.Net: Bug: OpenAIPromptExecutionSettings do not override kernel modelId

[mateuszmigas]

Describe the bug Function OpenAIPromptExecutionSettings do not override kernel modelId.

To Reproduce Steps to reproduce the behavior: Run the following code:

var kernel = Kernel.CreateBuilder()
    .AddOpenAIChatCompletion("gpt-3.5-turbo", configuration["API_KEY"], null,
        null, httpClient)
    .Build();
var gpt4 = kernel.CreateFunctionFromPrompt(..., new OpenAIPromptExecutionSettings { ModelId = "gpt-4" });    
var result = await gpt4.InvokeAsync(kernel, new KernelArguments { { "input", userInput  } });

2. Check traffic with OpenAI:

   "model": "gpt-3.5-turbo-0125",
   "choices": [
   {
     "index": 0,
     "message": {
       "role": "assistant",
       "content": "Hello there, greeting from GPT-4! How may I assist you today?"
     },
     "logprobs": null,
     "finish_reason": "stop"
   }
   ]

3. It uses kernel model instead of execution settings model.

Expected behavior It should use execution settings model.

Screenshots N/A

Platform

• OS: Mac, not sure about Windows
• IDE: N/A
• Language: C#
• Source: Nuget: Microsoft.SemanticKernel.dll 1.15.0.0

Additional context Code to run:

var kernel = Kernel.CreateBuilder()
    .AddOpenAIChatCompletion("gpt-3.5-turbo", configuration["API_KEY"], null,
        null, httpClient)
    .Build();

var gpt4Def = """
             Say hi from GPT-4
             """;
var gpt4 = kernel.CreateFunctionFromPrompt(gpt4Def, new OpenAIPromptExecutionSettings { ModelId = "gpt-4" });

var gpt3TurboDef = """
                Say hi from GPT-3.5 Turbo
                """;
var gpt3Turbo = kernel.CreateFunctionFromPrompt(gpt3TurboDef, new OpenAIPromptExecutionSettings { ModelId = "gpt-3.5-turbo" });

do
{
    Console.WriteLine("Please enter prompt:");
    var userInput = Console.ReadLine();
    var result1 = await gpt3Turbo.InvokeAsync(kernel, new KernelArguments { { "input", userInput } });
    var result2 = await gpt4.InvokeAsync(kernel, new KernelArguments { { "input", userInput  } });
    Console.WriteLine(result1.GetValue<string>()!);
    Console.WriteLine(result2.GetValue<string>()!);
} while (true);

[dmytrostruk]

Hi @mateuszmigas , thanks for reporting this issue!

You can define different functions which will use different models, but as first step you need to register all available models in your app in Kernel, and then during function initialization or invocation you can select which model to use based on registered models in Kernel:

// Build a kernel with multiple chat completion services
var builder = Kernel.CreateBuilder()
    .AddOpenAIChatCompletion(
        modelId: "gpt-4-0125-preview",
        apiKey: TestConfiguration.OpenAI.ApiKey)
    .AddOpenAIChatCompletion(
        modelId: "gpt-3.5-turbo-0125",
        apiKey: TestConfiguration.OpenAI.ApiKey);

Kernel kernel = builder.Build();

var prompt = "Hello AI, what can you do for me?";

var result = await kernel.InvokePromptAsync(prompt, new(new OpenAIPromptExecutionSettings { ModelId = "gpt-3.5-turbo-0125" }));

[mateuszmigas]

@dmytrostruk thanks for answer!