Update app.py

app.py

@@ -22,19 +22,23 @@ def query_model(model_name: str, messages: List[Dict[str, str]]) -> str:
         "Content-Type": "application/json"
     }
     
-    # Model-specific prompt formatting
+    # Build full conversation history for context
+    conversation = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages])
+    
+    # Model-specific prompt formatting with full history
     model_prompts = {
         "Qwen2.5-72B-Instruct": (
-            f"<|im_start|>user\n{messages[-1]['content']}<|im_end|>\n<|im_start|>assistant\n"
+            f"<|im_start|>system\nCollaborate with other experts. Previous discussion:\n{conversation}<|im_end|>\n"
+            "<|im_start|>assistant\nMy analysis:"
         ),
         "Llama3.3-70B-Instruct": (
-            "<|begin_of_text|>"
-            "<|start_header_id|>user<|end_header_id|>\n\n"
-            f"{messages[-1]['content']}<|eot_id|>"
-            "<|start_header_id|>assistant<|end_header_id|>\n\n"
+            "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n"
+            f"Build upon this discussion:\n{conversation}<|eot_id|>\n"
+            "<|start_header_id|>assistant<|end_header_id|>\nMy contribution:"
         ),
         "Qwen2.5-Coder-32B-Instruct": (
-            f"<|im_start|>user\n{messages[-1]['content']}<|im_end|>\n<|im_start|>assistant\n"
+            f"<|im_start|>system\nTechnical discussion context:\n{conversation}<|im_end|>\n"
+            "<|im_start|>assistant\nTechnical perspective:"
         )
     }
 
@@ -68,33 +72,31 @@ def query_model(model_name: str, messages: List[Dict[str, str]]) -> str:
         return f"{model_name} error: {str(e)}"
 
 def respond(message: str, history: List[List[str]]) -> str:
-    """Handle chat responses from all models"""
-    # Prepare messages in OpenAI format
+    """Handle sequential model responses with collaboration"""
     messages = [{"role": "user", "content": message}]
+    responses = []
     
-    # Create threads for concurrent model queries
-    threads = []
-    results = {}
-    
-    def get_model_response(model_name):
-        results[model_name] = query_model(model_name, messages)
-    
-    for model_name in MODEL_ENDPOINTS:
-        thread = threading.Thread(target=get_model_response, args=(model_name,))
-        thread.start()
-        threads.append(thread)
-    
-    # Wait for all threads to complete
-    for thread in threads:
-        thread.join()
+    # Define processing order
+    processing_order = [
+        "Qwen2.5-Coder-32B-Instruct",
+        "Qwen2.5-72B-Instruct", 
+        "Llama3.3-70B-Instruct"
+    ]
     
-    # Format responses from all models
-    responses = []
-    for model_name, response in results.items():
+    # Process models in sequence with accumulating context
+    for model_name in processing_order:
+        # Get current model's response
+        response = query_model(model_name, messages)
         responses.append(f"**{model_name}**:\n{response}")
+        
+        # Add model's response to message history for next model
+        messages.append({
+            "role": "assistant",
+            "content": f"{model_name} response: {response}"
+        })
     
-    # Format responses with clear separation
-    return "\n\n----------------------------------------\n\n".join(responses)
+    # Format output with collaboration timeline
+    return "\n\n→→→ Next Model Builds Upon This →→→\n\n".join(responses)
 
 # Create the Gradio interface
 chat_interface = gr.ChatInterface(