Update app.py

app.py

@@ -12,36 +12,39 @@ 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):
-    # Constrói o histórico completo com system message
-    messages = [{"role": "system", "content": system_message}]
-    messages.extend(history)
-    messages.append({"role": "user", "content": message})
+    try:
+        # Constrói o histórico completo com system message
+        messages = [{"role": "system", "content": system_message}]
+        messages.extend(history)
+        messages.append({"role": "user", "content": message})
 
-    full_prompt = "\n".join([m["content"] for m in messages])
+        full_prompt = "\n".join([m["content"] for m in messages])
 
-  # Llama 3.1
-result_main = client_main.text_generation(
-    inputs=full_prompt,
-    max_tokens=max_tokens,
-    temperature=temperature,
-    top_p=top_p
-)
-response_main = result_main[0]["generated_text"]
+        # --- Passo 1: Llama 3.1 ---
+        result_main = client_main.text_generation(
+            inputs=full_prompt,
+            max_tokens=max_tokens,
+            temperature=temperature,
+            top_p=top_p
+        )
+        response_main = result_main[0]["generated_text"]
 
-# FLAN-T5
-result_aux1 = client_aux1.text_generation(
-    inputs=f"Reformule este texto de forma clara e objetiva:\n{response_main}",
-    max_new_tokens=max_tokens
-)
-response_aux1 = result_aux1[0]["generated_text"]
+        # --- Passo 2: FLAN-T5 (reformulação) ---
+        result_aux1 = client_aux1.text_generation(
+            inputs=f"Reformule este texto de forma clara e objetiva:\n{response_main}",
+            max_new_tokens=max_tokens
+        )
+        response_aux1 = result_aux1[0]["generated_text"]
 
-# BART
-result_aux2 = client_aux2.text_generation(
-    inputs=f"Resuma este texto em 3 frases:\n{response_aux1}",
-    max_new_tokens=150
-)
-response_aux2 = result_aux2[0]["generated_text"]
+        # --- Passo 3: BART (resumo) ---
+        result_aux2 = client_aux2.text_generation(
+            inputs=f"Resuma este texto em 3 frases:\n{response_aux1}",
+            max_new_tokens=150
+        )
+        response_aux2 = result_aux2[0]["generated_text"]
 
+    except Exception as e:
+        response_aux2 = f"Erro ao gerar resposta: {e}"
 
     # Atualiza histórico do chat
     history.append({"role": "user", "content": message})
@@ -52,7 +55,10 @@ response_aux2 = result_aux2[0]["generated_text"]
 # Interface Gradio
 with gr.Blocks() as demo:
     gr.Markdown("## 🤖 Chatbot em Cascata (Llama 3.1 + FLAN-T5 + BART)")
-    system_message = gr.Textbox(value="Você é um chatbot amigável e prestativo.", label="System Message")
+    system_message = gr.Textbox(
+        value="Você é um chatbot amigável e prestativo.", 
+        label="System Message"
+    )
     chatbot = gr.Chatbot()
     msg = gr.Textbox(label="Digite sua mensagem")
     max_tokens = gr.Slider(50, 2048, 512, step=50, label="Max Tokens")
@@ -64,7 +70,11 @@ with gr.Blocks() as demo:
     def handle_submit(message, history, system_message, max_tokens, temperature, top_p):
         return respond(message, history, system_message, max_tokens, temperature, top_p)
 
-    msg.submit(handle_submit, inputs=[msg, history, system_message, max_tokens, temperature, top_p], outputs=[chatbot, history])
+    msg.submit(
+        handle_submit, 
+        inputs=[msg, history, system_message, max_tokens, temperature, top_p], 
+        outputs=[chatbot, history]
+    )
 
 if __name__ == "__main__":
     demo.launch()