refactor(types): replace Any with proven types in 6 files (#47268)

* refactor(types): replace Any with proven types in 39 files

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* test(anthropic): use valid response fixtures for strict parsing

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* refactor(types): revert anthropic chat response seam

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* refactor(types): revert validation seams that changed tolerated inputs

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* refactor(types): revert remaining validation seams to keep behavior unchanged

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* refactor(types): drop no-op search router annotation and helper docstring

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* refactor(types): keep ip address utils at base

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

---------

Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

litellm/integrations/datadog/datadog_llm_obs.py

@@ -12,7 +12,7 @@ import os
 from collections.abc import Mapping, Sequence
 from datetime import datetime
 from types import MappingProxyType
-from typing import Any, Final, Literal
+from typing import Final, Literal
 
 import httpx
 from pydantic import ConfigDict, TypeAdapter, ValidationError
@@ -54,6 +54,7 @@ from litellm.types.utils import (
     PROMPT_CARRYING_GUARDRAIL_FIELDS,
     PROMPT_QUOTING_ROUTING_DECISION_FIELDS,
     CallTypes,
+    StandardLoggingMetadata,
     StandardLoggingPayload,
     StandardLoggingPayloadErrorInformation,
 )
@@ -93,8 +94,12 @@ _ROUTER_DIMENSIONS: Final[tuple[str, ...]] = ("router_tier", "router_cause", "ro
 _COST_DIMENSIONS: Final[tuple[str, ...]] = ("team", "user", "key_alias", "model_group", *_ROUTER_DIMENSIONS)
 
 
-def _metadata_of(standard_logging_payload: StandardLoggingPayload) -> Mapping[str, Any]:
-    metadata: Final = standard_logging_payload.get("metadata")
+def _logging_metadata_of(standard_logging_payload: StandardLoggingPayload) -> StandardLoggingMetadata | None:
+    return standard_logging_payload.get("metadata")
+
+
+def _metadata_of(standard_logging_payload: StandardLoggingPayload) -> Mapping[str, object]:
+    metadata: Final = _logging_metadata_of(standard_logging_payload)
     return metadata or _EMPTY_MAPPING
 
 
@@ -1057,14 +1062,23 @@ class DataDogLLMObsLogger(CustomBatchLogger):
 
         # Get budget information from metadata
         metadata: Final = _metadata_of(standard_logging_payload)
+        logging_metadata: Final = _logging_metadata_of(standard_logging_payload)
 
         # API key max budget
-        user_api_key_max_budget: Final = metadata.get("user_api_key_max_budget")
+        user_api_key_max_budget: Final = (
+            logging_metadata["user_api_key_max_budget"]
+            if logging_metadata and "user_api_key_max_budget" in logging_metadata
+            else None
+        )
         if user_api_key_max_budget is not None:
             spend_metrics["user_api_key_max_budget"] = float(user_api_key_max_budget)
 
         # API key spend
-        user_api_key_spend: Final = metadata.get("user_api_key_spend")
+        user_api_key_spend: Final = (
+            logging_metadata["user_api_key_spend"]
+            if logging_metadata and "user_api_key_spend" in logging_metadata
+            else None
+        )
         if user_api_key_spend is not None:
             try:
                 spend_metrics["user_api_key_spend"] = float(user_api_key_spend)

litellm/llms/base_llm/chat/transformation.py

@@ -428,7 +428,7 @@ class BaseConfig(ABC):
         streaming_response: Union[Iterator[str], AsyncIterator[str], "ModelResponse"],
         sync_stream: bool,
         json_mode: bool | None = False,
-    ) -> Any:
+    ) -> object:
         pass
 
     def get_model_response_iterator_for_request(

litellm/models/organization_membership.py

@@ -6,7 +6,6 @@ Canonical definition for ``litellm_organizationmembership``. Re-exported from
 """
 
 from datetime import datetime
-from typing import Any
 
 from pydantic import ConfigDict, model_validator
 
@@ -24,7 +23,7 @@ class LiteLLM_OrganizationMembershipTable(LiteLLMPydanticObjectBase):
     budget_id: str | None = None
     created_at: datetime
     updated_at: datetime
-    user: Any | None = None
+    user: object | None = None
     litellm_budget_table: LiteLLM_BudgetTable | None = None
     user_email: str | None = None
 

litellm/proxy/auth/auth_checks.py

@@ -2663,7 +2663,7 @@ async def _load_team_membership_on_cache_miss(
     parent_otel_span: Span | None,
     proxy_logging_obj: ProxyLogging | None,
 ) -> LiteLLM_TeamMembership | None:
-    redis_cached: Final[object] = await user_api_key_cache.async_get_cache(key=cache_key)
+    redis_cached: Final = await _raw_cache(user_api_key_cache).async_get_cache(key=cache_key)
     redis_membership: Final = _membership_from_cached_payload(redis_cached)
     if not isinstance(redis_membership, _TeamMembershipCacheMiss):
         return redis_membership

litellm/proxy/management_endpoints/workflow_management_endpoints.py

@@ -147,6 +147,34 @@ class _MessageCreateData(TypedDict, total=False):
     session_id: str
 
