精简后端功能模块并补充教学互动

app/agents/scoring_agent.py

@@ -58,6 +58,162 @@ class ScoringAgent:
             }
             return data
 
+    async def score_teaching(
+        self,
+        *,
+        case: CaseBase,
+        teaching_payload: dict,
+        scoring_rules: list,
+        guideline_refs: list[dict],
+        score_type: str,
+    ) -> dict:
+        """教学互动评价：根据题目、标准答案、学生作答和评分规则生成结构化评价。"""
+        start = time.perf_counter()
+        messages = self._build_teaching_messages(case, teaching_payload, scoring_rules, guideline_refs, score_type)
+        try:
+            response = await self.llm.chat(
+                messages,
+                settings.llm_fast_model,
+                thinking_enabled=settings.llm_fast_thinking_enabled,
+                reasoning_effort=None,
+                response_format={"type": "json_object"} if settings.llm_scoring_json_response else None,
+                max_tokens=min(settings.llm_scoring_max_tokens, 1600),
+            )
+            data = json.loads(response.content)
+            data = self._normalize_score_payload(data, score_type, guideline_refs)
+            data["_llm_model"] = response.model
+            data["_latency_metrics"] = {"scoring_latency_ms": response.latency_ms, "fallback_used": False}
+            return data
+        except (AppError, json.JSONDecodeError, KeyError, TypeError, ValueError) as exc:
+            logger.warning("teaching_scoring_agent.fallback case_id=%s error=%s", case.id, exc.__class__.__name__)
+            data = self._fallback_teaching_score(score_type, guideline_refs, teaching_payload)
+            data["_llm_model"] = f"local-fallback-{settings.llm_fast_model}"
+            data["_latency_metrics"] = {
+                "scoring_latency_ms": int((time.perf_counter() - start) * 1000),
+                "fallback_used": True,
+                "fallback_reason": exc.__class__.__name__,
+            }
+            return data
+
+    def _build_teaching_messages(
+        self,
+        case: CaseBase,
+        teaching_payload: dict,
+        scoring_rules: list,
+        guideline_refs: list[dict],
+        score_type: str,
+    ) -> list[dict]:
+        """教学评分提示词：只传教学互动评价需要的病例、题目、答案和评分规则。"""
+        payload = {
+            "score_type": score_type,
+            "case": {
+                "case_id": case.id,
+                "title": case.title,
+                "chief_complaint": case.chief_complaint,
+                "description": self._truncate(case.description, 320),
+                "knowledge_points": case.knowledge_points or [],
+                "key_points": case.key_points or [],
+            },
+            "teaching": teaching_payload,
+            "scoring_rules": self._compact_scoring_rules(scoring_rules),
+            "guidelines": self._compact_guidelines(guideline_refs),
+        }
+        system = (
+            "你是医学教学互动评价专家，只输出合法 JSON。"
+            "请根据病例、教学目标、选择题、标准答案、解析文本、学生作答和评分规则生成教学评价。"
+            "输出字段固定为 score_type,total_score,dimension_scores,errors,improvement_plan,"
+            "evidence_summary,guideline_refs,overall_comment,score_details。"
+            "dimension_scores 包含 知识掌握、临床推理、检查理解、治疗决策、人文沟通 维度，"
+            "每项包含 dimension,score,max_score,comment,evidence,deductions,improvement。"
+            "score_details 对应 scoring_rules 或题目维度，每项包含 rule_id,dimension,score,"
+            "deducted_reason,evidence_message_ids,ai_confidence,comment。"
+            "必须指出答对题目、答错题目、错误原因、下一步学习重点。"
+            "本评价仅用于医学教学训练，不替代真实临床诊疗。"
+        )
+        return [{"role": "system", "content": system}, {"role": "user", "content": json.dumps(payload, ensure_ascii=False)}]
+
+    def _fallback_teaching_score(self, score_type: str, guideline_refs: list[dict], teaching_payload: dict) -> dict:
