AzureOpenAI

GGuru

Hi Guys...I am using Azure OpenAI in the below code and getting error message "InvalidRequestError: Must provide an 'engine' or 'deployment_id' parameter to create a <class 'openai.api_resources.chat_completion.ChatCompletion'>".
Below is the code- llm = AzureOpenAI(
model="gpt-35-turbo",
model_name="text-davinci-003",
deployment_id=os.getenv("OPENAI_API_DEPLOYMENT_NAME"),
api_key=os.getenv("OPENAI_API_KEY"),
api_base=os.getenv("OPENAI_API_BASED"),
api_type=os.getenv("OPENAI_API_TYPE"),
api_version=os.getenv("OPENAI_API_VERSION"),
)

service_context = ServiceContext.from_defaults(
llm=llm,
)
set_global_service_context(service_context)

node_parser = SimpleNodeParser.from_defaults(
text_splitter=text_splitter,
metadata_extractor=metadata_extractor
)

ls_spilt_document=[]
for document_page in doc_patient_visit:
ls_spilt_document=markdown_splitter.split_text(document_page.page_content)
for split_index in range(len(ls_spilt_document)):
if "Patient Demographics" in list(ls_spilt_document[split_index].metadata.keys()):
document = Document(text=ls_spilt_document[split_index].page_content)
nodes = node_parser.get_nodes_from_documents([document],show_progress=True)"

21 comments

WWhiteFang_Jr

Hi, did you follow this example?
https://gpt-index.readthedocs.io/en/latest/examples/llm/azure_openai.html

WWhiteFang_Jr

It uses engine while defining Azure llm. Could you try with that once

GGuru

Thanks for your response.Yes this works with this line of code 'llm.complete("what is life")'. However it fails at the line "nodes = node_parser.get_nodes_from_documents([document])"

GGuru

detailed error message "/usr/local/lib/python3.10/dist-packages/openai/api_resources/abstract/engine_api_resource.py in __prepare_create_request(cls, api_key, api_base, api_type, api_version, organization, **params)
83 if typed_api_type in (util.ApiType.AZURE, util.ApiType.AZURE_AD):
84 if deployment_id is None and engine is None:
---> 85 raise error.InvalidRequestError(
86 "Must provide an 'engine' or 'deployment_id' parameter to create a %s"
87 % cls,

InvalidRequestError: Must provide an 'engine' or 'deployment_id' parameter to create a <class 'openai.api_resources.chat_completion.ChatCompletion'>"

WWhiteFang_Jr

Can you show how you are importing AzureOpenAI, is it via llamaindex or langchain?

LLogan M

Yea, double check you followed that example

LLogan M

setting the global service context solves a lot of issues

GGuru

I am importing AzureOpenAI via llamaindex. from llama_index.llms import OpenAI,AzureOpenAI

GGuru

Yes that was my understanding as well. In spite of setting Global context, I am getting error...maybe I am missing something here..Here is the code from llama_index.llms import AzureOpenAI llm=AzureOpenAI(
model="gpt-35-turbo",
model_name="text-davinci-003",
deployment_id=os.getenv("OPENAI_API_DEPLOYMENT_NAME"),
api_key=os.getenv("OPENAI_API_KEY"),
api_base=os.getenv("OPENAI_API_BASED"),
api_type=os.getenv("OPENAI_API_TYPE"),
api_version=os.getenv("OPENAI_API_VERSION"),
)

service_context = ServiceContext.from_defaults(
llm=llm,
)
set_global_service_context(service_context)

WWhiteFang_Jr

Just tried with AzureOpenAI with the following code

Plain Text

import openai
openai.api_key = AZURE_API_KEY
openai.api_base = AZURE_API_BASE_URL
openai.api_type = API_TYPE
openai.api_version = API_VERSION
llm = AzureOpenAI(engine="engine_name", model="gpt-35-turbo", temperature=0.0)

GGuru

I am using MarvinMetadataExtractor, do I have to explicitly pass the Azure OpenAI parameters

GGuru

this is the code " @ai_model
class PatientDemographic(BaseModel):
Address:str=Field(...,description="Address of the patient")
Contact_number:str=Field(...,description="Contact number at work,home and office")
contact_email:str=Field(...,description="Email id of the patient")
Language:str=Field(...,description="Language of the patient")
Race:str=Field(...,description="Race or Ethinicity of the patient")"

GGuru

Attachment

GGuru

I had tried this earlier...still getting the same error. Not sure if this is related to Marvin

WWhiteFang_Jr

You have set OpenAI key for Marvin?
marvin.settings.openai.api_key = os.environ["OPENAI_API_KEY"]

WWhiteFang_Jr

For ref: https://gpt-index.readthedocs.io/en/latest/examples/metadata_extraction/MarvinMetadataExtractorDemo.html#metadata-extraction-and-augmentation-w-marvin

GGuru

Yes tried that as well

GGuru

marvin.settings.azure_openai.api_key=os.environ['OPENAI_API_KEY']
marvin.settings.azure_openai.api_base=os.environ['OPENAI_API_BASE']
marvin.settings.azure_openai.deployment_name=os.environ['OPENAI_API_DEPLOYMENT_NAME']
marvin.settings.azure_openai.api_type='azure'
marvin.settings.llm_model = 'azure_openai/gpt-3.5-turbo'

marvin.settings.azure_openai.api_version=os.environ['OPENAI_API_VERSION']

LLogan M

I'm not sure if marvin is compatible with azure tbh

GGuru

Thanks for your inputs

Add a reply

Find answers from the community

AzureOpenAI