mlrun 1.10.0rc15__py3-none-any.whl → 1.10.0rc17__py3-none-any.whl

mlrun/artifacts/llm_prompt.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 import json
 import tempfile
+from collections import defaultdict
 from typing import Optional, Union
 
 import mlrun
@@ -253,3 +254,8 @@ class LLMPromptArtifact(Artifact):
             self.spec.prompt_template = None
             self._src_is_temp = True
         super().before_log()
+
+
+class PlaceholderDefaultDict(defaultdict):
+    def __missing__(self, key):
+        return f"{{{key}}}"

mlrun/common/constants.py

@@ -83,6 +83,9 @@ class MLRunInternalLabels:
     mlrun_type = "mlrun__type"
     original_workflow_id = "original-workflow-id"
     workflow_id = "workflow-id"
+    retrying = "retrying"
+    rerun_counter = "rerun-counter"
+    rerun_index = "rerun-index"
 
     owner = "owner"
     v3io_user = "v3io_user"

mlrun/common/schemas/__init__.py

@@ -133,6 +133,7 @@ from .k8s import NodeSelectorOperator, Resources, ResourceSpec
 from .memory_reports import MostCommonObjectTypesReport, ObjectTypeReport
 from .model_monitoring import (
     DriftStatus,
+    EndpointMode,
     EndpointType,
     EndpointUID,
     EventFieldType,

mlrun/common/schemas/model_monitoring/__init__.py

@@ -16,6 +16,7 @@ from .constants import (
     INTERSECT_DICT_KEYS,
     ApplicationEvent,
     DriftStatus,
+    EndpointMode,
     EndpointType,
     EndpointUID,
     EventFieldType,

mlrun/common/schemas/model_monitoring/constants.py

@@ -205,6 +205,11 @@ class ControllerEvent(MonitoringStrEnum):
     FIRST_REQUEST = "first_request"
     FEATURE_SET_URI = "feature_set_uri"
     ENDPOINT_TYPE = "endpoint_type"
+
+    # first_timestamp and last_timestamp are used to batch completed events
+    FIRST_TIMESTAMP = "first_timestamp"
+    LAST_TIMESTAMP = "last_timestamp"
+
     ENDPOINT_POLICY = "endpoint_policy"
     # Note: currently under endpoint policy we will have a dictionary including the keys: "application_names"
     # "base_period", and "updated_endpoint" stand for when the MEP was updated
@@ -219,6 +224,7 @@ class ControllerEventEndpointPolicy(MonitoringStrEnum):
 class ControllerEventKind(MonitoringStrEnum):
     NOP_EVENT = "nop_event"
     REGULAR_EVENT = "regular_event"
+    BATCH_COMPLETE = "batch_complete"
 
 
 class MetricData(MonitoringStrEnum):
@@ -320,6 +326,19 @@ class EndpointType(IntEnum):
     def top_level_list(cls):
         return [cls.NODE_EP, cls.ROUTER, cls.BATCH_EP]
 
+    @classmethod
+    def real_time_list(cls):
+        return [cls.NODE_EP, cls.ROUTER, cls.LEAF_EP]
+
+    @classmethod
+    def batch_list(cls):
+        return [cls.BATCH_EP]
+
+
+class EndpointMode(StrEnum):
+    REAL_TIME = "real_time"
+    BATCH = "batch"
+
 
 class MonitoringFunctionNames(MonitoringStrEnum):
     STREAM = "model-monitoring-stream"

mlrun/common/schemas/serving.py

@@ -47,3 +47,6 @@ class MonitoringData(StrEnum):
 class ModelsData(enum.Enum):
     MODEL_CLASS = 0
     MODEL_PARAMETERS = 1
+
+
+MAX_BATCH_JOB_DURATION = "1w"

mlrun/common/schemas/workflow.py

@@ -51,6 +51,9 @@ class RerunWorkflowRequest(pydantic.v1.BaseModel):
     run_id: typing.Optional[str] = None
     notifications: typing.Optional[list[Notification]] = None
     workflow_runner_node_selector: typing.Optional[dict[str, str]] = None
+    original_workflow_runner_uid: typing.Optional[str] = None
+    original_workflow_name: typing.Optional[str] = None
+    rerun_index: typing.Optional[int] = None
 
 
 class WorkflowResponse(pydantic.v1.BaseModel):

mlrun/config.py

@@ -406,11 +406,7 @@ default_config = {
                 #
                 # if set to "nil" or "none", nothing would be set
                 "modes": (
-                    "STRICT_TRANS_TABLES"
-                    ",NO_ZERO_IN_DATE"
-                    ",NO_ZERO_DATE"
-                    ",ERROR_FOR_DIVISION_BY_ZERO"
-                    ",NO_ENGINE_SUBSTITUTION",
+                    "STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,NO_ENGINE_SUBSTITUTION"
                 )
             },
         },

mlrun/db/base.py

@@ -55,6 +55,12 @@ class RunDBInterface(ABC):
     def update_run(self, updates: dict, uid, project="", iter=0):
         pass
 
