Update app.py

app.py

@@ -13,26 +13,29 @@ 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:
-        # Prompt principal incluindo system_message
-        full_prompt = f"{system_message}\nUsuário: {message}"
+        # --- Monta contexto de conversa ---
+        conversation_context = "\n".join(
+            [f"Usuário: {h['content']}" if h['role']=='user' else f"Assistente: {h['content']}" for h in history]
+        )
+        full_prompt = f"{system_message}\n{conversation_context}\nUsuário: {message}"
 
-        # --- Passo 1: Llama 3.1 (conversational) ---
-        result_main = client_main.conversational(
-            prompt=full_prompt,        # string diretamente
+        # --- 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_responses[-1]  # pega a última resposta
+        response_main = result_main.generated_text
 
-        # --- Passo 2: FLAN-T5 ---
+        # --- Passo 2: FLAN-T5 (reformula o texto) ---
         result_aux1 = client_aux1.text_generation(
-            prompt=f"Reformule este texto de forma clara:\n{response_main}",
+            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 ---
+        # --- Passo 3: BART (resuma em 3 frases) ---
         result_aux2 = client_aux2.text_generation(
             prompt=f"Resuma este texto em 3 frases:\n{response_aux1}",
             max_new_tokens=150
@@ -42,7 +45,7 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
     except Exception as e:
         response_aux2 = f"Erro ao gerar resposta: {e}"
 
-    # Atualiza histórico no formato correto para o Gradio Chatbot
+    # Atualiza histórico no formato Gradio Chatbot
     history.append({"role": "user", "content": message})
     history.append({"role": "assistant", "content": response_aux2})