Guardrails support of AzureOpenAI with openai>1.0.0[bug]

[shima-khoshraftar]

Describe the bug
Does Guardrails-ai support AzureOpenAI with openai library>1.0 which has a different llm call api from openai==0.28? (openai.chat.completions.create instead of openai.ChatCompletion.create)

To Reproduce
openai.api_key = api_key
openai.azure_endpoint =azure_endpoint
openai.api_type = 'azure'
openai.api_version = api_version

raw_llm_response, validated_response,*rest = guard(
#openai.ChatCompletion.create,
openai.chat.completions.create,
prompt_params={"document": content[:6000]},
#engine="text-davinci-003",
model='gpt-35-turbo-1106',
max_tokens=2048,
temperature=0.3,
)

Expected behavior
I expect it to call llm just like it worked with openai==0.28 (openai.ChatCompletion.create) but I get this error instead:

TypeError: create() takes 1 argument(s) but 2 were given

I realized that this was fixed for openAI by setting the api_key through os.environ. But how can I do it for AzureOpenAI?

Library version:
Guardrails-ai 0.3.2
openai 1.12.0

Thanks.

[zsimjee]

This seems more like it's passing too many requests to the create constructor. Can you try removing temperature and max_tokens and see if it works?

[shima-khoshraftar]

Thanks for your reply. I removed temperature and max_tokens but still have the same error. This error is similar to these issues that were previously raised for using openai>1.x:

#514 (comment)
#504

but I am wondering what the solution is for azure openai where these parameters should be set in order to call openai.chat.completions:
openai.api_type = "azure"
openai.api_version = "2023-05-15"
openai.api_base = os.environ.get("AZURE_OPENAI_API_BASE")
openai.api_key = os.environ.get("AZURE_OPENAI_API_KEY")

according to https://www.guardrailsai.com/docs/integrations/azure_openai.
I even set these parameters like this but still get the same error.
os.environ["OPENAI_API_TYPE"] = 'azure'
os.environ["OPENAI_API_VERSION"] = api_version
os.environ["AZURE_OPENAI_API_KEY"] = api_key
os.environ["AZURE_OPENAI_ENDPOINT"] = api_base

Thanks.

[Aman0509]

@shima-khoshraftar - Facing the similar issue. Could you please tell me what version of guardrails is working with openai v0.28 currently?

[shima-khoshraftar]

@Aman0509 The latest release of guardrails works with Azure openai with openai==0.28.

[vijayoct27]

I am wondering if there is any update to this issue? I am also trying to use guardrails with Azure OpenAI and openai>1.0.0.

[thekaranacharya]

You can now use AzureOpenAI with guardrails using litellm. Please follow this example (just substitute with AzureOpenAI) instead. Please let us know here if there are any issues.

[vijayoct27]

Please follow this example (just substitute with AzureOpenAI) instead.
Sorry for the confusion, what example are you referring to here?

[thekaranacharya]

My bad. I thought I added the link! Just updated the comment.

[shima-khoshraftar]

Thanks for letting us know about the update and sending the link. However, the link does not seem to work, It can not find examples/litellm_example.ipynb and throws a page not found error. Could you please update the link? Thanks.

[jonathanbouchet]

@shima-khoshraftar I'm also looking at this example since I would like to use guardails-ai with openai > 1
It looks like @thekaranacharya is referring to litellm package (link).
They have a quick tutorial here

from litellm import completion
import os

## set ENV variables 
os.environ["OPENAI_API_KEY"] = "your-openai-key" 

messages = [{ "content": "Hello, how are you?","role": "user"}]

# openai call
response = completion(model="gpt-3.5-turbo", messages=messages)

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 14 days.

[github-actions]

This issue was closed because it has been stalled for 14 days with no activity.