From e87ae809a66852550a6639048952e0421eaebadb Mon Sep 17 00:00:00 2001
From: Maximilian Rink <maximilian.rink@telekom.de>
Date: Sat, 10 Oct 2026 18:56:23 +0200
Subject: [PATCH] fix(cache): keep typed cache helper within strict gates

---
 litellm/caching/caching_handler.py            | 20 ++++++++++++++-----
 .../vertex_ai_context_caching.py              |  6 +++++-
 2 files changed, 20 insertions(+), 6 deletions(-)

diff --git a/litellm/caching/caching_handler.py b/litellm/caching/caching_handler.py
index b1667921b22..2f3484ae12b 100644
--- a/litellm/caching/caching_handler.py
+++ b/litellm/caching/caching_handler.py
@@ -19,7 +19,14 @@ import datetime
 import inspect
 import time
 from collections.abc import AsyncGenerator, AsyncIterator, Awaitable, Callable, Generator, Mapping
-from typing import TYPE_CHECKING, Any, Final, Optional, TypeVar, cast
+from typing import (
+    TYPE_CHECKING,
+    Any,
+    Final,
+    Optional,
+    TypeVar,
+    cast,  # noqa: TID251  # typed boundary for dynamic cache request kwargs
+)
 
 from pydantic import ConfigDict, SkipValidation, TypeAdapter, ValidationError
 
@@ -106,6 +113,11 @@ def _drop_logging_obj_from_kwargs(request_kwargs: dict[str, object]) -> dict[str
     return {k: v for k, v in request_kwargs.items() if k != "litellm_logging_obj"}
 
 
+def _set_cache_key_if_available(request_kwargs: dict[str, object], cache_key: str | None) -> None:
+    if cache_key is not None:
+        request_kwargs["cache_key"] = cache_key
+
+
 def _is_response_cache_excluded(model: str | None, kwargs: Mapping[str, object]) -> bool:
     from litellm.llms.github_copilot.per_user_auth import is_github_copilot_per_user_request
 
@@ -465,8 +477,7 @@ class LLMCachingHandler:
                 new_kwargs.pop("metadata", None)
             if new_kwargs.get("stream") is True and "cache_key" not in new_kwargs:
                 derived_cache_key = litellm.cache.get_cache_key(**cast(dict[str, object], new_kwargs))
-                if derived_cache_key is not None:
-                    new_kwargs["cache_key"] = derived_cache_key
+                _set_cache_key_if_available(new_kwargs, derived_cache_key)
             self.request_kwargs = _drop_logging_obj_from_kwargs(new_kwargs)
             print_verbose("Checking Sync Cache")
             logging_obj.reset_caching_details()
@@ -888,8 +899,7 @@ class LLMCachingHandler:
             return None
         if new_kwargs.get("stream") is True and "cache_key" not in new_kwargs:
             derived_cache_key = litellm.cache.get_cache_key(**new_kwargs)
-            if derived_cache_key is not None:
-                new_kwargs["cache_key"] = derived_cache_key
+            _set_cache_key_if_available(new_kwargs, derived_cache_key)
         cached_result: object | None = None
         if call_type == CallTypes.aembedding.value:
             new_kwargs["input"] = self.handle_kwargs_input_list_or_str(new_kwargs)
diff --git a/litellm/llms/vertex_ai/context_caching/vertex_ai_context_caching.py b/litellm/llms/vertex_ai/context_caching/vertex_ai_context_caching.py
index 26ca1be77a2..f39194b72a0 100644
--- a/litellm/llms/vertex_ai/context_caching/vertex_ai_context_caching.py
+++ b/litellm/llms/vertex_ai/context_caching/vertex_ai_context_caching.py
@@ -1,5 +1,9 @@
 from collections.abc import Sequence
-from typing import Final, Literal, cast
+from typing import (
+    Final,
+    Literal,
+    cast,  # noqa: TID251  # preserve the runtime-compatible auth header union
+)
 
 import httpx
 from pydantic import ConfigDict, RootModel, with_config