improve ui

src/agent.py

@@ -6,7 +6,6 @@ from langchain_openai import ChatOpenAI
 import os
 
 class AgentOutput(BaseModel):
-    final_response: str = Field(description="The final response to the user.", default="")
     output_audio_files: list[str] = Field(description="The output audio files.", default=[])
 
 system_prompt = """You are an expert Audio Processing Assistant with specialized capabilities in audio manipulation, analysis, and editing. Your primary purpose is to help users with audio-related tasks and provide knowledgeable assistance in the audio domain.
@@ -144,4 +143,8 @@ class AudioAgent:
             {"messages": messages},
         )
 
-        return res["structured_response"]
+        output = {}
+        output["final_response"] = res["messages"][-1].content
+        output["output_audio_files"] = res["structured_response"].output_audio_files
+
+        return output

src/ui.py

@@ -89,8 +89,8 @@ async def bot_response(history, audio_file_urls, custom_history):
         result = await agent.run_agent(user_message, input_files, custom_history[:-1])
 
         # Extract the final response and audio files from the result
-        final_response = result.final_response
-        output_audio_files = result.output_audio_files
+        final_response = result["final_response"]
+        output_audio_files = result["output_audio_files"]
         
         # Add assistant response to history with output files
         history.append({
@@ -146,15 +146,12 @@ def create_interface():
                     show_share_button=False
                 )
 
-                with gr.Row(equal_height=True):
-                    msg = gr.Textbox(
-                        label="Describe what you want to do?",
-                        placeholder="e.g., 'Remove filler words and improve audio quality''",
-                        lines=3,
-                        scale=6
-                    )
-                    send_btn = gr.Button("Ask", variant="primary", scale=1, size="lg")
-
+                msg = gr.Textbox(
+                    label="Describe what you want to do?",
+                    placeholder="e.g., 'Remove filler words and improve audio quality''",
+                    lines=3,
+                    submit_btn=True
+                )
             
             with gr.Column(scale=1):
                 # Model Configuration
@@ -230,18 +227,7 @@ def create_interface():
             [chatbot, audio_urls_state, custom_history_state], 
             [chatbot, output_audio_files, custom_history_state]
         )
-        
-        send_btn.click(
-            handle_submit,
-            [msg, audio_files, chatbot, custom_history_state, model_name, temperature, api_key], 
-            [msg, audio_files, chatbot, custom_history_state], 
-            queue=False
-        ).then(
-            handle_bot_response, 
-            [chatbot, audio_urls_state, custom_history_state], 
-            [chatbot, output_audio_files, custom_history_state]
-        )
-            
+                    
     return interface
 
 if __name__ == "__main__":