例外對應
LiteLLM 會將所有提供者的例外對應至其 OpenAI 對應項目。
所有例外都可以從 litellm 匯入 - 例如 from litellm import BadRequestError
LiteLLM 例外
| 狀態碼 | 錯誤類型 | 繼承自 | 說明 |
|---|---|---|---|
| 400 | BadRequestError | openai.BadRequestError | |
| 400 | UnsupportedParamsError | litellm.BadRequestError | 當傳入不支援的參數時引發 |
| 400 | ContextWindowExceededError | litellm.BadRequestError | 用於上下文視窗超出錯誤訊息的特殊錯誤類型 - 可啟用上下文視窗備援 |
| 400 | ContentPolicyViolationError | litellm.BadRequestError | 用於內容政策違規錯誤訊息的特殊錯誤類型 - 可啟用內容政策備援 |
| 400 | ImageFetchError | litellm.BadRequestError | 當擷取或處理圖片時發生錯誤時引發 |
| 400 | InvalidRequestError | openai.BadRequestError | 已淘汰的錯誤，請改用 BadRequestError |
| 401 | AuthenticationError | openai.AuthenticationError | |
| 403 | PermissionDeniedError | openai.PermissionDeniedError | |
| 404 | NotFoundError | openai.NotFoundError | 在傳入無效模型時引發，例如 gpt-8 |
| 408 | Timeout | openai.APITimeoutError | 發生逾時時引發 |
| 422 | UnprocessableEntityError | openai.UnprocessableEntityError | |
| 429 | RateLimitError | openai.RateLimitError | |
| 500 | APIConnectionError | openai.APIConnectionError | 如果傳回任何未對應的錯誤，我們會回傳此錯誤 |
| 500 | APIError | openai.APIError | 一般 500 狀態碼錯誤 |
| 503 | ServiceUnavailableError | openai.APIStatusError | 如果提供者回傳服務無法使用錯誤，則會引發此錯誤 |
| >=500 | InternalServerError | openai.InternalServerError | 如果回傳任何未對應的 500 狀態碼錯誤，則會引發此錯誤 |
| N/A | APIResponseValidationError | openai.APIResponseValidationError | 如果使用 Rules，且請求/回應未通過某項規則，則會引發此錯誤 |
| N/A | BudgetExceededError | Exception | 在代理時，當超出預算時引發 |
| N/A | JSONSchemaValidationError | litellm.APIResponseValidationError | 當回應不符合預期的 json schema 時引發 - 若 response_schema 參數與 enforce_validation=True 一起傳入則會使用 |
| N/A | MockException | Exception | 內部例外，由 mock_completion 類別引發。請勿直接使用 |
| N/A | OpenAIError | openai.OpenAIError | 已淘汰的內部例外，繼承自 openai.OpenAIError。 |
基本情況下我們會回傳 APIConnectionError
我們所有的例外都繼承自 OpenAI 的例外類型，因此您已經為該類型撰寫的任何錯誤處理，都能在 LiteLLM 上直接運作。
在所有情況下，回傳的例外都繼承自原始的 OpenAI Exception，但會額外包含 3 個屬性：
- status_code - 例外的 http 狀態碼
- message - 錯誤訊息
- llm_provider - 引發例外的提供者
使用方式
import litellm
import openai
try:
response = litellm.completion(
model="gpt-4",
messages=[
{
"role": "user",
"content": "hello, write a 20 pageg essay"
}
],
timeout=0.01, # this will raise a timeout exception
)
except openai.APITimeoutError as e:
print("Passed: Raised correct exception. Got openai.APITimeoutError\nGood Job", e)
print(type(e))
pass
使用方式 - 擷取串流例外
import litellm
try:
response = litellm.completion(
model="gpt-3.5-turbo",
messages=[
{
"role": "user",
"content": "hello, write a 20 pg essay"
}
],
timeout=0.0001, # this will raise an exception
stream=True,
)
for chunk in response:
print(chunk)
except openai.APITimeoutError as e:
print("Passed: Raised correct exception. Got openai.APITimeoutError\nGood Job", e)
print(type(e))
pass
except Exception as e:
print(f"Did not raise error `openai.APITimeoutError`. Instead raised error type: {type(e)}, Error: {e}")
使用方式 - 是否應重試例外？
import litellm
import openai
try:
response = litellm.completion(
model="gpt-4",
messages=[
{
"role": "user",
"content": "hello, write a 20 pageg essay"
}
],
timeout=0.01, # this will raise a timeout exception
)
except openai.APITimeoutError as e:
should_retry = litellm._should_retry(e.status_code)
print(f"should_retry: {should_retry}")
進階
存取提供者專屬錯誤詳細資訊
LiteLLM 例外包含一個 provider_specific_fields 屬性，其中包含各提供者專屬的額外錯誤資訊。這對 Azure OpenAI 特別有用，因為它會提供詳細的內容篩選資訊。
Azure OpenAI - 內容政策違規內部錯誤存取
當 Azure OpenAI 傳回內容政策違規時，您可以透過 innererror 欄位存取詳細的內容篩選結果：
import litellm
from litellm.exceptions import ContentPolicyViolationError
try:
response = litellm.completion(
model="azure/gpt-4",
messages=[
{
"role": "user",
"content": "Some content that might violate policies"
}
]
)
except ContentPolicyViolationError as e:
# Access Azure-specific error details
if e.provider_specific_fields and "innererror" in e.provider_specific_fields:
