Update app.py

app.py

@@ -55,6 +55,7 @@ def auto_reset_state():
     time.sleep(2)
     return None, ""  # Reset the state and clear input text
 
+
 # Function to process audio input and transcribe it
 def transcribe_function(stream, new_chunk):
     try:
@@ -72,20 +73,30 @@ def transcribe_function(stream, new_chunk):
     if max_abs_y > 0:
         y = y / max_abs_y
 
-    # Ensure stream is also at least 1-dimensional before concatenation
-    if stream is not None and len(stream) > 0:
-        stream = np.concatenate([stream, y])
-    else:
-        stream = y
+    # Concatenate the new chunk to the existing stream if available
+    stream = y if stream is None else np.concatenate([stream, y])
+
+    # Check if the stream is at least 7 seconds long
+    duration = len(stream) / sr
+    if duration >= 7:
+        # Process the audio data for transcription
+        result = pipe_asr({"array": stream, "sampling_rate": sr}, return_timestamps=False)
+        full_text = result.get("text", "")
+        print(f"Transcription: {full_text}")
+
+        # Retrieve hybrid response using Neo4j and other methods
+        response_text = retriever(full_text)
+        print(f"Response: {response_text}")
+
+        # Generate audio from the response text
+        audio_path = generate_audio_elevenlabs(response_text)
+
+        # Reset the stream for the next chunk and return the generated audio
+        return None, audio_path, full_text  # Clear the stream after processing
 
-    # Process the audio data for transcription
-    result = pipe_asr({"array": stream, "sampling_rate": sr}, return_timestamps=False)
-    full_text = result.get("text", "")
+    return stream, None, ""  # Continue streaming until 7 seconds is reached
 
-    # Start a thread to reset the state after 10 seconds
-    threading.Thread(target=auto_reset_state).start()
 
-    return stream, full_text, full_text
 
 # Function to generate a full-text search query for Neo4j
 #def generate_full_text_query(input: str) -> str: