[bug] Handle failed OpenAI API requests about guardrails HOT 4 CLOSED

Looks like latest version 0.1.7 returns a readable error.
PyPI still lists 0.1.6 as latest version: https://pypi.org/project/guardrails-ai/, it would be great to update that.

Feel free to close this issue.

from guardrails.

@matejm can you check if you are able to call the API with the above config without wrapped by the guard object?

from guardrails.

Hey @matejm @ShreyaR @stchau4work

Would recommend wrapping the base openai call with reliableGPT - it'll handle the retries, model switching, etc. when OpenAI throws errors

from reliablegpt import reliableGPT
openai.ChatCompletion.create = reliableGPT(openai.ChatCompletion.create,...)

Source: https://github.com/BerriAI/reliableGPT

--
@ShreyaR happy to make any changes you need to make this useful for guardrails.

from guardrails.

Thanks for the issue @matejm.

@stchau4work not sure what you mean, please open a separate issue.

from guardrails.