Question: get started with AzureOpenAI

[WeihanLi]

Great thanks for the project, I'm trying to use it with Azure OpenAI service, but do not know how to find the DeploymentId, could you please help, thanks

[WeihanLi]

I'm using the following code for the AzureOpenAI

var options = new OpenAiOptions()
{
	BaseDomain = " https://spark.openai.azure.com",
	ApiKey = apiKey,
	DeploymentId = "gpt35",
	ResourceName = "spark",
	ProviderType = ProviderType.Azure,
};
var openAiService = new OpenAIService(options);
var completionResult = await openAiService.ChatCompletion.CreateCompletion(new ChatCompletionCreateRequest
{
	Messages = new List<ChatMessage>
	{
		ChatMessage.FromSystem("You are a helpful assistant."),
		ChatMessage.FromUser("What's new in .NET 6")
	},
	Model = Models.ChatGpt3_5Turbo // "gpt-35-turbo" also not working
});

completionResult has error as follows:

{"Code":"404","Message":"Resource not found"}

While I could call the AzureOpenAI API by following the rest api docs, after some research, I found that the API is not the same, wondering that maybe there's something wrong with my way calling AzureOpenAI.

For details, the raw Azure OpenAI request likes below:

POST https://spark.openai.azure.com/openai/deployments/gpt35/completions?api-version=2022-12-01
Host: spark.openai.azure.com
Content-Type: application/json
api-key: <api-key>

{
  "prompt": "<|im_start|>system\nThe system is a helpful AI assistant.\n<|im_end|>\n<|im_start|>user\nWhat's new with .NET 6?\n<|im_end|>\n<|im_start|>assistant",
  "max_tokens": 800,
  "temperature": 1,
  "frequency_penalty": 0,
  "presence_penalty": 0,
  "top_p": 0.95,
  "stop": ["<|im_end|>"]
}

The request fired with before code like follows:

POST https://spark.openai.azure.com/openai/deployments/gpt35/chat/completions?api-version=2022-12-01 HTTP/1.1
Host: spark.openai.azure.com
api-key: <api-key>
Content-Type: application/json; charset=utf-8

{"messages":[{"role":"system","content":"You are a helpful assistant."},{"role":"user","content":"What\u0027s new in .NET 6"}],"model":"gpt-3.5-turbo"}

And besides, the model deployment I created as follows:

[kayhantolga]

I recently got access to Azure OpenAI. I couldn't find a chance to try services before releasing the SDK. I will check this issue as soon as possible.
I have used this documentation as a reference, it may help you too.
https://learn.microsoft.com/en-us/azure/cognitive-services/openai/reference

[realchrisparker]

I am tracking this also as I started using this library today with Azure OpenAI service and have yet to get it working.

[kayhantolga]

It appears that Azure has decided to implement a different approach from that of the OpenAI team. Additionally, they have chosen different model names. 😮‍💨
I will address these issues in the next version.

https://learn.microsoft.com/en-gb/azure/cognitive-services/openai/how-to/chatgpt

[realchrisparker]

I did some quick test code with Azure OpenAI and this worked.

CompletionMessage completionMessage = new( message );
HttpRequestMessage request = new()
{
Method = HttpMethod.Post,
RequestUri = new Uri( $"https://{resourceName}.openai.azure.com/openai/deployments/{deploymentName}/completions?api-version=2022-12-01" ),
Content = new StringContent( completionMessage.ToString(), Encoding.UTF8, "application/json" ),
};
request.Headers.Clear();
request.Headers.Accept.Add( new MediaTypeWithQualityHeaderValue( "application/json" ) );
request.Headers.Add( "api-key", apiKey );

            var response = await _httpClient.SendAsync( request );

Not saying this is the code you should use but giving guidance on what normally works with httpclient.
The completionMessage is a quick class I created mimicking the following JSON.
{
"prompt": "",
"temperature": 0.7,
"top_p": 0.95,
"frequency_penalty": 0,
"presence_penalty": 0,
"max_tokens": 800,
"stop": [
"<|im_end|>"
]
}

[doggy8088]

@kayhantolga May I ask which Betalgo.OpenAI.GPT3 version can I use for AOAI?

Also, I think the ResourceName property is not necessary for OpenAiOptions in AOAI.

[magols]

Hello, I just got access to Azure OpenAI GPT-4 at work today and got my existing app (OpenAI API 3.5-turbo) working by changing the Program.cs to

builder.Services.AddOpenAIService(options =>
{
    options.ProviderType = ProviderType.Azure;
    options.ResourceName = builder.Configuration["OpenAIServiceOptions:ResourceName"];
    options.DeploymentId = builder.Configuration["OpenAIServiceOptions:DeploymentId"];
    options.ApiVersion = "2023-03-15-preview";
    options.ApiKey = builder.Configuration["OpenAIServiceOptions:ApiKey"];

});

and changing the model from Models.ChatGpt3_5Turbo to "gpt-4" (the 8K version) to like this.

 var stream =  Ai.ChatCompletion.CreateCompletionAsStream(new ChatCompletionCreateRequest()
                {
                    Model = "gpt-4",
                    MaxTokens = 2000,
                    Temperature = 0.9f,
                    Messages = conv.Messages.Select(m => new ChatMessage(m.Role, m.Content)).ToList()

                });

Initially I got a 404 Not Found until I spied on the Azure OpenAI Studio Chat playground and saw it was using a different API version. When plugging in the "2023-03-15-preview" as version and also using the new model string I now get chat completions!

[remixtedi]

If the client is configured to use Azure OpenAI service then Image generation doesn't work because it is not available on Azure, is there a way to have the client configured for both platforms? to use Azure for chat completions and OpenAI Dalle for images?

[kayhantolga]

Released a new version, Please have a look at the wiki page.
https://github.com/betalgo/openai/wiki/Azure-OpenAI