From fadc04fbe2790f417f474c9a15b4d75ebb9935b2 Mon Sep 17 00:00:00 2001
From: ryan-crabbe <128659760+ryan-crabbe@users.noreply.github.com>
Date: Mon, 2 Feb 2026 10:42:12 -0800
Subject: [PATCH] perf: optimize wrapper_async with CallTypes caching and
 reduced lookups (#20204)

- Cache CallTypes enum values as module-level dict to avoid repeated list
  comprehension and enum construction on every call
- Hoist update_response_metadata getattr lookup to top of function
- Guard verbose print_verbose call behind _is_debugging_on() check
---
 litellm/utils.py | 25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)

diff --git a/litellm/utils.py b/litellm/utils.py
index e67b967d75a..f3d14b455cd 100644
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -199,6 +199,8 @@ from litellm.types.utils import (
     all_litellm_params,
 )
 
+_CALL_TYPE_ENUM_MAP: dict = {ct.value: ct for ct in CallTypes}
+
 # +-----------------------------------------------+
 # |                                               |
 # |           Give Feedback / Get Help            |
@@ -1746,6 +1748,7 @@ def client(original_function):  # noqa: PLR0915
         print_args_passed_to_litellm(original_function, args, kwargs)
         start_time = datetime.datetime.now()
         result = None
+        _update_response_metadata = getattr(sys.modules[__name__], "update_response_metadata")
         logging_obj: Optional[LiteLLMLoggingObject] = kwargs.get(
             "litellm_logging_obj", None
         )
@@ -1786,9 +1789,10 @@ def client(original_function):  # noqa: PLR0915
                     )
 
             # [OPTIONAL] CHECK CACHE
-            print_verbose(
-                f"ASYNC kwargs[caching]: {kwargs.get('caching', False)}; litellm.cache: {litellm.cache}; kwargs.get('cache'): {kwargs.get('cache', None)}"
-            )
+            if _is_debugging_on():
+                print_verbose(
+                    f"ASYNC kwargs[caching]: {kwargs.get('caching', False)}; litellm.cache: {litellm.cache}; kwargs.get('cache'): {kwargs.get('cache', None)}"
+                )
             _caching_handler_response: "Optional[CachingHandlerResponse]" = (
                 await _llm_caching_handler._async_get_cache(
                     model=model or "",
@@ -1864,10 +1868,7 @@ def client(original_function):  # noqa: PLR0915
                         chunks, messages=kwargs.get("messages", None)
                     )
                 else:
-                    update_response_metadata = getattr(
-                        sys.modules[__name__], "update_response_metadata"
-                    )
-                    update_response_metadata(
+                    _update_response_metadata(
                         result=result,
                         logging_obj=logging_obj,
                         model=model,
@@ -1887,11 +1888,12 @@ def client(original_function):  # noqa: PLR0915
                 rules_obj=rules_obj,
             )
             # Only run if call_type is a valid value in CallTypes
-            if call_type in [ct.value for ct in CallTypes]:
+            _call_type_enum = _CALL_TYPE_ENUM_MAP.get(call_type)
+            if _call_type_enum is not None:
                 result = await async_post_call_success_deployment_hook(
                     request_data=kwargs,
                     response=result,
-                    call_type=CallTypes(call_type),
+                    call_type=_call_type_enum,
                 )
 
             ## Add response to cache
@@ -1931,10 +1933,7 @@ def client(original_function):  # noqa: PLR0915
                     end_time=end_time,
                 )
 
-            update_response_metadata = getattr(
-                sys.modules[__name__], "update_response_metadata"
-            )
-            update_response_metadata(
+            _update_response_metadata(
                 result=result,
                 logging_obj=logging_obj,
                 model=model,