Refactor chat history management in process_input function of app.py to utilize dictionaries for user and assistant messages, improving clarity and structure. Update error handling to maintain consistent message formatting in chat history.

app.py

@@ -189,11 +189,15 @@ def process_input(image, audio, video, text, chat_history, voice_type, enable_au
             if not isinstance(chat_history, list):
                 chat_history = []
             
+            # Convert the current messages to the proper format
+            user_message = {"role": "user", "content": user_message_for_display}
+            assistant_message = {"role": "assistant", "content": text_response}
+            
             # Find the last incomplete message pair if it exists
-            if chat_history and isinstance(chat_history[-1], list) and len(chat_history[-1]) == 2 and chat_history[-1][1] is None:
-                chat_history[-1][1] = text_response
+            if chat_history and isinstance(chat_history[-1], dict) and chat_history[-1]["role"] == "user":
+                chat_history.append(assistant_message)
             else:
-                chat_history.append([user_message_for_display, text_response])
+                chat_history.extend([user_message, assistant_message])
             
             # Clear GPU memory after processing
             if torch.cuda.is_available():
@@ -209,7 +213,9 @@ def process_input(image, audio, video, text, chat_history, voice_type, enable_au
         except Exception as e:
             print(f"Error during generation: {str(e)}")
             error_msg = "I apologize, but I encountered an error processing your request. Please try again."
-            chat_history.append([user_message_for_display, error_msg])
+            chat_history.append(
+                {"role": "assistant", "content": error_msg}
+            )
             return chat_history, error_msg, None
             
     except Exception as e:
@@ -217,7 +223,10 @@ def process_input(image, audio, video, text, chat_history, voice_type, enable_au
         if not isinstance(chat_history, list):
             chat_history = []
         error_msg = "I apologize, but I encountered an error processing your request. Please try again."
-        chat_history.append([user_message_for_display, error_msg])
+        chat_history.extend([
+            {"role": "user", "content": user_message_for_display},
+            {"role": "assistant", "content": error_msg}
+        ])
         return chat_history, error_msg, None
 
 def user_input_to_content(user_input):
@@ -364,7 +373,7 @@ def create_demo():
             if not display_message.strip():
                 display_message = "Multimodal content"
                 
-            return [[display_message, None]]
+            return [{"role": "user", "content": display_message}]
         
         multimodal_submit.click(
             fn=prepare_multimodal_input,