+        """教学评分兜底：LLM 不可用时按选择题正确率生成稳定评价结构。"""
+        results = teaching_payload.get("answer_results") or []
+        total = len(results)
+        correct = sum(1 for item in results if item.get("is_correct"))
+        accuracy = correct / total if total else 0
+        total_score = round(accuracy * 100, 1) if total else 0
+        incorrect = [item for item in results if not item.get("is_correct")]
+        incorrect_titles = [f"{item.get('question_id')}: {item.get('stem', '')}" for item in incorrect[:5]]
+        data = {
+            "score_type": "percentage",
+            "total_score": total_score,
+            "dimension_scores": [
+                {
+                    "dimension": "知识掌握",
+                    "score": round(total_score * 0.35, 1),
+                    "max_score": 35,
+                    "comment": f"共 {total} 题，答对 {correct} 题。",
+                    "evidence": [f"正确率 {round(accuracy * 100, 1)}%"],
+                    "deductions": incorrect_titles,
+                    "improvement": "复习错题对应知识点和病例解析。",
+                },
+                {
+                    "dimension": "临床推理",
+                    "score": round(total_score * 0.25, 1),
+                    "max_score": 25,
+                    "comment": "根据选择题表现评估临床判断链路。",
+                    "evidence": [item.get("stem", "") for item in results[:3]],
+                    "deductions": incorrect_titles,
+                    "improvement": "把题目选项与病例主诉、体征和检查结果逐项对应。",
+                },
+                {
+                    "dimension": "检查理解",
+                    "score": round(total_score * 0.15, 1),
+                    "max_score": 15,
+                    "comment": "重点关注检查项目与病情严重程度判断。",
+                    "evidence": teaching_payload.get("scoring_focus", "").split("、")[:3],
+                    "deductions": [],
+                    "improvement": "理解血氧、胸片和炎症指标在肺炎评估中的作用。",
+                },
+                {
+                    "dimension": "治疗决策",
+                    "score": round(total_score * 0.15, 1),
+                    "max_score": 15,
+                    "comment": "根据题目表现评估治疗原则掌握情况。",
+                    "evidence": teaching_payload.get("teaching_goal", "").split("、")[:3],
+                    "deductions": [],
+                    "improvement": "复习抗感染、平喘、氧合监测和风险预案。",
+                },
+                {
+                    "dimension": "人文沟通",
+                    "score": round(total_score * 0.10, 1),
+                    "max_score": 10,
+                    "comment": "教学互动中需继续强化家属沟通和健康教育。",
+                    "evidence": ["教学互动题包含沟通与健康教育相关内容。"],
+                    "deductions": [],
+                    "improvement": "向家属说明病情、观察指标、复诊指征和用药注意事项。",
+                },
+            ],
+            "score_details": [],
+            "errors": [
+                {
+                    "title": "教学题目答题错误",
+                    "description": "；".join(incorrect_titles) if incorrect_titles else "暂无明显错题。",
+                    "severity": "medium" if incorrect_titles else "low",
+                    "related_dimension": "知识掌握",
+                }
+            ],
+            "improvement_plan": [
+                "复盘错题解析，明确每个选项与病例证据的对应关系。",
+                "把病例中的主诉、体征、检查和治疗原则整理成一条临床推理链。",
+                "针对血氧、胸片、炎症指标和医患沟通进行专项复习。",
+            ],
+            "evidence_summary": [
+                f"教学互动共提交 {total} 题，答对 {correct} 题。",
+                "评分依据包括题目标准答案、解析文本、教学目标和评分规则。",
+            ],
+            "guideline_refs": guideline_refs,
+            "overall_comment": f"本次教学互动正确率为 {round(accuracy * 100, 1)}%，请结合错题解析继续巩固病例关键知识点。",
+        }
+        return self._convert_to_five_point(data) if score_type == "five_point" else data
+
     def _build_messages(
         self,
         session: TrainingSession,