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

updates to exception mapping

Browse source
This commit is contained in:
Krrish Dholakia 2023-09-06 10:36:19 -07:00
parent d0b16892e0
commit ef43141554
2 changed files with 121 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

32
litellm/exceptions.py
View file

@ -5,6 +5,9 @@ from openai.error import (
RateLimitError, RateLimitError,
ServiceUnavailableError, ServiceUnavailableError,
OpenAIError, OpenAIError,
APIError,
Timeout,
APIConnectionError,
) )
@ -29,6 +32,16 @@ class InvalidRequestError(InvalidRequestError): # type: ignore
self.message, f"{self.model}" self.message, f"{self.model}"
) # Call the base class constructor with the parameters it needs ) # Call the base class constructor with the parameters it needs
class Timeout(Timeout): # type: ignore
def __init__(self, message, model, llm_provider):
self.status_code = 408
self.message = message
self.model = model
self.llm_provider = llm_provider
super().__init__(
self.message, f"{self.model}"
) # Call the base class constructor with the parameters it needs
# sub class of invalid request error - meant to give more granularity for error handling context window exceeded errors # sub class of invalid request error - meant to give more granularity for error handling context window exceeded errors
class ContextWindowExceededError(InvalidRequestError): # type: ignore class ContextWindowExceededError(InvalidRequestError): # type: ignore
def __init__(self, message, model, llm_provider): def __init__(self, message, model, llm_provider):
@ -63,6 +76,25 @@ class ServiceUnavailableError(ServiceUnavailableError): # type: ignore
) # Call the base class constructor with the parameters it needs ) # Call the base class constructor with the parameters it needs
class APIError(APIError): # raise this when the API returns an invalid response object - https://github.com/openai/openai-python/blob/1be14ee34a0f8e42d3f9aa5451aa4cb161f1781f/openai/api_requestor.py#L401
def __init__(self, status_code, message, llm_provider, model):
self.status_code = status_code
self.message = message
self.llm_provider = llm_provider
self.model = model
super().__init__(
self.message
)
class APIConnectionError(APIConnectionError): # raised if an invalid request (not get, delete, put, post) is made
def __init__(self, message, llm_provider, model):
self.message = message
self.llm_provider = llm_provider
self.model = model
super().__init__(
self.message
)
class OpenAIError(OpenAIError): # type: ignore class OpenAIError(OpenAIError): # type: ignore
def __init__(self, original_exception): def __init__(self, original_exception):
self.status_code = original_exception.http_status self.status_code = original_exception.http_status

