Throw exception when streaming is used with unsupported whisper-1 model (#537)

Throw exception when streaming is used with unsupported whisper-1 model

Author: ShivangiReja

src/Custom/Audio/AudioClient.cs

@@ -229,6 +229,8 @@ public virtual AsyncCollectionResult<StreamingAudioTranscriptionUpdate> Transcri
         Argument.AssertNotNull(audio, nameof(audio));
         Argument.AssertNotNullOrEmpty(audioFilename, nameof(audioFilename));
 
+        EnsureModelSupportsStreaming();
+
         MultiPartFormDataBinaryContent content
             = CreatePerCallTranscriptionOptions(options, stream: true)
                 .ToMultipartContent(audio, audioFilename);
@@ -245,6 +247,8 @@ public virtual AsyncCollectionResult<StreamingAudioTranscriptionUpdate> Transcri
     {
         Argument.AssertNotNullOrEmpty(audioFilePath, nameof(audioFilePath));
 
+        EnsureModelSupportsStreaming();
+
         FileStream inputStream = File.OpenRead(audioFilePath);
 
         MultiPartFormDataBinaryContent content
@@ -266,6 +270,8 @@ public virtual CollectionResult<StreamingAudioTranscriptionUpdate> TranscribeAud
         Argument.AssertNotNull(audio, nameof(audio));
         Argument.AssertNotNullOrEmpty(audioFilename, nameof(audioFilename));
 
+        EnsureModelSupportsStreaming();
+
         MultiPartFormDataBinaryContent content
             = CreatePerCallTranscriptionOptions(options, stream: true)
                 .ToMultipartContent(audio, audioFilename);
@@ -282,6 +288,8 @@ public virtual CollectionResult<StreamingAudioTranscriptionUpdate> TranscribeAud
     {
         Argument.AssertNotNullOrEmpty(audioFilePath, nameof(audioFilePath));
 
+        EnsureModelSupportsStreaming();
+
         FileStream inputStream = File.OpenRead(audioFilePath);
 
         MultiPartFormDataBinaryContent content
@@ -296,6 +304,20 @@ MultiPartFormDataBinaryContent content
         return result;
     }
 
+    private void EnsureModelSupportsStreaming()
+    {
+        if (string.Equals(_model, "whisper-1", StringComparison.OrdinalIgnoreCase))
+        {
+            string isEnabled = Environment.GetEnvironmentVariable("OPENAI_ENABLE_WHISPER_1_STREAMING");
+            if (!string.Equals(isEnabled, "true", StringComparison.OrdinalIgnoreCase))
+            {
+                throw new NotSupportedException(
+                    "The selected model 'whisper-1' does not support streaming transcription. " +
+                    "Please use a compatible model or set the environment variable 'OPENAI_ENABLE_WHISPER_1_STREAMING=true' to bypass this check.");
+            }
+        }
+    }
+
     #endregion
 
     #region TranslateAudio

tests/Audio/TranscriptionTests.cs

@@ -322,4 +322,30 @@ public async Task StreamingTranscriptionWorks(AudioSourceKind audioSourceKind)
 
         inputStream?.Dispose();
     }
+
+    [Test]
+    [TestCase(AudioSourceKind.UsingStream)]
+    [TestCase(AudioSourceKind.UsingFilePath)]
+    public void StreamingTranscriptionThrowsForWhisperModel(AudioSourceKind audioSourceKind)
+    {
+        AudioClient client = GetTestClient<AudioClient>(TestScenario.Audio_Whisper);
+        string filename = "audio_hello_world.mp3";
+        string path = Path.Combine("Assets", filename);
+
+        if (audioSourceKind == AudioSourceKind.UsingStream)
+        {
+            using FileStream inputStream = File.OpenRead(path);
+            Assert.Throws<NotSupportedException>(() =>
+            {
+                _ = client.TranscribeAudioStreamingAsync(inputStream, filename);
+            });
+        }
+        else if (audioSourceKind == AudioSourceKind.UsingFilePath)
+        {
+            Assert.Throws<NotSupportedException>(() =>
+            {
+                _ = client.TranscribeAudioStreamingAsync(path);
+            });
+        }
+    }
 }