mlrun 1.10.0rc25__py3-none-any.whl → 1.10.0rc27__py3-none-any.whl

mlrun/serving/system_steps.py

@@ -11,7 +11,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
 import random
 from copy import copy
 from datetime import timedelta
@@ -25,10 +24,27 @@ import mlrun.artifacts
 import mlrun.common.schemas.model_monitoring as mm_schemas
 import mlrun.feature_store
 import mlrun.serving
+from mlrun.common.model_monitoring.helpers import (
+    get_model_endpoints_creation_task_status,
+)
 from mlrun.common.schemas import MonitoringData
 from mlrun.utils import get_data_from_path, logger
 
 
+class MatchingEndpointsState(mlrun.common.types.StrEnum):
+    all_matched = "all_matched"
+    not_all_matched = "not_all_matched"
+    no_check_needed = "no_check_needed"
+    not_yet_checked = "not_yet_matched"
+
+    @staticmethod
+    def success_states() -> list[str]:
+        return [
+            MatchingEndpointsState.all_matched,
+            MatchingEndpointsState.no_check_needed,
+        ]
+
+
 class MonitoringPreProcessor(storey.MapClass):
     """preprocess step, reconstructs the serving output event body to StreamProcessingEvent schema"""
 
@@ -317,6 +333,9 @@ class BackgroundTaskStatus(storey.MapClass):
 
     def __init__(self, **kwargs):
         super().__init__(**kwargs)
+        self.matching_endpoints = MatchingEndpointsState.not_yet_checked
+        self.graph_model_endpoint_uids: set = set()
+        self.listed_model_endpoint_uids: set = set()
         self.server: mlrun.serving.GraphServer = (
             getattr(self.context, "server", None) if self.context else None
         )
@@ -337,43 +356,47 @@ class BackgroundTaskStatus(storey.MapClass):
                 )
             )
         ):
-            background_task = mlrun.get_run_db().get_project_background_task(
-                self.server.project, self.server.model_endpoint_creation_task_name
-            )
-            self._background_task_check_timestamp = mlrun.utils.now_date()
-            self._log_background_task_state(background_task.status.state)
-            self._background_task_state = background_task.status.state
+            (
+                self._background_task_state,
+                self._background_task_check_timestamp,
+                self.listed_model_endpoint_uids,
+            ) = get_model_endpoints_creation_task_status(self.server)
+        if (
+            self.listed_model_endpoint_uids
+            and self.matching_endpoints == MatchingEndpointsState.not_yet_checked
+        ):
+            if not self.graph_model_endpoint_uids:
+                self.graph_model_endpoint_uids = collect_model_endpoint_uids(
+                    self.server
+                )
+
+            if self.graph_model_endpoint_uids.issubset(self.listed_model_endpoint_uids):
+                self.matching_endpoints = MatchingEndpointsState.all_matched
+        elif self.listed_model_endpoint_uids is None:
+            self.matching_endpoints = MatchingEndpointsState.no_check_needed
 
         if (
             self._background_task_state
             == mlrun.common.schemas.BackgroundTaskState.succeeded
+            and self.matching_endpoints in MatchingEndpointsState.success_states()
         ):
             return event
         else:
             return None
 
-    def _log_background_task_state(
-        self, background_task_state: mlrun.common.schemas.BackgroundTaskState
-    ):
-        logger.info(
-            "Checking model endpoint creation task status",
-            task_name=self.server.model_endpoint_creation_task_name,
-        )
-        if (
-            background_task_state
-            in mlrun.common.schemas.BackgroundTaskState.terminal_states()
-        ):
-            logger.info(
-                f"Model endpoint creation task completed with state {background_task_state}"
-            )
-        else:  # in progress
-            logger.info(
-                f"Model endpoint creation task is still in progress with the current state: "
-                f"{background_task_state}. Events will not be monitored for the next "
-                f"{mlrun.mlconf.model_endpoint_monitoring.model_endpoint_creation_check_period} seconds",
-                name=self.name,
-                background_task_check_timestamp=self._background_task_check_timestamp.isoformat(),
-            )
+
+def collect_model_endpoint_uids(server: mlrun.serving.GraphServer) -> set[str]:
+    """Collects all model endpoint UIDs from the server's graph steps."""
+    model_endpoint_uids = set()
+    for step in server.graph.steps.values():
+        if hasattr(step, "monitoring_data"):
+            for model in step.monitoring_data.keys():
+                uid = step.monitoring_data[model].get(
+                    mlrun.common.schemas.MonitoringData.MODEL_ENDPOINT_UID
+                )
+                if uid:
+                    model_endpoint_uids.add(uid)
+    return model_endpoint_uids
 
 
 class SamplingStep(storey.MapClass):

mlrun/serving/v2_serving.py

@@ -24,6 +24,9 @@ import mlrun.common.schemas.model_monitoring
 import mlrun.model_monitoring
 from mlrun.utils import logger, now_date
 
+from ..common.model_monitoring.helpers import (
+    get_model_endpoints_creation_task_status,
+)
 from .utils import StepToDict, _extract_input_data, _update_result_body
 
 
@@ -474,22 +477,18 @@ class V2ModelServer(StepToDict):
         ) or getattr(self.context, "server", None)
         if not self.context.is_mock or self.context.monitoring_mock:
             if server.model_endpoint_creation_task_name:
-                background_task = mlrun.get_run_db().get_project_background_task(
-                    server.project, server.model_endpoint_creation_task_name
-                )
-                logger.debug(
-                    "Checking model endpoint creation task status",
-                    task_name=server.model_endpoint_creation_task_name,
+                background_task_state, _, _ = get_model_endpoints_creation_task_status(
+                    server
                 )
                 if (
-                    background_task.status.state
+                    background_task_state
                     in mlrun.common.schemas.BackgroundTaskState.terminal_states()
                 ):
                     logger.debug(
-                        f"Model endpoint creation task completed with state {background_task.status.state}"
+                        f"Model endpoint creation task completed with state {background_task_state}"
                     )
                     if (
-                        background_task.status.state
+                        background_task_state
                         == mlrun.common.schemas.BackgroundTaskState.succeeded
                     ):
                         self._model_logger = (
@@ -504,7 +503,7 @@ class V2ModelServer(StepToDict):
                 else:  # in progress
                     logger.debug(
                         f"Model endpoint creation task is still in progress with the current state: "
-                        f"{background_task.status.state}.",
+                        f"{background_task_state}.",
                         name=self.name,
                     )
             else:

mlrun/utils/helpers.py

@@ -15,7 +15,6 @@
 import asyncio
 import base64
 import enum
-import functools
 import gzip
 import hashlib
 import inspect
@@ -804,7 +803,7 @@ def remove_tag_from_artifact_uri(uri: str) -> Optional[str]:
 
 def extend_hub_uri_if_needed(uri) -> tuple[str, bool]:
     """
-    Retrieve the full uri of the item's yaml in the hub.
+    Retrieve the full uri of the function's yaml in the hub.
 
     :param uri: structure: "hub://[<source>/]<item-name>[:<tag>]"
 
@@ -845,7 +844,10 @@ def extend_hub_uri_if_needed(uri) -> tuple[str, bool]:
     # hub function directory name are with underscores instead of hyphens
     name = name.replace("-", "_")
     function_suffix = f"{name}/{tag}/src/function.yaml"
-    return indexed_source.source.get_full_uri(function_suffix), is_hub_uri
+    function_type = mlrun.common.schemas.hub.HubSourceType.functions
+    return indexed_source.source.get_full_uri(
+        function_suffix, function_type
+    ), is_hub_uri
 
 
 def gen_md_table(header, rows=None):
@@ -912,12 +914,10 @@ def enrich_image_url(
     )
     mlrun_version = config.images_tag or client_version or server_version
     tag = mlrun_version or ""
-
-    # TODO: Remove condition when mlrun/mlrun-kfp image is also supported
-    if "mlrun-kfp" not in image_url:
-        tag += resolve_image_tag_suffix(
-            mlrun_version=mlrun_version, python_version=client_python_version
-        )
+    tag += resolve_image_tag_suffix(
+        mlrun_version=mlrun_version,
+        python_version=client_python_version,
+    )
 
     # it's an mlrun image if the repository is mlrun
     is_mlrun_image = image_url.startswith("mlrun/") or "/mlrun/" in image_url
@@ -1859,10 +1859,7 @@ async def run_in_threadpool(func, *args, **kwargs):
     Run a sync-function in the loop default thread pool executor pool and await its result.
     Note that this function is not suitable for CPU-bound tasks, as it will block the event loop.
     """
-    loop = asyncio.get_running_loop()
-    if kwargs:
-        func = functools.partial(func, **kwargs)
-    return await loop.run_in_executor(None, func, *args)
+    return await asyncio.to_thread(func, *args, **kwargs)
 
 
 def is_explicit_ack_supported(context):

mlrun/utils/notifications/notification/base.py

@@ -15,11 +15,29 @@
 import asyncio
 import typing
 from copy import deepcopy
+from typing import Optional
+
+import aiohttp
 
