORY']._serialized_end=693
+  _globals['_PYPIREPOSITORY']._serialized_start=695
+  _globals['_PYPIREPOSITORY']._serialized_end=711
+  _globals['_REPOSITORYTYPE']._serialized_start=714
+  _globals['_REPOSITORYTYPE']._serialized_end=1207
+# @@protoc_insertion_point(module_scope)

{frogml-1.2.21 → frogml-1.2.23}/frogml/_proto/qwak/model_group/model_group_repository_details_pb2.pyi

@@ -39,6 +39,8 @@ class RemoteRepositoryDetails(google.protobuf.message.Message):
 
     DOCKER_REPOSITORY_FIELD_NUMBER: builtins.int
     HUGGING_FACE_REPOSITORY_FIELD_NUMBER: builtins.int
+    NPM_REPOSITORY_FIELD_NUMBER: builtins.int
+    PYPI_REPOSITORY_FIELD_NUMBER: builtins.int
     REPOSITORY_REMOTE_URL_FIELD_NUMBER: builtins.int
     repository_remote_url: builtins.str
     """The remote repository URL"""
@@ -50,16 +52,26 @@ class RemoteRepositoryDetails(google.protobuf.message.Message):
     def hugging_face_repository(self) -> global___HuggingFaceRepository:
         """HuggingFace repository"""
 
+    @property
+    def npm_repository(self) -> global___NpmRepository:
+        """NPM repository"""
+
+    @property
+    def pypi_repository(self) -> global___PypiRepository:
+        """PyPI repository"""
+
     def __init__(
         self,
         *,
         docker_repository: global___DockerRepository | None = ...,
         hugging_face_repository: global___HuggingFaceRepository | None = ...,
+        npm_repository: global___NpmRepository | None = ...,
+        pypi_repository: global___PypiRepository | None = ...,
         repository_remote_url: builtins.str = ...,
     ) -> None: ...
-    def HasField(self, field_name: typing.Literal["docker_repository", b"docker_repository", "hugging_face_repository", b"hugging_face_repository", "repository_type", b"repository_type"]) -> builtins.bool: ...
-    def ClearField(self, field_name: typing.Literal["docker_repository", b"docker_repository", "hugging_face_repository", b"hugging_face_repository", "repository_remote_url", b"repository_remote_url", "repository_type", b"repository_type"]) -> None: ...
-    def WhichOneof(self, oneof_group: typing.Literal["repository_type", b"repository_type"]) -> typing.Literal["docker_repository", "hugging_face_repository"] | None: ...
+    def HasField(self, field_name: typing.Literal["docker_repository", b"docker_repository", "hugging_face_repository", b"hugging_face_repository", "npm_repository", b"npm_repository", "pypi_repository", b"pypi_repository", "repository_type", b"repository_type"]) -> builtins.bool: ...
+    def ClearField(self, field_name: typing.Literal["docker_repository", b"docker_repository", "hugging_face_repository", b"hugging_face_repository", "npm_repository", b"npm_repository", "pypi_repository", b"pypi_repository", "repository_remote_url", b"repository_remote_url", "repository_type", b"repository_type"]) -> None: ...
+    def WhichOneof(self, oneof_group: typing.Literal["repository_type", b"repository_type"]) -> typing.Literal["docker_repository", "hugging_face_repository", "npm_repository", "pypi_repository"] | None: ...
 
 global___RemoteRepositoryDetails = RemoteRepositoryDetails
 
@@ -103,6 +115,26 @@ class HuggingFaceRepository(google.protobuf.message.Message):
 
 global___HuggingFaceRepository = HuggingFaceRepository
 
+@typing.final
+class NpmRepository(google.protobuf.message.Message):
+    DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+    def __init__(
+        self,
+    ) -> None: ...
+
+global___NpmRepository = NpmRepository
+
+@typing.final
+class PypiRepository(google.protobuf.message.Message):
+    DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+    def __init__(
+        self,
+    ) -> None: ...
+
+global___PypiRepository = PypiRepository
+
 @typing.final
 class RepositoryType(google.protobuf.message.Message):
     """The repository type"""
@@ -113,6 +145,8 @@ class RepositoryType(google.protobuf.message.Message):
     DATASET_REPOSITORY_FIELD_NUMBER: builtins.int
     ARTIFACT_REPOSITORY_FIELD_NUMBER: builtins.int
     HUGGING_FACE_REPOSITORY_FIELD_NUMBER: builtins.int
+    NPM_REPOSITORY_FIELD_NUMBER: builtins.int
+    PYPI_REPOSITORY_FIELD_NUMBER: builtins.int
     @property
     def docker_repository(self) -> global___DockerRepository:
         """Docker repository"""
@@ -129,6 +163,14 @@ class RepositoryType(google.protobuf.message.Message):
     def hugging_face_repository(self) -> global___HuggingFaceRepository:
         """HuggingFace repository"""
 
