InvalidRequestError: Must provide an 'engine' or 'deployment_id' parameter to create a <class 'openai.api_resources.completion.Completion'>

[pawarbi]

Hi, I am using the notebook in Microsoft Fabric. I have python 3.10.10. I installed !pip install jupyter_ai_magics and saved the open ai key using %env OPENAI_API_KEY=sk-xxxxxx. After loading the extension, if I use any model from openai, I get below error message:

%%ai gpt3
write a pandas program to import csv data

Error:

File ~/cluster-env/trident_env/lib/python3.10/site-packages/openai/api_resources/abstract/engine_api_resource.py:83, in EngineAPIResource.__prepare_create_request(cls, api_key, api_base, api_type, api_version, organization, **params)
     81 if typed_api_type in (util.ApiType.AZURE, util.ApiType.AZURE_AD):
     82     if deployment_id is None and engine is None:
---> 83         raise error.InvalidRequestError(
     84             "Must provide an 'engine' or 'deployment_id' parameter to create a %s"
     85             % cls,
     86             "engine",
     87         )
     88 else:
     89     if model is None and engine is None:

InvalidRequestError: Must provide an 'engine' or 'deployment_id' parameter to create a <class 'openai.api_resources.completion.Completion'>

All other models from other providers (cohere, huggingface, ai21) work fine without any issues. it's only openai that throw this error. I also installed and upgraded openai and langchain but same error message. If I follow the same steps in my local jupyter notebook with Python 3.9, everything works fine and I get the expected output. I am stumped why I am getting Must provide an 'engine' or 'deployment_id' parameter message. Can you please help? Thanks.

Description

Reproduce

1. Go to '...'
2. Click on '...'
3. Scroll down to '...'
4. See error '...'

Expected behavior

Context

• Operating System and version:
• Browser and version:
• JupyterLab version:

Troubleshoot Output

Paste the output from running `jupyter troubleshoot` from the command line here.
You may want to sanitize the paths in the output.

Command Line Output

Paste the output from your command line running `jupyter lab` here, use `--debug` if possible.

Browser Output

Paste the output from your browser Javascript console here, if applicable.

[welcome[bot]]

Thank you for opening your first issue in this project! Engagement like this is essential for open source projects! :hugs:
If you haven't done so already, check out Jupyter's Code of Conduct. Also, please try to follow the issue template as it helps other other community members to contribute more effectively. You can meet the other Jovyans by joining our Discourse forum. There is also an intro thread there where you can stop by and say Hi! :wave:
Welcome to the Jupyter community! :tada:

[dlqqq]

@pawarbi Hm, it almost looks like you're somehow using the azure-chat-openai provider instead of the openai-chat provider? I did a bit more digging and there are several identical issues upstream in LangChain. For example: https://github.com/langchain-ai/langchain/issues/5422

It looks like the OpenAI SDK uses a singleton object that persists over the lifetime of a process. To clarify, if you used an Azure OpenAI provider previously in the same notebook kernel, subsequent calls to the "non-Azure" OpenAI provider will fail because the OpenAI SDK assumes you're still trying to make calls to the Azure API. The workaround recommended in the linked issue is to explicitly "reset" the singleton before calling the non-Azure OpenAI provider:

llmazure =  AzureChatOpenAI(**openaiazure_params)

# run the below 3 lines to reset the state after using an Azure OpenAI provider
openai.api_type = "open_ai"
openai.api_base = "https://api.openai.com/v1"
openai.api_version = None

llm = ChatOpenAI(**openai_params)

If this happens even if you didn't use an Azure OpenAI provider previously, it might be because the Microsoft Fabric environment is setting the OPENAI_API_TYPE environment variable to azure instead of open_ai. Try running:

echo $OPENAI_API_TYPE

in your shell. Let me know if the workaround helped, and if Microsoft Fabric is setting OPENAI_API_TYPE=azure by default. Thanks!

[dlqqq]

cc @3coins This might motivate an upstream fix to LangChain, or even to the OpenAI Python SDK. Here's a full list of identical issues on the LangChain repo:

https://github.com/langchain-ai/langchain/issues?q=Must+provide+an+%27engine%27+or+%27deployment_id%27+parameter

[pawarbi]

Thank you @dlqqq . Running !echo $OPENAI_API_TYPE did not return anything.

