phoenix/litellm-mirror
0
0
Fork 1
mirror of https://github.com/BerriAI/litellm.git synced 2025-04-25 10:44:24 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix(utils.py): fix cached responses - translate dict to objects

Browse source
This commit is contained in:
Krrish Dholakia 2023-11-10 10:38:20 -08:00
parent 84460b8222
commit a4c9e6bd46
4 changed files with 108 additions and 21 deletions
Download patch file Download diff file Expand all files Collapse all files

27
litellm/utils.py
View file

@ -907,6 +907,29 @@ def client(original_function):
# [Non-Blocking Error]
pass
def convert_to_model_response_object(response_object: Optional[dict]=None, model_response_object: Optional[ModelResponse]=None):
try:
if response_object is None or model_response_object is None:
raise OpenAIError(status_code=500, message="Error in response object format")
choice_list=[]
for idx, choice in enumerate(response_object["choices"]):
message = Message(content=choice["message"]["content"], role=choice["message"]["role"])
choice = Choices(finish_reason=choice["finish_reason"], index=idx, message=message)
choice_list.append(choice)
model_response_object.choices = choice_list
if "usage" in response_object:
model_response_object.usage = response_object["usage"]
if "id" in response_object:
model_response_object.id = response_object["id"]
if "model" in response_object:
model_response_object.model = response_object["model"]
return model_response_object
except:
OpenAIError(status_code=500, message="Invalid response object.")
def wrapper(*args, **kwargs):
start_time = datetime.datetime.now()
result = None
@ -932,7 +955,7 @@ def client(original_function):
# [OPTIONAL] CHECK CACHE
# remove this after deprecating litellm.caching
print_verbose(f"litellm.caching: {litellm.caching}; litellm.caching_with_models: {litellm.caching_with_models}")
print_verbose(f"litellm.caching: {litellm.caching}; litellm.caching_with_models: {litellm.caching_with_models}; litellm.cache: {litellm.cache}")
if (litellm.caching or litellm.caching_with_models) and litellm.cache is None:
litellm.cache = Cache()
@ -945,7 +968,7 @@ def client(original_function):
cached_result = litellm.cache.get_cache(*args, **kwargs)
if cached_result != None:
print_verbose(f"Cache Hit!")
return cached_result
return convert_to_model_response_object(response_object=cached_result, model_response_object=ModelResponse())
# MODEL CALL
result = original_function(*args, **kwargs)