Update app.py

app.py

@@ -1,6 +1,6 @@
 import os
 import gradio as gr
-from huggingface_hub import InferenceClient
+from huggingface_hub import InferenceClient, Conversation
 
 # Pega o token do Hugging Face dos Secrets
 HF_TOKEN = os.environ.get("HF_TOKEN")
@@ -16,14 +16,10 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
         # Prompt principal incluindo system_message
         full_prompt = f"{system_message}\nUsuário: {message}"
 
-        # --- Passo 1: Llama 3.1 ---
-        result_main = client_main.text_generation(
-            prompt=full_prompt,
-            max_new_tokens=max_tokens,
-            temperature=temperature,
-            top_p=top_p
-        )
-        response_main = result_main.generated_text
+        # --- Passo 1: Llama 3.1 (conversational) ---
+        conversation = Conversation(full_prompt)
+        result_main = client_main.conversational(conversation)
+        response_main = result_main.generated_responses[-1]  # pega a última resposta
 
         # --- Passo 2: FLAN-T5 ---
         result_aux1 = client_aux1.text_generation(