How to utilize Azure OpenAI with version 0.28.1

Other Questions OpenAI
1

I’m looking to integrate Azure OpenAI but need to stick with an older version of the OpenAI library (0.28.1), rather than any version above 1.0.

Currently, I have the following setup using the newer OpenAI library:

import os

os.environ["OPENAI_API_TYPE"] = "azure"
os.environ["OPENAI_API_VERSION"] = "2023-07-01-preview"
os.environ["OPENAI_API_BASE"] = "your-endpoint-here"
os.environ["OPENAI_API_KEY"] = "your-key-here"

from langchain_openai import AzureOpenAI

ai_model = AzureOpenAI(
    deployment_name="my-deployment",
    model_name="gpt-3.5-turbo-instruct",
)

response = ai_model("Give me a funny story")

Unfortunately, this setup doesn’t function with openai==0.28.1. I’m restricted to this older version due to my project’s specifications. How can I adjust my code to work with the legacy OpenAI library while maintaining the same straightforward usage, like using ai_model("my prompt here") to receive a response?

2

I hit this exact issue a few months ago on a legacy project. Version 0.28.1 uses a completely different API structure. Drop the langchain_openai import and use the older openai module directly.

Here’s what worked for me:

import openai

openai.api_type = "azure"
openai.api_base = "your-endpoint-here"
openai.api_version = "2023-07-01-preview"
openai.api_key = "your-key-here"

response = openai.Completion.create(
    engine="my-deployment",
    prompt="Give me a funny story",
    max_tokens=150
)

print(response.choices[0].text)

Use engine instead of deployment_name and the response structure is different. The older version returns a response object where you access .choices[0].text to get the actual text. Also, set max_tokens explicitly since it’s required in the legacy version.