Using Azure OpenAI with Legacy OpenAI Library Version 0.28.1

Other Questions OpenAI
1

I have this working code for Azure OpenAI integration using openai version 1.0 and above:

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

# Initialize Azure OpenAI instance
ai_model = AzureOpenAI(
    deployment_name="my-deployment",
    model_name="gpt-3.5-turbo-instruct",
)

# Execute the model
result = ai_model("Give me a funny story")

The problem is I need to use an older version of the openai package (specifically version 0.28.1) for compatibility reasons. When I try to run this code with openai==0.28.1, it fails completely.

Is there a way to make Azure OpenAI work with the legacy openai library? I want to keep the same simple calling pattern where I can just do ai_model("my prompt") but using the older package version. What changes do I need to make to get this working?

2

yup, i faced this too. with 0.28.1, u gotta stick to the old openai.Completion.create() method - the langchain wrapper won’t cut it. just set your azure configs and do openai.Completion.create(engine="your-deployment-name", prompt="your prompt"). langchain_openai is for newer versions only.

3

Been there. The version mess between old and new OpenAI packages is a nightmare.

You need a simple wrapper that mimics your current calls but uses legacy methods under the hood. I built something similar when our production system was stuck on 0.28.1 for months.

Here’s the pattern:

import openai

class LegacyAzureWrapper:
    def __init__(self, deployment_name):
        self.deployment_name = deployment_name
        openai.api_type = "azure"
        openai.api_version = "2023-07-01-preview" 
        openai.api_base = "your-endpoint-here"
        openai.api_key = "your-key-here"
    
    def __call__(self, prompt):
        response = openai.Completion.create(
            engine=self.deployment_name,
            prompt=prompt,
            max_tokens=150
        )
        return response.choices[0].text.strip()

ai_model = LegacyAzureWrapper("my-deployment")
result = ai_model("Give me a funny story")

Honestly though, managing these version conflicts manually is a pain. Rather than wrestling with legacy code, I’d automate the whole integration. Set up proper API orchestration that handles version compatibility, retries, and scaling automatically.

This is exactly why I use Latenode. It manages all the API complexity and version headaches, so you can focus on building instead of debugging compatibility issues.

4

Had the same issue last year when we were stuck on 0.28.1 because of legacy dependencies. The old version works completely differently - it uses direct openai imports instead of langchain_openai, and you can’t use environment variables the same way. You have to configure openai directly in your code by setting openai.api_type, openai.api_base, openai.api_key, and openai.api_version right on the openai module. For Azure with 0.28.1, you’ll use openai.Completion.create() with the engine parameter set to your deployment name. The response format’s different too - you get a response object where you need .choices[0].text to grab the actual output. If you want that clean ai_model("prompt") pattern, you’ll need to build a simple wrapper class that handles all the legacy API stuff behind the scenes.