Acknowledgements & Sound

The IAiConversationService provides four acknowledgement modes that control how the AI delivers responses — from silent text-only to full text-to-speech.

Acknowledgement Modes

aiService.Acknowledgement = AiAcknowledgement.Full;

Mode	Behavior
None	No audio feedback. Responses are text-only via the AiResponded event.
AudioBlip	Short sound effects play at state transitions (thinking, responding, ok, cancel, error). No text-to-speech.
LessWordy	Text-to-speech reads the response aloud. A system prompt is added requesting concise responses. A voice-chat system prompt is also injected so the AI knows it’s in a real-time voice conversation and will ask follow-up questions naturally.
Full	Text-to-speech reads the full, unmodified response aloud. A voice-chat system prompt is also injected so the AI knows it’s in a real-time voice conversation and will ask follow-up questions naturally.

Sound Effects

When Acknowledgement is set to AudioBlip, the service plays sounds at each state transition. Configure sounds with string file names and a resolver callback:

// Set the resolver that converts file names to streams
aiService.SoundResolver = name => FileSystem.OpenAppPackageFileAsync(name);

// Assign sound file names
aiService.OkSound = "ok.mp3";
aiService.ThinkSound = "think.mp3";
aiService.RespondingSound = "responding.mp3";
aiService.CancelSound = "cancel.mp3";
aiService.ErrorSound = "error.mp3";

Sound Trigger Points

Sound	When it plays
ThinkSound	AI begins processing (state → Thinking)
RespondingSound	AI starts streaming response (state → Responding)
OkSound	Request completes successfully
CancelSound	Operation is cancelled
ErrorSound	An error occurs during processing

SoundResolver

The SoundResolver property is a Func<string, Task<Stream>>? that the service calls with the sound file name to get a playable audio stream. This keeps the library platform-agnostic — you provide the resolver appropriate for your platform:

MAUI:

aiService.SoundResolver = name => FileSystem.OpenAppPackageFileAsync(name);

Blazor / ASP.NET:

var env = app.Services.GetRequiredService<IWebHostEnvironment>();
aiService.SoundResolver = name =>
    Task.FromResult<Stream>(File.OpenRead(Path.Combine(env.WebRootPath, "sounds", name)));

Note

Sound effects only play when Acknowledgement is AudioBlip and both SoundResolver is set and the corresponding sound property is non-null.

State Tracking

The service exposes its current state and fires events you can observe:

// Current state
var state = aiService.Status; // Idle, Listening, Thinking, Responding

// State change event (receives the new AiState)
aiService.StatusChanged += state =>
{
    Console.WriteLine($"State: {state}");
};

// AI response event (fired when the AI completes its response)
aiService.AiResponded += response =>
{
    if (response.Response.Text is { } text)
        Console.WriteLine($"AI: {text}");

    Console.WriteLine($"Was read aloud: {response.WasReadAloud}");
};

The AiResponse record is fired once when the AI completes its response:

• Response — the complete ChatResponse including text, tool calls, and usage details
• WasReadAloud — whether TTS was used, so your UI can decide whether to show a visual notification
• ExpectsResponse — true when the AI’s response ends with a question, indicating the service will automatically keep listening for a reply (conversation continuation)

Speech Options

Configure speech-to-text and text-to-speech options on the service:

// Speech-to-text options (culture, silence timeout, prefer on-device)
aiService.SpeechToTextOptions = new SpeechRecognitionOptions
{
    Culture = "en-US",
    SilenceTimeoutMs = 2000
};

// Text-to-speech options (culture, voice, speech rate, pitch, volume)
aiService.TextToSpeechOptions = new Shiny.Speech.TextToSpeechOptions
{
    SpeechRate = 1.2f
};

Voice Interruption

During TTS playback, the service can listen for voice interruptions using the QuietWords property:

// Default: cancel, quiet, shut up, stop, nevermind, never mind, hush
aiService.QuietWords = ["stop", "cancel", "quiet"];

// Disable entirely:
aiService.QuietWords = null;

Speech	Behavior
Quiet words	saying only a quiet word (e.g., “stop”) silences TTS and breaks the conversation loop
Other speech	saying anything else during TTS silences it and sends the new utterance to the AI as the next message

See Wake Word — Voice Interruption for more details.