Merge remote-tracking branch 'origin/main' into litellm_provider_throughput_chart

litellm/model_prices_and_context_window_backup.json

@@ -79845,5 +79845,27 @@
         "supports_tool_choice": true,
         "supports_vision": false,
         "supports_web_search": false
+    },
+    "azure_ai/kimi-k2-thinking": {
+        "input_cost_per_token": 6e-07,
+        "litellm_provider": "azure_ai",
+        "max_input_tokens": 262144,
+        "max_output_tokens": 262144,
+        "max_tokens": 262144,
+        "mode": "chat",
+        "output_cost_per_token": 2.5e-06,
+        "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/kimi/",
+        "supported_modalities": [
+            "text"
+        ],
+        "supported_output_modalities": [
+            "text"
+        ],
+        "supports_function_calling": true,
+        "supports_prompt_caching": false,
+        "supports_reasoning": true,
+        "supports_tool_choice": true,
+        "supports_video_input": false,
+        "supports_vision": false
     }
 }

litellm/proxy/db/log_db_metrics.py

@@ -13,6 +13,7 @@ from functools import wraps
 from types import MappingProxyType
 from typing import Final
 
+from litellm._logging import verbose_proxy_logger
 from litellm._service_logger import ServiceTypes
 
 _PRISMA_CLIENT_CRUD: Final = frozenset({"get_data", "update_data", "delete_data"})
@@ -200,14 +201,17 @@ async def _handle_logging_db_exception(
     if _is_exception_related_to_db(e) is not True:
         return False
 
-    await proxy_logging_obj.service_logging_obj.async_service_failure_hook(
-        error=e,
-        service=ServiceTypes.DB,
-        call_type=func.__name__,
-        parent_otel_span=kwargs.get("parent_otel_span"),
-        duration=(end_time - start_time).total_seconds(),
-        start_time=start_time,
-        end_time=end_time,
-        event_metadata=metadata_of(kwargs),
-    )
+    try:
+        await proxy_logging_obj.service_logging_obj.async_service_failure_hook(
+            error=e,
+            service=ServiceTypes.DB,
+            call_type=func.__name__,
+            parent_otel_span=kwargs.get("parent_otel_span"),
+            duration=(end_time - start_time).total_seconds(),
+            start_time=start_time,
+            end_time=end_time,
+            event_metadata=metadata_of(kwargs),
+        )
+    except Exception as hook_error:
+        verbose_proxy_logger.debug("log_db_metrics: failure hook raised for %s: %s", func.__name__, hook_error)
     return True

model_prices_and_context_window.json

@@ -79845,5 +79845,27 @@
         "supports_tool_choice": true,
         "supports_vision": false,
         "supports_web_search": false
+    },
+    "azure_ai/kimi-k2-thinking": {
+        "input_cost_per_token": 6e-07,
+        "litellm_provider": "azure_ai",
+        "max_input_tokens": 262144,
+        "max_output_tokens": 262144,
+        "max_tokens": 262144,
+        "mode": "chat",
+        "output_cost_per_token": 2.5e-06,
+        "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/kimi/",
+        "supported_modalities": [
+            "text"
+        ],
+        "supported_output_modalities": [
+            "text"
+        ],
+        "supports_function_calling": true,
+        "supports_prompt_caching": false,
+        "supports_reasoning": true,
+        "supports_tool_choice": true,
+        "supports_video_input": false,
+        "supports_vision": false
     }
 }

tests/unit/proxy/db/test_log_db_metrics.py

@@ -5,6 +5,7 @@ from typing import Final
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
+from prisma.errors import PrismaError
 
 from litellm._service_logger import ServiceTypes
 from litellm.proxy.db.db_lookup_gate import bounded_db_lookup
@@ -27,6 +28,16 @@ def success_hook() -> Iterator[AsyncMock]:
         yield hook
 
 
+@pytest.fixture
+def failure_hook() -> Iterator[AsyncMock]:
+    hook: Final = AsyncMock()
+    with patch(
+        "litellm.proxy.proxy_server.proxy_logging_obj",
+        MagicMock(service_logging_obj=MagicMock(async_service_failure_hook=hook)),
+    ):
+        yield hook
+
+
 async def _db_call_types(hook: AsyncMock) -> tuple[str, ...]:
     await asyncio.sleep(0)
     return tuple(call.kwargs["call_type"] for call in hook.await_args_list if call.kwargs["service"] == ServiceTypes.DB)
@@ -248,3 +259,18 @@ async def test_a_task_spawned_by_a_decorated_call_that_queries_after_it_returned
     await background
 
     assert await _db_call_types(success_hook) == ("read_key_row", "find_unique")
+
+
+@pytest.mark.asyncio
+async def test_a_raising_failure_hook_never_replaces_the_prisma_error(failure_hook: AsyncMock) -> None:
+    failure_hook.side_effect = RuntimeError("exporter down")
+
+    @log_db_metrics
+    async def insert_data(**kwargs: object) -> None:
+        raise PrismaError("connection reset")
+
+    with pytest.raises(PrismaError, match="connection reset"):
+        await insert_data(table_name="key")
+
+    assert failure_hook.await_count == 1
+    assert failure_hook.await_args_list[0].kwargs["call_type"] == "insert_data"