audio change

app.py

@@ -120,72 +120,65 @@ def process_audio(audio_url):
         logger.error(f"Error processing audio: {e}")
         return None
 
-def generate_response(audio_data, message, chat_history=[]):
+def generate_response(audio_url, message, chat_history=None):
     """Generate response using the model"""
     global model, processor
-    
+
     try:
-        # Load model if not already loaded
         if model is None or processor is None:
             model, processor = load_model()
-        
-        # Define system prompt
+
+        conversation = []
+        audios = []
+
         system_prompt = "You are an expert audio engineer assisting with music production and mixing. Provide clear, specific advice on audio engineering techniques, mixing adjustments, and production decisions based on the audio samples and the user's questions. Focus on practical, actionable guidance."
-        
-        # Build conversation history
-        conversation = [
-            {"role": "system", "content": system_prompt}
-        ]
-        
-        # Add chat history (limited to last 3 turns)
-        history_limit = min(len(chat_history), 3)
-        for user_msg, bot_msg in chat_history[-history_limit:]:
-            conversation.append({"role": "user", "content": user_msg})
-            if bot_msg:  # Skip None responses
-                conversation.append({"role": "assistant", "content": bot_msg})
-        
-        # Add current message with audio
-        if audio_data is not None:
-            # First message with audio - use proper format with audio_url
+        conversation.append({"role": "system", "content": system_prompt})
+
+        if chat_history:
+            history_limit = min(len(chat_history), 3)
+            for user_msg, bot_msg in chat_history[-history_limit:]:
+                conversation.append({"role": "user", "content": user_msg})
+                if bot_msg:
+                    conversation.append({"role": "assistant", "content": bot_msg})
+
+        if audio_url:
             conversation.append({
                 "role": "user",
                 "content": [
-                    {"type": "audio", "audio_url": "https://cdn.freesound.org/previews/92/92990_321967-lq.mp3"},  # Placeholder URL
+                    {"type": "audio", "audio_url": audio_url},
                     {"type": "text", "text": message}
                 ]
             })
+
+            audio_data = process_audio(audio_url)
+            if audio_data is not None:
+                audios.append(audio_data)
+
         else:
-            # Text-only follow-up message
-            conversation.append({
-                "role": "user", 
-                "content": message
-            })
-        
-        # Apply chat template
+            conversation.append({"role": "user", "content": message})
+
         logger.info("Applying chat template")
         text = processor.apply_chat_template(
-            conversation,
-            add_generation_prompt=True,
-            tokenize=False
+            conversation, add_generation_prompt=True, tokenize=False
         )
-        
-        # Process inputs
+
         logger.info("Processing inputs")
         inputs = processor(
             text=text,
-            audios=[audio_data] if audio_data is not None else None,
+            audios=audios if audios else None,
             return_tensors="pt",
             padding=True,
             truncation=True
         )
-        
+
         # Move inputs to the same device as model
         device = next(model.parameters()).device
         inputs = {k: v.to(device) for k, v in inputs.items()}
-        
+
+        logger.info(f"Processor output: {inputs}")
+
         log_gpu_memory("Before generation")
-        
-        # Generate response
+
         logger.info("Generating response")
         with torch.no_grad():
             output = model.generate(
@@ -197,22 +190,23 @@ def generate_response(audio_data, message, chat_history=[]):
                 use_cache=True,
                 pad_token_id=processor.tokenizer.pad_token_id
             )
-        
-        # Decode only the new tokens
+
+        logger.info(f"Model output: {output}")
+
         generated_text = processor.batch_decode(
-            output[:, inputs.input_ids.shape[1]:],
+            output[:, inputs["input_ids"].shape[1]:],
             skip_special_tokens=True
         )[0]
-        
-        # Clean up
+
         del inputs, output
         gc.collect()
         if torch.cuda.is_available():
             torch.cuda.empty_cache()
-        
+
         log_gpu_memory("After generation")
-        
+
         return generated_text
+
     except Exception as e:
         logger.error(f"Error generating response: {e}")
         return f"I encountered an error while processing your request: {str(e)}"
@@ -225,7 +219,7 @@ def create_interface():
         
