Sample Kernel Http Server /api/skills/funskill/invoke/joke - BadRequest logprobs, best_of and echo parameters are not available on gpt-35-turbo model.

[wanermiranda]

Describe the bug Considering the AzureOpenAI account and the gpt-35-turbo model. When executing the /api/skills/funskill/invoke/joke the kernel is returning the following message.

InvalidRequest: The request is not valid, HTTP status: 400 - Detail: logprobs, best_of and echo parameters are not available on gpt-35-turbo model. Please remove the parameter and try again. For more details, see https://go.microsoft.com/fwlink/?linkid=2227346.
Status: 400 (BadRequest)
ErrorCode: BadRequest

Content:
{"error":{"code":"BadRequest","message":"logprobs, best_of and echo parameters are not available on gpt-35-turbo model. Please remove the parameter and try again. For more details, see https://go.microsoft.com/fwlink/?linkid=2227346."}}

Headers:
apim-request-id: REDACTED
Strict-Transport-Security: REDACTED
X-Content-Type-Options: REDACTED
x-ms-region: REDACTED
Date: Fri, 21 Apr 2023 03:15:59 GMT
Content-Length: 236
Content-Type: application/json

I saw there was a similar problem in the past when using the GPT.

To Reproduce

curl 'http://localhost:7071/api/skills/funskill/invoke/joke' \
  -H 'Accept: */*' \
  -H 'Accept-Language: en-GB,en;q=0.9,en-US;q=0.8,pt;q=0.7' \
  -H 'Connection: keep-alive' \
  -H 'Content-Type: text/plain;charset=UTF-8' \
  -H 'Origin: http://localhost:3000' \
  -H 'Referer: http://localhost:3000/' \
  -H 'Sec-Fetch-Dest: empty' \
  -H 'Sec-Fetch-Mode: cors' \
  -H 'Sec-Fetch-Site: same-site' \
  -H 'User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/112.0.0.0 Safari/537.36 Edg/112.0.1722.48' \
  -H 'sec-ch-ua: "Chromium";v="112", "Microsoft Edge";v="112", "Not:A-Brand";v="99"' \
  -H 'sec-ch-ua-mobile: ?0' \
  -H 'sec-ch-ua-platform: "Windows"' \
  -H 'x-ms-sk-completion-backend: 0' \
  -H 'x-ms-sk-completion-endpoint: XXXXXXXX' \
  -H 'x-ms-sk-completion-key: XXXXXXXXX' \
  -H 'x-ms-sk-completion-model: completion-model' \
  --data-raw '{"value":"clippy","inputs":[{"key":"style","value":"Bill & Ted"}]}' \
  --compressed

Screenshots If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: Windows 11
• IDE: Visual Studio 2022
• NuGet Package Version

   <TargetFramework>net6.0</TargetFramework>
   <AzureFunctionsVersion>v4</AzureFunctionsVersion>
   <OutputType>Exe</OutputType>
   <LangVersion>10</LangVersion>
   <Nullable>enable</Nullable>
   <ImplicitUsings>disable</ImplicitUsings>
   <IsPackable>false</IsPackable>
 </PropertyGroup>

 <ItemGroup>
   <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.Http" Version="3.0.13" />
   <PackageReference Include="Microsoft.Azure.Functions.Worker.Sdk" Version="1.9.0" OutputItemType="Analyzer" />
   <PackageReference Include="Microsoft.Azure.Functions.Worker" Version="1.13.0" />
   <PackageReference Include="Microsoft.Graph" Version="4.54.0" />
 </ItemGroup>

[craigomatic]

Thank you for the report, we are in the process of cleaning up the earlier samples to default to chat completion endpoints. Currently they expect text completion endpoints.

Two options:

1. Use text-davinci-003

OR

2. Change this line in KernelHttpServer: https://github.com/microsoft/semantic-kernel/blob/77259db3e352db64be4b5f08f5c11a391a101e60/samples/dotnet/KernelHttpServer/SemanticKernelFactory.cs#L55

The change would look like this:

c.AddAzureChatCompletionService(config.CompletionConfig.ServiceId, config.CompletionConfig.DeploymentOrModelId,
                        config.CompletionConfig.Endpoint,
                        config.CompletionConfig.Key, 
                        true);

[wanermiranda]

Thank you @craigomatic :)

[wanermiranda]

The second option when adding a new parameter, the signature does not match the code. error CS1501: No overload for method 'AddAzureTextCompletionService' takes 5 arguments

[craigomatic]

Try with AddAzureChatCompletionService

[wanermiranda]

My mistake, tks. It worked like charm! Do you want to close this or will you we use it to create a PR?

[craigomatic]

Lets close this for now, we will take care of all the samples in a future effort. Thanks for trying SK :)