Does Azure Open AI GPT 4 Vision model supports stream options in the request ?

Question

Does Azure Open AI GPT 4 Vision model supports stream options in the request ?

Supraja Narra 0

Hi,

I am working with the Azure OpenAI Service using the Azure Open AI GPT 4 Vision model. When I include the stream options parameter in my API request (along with "stream: true") I receive the following error

"error": {

    "message": "1 validation error for Request\nbody -> stream_options\n  extra fields not permitted (type=value_error.extra)",

    "type": "invalid_request_error",

    "param": null,

    "code": null

}

The same request works correctly if I remove the stream options field.

Please suggest if any solution.

Thanks,Supraja

Deleted

This comment has been deleted due to a violation of our Code of Conduct. The comment was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.

Manas Mohanty 3,125 Microsoft External Staff

Hi Supraja Narra

Yes, "Stream_option = True" is not valid, but "stream = True" is valid one.

I have used turbo-2024-04-09 version of GPT 4 and api_version = '2024-02-15-preview' for my tests.

Streaming seems to be a valid parameter as per my below trial codes. Though parsing the streamed object does not give any output

Here is my sample code.

api_base = '<endpointurl' # your endpoint should look like the following https://YOUR_RESOURCE_NAME.openai.azure.com/
api_key="<apikey>"
deployment_name = 'gpt-4'
api_version = '2024-02-15-preview' # this might change in the future

client = AzureOpenAI(
    api_key=api_key,  
    api_version=api_version,
    base_url=f"{api_base}openai/deployments/{deployment_name}",



response = client.chat.completions.create(
    model=deployment_name,
    messages=[
        { "role": "system", "content": "You are a helpful assistant." },
        { "role": "user", "content": [  
           
            { 
                "type": "image_url",
                "image_url": {
                    "url": "https://www.pngkey.com/png/detail/137-1377870_canvas-sample-sample-image-url.png"
                }
            },
             { 
                "type": "text", 
                "text": "Describe this picture:" 
            }
        ] } 
    ],
    max_tokens=2000,
    stream=True, #Gives response when set false else Stream object

)


print(response)


# Handle the streamed response
for chunk in response:
    if 'choices' in chunk:
        for choice in chunk['choices']:
            print(choice['delta']['content'], end='', flush=True)

shall update you again after reviewing with team internally.

Thank you.

Manas Mohanty 3,125 Reputation points Microsoft External Staff

2025-04-14T08:04:24.0866667+00:00

Hi Supraja Narra

Could you please let us know if the pointers in below answer have been helpful to you.

Thank you.

1 answer

Your answer

Deleted

This comment has been deleted due to a violation of our Code of Conduct. The comment was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.
Manas Mohanty 3,125 Reputation points Microsoft External Staff

2025-04-10T15:30:23.9366667+00:00

Hi Supraja Narra

Yes, "Stream_option = True" is not valid, but "stream = True" is valid one.

I have used turbo-2024-04-09 version of GPT 4 and api_version = '2024-02-15-preview' for my tests.

Streaming seems to be a valid parameter as per my below trial codes. Though parsing the streamed object does not give any output

Here is my sample code.

api_base = '<endpointurl' # your endpoint should look like the following https://YOUR_RESOURCE_NAME.openai.azure.com/ api_key="<apikey>" deployment_name = 'gpt-4' api_version = '2024-02-15-preview' # this might change in the future client = AzureOpenAI( api_key=api_key, api_version=api_version, base_url=f"{api_base}openai/deployments/{deployment_name}", response = client.chat.completions.create( model=deployment_name, messages=[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": [ { "type": "image_url", "image_url": { "url": "https://www.pngkey.com/png/detail/137-1377870_canvas-sample-sample-image-url.png" } }, { "type": "text", "text": "Describe this picture:" } ] } ], max_tokens=2000, stream=True, #Gives response when set false else Stream object ) print(response) # Handle the streamed response for chunk in response: if 'choices' in chunk: for choice in chunk['choices']: print(choice['delta']['content'], end='', flush=True)

shall update you again after reviewing with team internally.

Thank you.
Manas Mohanty 3,125 Reputation points Microsoft External Staff

2025-04-14T08:04:24.0866667+00:00

Hi Supraja Narra

Could you please let us know if the pointers in below answer have been helpful to you.

Thank you.

Answer 1

Hi Supraja Narra

I am able to get streamed response with api_version = '2024-12-01-preview' and another chunking approach


import os
import json
import requests
import time
from openai import AzureOpenAI

api_base = 'https://resourcename.openai.azure.com/' # your endpoint should look like the following https://YOUR_RESOURCE_NAME.openai.azure.com/
api_key="<apikey>"
deployment_name = 'gpt-4'
api_version = '2024-12-01-preview' # this might change in the future

client = AzureOpenAI(
    api_key=api_key,  
    api_version=api_version,
    base_url=f"{api_base}openai/deployments/{deployment_name}",
response = client.chat.completions.create(
    model="gpt-4",
    messages=[
        { "role": "system", "content": "You are a helpful assistant." },
        { "role": "user", "content": [  
            {
                "type": "text",
                "text": "Describe this picture:"
            },
            {
                "type": "image_url",
                "image_url": {
                    "url": "https://th.bing.com/th/id/OIP.JQEaFD6izhB9jFmJlO1_NgHaFL?rs=1&pid=ImgDetMain"
                }
            }
        ] }
    ],
    stream=True
   # stream_options={"include_usage": True}
)
 
for i in response:
    if len(i.choices)==0:
        continue
    print(i.choices[0].delta.content)

Output

None The image shows a close -up photograph of vibrant pink flowers , possibly dah lias , None

Hope it helps

Thank you.

Share via

Does Azure Open AI GPT 4 Vision model supports stream options in the request ?

1 answer

Your answer