Update app.py

app.py

@@ -13,28 +13,32 @@ client_aux2 = InferenceClient(token=HF_TOKEN, model="facebook/bart-large-cnn")
 # Função principal de resposta
 def respond(message, history, system_message, max_tokens, temperature, top_p):
     try:
-        # --- Passo 1: Llama 3.1 (chat) ---
-        chat_messages = [{"role": "system", "content": system_message}]
+        # --- Passo 1: Llama 3.1 via ProxyClientChat ---
+        chat = client_main.chat  # objeto de chat, não chamável
+        chat.clear_messages()  # limpa mensagens anteriores do objeto (opcional)
+
+        # Adiciona mensagens do histórico
+        chat.add_message("system", system_message)
         for h in history:
-            chat_messages.append({"role": h['role'], "content": h['content']})
-        chat_messages.append({"role": "user", "content": message})
+            chat.add_message(h['role'], h['content'])
+        chat.add_message("user", message)
 
-        result_main = client_main.chat(
-            messages=chat_messages,
+        # Gera resposta
+        response_main_obj = chat.send_message(
             max_new_tokens=max_tokens,
             temperature=temperature,
             top_p=top_p
         )
-        response_main = result_main.choices[0].message["content"]
+        response_main = response_main_obj.content  # pega o texto gerado
 
-        # --- Passo 2: FLAN-T5 (reformula o texto) ---
+        # --- Passo 2: FLAN-T5 (reformulação) ---
         result_aux1 = client_aux1.text_generation(
             prompt=f"Reformule este texto de forma clara e concisa:\n{response_main}",
             max_new_tokens=max_tokens
         )
         response_aux1 = result_aux1.generated_text
 
-        # --- Passo 3: BART (resuma em 3 frases) ---
+        # --- Passo 3: BART (resumo em 3 frases) ---
         result_aux2 = client_aux2.text_generation(
             prompt=f"Resuma este texto em 3 frases:\n{response_aux1}",
             max_new_tokens=150