innererror = e.provider_specific_fields["innererror"]
# Access content filter results
content_filter_result = innererror.get("content_filter_result", {})
print(f"Content filter code: {innererror.get('code')}")
print(f"Hate filtered: {content_filter_result.get('hate', {}).get('filtered')}")
print(f"Violence severity: {content_filter_result.get('violence', {}).get('severity')}")
print(f"Sexual content filtered: {content_filter_result.get('sexual', {}).get('filtered')}")
範例回應結構：
當呼叫 LiteLLM 代理時，內容政策違規會傳回詳細的篩選資訊：
{
"error": {
"message": "litellm.ContentPolicyViolationError: AzureException - The response was filtered due to the prompt triggering Azure OpenAI's content management policy...",
"type": null,
"param": null,
"code": "400",
"provider_specific_fields": {
"innererror": {
"code": "ResponsibleAIPolicyViolation",
"content_filter_result": {
"hate": {
"filtered": true,
"severity": "high"
},
"jailbreak": {
"filtered": false,
"detected": false
},
"self_harm": {
"filtered": false,
"severity": "safe"
},
"sexual": {
"filtered": false,
"severity": "safe"
},
"violence": {
"filtered": true,
"severity": "medium"
}
}
}
}
}
}
## 詳細資訊 \{#details}
若要了解其實作方式 - [請查看程式碼](https://github.com/BerriAI/litellm/blob/a42c197e5a6de56ea576c73715e6c7c6b19fa249/litellm/utils.py#L1217)
如果您想改進例外對應，請 [建立 issue](https://github.com/BerriAI/litellm/issues/new) **或** [提出 PR](https://github.com/BerriAI/litellm/pulls)。
**注意** 對於 OpenAI 和 Azure，我們會回傳原始例外（因為它們屬於 OpenAI Error 類型）。但我們會為它們加上 'llm_provider' 屬性。[查看程式碼](https://github.com/BerriAI/litellm/blob/a42c197e5a6de56ea576c73715e6c7c6b19fa249/litellm/utils.py#L1221)
## 自訂對應清單 \{#custom-mapping-list}
基本情況下 - 我們會回傳 `litellm.APIConnectionError` 例外（繼承自 openai 的 APIConnectionError 例外）。
| custom_llm_provider | 逾時 | ContextWindowExceededError | BadRequestError | NotFoundError | ContentPolicyViolationError | AuthenticationError | APIError | RateLimitError | ServiceUnavailableError | PermissionDeniedError | UnprocessableEntityError |
|----------------------------|---------|----------------------------|------------------|---------------|-----------------------------|---------------------|----------|----------------|-------------------------|-----------------------|-------------------------|
| openai | ✓ | ✓ | ✓ | | ✓ | ✓ | | | | | |
| watsonx | | | | | | | |✓| | | |
| text-completion-openai | ✓ | ✓ | ✓ | | ✓ | ✓ | | | | | |
| custom_openai | ✓ | ✓ | ✓ | | ✓ | ✓ | | | | | |
| openai_compatible_providers| ✓ | ✓ | ✓ | | ✓ | ✓ | | | | | |
| anthropic | ✓ | ✓ | ✓ | ✓ | | ✓ | | | ✓ | ✓ | |
| replicate | ✓ | ✓ | ✓ | ✓ | | ✓ | | ✓ | ✓ | | |
| bedrock | ✓ | ✓ | ✓ | ✓ | | ✓ | | ✓ | ✓ | ✓ | |
| sagemaker | | ✓ | ✓ | | | | | | | | |
| vertex_ai | ✓ | | ✓ | | | | ✓ | | | | ✓ |
| palm | ✓ | ✓ | | | | | ✓ | | | | |
| gemini | ✓ | ✓ | | | | | ✓ | | | | |
| cloudflare | | | ✓ | | | ✓ | | | | | |
| cohere | | ✓ | ✓ | | | ✓ | | | ✓ | | |
| cohere_chat | | ✓ | ✓ | | | ✓ | | | ✓ | | |
| huggingface | ✓ | ✓ | ✓ | | | ✓ | | ✓ | ✓ | | |
| ai21 | ✓ | ✓ | ✓ | ✓ | | ✓ | | ✓ | | | |
| nlp_cloud | ✓ | ✓ | ✓ | | | ✓ | ✓ | ✓ | ✓ | | |
| together_ai | ✓ | ✓ | ✓ | | | ✓ | | | | | |
| aleph_alpha | | | ✓ | | | ✓ | | | | | |
| ollama | ✓ | | ✓ | | | | | | ✓ | | |
| ollama_chat | ✓ | | ✓ | | | | | | ✓ | | |
| vllm | | | | | | ✓ | ✓ | | | | |
| azure | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | | | ✓ | | |
- "✓" 表示指定的 `custom_llm_provider` 可以引發對應的例外。
- 空白儲存格表示沒有關聯，或如函式所示，該提供者不會引發該特定例外類型。
> 若要更深入了解這些例外，您可以查看 [此處](https://github.com/BerriAI/litellm/blob/d7e58d13bf9ba9edbab2ab2f096f3de7547f35fa/litellm/utils.py#L1544) 的實作以取得更多見解。
`ContextWindowExceededError` 是 `InvalidRequestError` 的子類別。引入它是為了讓例外處理情境有更細緻的區分。請參閱 [此 issue 以了解更多](https://github.com/BerriAI/litellm/issues/228)。
歡迎對改善例外對應提出貢獻 [歡迎](https://github.com/BerriAI/litellm#contributing)