Fix team-based logging to langfuse + allow custom tokenizer on /token_counter endpoint (#7493)

* fix(langfuse_prompt_management.py): migrate dynamic logging to langfuse custom logger compatible class

* fix(langfuse_prompt_management.py): support failure callback logging to langfuse as well

* feat(proxy_server.py): support setting custom tokenizer on config.yaml

Allows customizing value for `/utils/token_counter`

* fix(proxy_server.py): fix linting errors

* test: skip if file not found

* style: cleanup unused import

* docs(configs.md): add docs on setting custom tokenizer

litellm/integrations/langfuse/langfuse_prompt_management.py

@@ -16,7 +16,11 @@ from litellm.proxy._types import UserAPIKeyAuth
 from litellm.types.llms.openai import AllMessageValues
 from litellm.types.utils import StandardCallbackDynamicParams, StandardLoggingPayload
 
+from ...litellm_core_utils.specialty_caches.dynamic_logging_cache import (
+    DynamicLoggingCache,
+)
 from .langfuse import LangFuseLogger
+from .langfuse_handler import LangFuseHandler
 
 if TYPE_CHECKING:
     from langfuse import Langfuse
@@ -29,6 +33,8 @@ else:
     PROMPT_CLIENT = Any
     LangfuseClass = Any
 
+in_memory_dynamic_logger_cache = DynamicLoggingCache()
+
 
 @lru_cache(maxsize=10)
 def langfuse_client_init(
@@ -252,7 +258,15 @@ class LangfusePromptManagement(LangFuseLogger, CustomLogger):
         return model, messages, non_default_params
 
     async def async_log_success_event(self, kwargs, response_obj, start_time, end_time):
-        self._old_log_event(
+        standard_callback_dynamic_params = kwargs.get(
+            "standard_callback_dynamic_params"
+        )
+        langfuse_logger_to_use = LangFuseHandler.get_langfuse_logger_for_request(
+            globalLangfuseLogger=self,
+            standard_callback_dynamic_params=standard_callback_dynamic_params,
+            in_memory_dynamic_logger_cache=in_memory_dynamic_logger_cache,
+        )
+        langfuse_logger_to_use._old_log_event(
             kwargs=kwargs,
             response_obj=response_obj,
             start_time=start_time,
@@ -262,13 +276,21 @@ class LangfusePromptManagement(LangFuseLogger, CustomLogger):
         )
 
     async def async_log_failure_event(self, kwargs, response_obj, start_time, end_time):
+        standard_callback_dynamic_params = kwargs.get(
+            "standard_callback_dynamic_params"
+        )
+        langfuse_logger_to_use = LangFuseHandler.get_langfuse_logger_for_request(
+            globalLangfuseLogger=self,
+            standard_callback_dynamic_params=standard_callback_dynamic_params,
+            in_memory_dynamic_logger_cache=in_memory_dynamic_logger_cache,
+        )
         standard_logging_object = cast(
             Optional[StandardLoggingPayload],
             kwargs.get("standard_logging_object", None),
         )
         if standard_logging_object is None:
             return
-        self._old_log_event(
+        langfuse_logger_to_use._old_log_event(
             start_time=start_time,
             end_time=end_time,
             response_obj=None,