Azure Function for Transcribing Speech

run.cs

#r "Microsoft.WindowsAzure.Storage"
#r "Newtonsoft.Json"
using System.Net;
using Microsoft.CognitiveServices.SpeechRecognition;

public class ReportItem{
    public string id {get;set;}
    public string audioFile {get;set;}
    public string audioText {get;set;}
}

public class Report
{
    public string id {get;set;}
    public string title {get;set;}
    public List<ReportItem> items {get; set;}
}

public static async Task<HttpResponseMessage> Run(Report report, IAsyncCollector<object> outputDocument, TraceWriter log)
{
    log.Info($"C# HTTP trigger function processing a request...");
    //var json = await req.Content.ReadAsStringAsync();
    //var report = Newtonsoft.Json.JsonConvert.DeserializeObject<report>(json);
    if (report?.items?.Count > 0) {
        
        log.Info($"{report.items.Count} items found.");
        var storageAccount = Microsoft.WindowsAzure.Storage.CloudStorageAccount.Parse(AzureStorageAccount.ConnectionString);
        var blobClient = storageAccount.CreateCloudBlobClient();
        var container = blobClient.GetContainerReference(AzureStorageAccount.ContainerName);
        
        var audioItems = from i in report.items where !string.IsNullOrWhiteSpace(i.audioFile) select i;
        foreach(var item in audioItems)
        {
            log.Info($"Processing {item.id}...");
            var audioHandler = new AudioHandler();
            item.audioText = await audioHandler.ProcessBlob(container, item.audioFile, log);   
            log.Info($"[{item.audioFile}] transcribed: {item.audioText}");        
        }

        await outputDocument.AddAsync(report);

        return  new HttpResponseMessage(HttpStatusCode.OK);
    }

    return new HttpResponseMessage(HttpStatusCode.BadRequest);
}

public abstract class AzureStorageAccount
{
    public static string ConnectionString = "DefaultEndpointsProtocol=https;AccountName=<your-account-name>;AccountKey=<your-account-key>";
    public static string ContainerName = "upload";
}

public class AudioHandler {

TaskCompletionSource<string> _tcs;
static DataRecognitionClient _dataClient;

static AudioHandler()
{
     _dataClient = SpeechRecognitionServiceFactory.CreateDataClient(
                SpeechRecognitionMode.LongDictation,
                "en-US",
                "<your-api-key>");
}

public AudioHandler()
{
  _dataClient.OnResponseReceived += responseHandler;
}

private void responseHandler(object sender, SpeechResponseEventArgs args){

      
        if (args.PhraseResponse.Results.Length == 0)
            _tcs.SetResult("ERROR: Bad audio");
        else
            _tcs.SetResult(args.PhraseResponse.Results[0].DisplayText);
        var client = sender as DataRecognitionClient;
        client.OnResponseReceived -= responseHandler;

}

public   Task<string>  ProcessBlob(Microsoft.WindowsAzure.Storage.Blob.CloudBlobContainer container, string blobName, TraceWriter log)
{
     _tcs = new TaskCompletionSource<string>();
    var mem = new System.IO.MemoryStream();
    log.Info("Ready to read blob");
    var blockBlob = container.GetBlockBlobReference(blobName);
    blockBlob.DownloadToStream(mem);
    log.Info("Blob read - size=" + mem.Length);
    mem.Position = 0;

    int bytesRead = 0;
    byte[] buffer = new byte[1024];

    try
    {
        do
        {           
            bytesRead = mem.Read(buffer, 0, buffer.Length);
            
            _dataClient.SendAudio(buffer, bytesRead);
        }
        while (bytesRead > 0);
         log.Info("Done Reading bytes");
      
    }
    finally
    {
       
        _dataClient.EndAudio();
         log.Info("Finished");
         
    }
     log.Info("Returning");
    return _tcs.Task;
}
}