[Bug]: Error code 4, StatusCode: 1006 about cognitive-services-speech-sdk-js

[glharper]

@matt-carbon, this is a separate issue that updating to v1.32 may fix. If you still see this after updating, please open a new issue.

[glharper]

@znwhite Thank you for using JS Speech SDK, and writing this issue up. I copied your sample code and was able to get a result with correct credentials. Can you tell me more about the Azure resource you're getting the key/region pair from? Is it a Speech Service, or is it another kind of Azure resource? For the Azure Speech SDK, you have to use keys specifically from a Speech Service.

[matt-carbon]

I'm seeing the same issue with some cadence reported in our logging system (but I'm unable to reproduce on my own device). We're using eastus. We get the token just fine. Transcription works normally for most users. But every once in awhile we get errors. Here's one of the logs we have:

errorCode | 4
errorDetails | Unable to contact server. StatusCode: 0, undefined Reason:
origError | 0
sessionId | 1B262BF6100A49E09CBA075231C371FD

Which comes directly form our code that logs this error:

ApptAudioTranscriber.recognizer.canceled = (
        s: Recognizer,
        e: SpeechRecognitionCanceledEventArgs,
      ) => {
        logError(e.reason, "ApptAudioTranscriber canceled", {
          data: {
            errorDetails: e.errorDetails,
            errorCode: e.errorCode,
            sessionId: e.sessionId,
          },
        });
      };

Is there any other information we can debug to understand why it go canceled? We are seeing some other errors that get logged at different times but may be correlated. Specifically Another turn started before current completed. and The number of parallel requests exceeded the number of allowed concurrent transcriptions..

Happy to open separate issues if needed.