Update app.py

app.py

@@ -24,12 +24,9 @@ def query_model(model_name: str, messages: List[Dict[str, str]]) -> Generator[st
     # 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|>system\nCollaborate with other experts. Previous discussion:\n{conversation}<|im_end|>\n"
-            "<|im_start|>assistant\nMy analysis:"
-        ),
+    # System prompt configuration
+    system_prompts = {
+        "Qwen2.5-72B-Instruct": "Collaborate with other experts. Previous discussion:\n{conversation}",
         "Llama3.3-70B-Instruct": (
             "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n"
             f"Build upon this discussion:\n{conversation}<|eot_id|>\n"
@@ -44,11 +41,17 @@ def query_model(model_name: str, messages: List[Dict[str, str]]) -> Generator[st
     client = InferenceClient(base_url=endpoint, token=HF_API_KEY)
 
     try:
+        messages = [
+            {"role": "system", "content": system_prompts[model_name].format(conversation=conversation)},
+            {"role": "user", "content": "Continue the expert discussion"}
+        ]
+        
         stream = client.chat.completions.create(
-            messages=[{"role": "system", "content": model_prompts[model_name]}],
+            messages=messages,
             stream=True,
             max_tokens=2048,
-            temperature=0.7,
+            temperature=0.5,
+            top_p=0.7
         )
 
         for chunk in stream: