How to Migrate Your Custom Code for Chat Completions to a New Version of the OpenAI Python API library on Azure OpenAI.

OpenAI released a new version of the OpenAI Python API library.

Starting on November 6, 2023 pip install openai and pip install openai --upgrade will install version 1.x of the OpenAI Python library.

You can now instantiate a client, instead of using a global default.

Switch to explicit client instantiation.

Upgrade OpenAi, or install it anew if you don’t already have it installed.

pip install openai --upgrade

Import openai libraries:

import os
from openai import AzureOpenAI

Create a client:

client = AzureOpenAI(
    azure_endpoint=openai.api_base,
    api_key=openai.api_key,
    api_version="2023-05-15"
)

Create a query:

query = "What is the median value for acidity in this file?"

Create a response. Use exact name of the deployment you created to set the model variable:

response = client.chat.completions.create(
    model="gpt-4-32k-custom",  # model = "deployment_name".
    messages=[

        {"role": "user", "content": query},

    ]
)

Display query and response:

response = response.choices[0].message.content
print('\n\n-------Your question to LLM -----\n\n', query)
print("\n\n -------OpenAI response: ------\n\n", response)

You will need to use a different date for Completions. Update syntax for response.

client = AzureOpenAI(
    azure_endpoint=openai.api_base,
    api_key=openai.api_key,
    api_version="2023-10-01-preview",
)

deployment_name = 'REPLACE_WITH_YOUR_DEPLOYMENT_NAME'  # This will correspond to the custom name you chose for your deployment when you deployed a model.

# Send a completion call to generate an answer
print('Sending a test completion job')
start_phrase = 'Write a tagline for an ice cream shop. '
response = client.completions.create(model=deployment_name, prompt=start_phrase, max_tokens=10)
print(response.choices[0].text)

As of June 2024, GPT-4 Omni is available in many regions. Although Azure documentation provides the release date: “2024–05–13”, this is not the date you should use for the api_version. Using it will result in a ‘Resource not found’ error.

Instead, use api_version = “2024–02–15-preview”

To use embeddings, make sure you have a deployment for embeddings:

import os
from openai import AzureOpenAI

client = AzureOpenAI(
  azure_endpoint=openai.api_base,
  api_key=openai.api_key,
  api_version="2023-05-15"
)

response = client.embeddings.create(
    input = "Your text string goes here",
    model= "text-embedding-ada-002"
)

print(response.model_dump_json(indent=2))

References: https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/migration?tabs=python-new%2Cdalle-fix