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

bug fixes and updates

Browse source
This commit is contained in:
Krrish Dholakia 2023-08-02 13:27:10 -07:00
parent efe82c6bb6
commit 85f1591413
17 changed files with 646 additions and 330 deletions
Download patch file Download diff file Expand all files Collapse all files

79
litellm/utils.py
View file

@ -234,41 +234,45 @@ def handle_success(*args, **kwargs):
def exception_type(model, original_exception):
if isinstance(original_exception, OpenAIError):
# Handle the OpenAIError
raise original_exception
elif model:
error_str = str(original_exception)
if isinstance(original_exception, BaseException):
exception_type = type(original_exception).__name__
try:
if isinstance(original_exception, OpenAIError):
# Handle the OpenAIError
raise original_exception
elif model:
error_str = str(original_exception)
if isinstance(original_exception, BaseException):
exception_type = type(original_exception).__name__
else:
exception_type = ""
if "claude" in model: #one of the anthropics
if "status_code" in original_exception:
print_verbose(f"status_code: {original_exception.status_code}")
if original_exception.status_code == 401:
raise AuthenticationError(f"AnthropicException - {original_exception.message}")
elif original_exception.status_code == 400:
raise InvalidRequestError(f"AnthropicException - {original_exception.message}", f"{model}")
elif original_exception.status_code == 429:
raise RateLimitError(f"AnthropicException - {original_exception.message}")
elif "replicate" in model:
if "Incorrect authentication token" in error_str:
raise AuthenticationError(f"ReplicateException - {error_str}")
elif exception_type == "ModelError":
raise InvalidRequestError(f"ReplicateException - {error_str}", f"{model}")
elif "Request was throttled" in error_str:
raise RateLimitError(f"ReplicateException - {error_str}")
elif exception_type == "ReplicateError": ## ReplicateError implies an error on Replicate server side, not user side
raise ServiceUnavailableError(f"ReplicateException - {error_str}")
elif model == "command-nightly": #Cohere
if "invalid api token" in error_str or "No API key provided." in error_str:
raise AuthenticationError(f"CohereException - {error_str}")
elif "too many tokens" in error_str:
raise InvalidRequestError(f"CohereException - {error_str}", f"{model}")
elif "CohereConnectionError" in exception_type: # cohere seems to fire these errors when we load test it (1k+ messages / min)
raise RateLimitError(f"CohereException - {original_exception.message}")
raise original_exception # base case - return the original exception
else:
exception_type = ""
if "claude" in model: #one of the anthropics
print_verbose(f"status_code: {original_exception.status_code}")
if original_exception.status_code == 401:
raise AuthenticationError(f"AnthropicException - {original_exception.message}")
elif original_exception.status_code == 400:
raise InvalidRequestError(f"AnthropicException - {original_exception.message}", f"{model}")
elif original_exception.status_code == 429:
raise RateLimitError(f"AnthropicException - {original_exception.message}")
elif "replicate" in model:
if "Incorrect authentication token" in error_str:
raise AuthenticationError(f"ReplicateException - {error_str}")
elif exception_type == "ModelError":
raise InvalidRequestError(f"ReplicateException - {error_str}", f"{model}")
elif "Request was throttled" in error_str:
raise RateLimitError(f"ReplicateException - {error_str}")
elif exception_type == "ReplicateError": ## ReplicateError implies an error on Replicate server side, not user side
raise ServiceUnavailableError(f"ReplicateException - {error_str}")
elif model == "command-nightly": #Cohere
if "invalid api token" in error_str or "No API key provided." in error_str:
raise AuthenticationError(f"CohereException - {error_str}")
elif "too many tokens" in error_str:
raise InvalidRequestError(f"CohereException - {error_str}", f"{model}")
elif "CohereConnectionError" in exception_type: # cohere seems to fire these errors when we load test it (1k+ messages / min)
raise RateLimitError(f"CohereException - {original_exception.message}")
raise original_exception # base case - return the original exception
else:
raise original_exception
except:
raise original_exception
def safe_crash_reporting(model=None, exception=None, azure=None):
@ -277,11 +281,9 @@ def safe_crash_reporting(model=None, exception=None, azure=None):
"exception": str(exception),
"azure": azure
}
print(f"data in crash reporting: {data}")
threading.Thread(target=litellm_telemetry, args=(data,), daemon=True).start()
def litellm_telemetry(data):
print(f"data in in litellm telemetry: {data}")
# Load or generate the UUID
uuid_file = 'litellm_uuid.txt'
try:
@ -290,7 +292,6 @@ def litellm_telemetry(data):
uuid_value = file.read()
if uuid_value:
uuid_value = uuid_value.strip()
print(f"Loaded UUID: {uuid_value}")
else:
raise FileNotFoundError
except FileNotFoundError:
@ -299,7 +300,6 @@ def litellm_telemetry(data):
uuid_value = str(new_uuid)
with open(uuid_file, 'w') as file:
file.write(uuid_value)
print(f"Generated and stored UUID: {uuid_value}")
# Prepare the data to send to localhost:3000
payload = {
@ -311,7 +311,6 @@ def litellm_telemetry(data):
# Make the POST request to localhost:3000
response = requests.post('https://litellm.berri.ai/logging', json=payload)
response.raise_for_status() # Raise an exception for HTTP errors
print('Request successfully sent!')
except requests.exceptions.RequestException as e:
# Handle any errors in the request
print(f'Error: {e}')
pass