fix(exam): 修复出题批卷 API 契约一致性问题

P0:
- 修复 V2 降级路径响应格式，fallback 输出经过 enrich + dedup + balance
- 补充主观题 format example，要求 LLM 生成 scoring_points 评分标准

P1:
- 补充多选题、判断题 format example，明确 answer 为列表格式
- 统一批卷响应格式，所有题型增加 grading_status 字段
- 出题数量不足时返回 requested_types/actual_types/warnings
- 修复 _call_llm 未配置时抛异常而非静默返回 0 分

P2:
- API 层增加 question_types 和 difficulty 入参校验
- 修复 generate_questions_from_content 方法名和参数类型 bug

exam_pkg/grader.py

@@ -97,8 +97,13 @@ def grade_objective(answer: Dict) -> Dict:
         "question_id": answer.get('question_id'),
         "score": max_score if correct else 0,
         "max_score": max_score,
-        "correct": correct,
-        "feedback": f"正确答案: {correct_answer}" if not correct else "正确！"
+        "grading_status": "success",
+        "details": {
+            "correct": correct,
+            "student_answer": student_answer,
+            "correct_answer": correct_answer,
+            "feedback": f"正确答案: {correct_answer}" if not correct else "正确！"
+        }
     }
 
 
@@ -119,7 +124,8 @@ def grade_fill_blank(answer: Dict) -> Dict:
             "question_id": answer.get('question_id'),
             "score": 0,
             "max_score": max_score,
-            "details": {"error": "答案格式错误"}
+            "grading_status": "failed",
+            "details": {"error": "答案格式错误，correct_answers 或 student_answer 为空"}
         }
 
     # 计算每空分数
@@ -149,9 +155,11 @@ def grade_fill_blank(answer: Dict) -> Dict:
         "question_id": answer.get('question_id'),
         "score": round(total_score, 1),
         "max_score": max_score,
+        "grading_status": "success",
         "details": {
             "blank_scores": blank_scores,
-            "correct_answers": correct_answers
+            "total_blanks": len(correct_answers),
+            "correct_blanks": sum(1 for s in blank_scores if s > 0)
         }
     }
 
@@ -243,22 +251,38 @@ class AnswerGrader:
             }
 
             # 收集结果
-            for future in as_completed(future_to_qid, timeout=60):
-                qid = future_to_qid[future]
-                try:
-                    result = future.result(timeout=15)
-                    results_map[qid] = result
-                except Exception as e:
-                    # 失败时返回默认结果
-                    results_map[qid] = {
-                        "question_id": qid,
-                        "score": 0,
-                        "max_score": next(
-                            (a.get('max_score', 10) for a in questions if a.get('question_id') == qid),
-                            10
-                        ),
-                        "error": str(e)
-                    }
+            try:
+                for future in as_completed(future_to_qid, timeout=60):
+                    qid = future_to_qid[future]
+                    try:
+                        result = future.result(timeout=15)
+                        results_map[qid] = result
+                    except Exception as e:
+                        # 失败时返回默认结果
+                        results_map[qid] = {
+                            "question_id": qid,
+                            "score": 0,
+                            "max_score": next(
+                                (a.get('max_score', 10) for a in questions if a.get('question_id') == qid),
+                                10
+                            ),
+                            "grading_status": "failed",
+                            "details": {"error": str(e)}
+                        }
+            except TimeoutError:
+                # 超时：为未完成的任务设置失败状态
+                for future, qid in future_to_qid.items():
+                    if qid not in results_map:
+                        results_map[qid] = {
+                            "question_id": qid,
+                            "score": 0,
+                            "max_score": next(
+                                (a.get('max_score', 10) for a in questions if a.get('question_id') == qid),
+                                10
+                            ),
+                            "grading_status": "failed",
+                            "details": {"error": "批阅超时"}
+                        }
 
     @retry(times=2, delay=1)
     def _grade_subjective(self, answer: Dict) -> Dict:
@@ -315,8 +339,7 @@ class AnswerGrader:
     def _call_llm(self, prompt: str) -> str:
         """调用本地 LLM"""
         if not self.client:
-            # 无 LLM 客户端，返回默认评分
-            return json.dumps({"score": 0, "overall_feedback": "LLM 未配置"})
+            raise ValueError("LLM 未配置，无法进行主观题评分")
 
         messages = [
             {"role": "system", "content": "你是一个专业的阅卷老师，请严格按照JSON格式输出评分结果。"},
@@ -347,6 +370,7 @@ class AnswerGrader:
                 "question_id": answer.get('question_id'),
                 "score": score,
                 "max_score": max_score,
+                "grading_status": "success",
                 "details": {
                     "scoring_breakdown": result.get('scoring_breakdown', []),
                     "highlights": result.get('highlights', []),
@@ -360,6 +384,7 @@ class AnswerGrader:
                 "question_id": answer.get('question_id'),
                 "score": 0,
                 "max_score": max_score,
+                "grading_status": "failed",
                 "details": {"error": "评分结果解析失败"}
             }