AiFail

Python库，用于重试openai/gcp API调用

安装

pip install aifail

使用

要开始使用，只需将您的函数用 @retry_if 包装，指定重试的条件。

https://github.com/GooeyAI/aifail/blob/a540c05a2a9436c0b6b1caab8ed823387999d5f9/examples/basic_openai.py#L5

https://github.com/GooeyAI/aifail/blob/a540c05a2a9436c0b6b1caab8ed823387999d5f9/examples/basic_openai.py#L13-L30

自定义逻辑

您可以用这个来处理任何需要重试的东西，例如google sheets -

def sheets_api_should_retry(e: Exception) -> bool:
    return isinstance(e, HttpError) and (
        e.resp.status in (408, 429) or e.resp.status > 500
    )


@retry_if(sheets_api_should_retry)
def update_cell(spreadsheet_id: str, row: int, col: int, value: str):
    get_spreadsheet_service().values().update(
        spreadsheetId=spreadsheet_id,
        range=f"{col_i2a(col)}{row}:{col_i2a(col)}{row}",
        body={"values": [[value]]},
        valueInputOption="RAW",
    ).execute()

高级使用

此库由GooeyAI在生产中使用，每天处理数千次API调用。为了节省成本和处理速率限制，您可以智能地指定快速回退（例如azure openai） -

https://github.com/GooeyAI/aifail/blob/a540c05a2a9436c0b6b1caab8ed823387999d5f9/examples/azure_openai_fallback.py#L7-L16

https://github.com/GooeyAI/aifail/blob/a540c05a2a9436c0b6b1caab8ed823387999d5f9/examples/azure_openai_fallback.py#L19-L35

可追踪的错误

AiFail附带内置的记录器，并输出完整的堆栈跟踪，将错误追溯到原始调用位置。

# python examples/azure_openai_fallback.py 
2023-11-18 04:36:01.364 | WARNING  | aifail.aifail:try_all:63 - [2/2] tyring next fn, prev_exc=NotFoundError("Error code: 404 - {'error': {'code': 'DeploymentNotFound', 'message': 'The API deployment for this resource does not exist. If you created the deployment within the last 5 minutes, please wait a moment and try again.'}}")
2023-11-18 04:36:01.778 | WARNING  | aifail.aifail:wrapper:98 - [1/1] captured error, retry_delay=0.4117675457681431s, exc=NotFoundError("Error code: 404 - {'error': {'message': 'The model `gpt-4-x` does not exist', 'type': 'invalid_request_error', 'param': None, 'code': 'model_not_found'}}")
2023-11-18 04:36:02.483 | WARNING  | aifail.aifail:try_all:63 - [2/2] tyring next fn, prev_exc=NotFoundError("Error code: 404 - {'error': {'code': 'DeploymentNotFound', 'message': 'The API deployment for this resource does not exist. If you created the deployment within the last 5 minutes, please wait a moment and try again.'}}")
2023-11-18 04:36:04.093 | WARNING  | aifail.aifail:wrapper:98 - [2/1] captured error, retry_delay=0.9974197744911488s, exc=NotFoundError("Error code: 404 - {'error': {'message': 'The model `gpt-4-x` does not exist', 'type': 'invalid_request_error', 'param': None, 'code': 'model_not_found'}}")
Traceback (most recent call last):
  File "/Users/dev/Projects/dara/aifail/aifail/aifail.py", line 65, in try_all
    return fn()
  File "/Users/dev/Projects/dara/aifail/examples/azure_openai_fallback.py", line 28, in <lambda>
    lambda: azure_client.chat.completions.create(
  ...
openai.NotFoundError: Error code: 404 - {'error': {'code': 'DeploymentNotFound', 'message': 'The API deployment for this resource does not exist. If you created the deployment within the last 5 minutes, please wait a moment and try again.'}}

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Users/dev/Projects/dara/aifail/aifail/aifail.py", line 86, in wrapper
    return fn(*args, **kwargs)
  File "/Users/dev/Projects/dara/aifail/examples/azure_openai_fallback.py", line 26, in chad_gpt4
    response = try_all(
  File "/Users/dev/Projects/dara/aifail/aifail/aifail.py", line 69, in try_all
    raise prev_exc
  File "/Users/dev/Projects/dara/aifail/aifail/aifail.py", line 65, in try_all
    return fn()
  File "/Users/dev/Projects/dara/aifail/examples/azure_openai_fallback.py", line 34, in <lambda>
    lambda: openai_client.chat.completions.create(
  ...
openai.NotFoundError: Error code: 404 - {'error': {'message': 'The model `gpt-4-x` does not exist', 'type': 'invalid_request_error', 'param': None, 'code': 'model_not_found'}}

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Users/dev/Projects/dara/aifail/aifail/aifail.py", line 65, in try_all
    return fn()
  File "/Users/dev/Projects/dara/aifail/examples/azure_openai_fallback.py", line 28, in <lambda>
    lambda: azure_client.chat.completions.create(
  ...
openai.NotFoundError: Error code: 404 - {'error': {'code': 'DeploymentNotFound', 'message': 'The API deployment for this resource does not exist. If you created the deployment within the last 5 minutes, please wait a moment and try again.'}}

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Users/dev/Projects/dara/aifail/examples/azure_openai_fallback.py", line 43, in <module>
    chad_gpt4(
  File "/Users/dev/Projects/dara/aifail/aifail/aifail.py", line 102, in wrapper
    raise prev_exc
  File "/Users/dev/Projects/dara/aifail/aifail/aifail.py", line 86, in wrapper
    return fn(*args, **kwargs)
  File "/Users/dev/Projects/dara/aifail/examples/azure_openai_fallback.py", line 26, in chad_gpt4
    response = try_all(
  File "/Users/dev/Projects/dara/aifail/aifail/aifail.py", line 69, in try_all
    raise prev_exc
  File "/Users/dev/Projects/dara/aifail/aifail/aifail.py", line 65, in try_all
    return fn()
  File "/Users/dev/Projects/dara/aifail/examples/azure_openai_fallback.py", line 34, in <lambda>
    lambda: openai_client.chat.completions.create(
  ...
openai.NotFoundError: Error code: 404 - {'error': {'message': 'The model `gpt-4-x` does not exist', 'type': 'invalid_request_error', 'param': None, 'code': 'model_not_found'}}

Process finished with exit code 1

Sentry也会捕获整个重试循环