Update app.py

app.py

@@ -5,10 +5,10 @@ import torch
 import spaces
 import numpy as np
 
-# Maintain conversation history globally
+# Initialize the conversation history globally
 conversation_history = []
 
-@spaces.GPU(duration=15)
+@spaces.GPU(duration=20)
 def transcribe_and_respond(audio_file, chat_history):
     try:
         pipe = transformers.pipeline(
@@ -21,42 +21,53 @@ def transcribe_and_respond(audio_file, chat_history):
         # Load the audio file
         audio, sr = librosa.load(audio_file, sr=16000)
 
-        # Print audio properties for debugging
+        # Debug: 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 existing chat history
-
-        # Add the audio to the current turn
+        turns = chat_history.copy()  # Take the existing chat history and append user input
         turns.append({'role': 'user', 'content': '<|audio|>'})
 
-        # Debug: Print the updated turns
+        # Debug: Print the updated turns for debugging purposes
         print(f"Updated turns: {turns}")
 
         # Call the model with the updated conversation turns and audio
         output = pipe({'audio': audio, 'turns': turns, 'sampling_rate': sr}, max_new_tokens=512)
 
-        # Append the model's response to the chat history
+        # Append the model's response to the conversation history
         turns.append({'role': 'system', 'content': output})
 
         # Debug: Print the model's response
         print(f"Model output: {output}")
 
-        return turns, turns  # Return updated history to display
+        # 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
 
     except Exception as e:
         return f"Error: {str(e)}"
 
-# Chat interface setup
+# Define the Gradio interface
 iface = gr.Interface(
     fn=transcribe_and_respond,
-    inputs=[gr.Audio(sources="microphone", type="filepath", label="Your Audio (Microphone)"), gr.Chatbot(label="Conversation")],
-    outputs=gr.Chatbot(label="Conversation"),
+    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="🎙️ AI Chat with Live Transcription",
-    description="Talk to the AI through your microphone, and it will respond conversationally based on the ongoing chat. Keep the conversation going!",
+    description="Talk to the AI through your microphone, and it will respond conversationally, keeping the entire conversation visible.",
     live=True,
     allow_flagging="auto",
-    enable_queue=True
 )
 
 if __name__ == "__main__":