Update app.py

app.py

@@ -74,7 +74,7 @@ def query_model(model_name: str, messages: List[Dict[str, str]]) -> str:
     except Exception as e:
         return f"{model_name} error: {str(e)}"
 
-def respond(message: str, history: List[List[str]], session_id: str) -> str:
+def respond(message: str, history: List[List[str]], session_id: str) -> Generator[str, None, None]:
     """Handle sequential model responses with context preservation"""
     # Load or initialize session
     session = session_manager.load_session(session_id)
@@ -97,9 +97,8 @@ def respond(message: str, history: List[List[str]], session_id: str) -> str:
         "content": message
     })
     
-    responses = []
-    
-    # First model response
+    # First model
+    yield "🔵 Qwen2.5-Coder-32B-Instruct is thinking..."
     response1 = query_model("Qwen2.5-Coder-32B-Instruct", messages)
     session["history"].append({
         "timestamp": datetime.now().isoformat(),
@@ -108,9 +107,10 @@ def respond(message: str, history: List[List[str]], session_id: str) -> str:
         "content": response1
     })
     messages.append({"role": "assistant", "content": f"Qwen2.5-Coder-32B-Instruct: {response1}"})
-    responses.append(f"🔵 **Qwen2.5-Coder-32B-Instruct**\n{response1}")
+    yield f"🔵 **Qwen2.5-Coder-32B-Instruct**\n{response1}"
     
-    # Second model response
+    # Second model
+    yield f"🔵 **Qwen2.5-Coder-32B-Instruct**\n{response1}\n\n🟣 Qwen2.5-72B-Instruct is thinking..."
     response2 = query_model("Qwen2.5-72B-Instruct", messages)
     session["history"].append({
         "timestamp": datetime.now().isoformat(),
@@ -119,9 +119,10 @@ def respond(message: str, history: List[List[str]], session_id: str) -> str:
         "content": response2
     })
     messages.append({"role": "assistant", "content": f"Qwen2.5-72B-Instruct: {response2}"})
-    responses.append(f"🟣 **Qwen2.5-72B-Instruct**\n{response2}")
+    yield f"🔵 **Qwen2.5-Coder-32B-Instruct**\n{response1}\n\n🟣 **Qwen2.5-72B-Instruct**\n{response2}"
     
-    # Final model response
+    # Final model
+    yield f"🔵 **Qwen2.5-Coder-32B-Instruct**\n{response1}\n\n🟣 **Qwen2.5-72B-Instruct**\n{response2}\n\n🟡 Llama3.3-70B-Instruct is thinking..."
     response3 = query_model("Llama3.3-70B-Instruct", messages)
     session["history"].append({
         "timestamp": datetime.now().isoformat(),
@@ -130,50 +131,42 @@ def respond(message: str, history: List[List[str]], session_id: str) -> str:
         "content": response3
     })
     messages.append({"role": "assistant", "content": f"Llama3.3-70B-Instruct: {response3}"})
-    responses.append(f"🟡 **Llama3.3-70B-Instruct**\n{response3}")
     
     # Save final session state
     session_manager.save_session(session_id, session)
     
-    # Return responses
-    return "\n\n".join(responses)
-
-# Custom CSS for styling
-css = """
-.message { padding: 15px; margin: 10px 0; border-radius: 10px; }
-.assistant { background: #f8fafc; border-left: 4px solid #3b82f6; }
-.user { background: #eff6ff; border-left: 4px solid #60a5fa; }
-.model-name { font-weight: 600; color: #1e40af; margin-bottom: 8px; }
-.thinking { color: #6b7280; font-style: italic; }
-"""
+    # Return final combined response
+    yield f"🔵 **Qwen2.5-Coder-32B-Instruct**\n{response1}\n\n🟣 **Qwen2.5-72B-Instruct**\n{response2}\n\n🟡 **Llama3.3-70B-Instruct**\n{response3}"
 
 # Create the Gradio interface
-demo = gr.ChatInterface(
-    fn=respond,
-    title="Multi-LLM Collaboration Chat",
-    description="Experience collaborative AI thinking with three powerful language models",
-    examples=[
-        ["Explain how quantum computing works"],
-        ["Write a Python function to find prime numbers"],
-    ],
-    additional_inputs=[gr.State(session_manager.create_session)],
-    chatbot=gr.Chatbot(
-        height=600,
-        show_label=False,
-        bubble_full_width=False,
-        show_copy_button=True,
-        container=True,
-        sanitize_html=False,
-        render_markdown=True
-    ),
-    theme=gr.themes.Soft(
-        primary_hue="blue",
-        secondary_hue="indigo",
-        neutral_hue="slate",
-        font=("Inter", "sans-serif"),
-    ),
-    css=css,
-)
+with gr.Blocks() as demo:
+    gr.Markdown("## Multi-LLM Collaboration Chat")
+    
+    with gr.Row():
+        session_id = gr.State(session_manager.create_session)
+        new_session = gr.Button("🔄 New Session")
+    
+    chatbot = gr.Chatbot(height=600)
+    msg = gr.Textbox(label="Message")
+    
+    def on_new_session():
+        new_id = session_manager.create_session()
+        return new_id, []
+    
+    def user(message, history, session_id):
+        return "", history + [[message, None]]
+    
+    def bot(history, session_id):
+        if history and history[-1][1] is None:
+            message = history[-1][0]
+            for response in respond(message, history[:-1], session_id):
+                history[-1][1] = response
+                yield history
+    
+    msg.submit(user, [msg, chatbot, session_id], [msg, chatbot]).then(
+        bot, [chatbot, session_id], [chatbot]
+    )
+    new_session.click(on_new_session, None, [session_id, chatbot])
 
 if __name__ == "__main__":
     demo.launch(share=True)