+    @property
+    def npm_repository(self) -> global___NpmRepository:
+        """NPM repository"""
+
+    @property
+    def pypi_repository(self) -> global___PypiRepository:
+        """PyPI repository"""
+
     def __init__(
         self,
         *,
@@ -136,9 +178,11 @@ class RepositoryType(google.protobuf.message.Message):
         dataset_repository: global___DatasetRepository | None = ...,
         artifact_repository: global___ArtifactRepository | None = ...,
         hugging_face_repository: global___HuggingFaceRepository | None = ...,
+        npm_repository: global___NpmRepository | None = ...,
+        pypi_repository: global___PypiRepository | None = ...,
     ) -> None: ...
-    def HasField(self, field_name: typing.Literal["artifact_repository", b"artifact_repository", "dataset_repository", b"dataset_repository", "docker_repository", b"docker_repository", "hugging_face_repository", b"hugging_face_repository", "type", b"type"]) -> builtins.bool: ...
-    def ClearField(self, field_name: typing.Literal["artifact_repository", b"artifact_repository", "dataset_repository", b"dataset_repository", "docker_repository", b"docker_repository", "hugging_face_repository", b"hugging_face_repository", "type", b"type"]) -> None: ...
-    def WhichOneof(self, oneof_group: typing.Literal["type", b"type"]) -> typing.Literal["docker_repository", "dataset_repository", "artifact_repository", "hugging_face_repository"] | None: ...
+    def HasField(self, field_name: typing.Literal["artifact_repository", b"artifact_repository", "dataset_repository", b"dataset_repository", "docker_repository", b"docker_repository", "hugging_face_repository", b"hugging_face_repository", "npm_repository", b"npm_repository", "pypi_repository", b"pypi_repository", "type", b"type"]) -> builtins.bool: ...
+    def ClearField(self, field_name: typing.Literal["artifact_repository", b"artifact_repository", "dataset_repository", b"dataset_repository", "docker_repository", b"docker_repository", "hugging_face_repository", b"hugging_face_repository", "npm_repository", b"npm_repository", "pypi_repository", b"pypi_repository", "type", b"type"]) -> None: ...
+    def WhichOneof(self, oneof_group: typing.Literal["type", b"type"]) -> typing.Literal["docker_repository", "dataset_repository", "artifact_repository", "hugging_face_repository", "npm_repository", "pypi_repository"] | None: ...
 
 global___RepositoryType = RepositoryType

{frogml-1.2.21 → frogml-1.2.23}/frogml/core/clients/feature_store/execution_management_client.py

@@ -3,6 +3,9 @@ from grpc import RpcError
 
 from frogml._proto.qwak.execution.v1.backfill_pb2 import BackfillSpec
 from frogml._proto.qwak.execution.v1.batch_pb2 import BatchIngestion
+from frogml._proto.qwak.execution.v1.streaming_aggregation_pb2 import (
+    StreamingAggregationBackfillIngestion,
+)
 from frogml._proto.qwak.execution.v1.execution_service_pb2 import (
     GetExecutionEntryRequest,
     GetExecutionEntryResponse,
@@ -12,6 +15,8 @@ from frogml._proto.qwak.execution.v1.execution_service_pb2 import (
     TriggerBackfillResponse,
     TriggerBatchFeaturesetRequest,
     TriggerBatchFeaturesetResponse,
+    TriggerStreamingAggregationBackfillRequest,
+    TriggerStreamingAggregationBackfillResponse,
 )
 from frogml._proto.qwak.execution.v1.execution_service_pb2_grpc import (
     FeatureStoreExecutionServiceStub,
@@ -100,3 +105,25 @@ class ExecutionManagementClient:
             raise FrogmlException(
                 f"Failed to get execution entry, error encountered {e}"
             )
+
+    def trigger_streaming_aggregation_backfill(
+        self, backfill_ingestion: StreamingAggregationBackfillIngestion
+    ) -> TriggerStreamingAggregationBackfillResponse:
+        """
+        Receives a configured streaming aggregation backfill proto and triggers a streaming aggregation backfill against the execution manager
+
+        Args:
+            backfill_ingestion (StreamingAggregationBackfillIngestion): A protobuf message
+                containing the backfill specification details
+
+        Returns:
+            TriggerStreamingAggregationBackfillResponse: response object from the execution manager
+        """
+        try:
+            return self._feature_store_execution_service.TriggerStreamingAggregationBackfill(
+                TriggerStreamingAggregationBackfillRequest(backfill=backfill_ingestion)
+            )
+        except RpcError as e:
+            raise FrogmlException(
+                f"Failed to trigger streaming aggregation backfill job, error encountered {e}"
+            )

frogml-1.2.23/frogml/core/feature_store/execution/streaming_backfill.py

@@ -0,0 +1,49 @@
+import pathlib
+
+from frogml.core.clients.feature_store.execution_management_client import (
+    ExecutionManagementClient,
+)
+from frogml.feature_store.feature_sets.streaming_backfill import StreamingBackfill
+
+
+class StreamingAggregationBackfill:
+
+    def __init__(
+        self,
+        streaming_backfill: StreamingBackfill,
+        source_definition_path: pathlib.Path,
+    ):
+        """
+        Initialize the streaming aggregation backfill executor.
+
+        Args:
+            streaming_backfill (StreamingBackfill): Specification containing the
+                              featureset name, time range, data sources, and transformation
+            source_definition_path (Path): Path to the Python file containing the backfill
+                                   definition. Required for locating UDF artifacts.
+        """
+        self._streaming_backfill = streaming_backfill
+        self._source_definition_path = source_definition_path
+
+    def trigger(self) -> str:
+        """
+        Triggers the streaming aggregation backfill execution.
+
+        Converts the backfill specification to proto format and sends it to
+        the execution manager to start the backfill job.
+
+        Returns:
+            str: The execution ID for tracking the backfill job status
+
+        Raises:
+            FrogmlException: If the execution manager request fails
+        """
+        backfill_proto = self._streaming_backfill._to_proto(
+            str(self._source_definition_path)
+        )
+
+        execution_client = ExecutionManagementClient()
+        response = execution_client.trigger_streaming_aggregation_backfill(
+            backfill_proto
+        )
+        return response.execution_id

{frogml-1.2.21 → frogml-1.2.23}/frogml/feature_store/feature_sets/streaming.py

@@ -2,7 +2,7 @@ import collections
 import functools
 import inspect
 from dataclasses import dataclass, field
-from datetime import datetime
+from datetime import datetime, timezone
 from typing import TYPE_CHECKING, List, Optional, Tuple, Union
 
 from typeguard import typechecked
@@ -29,6 +29,7 @@ from frogml._proto.qwak.feature_store.sources.streaming_pb2 import StreamingSour
 from frogml._proto.qwak.feature_store.sources.streaming_pb2 import (
     StreamingSource as ProtoStreamingSource,
 )
+from google.protobuf.timestamp_pb2 import Timestamp as ProtoTimestamp
 from frogml.core.clients.feature_store import FeatureRegistryClient
 from frogml.core.exceptions import FrogmlException
 from frogml.feature_store._common.artifact_utils import (
@@ -47,7 +48,7 @@ from frogml.core.feature_store.feature_sets.metadata import (
     set_metadata_on_function,
 )
 from frogml.feature_store.feature_sets.streaming_backfill import (
-    BackfillBatchDataSourceSpec,
+    BackfillDataSource,
     StreamingBackfill,
 )
 from frogml.core.feature_store.feature_sets.transformations import (
@@ -87,6 +88,7 @@ def feature_set(
     key: Optional[str] = None,
     auxiliary_sinks: List[BaseSink] = [],
     repository: Optional[str] = None,
+    backfill_max_timestamp: Optional[datetime] = None,
 ):
     """
     Creates a streaming feature set for the specified entity using the given streaming data sources.
@@ -122,6 +124,11 @@ def feature_set(
     """
 
     def decorator(function):
+        if isinstance(function, StreamingBackfill):
+            raise FrogmlException(
+                "Backfill can no longer be defined as a decorator on the feature set, it must be triggered after feature set creation."
+            )
+
         user_transformation = function()
         FeaturesetUtils.validate_base_featureset_decorator(
             user_transformation=user_transformation, entity=entity, key=key
@@ -132,10 +139,6 @@ def feature_set(
             offline_scheduling_policy=offline_scheduling_policy,
         )
 
-        streaming_backfill: Optional[StreamingBackfill] = (
-            StreamingBackfill.get_streaming_backfill_from_function(function=function)
-        )
-
         fs_name = name or function.__name__
         streaming_feature_set = StreamingFeatureSet(
             name=fs_name,
@@ -162,7 +165,7 @@ def feature_set(
             online_cluster_template=getattr(
                 function, _ONLINE_CLUSTER_SPEC, ClusterTemplate.SMALL
             ),
-            backfill=streaming_backfill,
+            backfill_max_timestamp=backfill_max_timestamp,
             __instance_module_path__=inspect.stack()[1].filename,
             auxiliary_sinks=auxiliary_sinks,
         )
@@ -209,55 +212,68 @@ def execution_specification(
 @typechecked
 def backfill(
     *,
-    start_date: datetime,
-    end_date: datetime,
-    data_sources: Union[List[str], List[BackfillBatchDataSourceSpec]],
-    backfill_transformation: SparkSqlTransformation,
+    feature_set_name: str,
+    start_date: Optional[datetime],
+    end_date: Optional[datetime],
+    data_sources: Union[List[str], List[BackfillDataSource]],
     backfill_cluster_template: Optional[ClusterTemplate] = ClusterTemplate.SMALL,
 ):
     """
-    Configures optional backfill specification for a streaming featureset. Currently available for streaming
+    Triggers a backfill execution for an existing streaming featureset. Currently available for streaming
     aggregation featuresets only.
 
-    :param start_date: backfill start date
-    :param end_date: backfill end date
-    :param data_sources: a list of dict representations connecting each batch source name to its requested time range
-    :param backfill_transformation: a backfill SQL transformation. Only SQL transformations are supported for backfill
-    :param backfill_cluster_template: an optional cluster specification for the backfill job. Defaults to SMALL.
-
-    Example:
-
-    ... code-block:: python
-
-        @streaming.feature_set(
-            entity="users",
-            data_sources=["users_registration_stream"],
-            timestamp_column_name="reg_date"
-        )
+    Args:
+        feature_set_name (str): Name of the FeatureSet to trigger a backfill for.
+        start_date (datetime): Backfill start date, on Streaming Aggregation Feature Sets,
+            needs to align with the FeatureSet tiles.
+        end_date (datetime): Backfill end date, on Streaming Aggregation Feature Sets,
+            needs to align with the FeatureSet tiles and be smaller than the Feature Set's backfill_max_timestamp.
+        data_sources (list[BackfillDataSource] | list[str]): A list of BackfillDataSource objects containing
+            batch source name and optional time range, or a list of batch source names (with no time range limits).
+        backfill_cluster_template (ClusterTemplate, optional): An optional cluster specification for the backfill job.
+            Defaults to SMALL.
+
+    Examples:
         @streaming.backfill(
-            start_date=start_date=(2022,01,01,0,0,0),
-            end_date=start_date=(2023,09,01,0,0,0),
-            data_sources=[{"batch_backfill_source_name": BackfillSourceRange(start_date=(2023,01,01,0,0,0),
-                                                                             end_date=(2023,08,01,0,0,0))}],
+            feature_set_name="user_streaming_agg_features",
+            start_date=datetime(2022,1,1,0,0,0),
+            end_date=datetime(2023,9,1,0,0,0),
+            data_sources=[BackfillDataSource(data_source_name="backfill_data_source",
+                                             start_datetime=datetime(2023,1,1,0,0,0),
+                                             end_datetime=datetime(2023,8,1,0,0,0))],
             backfill_cluster_template=ClusterTemplate.SMALL
-            backfill_transformation=SparkSqlTransformation("SELECT user_id, reg_country, reg_date FROM backfill_data_source")
         )
-        def user_streaming_features():
-            return SparkSqlTransformation("SELECT user_id, reg_country, reg_date FROM data_source")
+        def backfill_transformation():
+            return SparkSqlTransformation("SELECT user_id, reg_country, reg_date FROM backfill_data_source")
     """
 
     def decorator(function):
-        _validate_decorator_ordering(function)
-        StreamingBackfill.set_streaming_backfill_on_function(
-            function=function,
-            start_date=start_date,
-            end_date=end_date,
-            data_sources=data_sources,
-            backfill_transformation=backfill_transformation,
-            backfill_cluster_template=backfill_cluster_template,
+        if isinstance(function, StreamingFeatureSet):
+            raise FrogmlException(
+                "Backfill can no longer be defined as a decorator on the feature set, it must be triggered after feature set creation."
+            )
+
+        backfill_transformation: SparkSqlTransformation = function()
+
+        if not isinstance(backfill_transformation, SparkSqlTransformation):
+            raise FrogmlException(
+                "Backfill must defined on a method returning a SparkSqlTransformation"
+            )
+
+        streaming_backfill = StreamingBackfill(
+            featureset_name=feature_set_name,
+            start_datetime=start_date,
+            end_datetime=end_date,
+            data_sources=StreamingBackfill._get_normalized_backfill_sources_spec(
+                data_sources
+            ),
+            transform=backfill_transformation,
+            cluster_template=backfill_cluster_template,
         )
 