 import mlrun.common.schemas
 import mlrun.lists
 
 
+class TimedHTTPClient:
+    def __init__(self, timeout: Optional[float] = 30.0):
+        """
+        HTTP client wrapper with built-in timeout.
+
+        Args:
+            timeout: Request timeout in seconds (default: 30.0)
+        """
+        self.timeout = aiohttp.ClientTimeout(total=timeout)
+
+    def session(self, **kwargs) -> aiohttp.ClientSession:
+        """Create a new ClientSession with the configured timeout and additional parameters."""
+        return aiohttp.ClientSession(timeout=self.timeout, **kwargs)
+
+
 class NotificationBase:
     def __init__(
         self,

mlrun/utils/notifications/notification/git.py

@@ -16,13 +16,11 @@ import json
 import os
 import typing
 
-import aiohttp
-
 import mlrun.common.schemas
 import mlrun.errors
 import mlrun.lists
 
-from .base import NotificationBase
+from .base import NotificationBase, TimedHTTPClient
 
 
 class GitNotification(NotificationBase):
@@ -148,7 +146,7 @@ class GitNotification(NotificationBase):
             }
             url = f"https://{server}/repos/{repo}/issues/{issue}/comments"
 
-        async with aiohttp.ClientSession() as session:
+        async with TimedHTTPClient().session() as session:
             resp = await session.post(url, headers=headers, json={"body": message})
             if not resp.ok:
                 resp_text = await resp.text()

mlrun/utils/notifications/notification/slack.py

@@ -14,14 +14,12 @@
 
 import typing
 
-import aiohttp
-
 import mlrun.common.runtimes.constants as runtimes_constants
 import mlrun.common.schemas
 import mlrun.lists
 import mlrun.utils.helpers
 
-from .base import NotificationBase
+from .base import NotificationBase, TimedHTTPClient
 
 
 class SlackNotification(NotificationBase):
@@ -67,7 +65,7 @@ class SlackNotification(NotificationBase):
 
         data = self._generate_slack_data(message, severity, runs, alert, event_data)
 
-        async with aiohttp.ClientSession() as session:
+        async with TimedHTTPClient().session() as session:
             async with session.post(webhook, json=data) as response:
                 response.raise_for_status()
 

mlrun/utils/notifications/notification/webhook.py

@@ -15,14 +15,13 @@
 import re
 import typing
 
-import aiohttp
 import orjson
 
 import mlrun.common.schemas
 import mlrun.lists
 import mlrun.utils.helpers
 
-from .base import NotificationBase
+from .base import NotificationBase, TimedHTTPClient
 
 
 class WebhookNotification(NotificationBase):
@@ -87,9 +86,7 @@ class WebhookNotification(NotificationBase):
         # we automatically handle it as `ssl=None` for their convenience.
         verify_ssl = verify_ssl and None if url.startswith("https") else None
 
-        async with aiohttp.ClientSession(
-            json_serialize=self._encoder,
-        ) as session:
+        async with TimedHTTPClient().session(json_serialize=self._encoder) as session:
             response = await getattr(session, method)(
                 url,
                 headers=headers,

mlrun/utils/version/version.json

@@ -1,4 +1,4 @@
 {
-  "git_commit": "93eae062e788304f5adff8f128f49db8f1f346af",
-  "version": "1.10.0-rc25"
+  "git_commit": "8b5d7492c858f675bcf8fce28d536c9af13a8d86",
+  "version": "1.10.0-rc27"
 }

