Update app.py

app.py

@@ -452,13 +452,22 @@ def run_current_evaluation_question(
 
     try:
         answer = BasicAgent()(question, task_id)
+        had_previous_answer = bool(
+            str((answers or {}).get(str(task_id), "")).strip()
+        )
         updated = dict(answers or {})
         updated[str(task_id)] = answer
         dataframe = review_dataframe(questions, updated)
+        action_message = (
+            "A resposta anterior foi substituída pela nova resposta."
+            if had_previous_answer
+            else "A primeira resposta desta questão foi salva."
+        )
         return (
             answer,
             updated,
-            "Questão executada e resposta salva. Revise o conteúdo antes de avançar.",
+            f"Questão executada. {action_message} "
+            "Revise o conteúdo antes de avançar.",
             progress_summary(questions, updated),
             readiness_summary(dataframe),
             dataframe,
@@ -475,6 +484,69 @@ def run_current_evaluation_question(
         )
 
 
+def run_all_evaluation_questions(
+    profile: gr.OAuthProfile | None, questions: list, answers: dict
+):
+    """Executa todas as questões carregadas e reúne as respostas para revisão."""
+    if not profile:
+        dataframe = review_dataframe(questions, answers)
+        return (
+            answers or {},
+            "Faça login no Hugging Face primeiro.",
+            progress_summary(questions, answers),
+            readiness_summary(dataframe),
+            dataframe,
+        )
+
+    if not questions:
+        dataframe = review_dataframe(questions, answers)
+        return (
+            answers or {},
+            "Primeiro clique em 'Carregar as 20 questões'.",
+            progress_summary(questions, answers),
+            readiness_summary(dataframe),
+            dataframe,
+        )
+
+    try:
+        agent = BasicAgent()
+    except Exception as exc:
+        dataframe = review_dataframe(questions, answers)
+        return (
+            answers or {},
+            f"Erro ao inicializar o agente: {exc}",
+            progress_summary(questions, answers),
+            readiness_summary(dataframe),
+            dataframe,
+        )
+
+    updated = dict(answers or {})
+    failures = 0
+    for item in questions:
+        task_id = str(item.get("task_id", "")).strip()
+        question = str(item.get("question", "")).strip()
+        if not task_id or not question:
+            continue
+        try:
+            updated[task_id] = agent(question, task_id)
+        except Exception as exc:
+            updated[task_id] = f"ERROR: {exc}"
+            failures += 1
+
+    dataframe = review_dataframe(questions, updated)
+    status = (
+        f"Execução das {len(questions)} questões concluída. "
+        f"Falhas encontradas: {failures}. Revise as respostas antes de enviar."
+    )
+    return (
+        updated,
+        status,
+        progress_summary(questions, updated),
+        readiness_summary(dataframe),
+        dataframe,
+    )
+
+
 def fetch_random_question():
     """Busca somente uma questão oficial aleatória, sem executar ou enviar."""
     try:
@@ -738,9 +810,13 @@ with gr.Blocks(theme=gr.themes.Soft(), title="GAIA Agent Evaluation") as demo:
             evaluation_answers = gr.State({})
             current_task_id = gr.State("")
 
-            load_questions_button = gr.Button(
-                "Carregar as 20 questões", variant="secondary"
-            )
+            with gr.Row():
+                load_questions_button = gr.Button(
+                    "Carregar as 20 questões", variant="secondary"
+                )
+                run_all_button = gr.Button(
+                    "Executar as 20 questões", variant="primary"
+                )
             question_selector = gr.Dropdown(
                 label="Escolha a questão",
                 choices=[],
@@ -761,7 +837,7 @@ with gr.Blocks(theme=gr.themes.Soft(), title="GAIA Agent Evaluation") as demo:
             )
             with gr.Row():
                 run_current_button = gr.Button(
-                    "Executar somente esta questão", variant="primary"
+                    "Responder novamente esta questão", variant="primary"
                 )
                 save_answer_button = gr.Button("Salvar resposta revisada")
 
@@ -801,6 +877,17 @@ with gr.Blocks(theme=gr.themes.Soft(), title="GAIA Agent Evaluation") as demo:
                     results_table,
                 ],
             )
+            run_all_button.click(
+                fn=run_all_evaluation_questions,
+                inputs=[evaluation_questions, evaluation_answers],
+                outputs=[
+                    evaluation_answers,
+                    run_status,
+                    evaluation_progress,
+                    evaluation_readiness,
+                    results_table,
+                ],
+            )
             question_selector.change(
                 fn=select_evaluation_question,
                 inputs=[