90
litellm/utils.py
View file

@ -26,7 +26,10 @@ from .exceptions import (
RateLimitError, RateLimitError,
ServiceUnavailableError, ServiceUnavailableError,
OpenAIError, OpenAIError,
ContextWindowExceededError ContextWindowExceededError,
Timeout,
APIConnectionError,
APIError
) )
from typing import List, Dict, Union, Optional from typing import List, Dict, Union, Optional
from .caching import Cache from .caching import Cache
@ -1495,6 +1498,13 @@ def exception_type(model, original_exception, custom_llm_provider):
model=model, model=model,
llm_provider="anthropic", llm_provider="anthropic",
) )
elif original_exception.status_code == 408:
exception_mapping_worked = True
raise Timeout(
message=f"AnthropicException - {original_exception.message}",
model=model,
llm_provider="anthropic"
)
elif original_exception.status_code == 413: elif original_exception.status_code == 413:
exception_mapping_worked = True exception_mapping_worked = True
raise InvalidRequestError( raise InvalidRequestError(
@ -1509,12 +1519,17 @@ def exception_type(model, original_exception, custom_llm_provider):
llm_provider="anthropic", llm_provider="anthropic",
model=model model=model
) )
elif ( elif original_exception.status_code == 500:
"Could not resolve authentication method. Expected either api_key or auth_token to be set."
in error_str
):
exception_mapping_worked = True exception_mapping_worked = True
raise AuthenticationError( raise ServiceUnavailableError(
message=f"AnthropicException - {original_exception.message}",
llm_provider="anthropic",
model=model
)
else:
exception_mapping_worked = True
raise APIError(
status_code=original_exception.status_code,
message=f"AnthropicException - {original_exception.message}", message=f"AnthropicException - {original_exception.message}",
llm_provider="anthropic", llm_provider="anthropic",
model=model model=model
@ -1551,11 +1566,20 @@ def exception_type(model, original_exception, custom_llm_provider):
elif ( elif (
exception_type == "ReplicateError" exception_type == "ReplicateError"
): # ReplicateError implies an error on Replicate server side, not user side ): # ReplicateError implies an error on Replicate server side, not user side
exception_mapping_worked = True
raise ServiceUnavailableError( raise ServiceUnavailableError(
message=f"ReplicateException - {error_str}", message=f"ReplicateException - {error_str}",
llm_provider="replicate", llm_provider="replicate",
model=model model=model
) )
else:
exception_mapping_worked = True
raise APIError(
status_code=original_exception.status_code,
message=f"ReplicateException - {original_exception.message}",
llm_provider="replicate",
model=model
)
elif model in litellm.cohere_models: # Cohere elif model in litellm.cohere_models: # Cohere
if ( if (
"invalid api token" in error_str "invalid api token" in error_str
@ -1583,6 +1607,14 @@ def exception_type(model, original_exception, custom_llm_provider):
llm_provider="cohere", llm_provider="cohere",
model=model model=model
) )
else:
exception_mapping_worked = True
raise APIError(
status_code=original_exception.status_code,
message=f"CohereException - {original_exception.message}",
llm_provider="cohere",
model=model
)
elif custom_llm_provider == "huggingface": elif custom_llm_provider == "huggingface":
if "length limit exceeded" in error_str: if "length limit exceeded" in error_str:
exception_mapping_worked = True exception_mapping_worked = True
@ -1606,6 +1638,13 @@ def exception_type(model, original_exception, custom_llm_provider):
model=model, model=model,
llm_provider="huggingface", llm_provider="huggingface",
) )
elif original_exception.status_code == 408:
exception_mapping_worked = True
raise Timeout(
message=f"HuggingfaceException - {original_exception.message}",
model=model,
llm_provider="huggingface"
)
elif original_exception.status_code == 429: elif original_exception.status_code == 429:
exception_mapping_worked = True exception_mapping_worked = True
raise RateLimitError( raise RateLimitError(
@ -1613,6 +1652,14 @@ def exception_type(model, original_exception, custom_llm_provider):
llm_provider="huggingface", llm_provider="huggingface",
model=model model=model
) )
else:
exception_mapping_worked = True
raise APIError(
status_code=original_exception.status_code,
message=f"HuggingfaceException - {original_exception.message}",
llm_provider="huggingface",
model=model
)
elif custom_llm_provider == "ai21": elif custom_llm_provider == "ai21":
if hasattr(original_exception, "message"): if hasattr(original_exception, "message"):
if "Prompt has too many tokens" in original_exception.message: if "Prompt has too many tokens" in original_exception.message:
@ -1630,6 +1677,13 @@ def exception_type(model, original_exception, custom_llm_provider):
llm_provider="ai21", llm_provider="ai21",
model=model model=model
) )
elif original_exception.status_code == 408:
exception_mapping_worked = True
raise Timeout(
message=f"AI21Exception - {original_exception.message}",
model=model,
llm_provider="ai21"
)
if original_exception.status_code == 422: if original_exception.status_code == 422:
exception_mapping_worked = True exception_mapping_worked = True
raise InvalidRequestError( raise InvalidRequestError(
@ -1643,6 +1697,14 @@ def exception_type(model, original_exception, custom_llm_provider):
message=f"AI21Exception - {original_exception.message}", message=f"AI21Exception - {original_exception.message}",
llm_provider="ai21", llm_provider="ai21",
) )
else:
exception_mapping_worked = True
raise APIError(
status_code=original_exception.status_code,
message=f"AI21Exception - {original_exception.message}",
llm_provider="ai21",
model=model
)
elif custom_llm_provider == "together_ai": elif custom_llm_provider == "together_ai":
error_response = json.loads(error_str) error_response = json.loads(error_str)
if "error" in error_response and "`inputs` tokens + `max_new_tokens` must be <=" in error_response["error"]: if "error" in error_response and "`inputs` tokens + `max_new_tokens` must be <=" in error_response["error"]:
@ -1673,6 +1735,13 @@ def exception_type(model, original_exception, custom_llm_provider):
model=model, model=model,
llm_provider="together_ai" llm_provider="together_ai"
) )
elif original_exception.status_code == 408:
exception_mapping_worked = True
raise Timeout(
message=f"TogetherAIException - {original_exception.message}",
model=model,
llm_provider="together_ai"
)
elif original_exception.status_code == 429: elif original_exception.status_code == 429:
exception_mapping_worked = True exception_mapping_worked = True
raise RateLimitError( raise RateLimitError(
@ -1680,7 +1749,14 @@ def exception_type(model, original_exception, custom_llm_provider):
llm_provider="together_ai", llm_provider="together_ai",
model=model model=model
) )
raise original_exception # base case - return the original exception else:
exception_mapping_worked = True
raise APIError(
status_code=original_exception.status_code,
message=f"TogetherAIException - {original_exception.message}",
llm_provider="together_ai",
model=model
)
else: else:
raise original_exception raise original_exception
except Exception as e: except Exception as e: