\Enable Azure Whisper model support + prompt hints"

app.py

@@ -129,9 +129,10 @@ async def transcribe_audio(
 @app.post("/api/transcribe/azure", response_model=TranscriptionResponse)
 async def transcribe_azure(
     file: UploadFile = File(...),
-    language: Optional[str] = Form(None)
+    language: Optional[str] = Form(None),
+    initial_prompt: Optional[str] = Form(None)
 ):
-    """Transcribes audio using Azure AI Speech (Whisper)"""
+    """Transcribes audio using Azure AI Speech (specifically Whisper model)"""
     key = os.getenv("AZURE_SPEECH_KEY")
     region = os.getenv("AZURE_SPEECH_REGION")
     
@@ -153,16 +154,29 @@ async def transcribe_azure(
 
         # Configure Azure Speech
         speech_config = speechsdk.SpeechConfig(subscription=key, region=region)
-        # Use your specific Whisper deployment if needed, otherwise Azure selects the best engine
-        # For Azure OpenAI Whisper specifically, you'd use a different REST path, 
-        # but Speech SDK is usually more robust for streaming.
+        
+        # MANDATORY: Set model to whisper for using the Whisper engine on Azure Speech Service
+        speech_config.set_service_property(
+            name='speech.model',
+            value='whisper',
+            strategy=speechsdk.ServicePropertyChannel.UriQueryParameter
+        )
+
         if language:
             speech_config.speech_recognition_language = language
+            
+        if initial_prompt:
+            # Pass the Whisper prompt to help with specific vocabulary or language hints
+            speech_config.set_service_property(
+                name='whisper.prompt',
+                value=initial_prompt,
+                strategy=speechsdk.ServicePropertyChannel.UriQueryParameter
+            )
         
         audio_config = speechsdk.audio.AudioConfig(filename=wav_path)
         speech_recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config, audio_config=audio_config)
 
-        logger.info(f"Azure Transcribing {file.filename} | Lang: {language}")
+        logger.info(f"Azure (Whisper) Transcribing {file.filename} | Lang: {language}")
         
         # Perform recognition
         result = speech_recognizer.recognize_once_async().get()