 
+class _EventTable(Protocol):
+    async def create(self, *, data: _EventCreateData) -> object: ...
+
+
+class _RunTable(Protocol):
+    async def update(self, *, where: Mapping[str, str], data: Mapping[str, str]) -> object: ...
+
+
+class _WorkflowTx(Protocol):
+    @property
+    def litellm_workflowevent(self) -> _EventTable: ...
+
+    @property
+    def litellm_workflowrun(self) -> _RunTable: ...
+
+
+async def _append_event_in_tx(
+    tx: _WorkflowTx, run_id: str, event_data: _EventCreateData, new_status: str | None
+) -> object:
+    event: Final = await tx.litellm_workflowevent.create(data=event_data)
+    if new_status:
+        await tx.litellm_workflowrun.update(
+            where={"run_id": run_id},
+            data={"status": new_status},
+        )
+    return event
+
+
 # ---------------------------------------------------------------------------
 # Helpers
 # ---------------------------------------------------------------------------
@@ -387,12 +415,7 @@ async def append_workflow_event(
                 event_data["data"] = _json(data.data)
 
             async with prisma_client.db.tx() as tx:
-                event: object = await tx.litellm_workflowevent.create(data=event_data)
-                if new_status:
-                    await tx.litellm_workflowrun.update(
-                        where={"run_id": run_id},
-                        data={"status": new_status},
-                    )
+                event: object = await _append_event_in_tx(tx, run_id, event_data, new_status)
 
             return event
 

litellm/router_strategy/complexity_router/complexity_router.py

@@ -22,7 +22,7 @@ import json
 import random
 import re
 import time
-from collections.abc import Callable, Iterator, Mapping, Sequence
+from collections.abc import Awaitable, Callable, Iterator, Mapping, Sequence
 from itertools import accumulate, chain, islice, takewhile
 from threading import Lock
 from types import MappingProxyType
@@ -30,6 +30,7 @@ from typing import TYPE_CHECKING, Any, Final, Literal, NamedTuple, cast
 
 from pydantic import BaseModel, TypeAdapter, ValidationError, create_model
 from pydantic_core import ErrorDetails
+from typing_extensions import ReadOnly, TypedDict
 
 from litellm._internal_context import with_service_target
 from litellm._logging import verbose_router_logger
@@ -411,9 +412,19 @@ def _append_custom_keywords(base_keywords: list[str], custom_keywords: list[str]
     return [*base_keywords, *deduped_custom.values()]
 
 
-def _parent_session_kwargs(request_kwargs: Mapping[str, object] | None) -> Mapping[str, Any]:
+class _ParentSessionKwargs(TypedDict, total=False):
+    litellm_session_id: ReadOnly[object]
+    litellm_trace_id: ReadOnly[object]
+
+
+def _parent_session_kwargs(request_kwargs: Mapping[str, object] | None) -> _ParentSessionKwargs:
     kwargs: Final = request_kwargs or {}
-    return {k: kwargs[k] for k in ("litellm_session_id", "litellm_trace_id") if kwargs.get(k) is not None}
+    session_id: Final = kwargs.get("litellm_session_id")
+    trace_id: Final = kwargs.get("litellm_trace_id")
+    return _ParentSessionKwargs(
+        **({"litellm_session_id": session_id} if session_id is not None else {}),
+        **({"litellm_trace_id": trace_id} if trace_id is not None else {}),
+    )
 
 
 def _response_cost_or_none(response: object) -> float | None:
@@ -2828,7 +2839,7 @@ class ComplexityRouter(CustomLogger):
             "originating_request_masked": masked_originating_request(request_kwargs),
             "body": {"model": llm_config.model, **payload},
         }
-        classify: Final = (
+        classify: Final[Callable[..., Awaitable[ModelResponse | ResponsesAPIResponse]]] = (
             self.litellm_router_instance.aresponses
             if encrypted_task is not None
             else self.litellm_router_instance.acompletion