Universal.xAI.Realtime.Client 1.0.0

Universal.xAI.Realtime.Client

Class library for interacting with xAI's Grok Voice Agent API over its realtime WebSocket event protocol (wss://api.x.ai/v1/realtime).

Getting Started

using Universal.xAI.Realtime.Client;

using RealtimeClient client = await RealtimeClient.ConnectAsync("YOUR_API_KEY", Models.GrokVoiceLatest);

await client.SendAsync(new SessionUpdateEvent(new RealtimeSessionConfiguration
{
    Voice = Voices.Eve,
    Instructions = "You are a helpful voice assistant.",
    TurnDetection = new TurnDetection(TurnDetectionTypes.ServerVad),
    Audio = new RealtimeAudioConfiguration
    {
        Input = new RealtimeAudioInputConfiguration { Format = RealtimeAudioFormat.Pcm(RealtimeAudioSampleRates.Hz24000) },
        Output = new RealtimeAudioOutputConfiguration { Format = RealtimeAudioFormat.Pcm(RealtimeAudioSampleRates.Hz24000) }
    },
    Tools = new RealtimeTool[]
    {
        new FunctionTool("get_weather", "Look up the current weather for a city.",
            JToken.Parse("""{ "type": "object", "properties": { "city": { "type": "string" } } }"""))
    }
}));

await foreach (RealtimeEvent serverEvent in client.ReceiveEventsAsync())
{
    switch (serverEvent)
    {
        case SessionCreatedEvent:
            await client.SendAsync(new InputAudioBufferAppendEvent(Convert.ToBase64String(micChunk)));
            break;

        case ResponseOutputAudioDeltaEvent audio:
            Play(Convert.FromBase64String(audio.Base64Audio));
            break;

        case ConversationItemInputAudioTranscriptionUpdatedEvent user:
            Console.WriteLine($"user: {user.Transcript}");
            break;

        case ResponseOutputAudioTranscriptDoneEvent assistant:
            Console.WriteLine($"assistant: {assistant.Transcript}");
            break;

        case ResponseFunctionCallArgumentsDoneEvent call:
            string result = await RunAsync(call.Name, call.Arguments);
            await client.SendAsync(new ConversationItemCreateEvent(new FunctionCallOutputItem(call.CallId, result)));
            await client.SendAsync(new ResponseCreateEvent());
            break;

        case RealtimeErrorEvent error:
            Console.WriteLine(error.Error?.Message);
            break;
    }
}

Server events are dispatched on their wire type, and anything this library does not model arrives as UnknownRealtimeEvent. Every event keeps the payload it was parsed from on RawJson.