{mlrun-1.10.0rc25.dist-info → mlrun-1.10.0rc27.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mlrun
-Version: 1.10.0rc25
+Version: 1.10.0rc27
 Summary: Tracking and config of machine learning runs
 Home-page: https://github.com/mlrun/mlrun
 Author: Yaron Haviv
@@ -21,8 +21,8 @@ Classifier: Topic :: Software Development :: Libraries
 Requires-Python: >=3.9, <3.12
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: urllib3>=1.26.20; python_version < "3.11"
-Requires-Dist: urllib3>=2.5.0; python_version >= "3.11"
+Requires-Dist: urllib3>=1.26.20
+Requires-Dist: v3io-frames>=0.10.15
 Requires-Dist: GitPython>=3.1.41,~=3.1
 Requires-Dist: aiohttp~=3.11
 Requires-Dist: aiohttp-retry~=2.9
@@ -39,13 +39,11 @@ Requires-Dist: tabulate~=0.8.6
 Requires-Dist: v3io~=0.7.0
 Requires-Dist: pydantic>=1.10.15
 Requires-Dist: mergedeep~=1.3
-Requires-Dist: v3io-frames~=0.10.15; python_version < "3.11"
-Requires-Dist: v3io-frames>=0.13.0; python_version >= "3.11"
 Requires-Dist: semver~=3.0
 Requires-Dist: dependency-injector~=4.41
 Requires-Dist: fsspec<=2025.7.0,>=2025.5.1
 Requires-Dist: v3iofs~=0.1.17
-Requires-Dist: storey~=1.10.11
+Requires-Dist: storey~=1.10.13
 Requires-Dist: inflection~=0.5.0
 Requires-Dist: python-dotenv~=1.0
 Requires-Dist: setuptools>=75.2
@@ -56,7 +54,7 @@ Requires-Dist: mlrun-pipelines-kfp-common~=0.5.8
 Requires-Dist: mlrun-pipelines-kfp-v1-8~=0.5.7
 Requires-Dist: docstring_parser~=0.16
 Requires-Dist: aiosmtplib~=3.0
-Requires-Dist: deepdiff~=7.0
+Requires-Dist: deepdiff<9.0.0,>=8.6.1
 Provides-Extra: s3
 Requires-Dist: boto3<1.36,>=1.28.0; extra == "s3"
 Requires-Dist: aiobotocore<2.16,>=2.5.0; extra == "s3"
@@ -92,10 +90,10 @@ Requires-Dist: databricks-sdk~=0.20.0; extra == "databricks-sdk"
 Provides-Extra: sqlalchemy
 Requires-Dist: sqlalchemy~=2.0; extra == "sqlalchemy"
 Provides-Extra: dask
-Requires-Dist: dask~=2024.12.1; python_version >= "3.11" and extra == "dask"
-Requires-Dist: distributed~=2024.12.1; python_version >= "3.11" and extra == "dask"
-Requires-Dist: dask~=2023.12.1; python_version < "3.11" and extra == "dask"
-Requires-Dist: distributed~=2023.12.1; python_version < "3.11" and extra == "dask"
+Requires-Dist: dask>=2023.12.1; python_version < "3.11" and extra == "dask"
+Requires-Dist: dask>=2024.8; python_version >= "3.11" and extra == "dask"
+Requires-Dist: distributed>=2023.12.1; python_version < "3.11" and extra == "dask"
+Requires-Dist: distributed>=2024.8; python_version >= "3.11" and extra == "dask"
 Provides-Extra: alibaba-oss
 Requires-Dist: ossfs==2025.5.0; extra == "alibaba-oss"
 Requires-Dist: oss2==2.18.4; extra == "alibaba-oss"
@@ -106,7 +104,7 @@ Requires-Dist: snowflake-connector-python~=3.7; extra == "snowflake"
 Provides-Extra: dev-postgres
 Requires-Dist: pytest-mock-resources[postgres]~=2.12; extra == "dev-postgres"
 Provides-Extra: kfp18
-Requires-Dist: mlrun_pipelines_kfp_v1_8[kfp]>=0.5.0; python_version < "3.11" and extra == "kfp18"
+Requires-Dist: mlrun_pipelines_kfp_v1_8[kfp]>=0.5.7; extra == "kfp18"
 Provides-Extra: api
 Requires-Dist: uvicorn~=0.32.1; extra == "api"
 Requires-Dist: dask-kubernetes~=0.11.0; extra == "api"
@@ -125,7 +123,6 @@ Requires-Dist: memray~=1.12; sys_platform != "win32" and extra == "api"
 Requires-Dist: aiosmtplib~=3.0; extra == "api"
 Requires-Dist: pydantic<2,>=1; extra == "api"
 Requires-Dist: mlrun-pipelines-kfp-v1-8~=0.5.7; extra == "api"
-Requires-Dist: grpcio~=1.70.0; extra == "api"
 Provides-Extra: all
 Requires-Dist: adlfs==2024.12.0; extra == "all"
 Requires-Dist: aiobotocore<2.16,>=2.5.0; extra == "all"
@@ -134,11 +131,11 @@ Requires-Dist: azure-core~=1.24; extra == "all"
 Requires-Dist: azure-identity~=1.5; extra == "all"
 Requires-Dist: azure-keyvault-secrets~=4.2; extra == "all"
 Requires-Dist: boto3<1.36,>=1.28.0; extra == "all"
