From f5cd76b047cfb64aa2e0e0250423305c09a79abd Mon Sep 17 00:00:00 2001
From: nuernber <benjamin.nuernberger@jpl.nasa.gov>
Date: Fri, 11 Sep 2026 14:04:01 -0700
Subject: [PATCH] fix(prompt-caching): suppress LIT lint violations with proper
 reason comments

  Swap ineffective `# noqa: LIT002` comments for `# mutable-ok` (the rule LIT002
  actually checks for) and add `# cast-ok` reasons on the casts in the deployment
  check, clearing the type-discipline gate.
---
 .../prompt_caching_deployment_check.py        | 10 ++++++---
 litellm/router_utils/prompt_caching_cache.py  | 22 +++++++++++++------
 2 files changed, 22 insertions(+), 10 deletions(-)

diff --git a/litellm/router_utils/pre_call_checks/prompt_caching_deployment_check.py b/litellm/router_utils/pre_call_checks/prompt_caching_deployment_check.py
index 514fbd18e8c..8892b4efdb3 100644
--- a/litellm/router_utils/pre_call_checks/prompt_caching_deployment_check.py
+++ b/litellm/router_utils/pre_call_checks/prompt_caching_deployment_check.py
@@ -102,7 +102,9 @@ class PromptCachingDeploymentCheck(CustomLogger):
             model_id_dict: Final = await prompt_cache.async_get_model_id(
                 messages=affinity_messages,
                 tools=(
-                    cast(list[AllToolParamValues] | None, request_kwargs.get("tools"))
+                    cast(  # cast-ok: request kwargs are untyped
+                        list[AllToolParamValues] | None, request_kwargs.get("tools")
+                    )  # cast-ok: request kwargs are untyped
                     if request_kwargs is not None
                     else None
                 ),
@@ -149,7 +151,7 @@ class PromptCachingDeploymentCheck(CustomLogger):
             return
 
         logged_messages: Final = PromptCachingCache.prepend_system_prompt(
-            cast(list[AllMessageValues], messages),
+            cast(list[AllMessageValues], messages),  # cast-ok: standard logging payload is partially typed
             kwargs.get("system"),
         )
 
@@ -164,7 +166,9 @@ class PromptCachingDeploymentCheck(CustomLogger):
             await cache.async_add_model_id(
                 model_id=model_id,
                 messages=logged_messages,
-                tools=cast(list[AllToolParamValues] | None, kwargs.get("tools")),
+                tools=cast(  # cast-ok: callback kwargs are untyped
+                    list[AllToolParamValues] | None, kwargs.get("tools")
+                ),
             )
 
         return
diff --git a/litellm/router_utils/prompt_caching_cache.py b/litellm/router_utils/prompt_caching_cache.py
index 5ef5f050deb..ef55d4e2735 100644
--- a/litellm/router_utils/prompt_caching_cache.py
+++ b/litellm/router_utils/prompt_caching_cache.py
@@ -153,7 +153,7 @@ class PromptCachingCache:
             None,
         )
         # Match the provider prefix exactly instead of pinning on uncached trailing tools
-        return tools[: cacheable_tool_index + 1] if cacheable_tool_index is not None else []
+        return tools[: cacheable_tool_index + 1] if cacheable_tool_index is not None else tools[:0]
 
     @staticmethod
     def prepend_system_prompt(
@@ -162,9 +162,9 @@ class PromptCachingCache:
     ) -> list[AllMessageValues]:
         if system is None:
             return messages
-        return cast(
+        return cast(  # cast-ok: system content is validated by the provider payload
             list[AllMessageValues],
-            [{"role": "system", "content": system}, *messages],
+            [{"role": "system", "content": system}, *messages],  # mutable-ok: cast target requires a concrete list
         )
 
     @staticmethod
@@ -172,7 +172,9 @@ class PromptCachingCache:
         messages: list[AllMessageValues] | None,
         tools: list[ChatCompletionToolParam] | None = None,
     ) -> int:
-        cacheable_prefix: Final = PromptCachingCache.extract_cacheable_prefix(messages) if messages is not None else []
+        cacheable_prefix: Final = (
+            PromptCachingCache.extract_cacheable_prefix(messages) if messages is not None else []  # mutable-ok: TTL helper requires a concrete list
+        )
         return PromptCachingCache.get_prompt_caching_ttl_from_prefix(cacheable_prefix, tools)
 
     @staticmethod
@@ -180,7 +182,7 @@ class PromptCachingCache:
         cacheable_prefix: list[AllMessageValues],
         tools: list[ChatCompletionToolParam] | None,
     ) -> int:
-        cacheable_tools: Final = PromptCachingCache.extract_cacheable_tools(tools or [])
+        cacheable_tools: Final = PromptCachingCache.extract_cacheable_tools(tools or [])  # mutable-ok: tool API requires a concrete list
         cache_control_values: Final = tuple(
             cache_control
             for message in cacheable_prefix
@@ -259,7 +261,10 @@ class PromptCachingCache:
         self.cache.set_cache(
             cache_key,
             PromptCachingCacheValue(model_id=model_id),
-            ttl=PromptCachingCache.get_prompt_caching_ttl_from_prefix(cacheable_prefix or [], tools),
+            ttl=PromptCachingCache.get_prompt_caching_ttl_from_prefix(
+                cacheable_prefix or [],  # mutable-ok: TTL helper requires a concrete list
+                tools,
+            ),
         )
         return
 
@@ -282,7 +287,10 @@ class PromptCachingCache:
         await self.cache.async_set_cache(
             cache_key,
             PromptCachingCacheValue(model_id=model_id),
-            ttl=PromptCachingCache.get_prompt_caching_ttl_from_prefix(cacheable_prefix or [], tools),
+            ttl=PromptCachingCache.get_prompt_caching_ttl_from_prefix(
+                cacheable_prefix or [],  # mutable-ok: TTL helper requires a concrete list
+                tools,
+            ),
         )
         return