/// <summary>
/// Establishes socket connected with the server
/// </summary>
/// <returns>Whether the socket can attempt to connect. Usually returns true when the service has not been authenticated</returns>
public bool Connect()
{
if (Status == BingStatus.Idle || !Auth.IsValid())
{
TryLogError("Could not connect. Not authorized");
return(false);
}
m_Socket.CustomHeaders = new Dictionary <string, string> {
{ "X-ConnectionId", BingUtils.GetNewRequestID() },
{ "Authorization", "Bearer " + Auth.Token }
};
TryLog("Connecting...");
Status = BingStatus.Connecting;
m_Socket.ConnectAsync(
() => {
TryLog("Connected");
Status = BingStatus.Connected;
},
exception => {
TryLogError("Could not connect: " + exception);
Status = BingStatus.Authenticated;
}
);
return(true);
}
void OnSocketMessage(object sender, MessageEventArgs e)
{
var parser = new MessageParser(e.Data);
var message = parser.GetObject();
// All events are fired from the main thread
Dispatcher.Enqueue(() => {
OnGetMessage.TryInvoke(message);
switch (parser.Path.ToLower())
{
case "turn.start":
OnTurnStart.TryInvoke(message as TurnStartMessage);
break;
case "turn.end":
m_Socket.Send(BingUtils.TurnEndAcknowledgement());
m_Buffer = new byte[0];
OnTurnEnd.TryInvoke(message as TurnEndMessage);
Status = BingStatus.Disconnecting;
m_Socket.CloseAsync(() => {
Connect();
});
break;
case "speech.enddetected":
OnSpeechEndDetected.TryInvoke(message as SpeechEndDetectedMessage);
break;
case "speech.phrase":
OnSpeechPhrase.TryInvoke(message as SpeechPhraseMessage);
break;
case "speech.hypothesis":
OnSpeechHypothesis.TryInvoke(message as SpeechHypothesisMessage);
break;
case "speech.startdetected":
OnSpeechStartDetected.TryInvoke(message as SpeechStartDetectedMessage);
break;
case "speech.fragment":
OnSpeechFragment.TryInvoke(message as SpeechFragmentMessage);
break;
}
});
}
/// <summary>
/// Adds the audio sample to the packet for streaming.
/// </summary>
/// <param name="sample">The audio sample to be streamed</param>
/// <returns>True if this frame lead to completion of a packet that got sent</returns>
public bool Stream(byte[] sample)
{
if (sample.Length == 0)
{
return(false);
}
if (Status < BingStatus.Connected)
{
return(false);
}
if (Status == BingStatus.Connected)
{
Status = BingStatus.Streaming;
// Before we stream the audio, we must once send the speechConfig
var config = BingUtils.GetConfig(m_RequestId);
m_Socket.Send(config);
}
var freeBytes = 8192 - m_Header.Length;
m_Buffer = m_Buffer.Concat(sample).ToArray();
// If we can not add more to the buffer
if (m_Buffer.Length + sample.Length < freeBytes)
{
return(true);
}
// We require 8192 bytes to data. After adding header and audio data, we may still have some bytes left
// We fill them up with empty/blank bytes and send via the socket. The blank is small enough that the service will
// not think of it as microphone silence
var blankLen = 8192 - m_Header.Length - m_Buffer.Length;
var packet = m_Header.Concat(m_Buffer).Concat(new byte[blankLen]).ToArray();
try { m_Socket.Send(packet); } catch { }
m_Buffer = new byte[0];
return(true);
}
/// <summary>
/// Request authentication so that the client can communicate with the server
/// </summary>
/// <param name="key">The Bing Speech to Text API key</param>
/// <param name="callback">Bool callback for whether the authentication was successful</param>
public void Authenticate(string key, Action <string> onSuccess, Action <Exception> onFailure)
{
TryLog("Authenticating");
Auth = new BingAuthorization(key);
Status = BingStatus.Authenticating;
Auth.FetchToken(
token => {
TryLog("Authorized: " + token);
Status = BingStatus.Authenticated;
m_RequestId = BingUtils.GetNewRequestID();
m_Header = BingUtils.GetHeader(m_RequestId);
onSuccess.TryInvoke(token);
},
exception => {
TryLogError(exception);
Status = BingStatus.Idle;
onFailure.TryInvoke(exception);
}
);
}