I ran below as you suggested :

from langchain.chat_models import AzureChatOpenAI, ChatOpenAI
import openai

openai.api_key = "sk-xxxxxxxxxxxxxxxxxxxx"

openai_params = {
    "api_type": "open_ai",
    "api_base": "https://api.openai.com/v1",
    "api_version": None
}

llm = ChatOpenAI(**openai_params)

%%ai gpt3
who are you

This led to a different error message :

--> 298     resp, got_stream = self._interpret_response(result, stream)
    299     return resp, got_stream, self.api_key

File ~/cluster-env/clonedenv/lib/python3.10/site-packages/openai/api_requestor.py:700, in APIRequestor._interpret_response(self, result, stream)
    692     return (
    693         self._interpret_response_line(
    694             line, result.status_code, result.headers, stream=True
    695         )
    696         for line in parse_stream(result.iter_lines())
    697     ), True
    698 else:
    699     return (
--> 700         self._interpret_response_line(
    701             result.content.decode("utf-8"),
    702             result.status_code,
    703             result.headers,
    704             stream=False,
    705         ),
    706         False,
    707     )

File ~/cluster-env/clonedenv/lib/python3.10/site-packages/openai/api_requestor.py:765, in APIRequestor._interpret_response_line(self, rbody, rcode, rheaders, stream)
    763 stream_error = stream and "error" in resp.data
    764 if stream_error or not 200 <= rcode < 300:
--> 765     raise self.handle_error_response(
    766         rbody, rcode, resp.data, rheaders, stream_error=stream_error
    767     )
    768 return resp

AuthenticationError: Your authentication token is not from a valid issuer.

I created a brand new token which worked fine locally but gave above error in fabric. I wasn;t sure what params to pass to : llmazure = AzureChatOpenAI(**openaiazure_params)

Please advise. Thank you!

[dlqqq]

@pawarbi You have to set these attributes on the openai object directly:

import openai

openai.api_type = "open_ai"
openai.api_base = "https://api.openai.com/v1"
openai.api_version = None

This block of code that you ran:

openai_params = {
    "api_type": "open_ai",
    "api_base": "https://api.openai.com/v1",
    "api_version": None
}

llm = ChatOpenAI(**openai_params)

doesn't do anything except define two unused variables, openai_params and llm.

[pawarbi]

Thanks @dlqqq

Same error message:

    150         api_key, api_base, api_type, api_version, organization, **params
    151     )
--> 153     response, _, api_key = requestor.request(
    154         "post",
    155         url,
    156         params=params,
    157         headers=headers,
    158         stream=stream,
    159         request_id=request_id,
    160         request_timeout=request_timeout,
    161     )
    163     if stream:
    164         # must be an iterator
    165         assert not isinstance(response, OpenAIResponse)

--> 700         self._interpret_response_line(
    701             result.content.decode("utf-8"),
    702             result.status_code,
    703             result.headers,
    704             stream=False,
    705         ),
    706         False,
    707     )

File ~/cluster-env/clonedenv/lib/python3.10/site-packages/openai/api_requestor.py:765, in APIRequestor._interpret_response_line(self, rbody, rcode, rheaders, stream)
    763 stream_error = stream and "error" in resp.data
    764 if stream_error or not 200 <= rcode < 300:
--> 765     raise self.handle_error_response(
    766         rbody, rcode, resp.data, rheaders, stream_error=stream_error
    767     )
    768 return resp

AuthenticationError: Your authentication token is not from a valid issuer.

My code:

!pip install jupyter_ai_magics --quiet
!pip install openai --quiet
import openai

openai.api_type = "open_ai"
openai.api_base = "https://api.openai.com/v1"
openai.api_version = None

%env OPENAI_API_KEY=sk-xxxx
%reload_ext jupyter_ai_magics

%%ai gpt4
write a pandas program to import csv data

This exact code and API token works without any errors locally in Jupyter Notebook. I tried openai.api_key = "sk-xxx" as well.

Thanks again for your help.