+    @abstractmethod
+    def set_run_retrying_status(
+        self, project: str, name: str, run_id: str, retrying: bool
+    ):
+        pass
+
     @abstractmethod
     def abort_run(self, uid, project="", iter=0, timeout=45, status_text=""):
         pass
@@ -735,6 +741,7 @@ class RunDBInterface(ABC):
         tsdb_metrics: bool = False,
         metric_list: Optional[list[str]] = None,
         top_level: bool = False,
+        mode: Optional[mlrun.common.schemas.EndpointMode] = None,
         uids: Optional[list[str]] = None,
         latest_only: bool = False,
     ) -> mlrun.common.schemas.ModelEndpointList:

mlrun/db/httpdb.py

@@ -3813,6 +3813,7 @@ class HTTPRunDB(RunDBInterface):
         tsdb_metrics: bool = False,
         metric_list: Optional[list[str]] = None,
         top_level: bool = False,
+        mode: mm_constants.EndpointMode = None,
         uids: Optional[list[str]] = None,
         latest_only: bool = False,
     ) -> mlrun.common.schemas.ModelEndpointList:
@@ -3833,6 +3834,8 @@ class HTTPRunDB(RunDBInterface):
                                 If tsdb_metrics=False, this parameter will be ignored and no tsdb metrics
                                 will be included.
         :param top_level:       Whether to return only top level model endpoints.
+        :param mode:            Specifies the mode of the model endpoint. Can be "real-time", "batch", or both if set
+                                to None.
         :param uids:            A list of unique ids to filter by.
         :param latest_only:     Whether to return only the latest model endpoint version.
         :return:                A list of model endpoints.
@@ -3856,6 +3859,7 @@ class HTTPRunDB(RunDBInterface):
                 "tsdb-metrics": tsdb_metrics,
                 "metric": metric_list,
                 "top-level": top_level,
+                "mode": mode,
                 "uid": uids,
                 "latest-only": latest_only,
             },
@@ -4741,6 +4745,28 @@ class HTTPRunDB(RunDBInterface):
         )
         return mlrun.common.schemas.GetWorkflowResponse(**response.json())
 
+    def set_run_retrying_status(
+        self, project: str, name: str, run_id: str, retrying: bool = False
+    ):
+        """
+        Toggle the “retrying” label on a workflow-runner run.
+
+        This will POST to the workflows endpoint to either add or remove the
+        `retrying` flag on a specific run, which prevents parallel retries.
+
+        :param project:   The project name under which the workflow is defined.
+        :param name:      The workflow name (as in the URL path).
+        :param run_id:    The UID of the workflow-runner run to update.
+        :param retrying:  True to add the `retrying` label, False to remove it.
+
+        :raises MLRunHTTPError: If the HTTP request fails or returns an error status.
+        """
+        path = f"projects/{project}/workflows/{name}/runs/{run_id}/set-retry-status"
+        params = {"retrying": retrying}
+        self.api_call(
+            "POST", path, f"set retrying on {project}/{run_id}", params=params
+        )
+
     def load_project(
         self,
         name: str,

mlrun/db/nopdb.py

@@ -72,6 +72,11 @@ class NopDB(RunDBInterface):
     def update_run(self, updates: dict, uid, project="", iter=0):
         pass
 
+    def set_run_retrying_status(
+        self, project: str, name: str, run_id: str, retrying: bool
+    ):
+        pass
+
     def abort_run(self, uid, project="", iter=0, timeout=45, status_text=""):
         pass
 

mlrun/launcher/local.py

@@ -276,6 +276,19 @@ class ClientLocalLauncher(launcher.ClientBaseLauncher):
                     args = sp[1:]
         return command, args
 
+    def _validate_run(
+        self,
+        runtime: "mlrun.runtimes.BaseRuntime",
+        run: "mlrun.run.RunObject",
+    ):
+        super()._validate_run(runtime, run)
+        if self._is_run_local and run.spec.retry.count:
+            logger.warning(
+                "Retry is not supported for local runs, ignoring retry settings",
+                retry=run.spec.retry.to_dict(),
+            )
+            run.spec.retry.count = 0
+
     def _push_notifications(
         self, runobj: "mlrun.run.RunObject", runtime: "mlrun.runtimes.BaseRuntime"
     ):