Update app.py

app.py

@@ -5,11 +5,8 @@ import torch
 import spaces
 import numpy as np
 
-# Initialize the conversation history globally
-conversation_history = []
-
-@spaces.GPU(duration=120)
-def transcribe_and_respond(audio_file, chat_history):
+@spaces.GPU(duration=20)
+def transcribe_and_respond(audio_file):
     try:
         pipe = transformers.pipeline(
             model='sarvamai/shuka_v1',
@@ -21,54 +18,35 @@ def transcribe_and_respond(audio_file, chat_history):
         # Load the audio file
         audio, sr = librosa.load(audio_file, sr=16000)
 
-        # Debug: Print audio properties for debugging
+        # Print audio properties for debugging
         print(f"Audio dtype: {audio.dtype}, Audio shape: {audio.shape}, Sample rate: {sr}")
 
-        # Prepare conversation turns
-        turns = chat_history.copy()  # Take the existing chat history and append user input
-        turns.append({'role': 'user', 'content': '<|audio|>'})
+        turns = [
+            {'role': 'system', 'content': 'Respond naturally and informatively.'},
+            {'role': 'user', 'content': '<|audio|>'}
+        ]
 
-        # Debug: Print the updated turns for debugging purposes
-        print(f"Updated turns: {turns}")
+        # Debug: Print the initial turns
+        print(f"Initial turns: {turns}")
 
-        # Call the model with the updated conversation turns and audio
+        # Call the model with the audio and prompt
         output = pipe({'audio': audio, 'turns': turns, 'sampling_rate': sr}, max_new_tokens=512)
 
-        # Append the model's response to the conversation history
-        turns.append({'role': 'system', 'content': output})
-
-        # Debug: Print the model's response
+        # Debug: Print the final output from the model
         print(f"Model output: {output}")
 
-        # Format the chat history for Gradio's Chatbot
-        chat_history_for_display = []
-        for turn in turns:
-            if turn['role'] == 'user':
-                chat_history_for_display.append(("User", "🗣️ (Spoken Audio)"))
-            else:
-                chat_history_for_display.append(("AI", turn['content']))
-
-        return chat_history_for_display, turns  # Return the formatted chat history for display and the updated history
+        return output
 
     except Exception as e:
-        return f"Error: {str(e)}", chat_history  # Ensure history is returned even on error
+        return f"Error: {str(e)}"
 
-# Define the Gradio interface
 iface = gr.Interface(
     fn=transcribe_and_respond,
-    inputs=[
-        gr.Audio(sources="microphone", type="filepath", label="Your Audio (Microphone)"), 
-        gr.State([])  # Hidden state to maintain conversation history
-    ],
-    outputs=[
-        gr.Chatbot(label="Conversation History"),  # Display the conversation
-        gr.State([])  # Hidden state to keep track of the updated conversation history
-    ],
-    title="Shuka demo",
-    description="shuka live demo",
-    live=True,  # Enable live mode for real-time interaction
-    allow_flagging="auto",
-#    enable_queue=True
+    inputs=gr.Audio(sources="microphone", type="filepath"),
+    outputs="text",
+    title="Live Transcription and Response",
+    description="Speak into your microphone, and the model will respond naturally and informatively.",
+    live=True
 )
 
 if __name__ == "__main__":