-Requires-Dist: dask~=2023.12.1; python_version < "3.11" and extra == "all"
-Requires-Dist: dask~=2024.12.1; python_version >= "3.11" and extra == "all"
+Requires-Dist: dask>=2023.12.1; python_version < "3.11" and extra == "all"
+Requires-Dist: dask>=2024.8; python_version >= "3.11" and extra == "all"
 Requires-Dist: databricks-sdk~=0.20.0; extra == "all"
-Requires-Dist: distributed~=2023.12.1; python_version < "3.11" and extra == "all"
-Requires-Dist: distributed~=2024.12.1; python_version >= "3.11" and extra == "all"
+Requires-Dist: distributed>=2023.12.1; python_version < "3.11" and extra == "all"
+Requires-Dist: distributed>=2024.8; python_version >= "3.11" and extra == "all"
 Requires-Dist: gcsfs<=2025.7.0,>=2025.5.1; extra == "all"
 Requires-Dist: google-cloud-bigquery-storage~=2.17; extra == "all"
 Requires-Dist: google-cloud-bigquery[bqstorage,pandas]==3.14.1; extra == "all"
@@ -165,11 +162,11 @@ Requires-Dist: azure-core~=1.24; extra == "complete"
 Requires-Dist: azure-identity~=1.5; extra == "complete"
 Requires-Dist: azure-keyvault-secrets~=4.2; extra == "complete"
 Requires-Dist: boto3<1.36,>=1.28.0; extra == "complete"
-Requires-Dist: dask~=2023.12.1; python_version < "3.11" and extra == "complete"
-Requires-Dist: dask~=2024.12.1; python_version >= "3.11" and extra == "complete"
+Requires-Dist: dask>=2023.12.1; python_version < "3.11" and extra == "complete"
+Requires-Dist: dask>=2024.8; python_version >= "3.11" and extra == "complete"
 Requires-Dist: databricks-sdk~=0.20.0; extra == "complete"
-Requires-Dist: distributed~=2023.12.1; python_version < "3.11" and extra == "complete"
-Requires-Dist: distributed~=2024.12.1; python_version >= "3.11" and extra == "complete"
+Requires-Dist: distributed>=2023.12.1; python_version < "3.11" and extra == "complete"
+Requires-Dist: distributed>=2024.8; python_version >= "3.11" and extra == "complete"
 Requires-Dist: gcsfs<=2025.7.0,>=2025.5.1; extra == "complete"
 Requires-Dist: google-cloud-bigquery-storage~=2.17; extra == "complete"
 Requires-Dist: google-cloud-bigquery[bqstorage,pandas]==3.14.1; extra == "complete"
@@ -200,11 +197,11 @@ Requires-Dist: azure-identity~=1.5; extra == "complete-api"
 Requires-Dist: azure-keyvault-secrets~=4.2; extra == "complete-api"
 Requires-Dist: boto3<1.36,>=1.28.0; extra == "complete-api"
 Requires-Dist: dask-kubernetes~=0.11.0; extra == "complete-api"
-Requires-Dist: dask~=2023.12.1; python_version < "3.11" and extra == "complete-api"
-Requires-Dist: dask~=2024.12.1; python_version >= "3.11" and extra == "complete-api"
+Requires-Dist: dask>=2023.12.1; python_version < "3.11" and extra == "complete-api"
+Requires-Dist: dask>=2024.8; python_version >= "3.11" and extra == "complete-api"
 Requires-Dist: databricks-sdk~=0.20.0; extra == "complete-api"
-Requires-Dist: distributed~=2023.12.1; python_version < "3.11" and extra == "complete-api"
-Requires-Dist: distributed~=2024.12.1; python_version >= "3.11" and extra == "complete-api"
+Requires-Dist: distributed>=2023.12.1; python_version < "3.11" and extra == "complete-api"
+Requires-Dist: distributed>=2024.8; python_version >= "3.11" and extra == "complete-api"
 Requires-Dist: fastapi~=0.116.0; extra == "complete-api"
 Requires-Dist: gcsfs<=2025.7.0,>=2025.5.1; extra == "complete-api"
 Requires-Dist: google-cloud-bigquery-storage~=2.17; extra == "complete-api"
@@ -212,7 +209,6 @@ Requires-Dist: google-cloud-bigquery[bqstorage,pandas]==3.14.1; extra == "comple
 Requires-Dist: google-cloud-storage==2.14.0; extra == "complete-api"
 Requires-Dist: google-cloud==0.34; extra == "complete-api"
 Requires-Dist: graphviz~=0.20.0; extra == "complete-api"
-Requires-Dist: grpcio~=1.70.0; extra == "complete-api"
 Requires-Dist: humanfriendly~=10.0; extra == "complete-api"
 Requires-Dist: igz-mgmt~=0.4.1; extra == "complete-api"
 Requires-Dist: kafka-python~=2.1.0; extra == "complete-api"