         # Chat state
         audio_url_state = gr.State("")
-        audio_processed_state = gr.State(None)
+        audio_loaded_state = gr.State(False)
         
         with gr.Row():
             with gr.Column(scale=2):
@@ -236,9 +230,10 @@ def create_interface():
                     msg = gr.Textbox(
                         placeholder="Ask about your mix...",
                         show_label=False,
-                        container=False
+                        container=False,
+                        interactive=False  # Disabled until audio is loaded
                     )
-                    submit_btn = gr.Button("Send", variant="primary")
+                    submit_btn = gr.Button("Send", variant="primary", interactive=False)  # Disabled until audio is loaded
                 
                 with gr.Row():
                     clear_btn = gr.Button("Clear Chat")
@@ -258,44 +253,49 @@ def create_interface():
                 audio_preview = gr.Audio(label="Audio Preview", type="filepath")
                 
                 # Status display
-                status = gr.Markdown("*Ready to assist with your mix*")
+                status = gr.Markdown("*⚠️ Please load an audio file before chatting*")
         
         # Set audio handler
         def set_audio(url):
             """Set the audio URL and process audio data"""
             if not url or not url.strip():
-                return url, None, gr.update(value=None), "*Please enter a valid audio URL*"
+                return url, gr.update(value=None), "*⚠️ Please enter a valid audio URL*", False, gr.update(interactive=False), gr.update(interactive=False)
             
             try:
                 # Try processing audio
-                audio_data = process_audio(url)
-                if audio_data is None:
-                    return url, None, gr.update(value=None), "*Failed to process audio file*"
+                process_audio(url)  # Process audio just to test if the url is valid
                 
-                # Return success
-                return url, audio_data, gr.update(value=url), "*Audio loaded successfully!*"
+                # Return success and enable chat input
+                return url, gr.update(value=url), "*✅ Audio loaded successfully! You can start chatting now.*", True, gr.update(interactive=True), gr.update(interactive=True)
             except Exception as e:
-                return url, None, gr.update(value=None), f"*Error: {str(e)}*"
+                return url, gr.update(value=None), f"*❌ Error: {str(e)}*", False, gr.update(interactive=False), gr.update(interactive=False)
         
         set_audio_btn.click(
             set_audio,
             inputs=[audio_input],
-            outputs=[audio_url_state, audio_processed_state, audio_preview, status]
+            outputs=[audio_url_state, audio_preview, status, audio_loaded_state, msg, submit_btn]
         )
         
         # Chat response handler
-        def chat_response(message, chat_history, audio_data):
+        def chat_response(message, chat_history, audio_url, audio_loaded):
             """Handle chat message and generate response"""
             if not message or not message.strip():
                 return chat_history, ""
             
+            if not audio_loaded:
+                chat_history.append((message, "Please load an audio file before chatting."))
+                return chat_history, ""
+            
             # Add user message to history
             chat_history.append((message, None))
             yield chat_history, ""
             
+            # Update status to show loading
+            yield chat_history, ""
+            
             try:
                 # Generate response
-                response = generate_response(audio_data, message, chat_history[:-1])
+                response = generate_response(audio_url, message, chat_history[:-1])
                 
                 # Update history with response
                 chat_history[-1] = (message, response)
@@ -307,20 +307,26 @@ def create_interface():
         # Connect submit button
         submit_btn.click(
             chat_response,
-            inputs=[msg, chatbot, audio_processed_state],
-            outputs=[chatbot, msg]
+            inputs=[msg, chatbot, audio_url_state, audio_loaded_state],
+            outputs=[chatbot, msg],
+            show_progress="full"  # Show loading indicator during processing
         )
         
         # Connect message box submit
         msg.submit(
             chat_response,
-            inputs=[msg, chatbot, audio_processed_state],
-            outputs=[chatbot, msg]
+            inputs=[msg, chatbot, audio_url_state, audio_loaded_state],
+            outputs=[chatbot, msg],
+            show_progress="full"  # Show loading indicator during processing
         )
         
         # Clear button
+        def clear_chat():
+            """Clear the chat and keep audio loaded state"""
+            return [], "", "*✅ Chat cleared. Audio file remains loaded.*"
+        
         clear_btn.click(
-            lambda: ([], "", "*Chat cleared*"),
+            clear_chat,
             outputs=[chatbot, msg, status]
         )