Enable Semantic Voice Activity detection using the SDK

[utkarsh-jha]

I was in process of migrating my application RealtimeClient layer from Websocket to OpenAI .NET SDK,

When configuring the session turn detection, I found the method TurnDetectionOptions.CreateSemanticVoiceActivityTurnDetectionOptions to be internal and because of which I can't use the following configuration.

ConversationSessionOptions sessionOptions = new()
{
    Instructions = _sessionConfig.SessionParams.Instruction,
    Voice = "alloy",
    InputAudioFormat = "pcm16",
    OutputAudioFormat = "pcm16",
    InputTranscriptionOptions = new InputTranscriptionOptions
    {
        Model = "whisper-1"
    },
    Temperature = 0.6f,
    // library currently has internal implementaions of 
    //TurnDetectionOptions.CreateSemanticVoiceActivityTurnDetectionOptions(
    //    eagerness: EagernessLevel.Medium,
    //    createResponse: true,
    //    interruptResponse: true),
};

[ShivangiReja]

Hi @utkarsh-jha, thanks for raising this! The method TurnDetectionOptions.CreateSemanticVoiceActivityTurnDetectionOptions was initially marked internal during early development to simplify the previous release process.

I’ve opened a PR to make this method public: #545. Really appreciate you surfacing this! Feel free to share any other feedback or scenarios you're exploring. We’re always happy to hear how the SDK is being used.