fix(alerts): satisfy strict lint for model usage timing

litellm/proxy/db/model_usage_rollup.py

@@ -103,7 +103,7 @@ def build_model_usage_transaction(payload: SpendLogsPayload) -> ModelUsageTransa
         return None
     model: Final = payload["model"] or "unknown"
     successful: Final = payload["status"] == "success"
-    duration: Final = payload["request_duration_ms"] if "request_duration_ms" in payload else None
+    duration: Final = payload["request_duration_ms"]
     return ModelUsageTransaction(
         key=ModelUsageKey(
             date=date,

litellm/repositories/alert_repository.py

@@ -1,11 +1,11 @@
 import hashlib
 import json
 from collections.abc import Sequence
-from typing import (  # noqa: TID251  # Prisma exposes generated table actions dynamically
+from typing import (
     TYPE_CHECKING,
     Final,
     Protocol,
-    cast,
+    cast,  # noqa: TID251  # Prisma exposes generated table actions dynamically
 )
 
 from litellm.models.alert import Alert, AlertDetection, AlertPage, AlertStatus

tests/unit/proxy/db/test_model_usage_rollup.py

@@ -51,6 +51,7 @@ def _payload(**overrides: Any) -> dict[str, Any]:
         "request_tags": "[]",
         "custom_llm_provider": "openai",
         "status": "success",
+        "request_duration_ms": None,
         **overrides,
     }