-        return function
+        functools.update_wrapper(streaming_backfill, backfill_transformation)
+
+        return streaming_backfill
 
     return decorator
 
@@ -328,7 +344,7 @@ class StreamingFeatureSet(BaseFeatureSet):
     offline_cluster_template: Optional[ClusterTemplate] = None
     online_cluster_template: Optional[ClusterTemplate] = None
     metadata: Optional[Metadata] = None
-    backfill: Optional[StreamingBackfill] = None
+    backfill_max_timestamp: Optional[StreamingBackfill] = None
     auxiliary_sinks: List[BaseSink] = field(default_factory=lambda: [])
 
     def __post_init__(self):
@@ -411,7 +427,6 @@ class StreamingFeatureSet(BaseFeatureSet):
             proto_featureset_type = self._get_streaming_aggregation_featureset_proto(
                 artifact_url=artifact_url,
                 streaming_sources=data_sources,
-                feature_registry=feature_registry,
                 initial_tile_size=maybe_initial_tile_size,
             )
 
@@ -465,10 +480,9 @@ class StreamingFeatureSet(BaseFeatureSet):
                 "Auxiliary Sinks Are not supported in Streaming Aggregation Feature Sets"
             )
 
-        # streaming backfill not yet supported for row-level streaming
-        if self.backfill and not is_streaming_agg:
+        if self.backfill_max_timestamp and not is_streaming_agg:
             raise FrogmlException(
-                "Streaming backfill is only supported for streaming aggregation feature sets at the moment"
+                "backfill_max_timestamp can only be set for Streaming Aggregation FeatureSet."
             )
 
         # Validate transformation is PySpark when multiple data sources are used
@@ -527,18 +541,29 @@ class StreamingFeatureSet(BaseFeatureSet):
                 )
             raise FrogmlException(error_message_str)
 
-        # Validate the backfill, if defined
-        if self.backfill:
-            self.backfill._validate_tile_size(initial_tile_size)
+        if not self.backfill_max_timestamp:
+            raise FrogmlException(
+                """
+                backfill_max_timestamp must be set for Streaming Aggregation FeatureSet.
+                 Events earlier than this timestamp can only be processed by triggering backfill,
+                 the Streaming job will not process events that are earlier than this timestamp.
+            """
+            )
+
+        self._validate_streaming_aggregation_backfill_max_timestamp()
 
         return initial_tile_size
 
-    def _validate_streaming_aggregation_backfill(self):
+    def _validate_streaming_aggregation_backfill_max_timestamp(self):
         initial_tile_size, _ = StreamingFeatureSet._get_default_slide_period(
             self.transformation.windows
         )
 
-        self.backfill._validate_tile_size(initial_tile_size)
+        if self.backfill_max_timestamp.timestamp() % initial_tile_size != 0:
+            raise FrogmlException(
+                f"Chosen backfill max timestamp is invalid,"
+                f" it has to be exactly dividable by slice size of {initial_tile_size} seconds."
+            )
 
     @staticmethod
     def _get_default_slide_period(
@@ -608,9 +633,12 @@ class StreamingFeatureSet(BaseFeatureSet):
         self,
         artifact_url: Optional[str],
         streaming_sources: List[StreamingSource],
-        feature_registry: FeatureRegistryClient,
         initial_tile_size: int,
     ) -> ProtoFeatureSetType:
+        backfill_max_timestamp = ProtoTimestamp()
+        backfill_max_timestamp.FromDatetime(
+            self.backfill_max_timestamp.astimezone(timezone.utc)
+        )
         return ProtoFeatureSetType(
             streaming_aggregation_feature_set=ProtoStreamingAggregationFeatureSet(
                 transformation=self.transformation._to_proto(
@@ -633,14 +661,7 @@ class StreamingFeatureSet(BaseFeatureSet):
                     allowed_late_arrival_seconds=60 * 10,
                     aggregations=self.transformation._get_aggregations_proto(),
                 ),
-                backfill_spec=(
-                    self.backfill._to_proto(
-                        feature_registry=feature_registry,
-                        original_instance_module_path=self.__instance_module_path__,
-                        featureset_name=self.name,
-                    )
-                    if self.backfill
-                    else None
-                ),
+                backfill_spec=None,
+                backfill_max_timestamp=backfill_max_timestamp,
             )
         )