phoenix/litellm
1
0
Fork 0
forked from phoenix/litellm-mirror
Code Pull requests Activity

fix(utils.py): ensure consistent cost calc b/w returned header and logged object

Browse source
This commit is contained in:
Krrish Dholakia 2024-08-20 18:03:35 -07:00
parent f51f7750c0
commit 8e9117f701
3 changed files with 20 additions and 26 deletions
Download patch file Download diff file Expand all files Collapse all files

28
litellm/utils.py
View file

@ -837,7 +837,7 @@ def client(original_function):
and kwargs.get("atranscription", False) != True
): # allow users to control returning cached responses from the completion function
# checking cache
print_verbose(f"INSIDE CHECKING CACHE")
print_verbose("INSIDE CHECKING CACHE")
if (
litellm.cache is not None
and str(original_function.__name__)
@ -965,10 +965,10 @@ def client(original_function):
# MODEL CALL
result = original_function(*args, **kwargs)
end_time = datetime.datetime.now()
if "stream" in kwargs and kwargs["stream"] == True:
if "stream" in kwargs and kwargs["stream"] is True:
if (
"complete_response" in kwargs
and kwargs["complete_response"] == True
and kwargs["complete_response"] is True
):
chunks = []
for idx, chunk in enumerate(result):
@ -978,15 +978,15 @@ def client(original_function):
)
else:
return result
elif "acompletion" in kwargs and kwargs["acompletion"] == True:
elif "acompletion" in kwargs and kwargs["acompletion"] is True:
return result
elif "aembedding" in kwargs and kwargs["aembedding"] == True:
elif "aembedding" in kwargs and kwargs["aembedding"] is True:
return result
elif "aimg_generation" in kwargs and kwargs["aimg_generation"] == True:
elif "aimg_generation" in kwargs and kwargs["aimg_generation"] is True:
return result
elif "atranscription" in kwargs and kwargs["atranscription"] == True:
elif "atranscription" in kwargs and kwargs["atranscription"] is True:
return result
elif "aspeech" in kwargs and kwargs["aspeech"] == True:
elif "aspeech" in kwargs and kwargs["aspeech"] is True:
return result
### POST-CALL RULES ###
@ -1005,7 +1005,7 @@ def client(original_function):
litellm.cache.add_cache(result, *args, **kwargs)
# LOG SUCCESS - handle streaming success logging in the _next_ object, remove `handle_success` once it's deprecated
verbose_logger.info(f"Wrapper: Completed Call, calling success_handler")
verbose_logger.info("Wrapper: Completed Call, calling success_handler")
threading.Thread(
target=logging_obj.success_handler, args=(result, start_time, end_time)
).start()
@ -1019,15 +1019,7 @@ def client(original_function):
optional_params=getattr(logging_obj, "optional_params", {}),
)
result._hidden_params["response_cost"] = (
litellm.response_cost_calculator(
response_object=result,
model=getattr(logging_obj, "model", ""),
custom_llm_provider=getattr(
logging_obj, "custom_llm_provider", None
),
call_type=getattr(logging_obj, "call_type", "completion"),
optional_params=getattr(logging_obj, "optional_params", {}),
)
logging_obj._response_cost_calculator(result=result)
)
result._response_ms = (
end_time - start_time