finalize medical consultation agent backend

backend/app/agents/report_agent.py

@@ -9,6 +9,7 @@ class ReportAgent:
             "score_type": scoring_result.get("score_type", "percentage"),
             "total_score": total_score,
             "dimension_scores": dimension_scores,
+            "score_details": self._normalize_score_details(scoring_result.get("score_details", []), dimension_scores),
             "errors": self._ensure_list(scoring_result.get("errors")),
             "improvement_plan": self._ensure_list(scoring_result.get("improvement_plan")),
             "evidence_summary": self._ensure_list(scoring_result.get("evidence_summary")),
@@ -32,6 +33,30 @@ class ReportAgent:
                     "score": self._safe_float(item.get("score"), 0),
                     "max_score": self._safe_float(item.get("max_score"), 0),
                     "comment": str(item.get("comment", "")),
+                    "evidence": self._ensure_list(item.get("evidence")),
+                    "deductions": self._ensure_list(item.get("deductions")),
+                    "improvement": str(item.get("improvement", "")),
+                }
+            )
+        return normalized
+
+    def _normalize_score_details(self, raw_details: object, dimension_scores: list[dict]) -> list[dict]:
+        """评分明细校验：保留可写入 training_score_detail 的细粒度字段。"""
+        source = raw_details if isinstance(raw_details, list) and raw_details else dimension_scores
+        normalized: list[dict] = []
+        for item in source:
+            if not isinstance(item, dict):
+                continue
+            deductions = self._ensure_list(item.get("deductions"))
+            normalized.append(
+                {
+                    "rule_id": item.get("rule_id"),
+                    "dimension": str(item.get("dimension", "综合表现")),
+                    "score": self._safe_float(item.get("score"), 0),
+                    "deducted_reason": str(item.get("deducted_reason") or "；".join(str(value) for value in deductions)),
+                    "evidence_message_ids": self._ensure_list(item.get("evidence_message_ids") or item.get("evidence")),
+                    "ai_confidence": self._safe_float(item.get("ai_confidence"), 0.85),
+                    "comment": str(item.get("comment") or item.get("improvement") or ""),
                 }
             )
         return normalized

backend/app/agents/scoring_agent.py

@@ -114,8 +114,9 @@ class ScoringAgent:
             "你是医学教学问诊评分专家，只输出合法 JSON。"
             "请结合病例、问诊过程、检查申请、诊断和治疗提交进行教学评价。"
             "输出字段固定为 score_type,total_score,dimension_scores,errors,improvement_plan,"
-            "evidence_summary,guideline_refs,overall_comment。"
+            "evidence_summary,guideline_refs,overall_comment,score_details。"
             "dimension_scores 为 5-6 项，每项包含 dimension,score,max_score,comment,evidence,deductions,improvement。"
+            "score_details 对应 scoring_rules，每项包含 rule_id,dimension,score,deducted_reason,evidence_message_ids,ai_confidence,comment。"
             "evidence、deductions、improvement_plan、evidence_summary 必须是数组，每个元素一句话。"
             "errors 每项包含 title,description,severity,related_dimension。"
             "评价必须具体指出用户问了什么、申请了什么检查、诊断治疗哪里充分或不足。"
@@ -129,6 +130,7 @@ class ScoringAgent:
         for item in scoring_rules[:12]:
             compact.append(
                 {
+                    "rule_id": getattr(item, "id", None),
                     "dimension": getattr(item, "dimension", ""),
                     "competency_dimension": getattr(item, "competency_dimension", ""),
                     "score_weight": float(getattr(item, "score_weight", 0) or 0),
@@ -161,6 +163,7 @@ class ScoringAgent:
         data.setdefault("score_type", "percentage")
         data.setdefault("total_score", 0)
         data.setdefault("dimension_scores", [])
+        data.setdefault("score_details", [])
         data.setdefault("errors", [])
         data.setdefault("improvement_plan", [])
         data.setdefault("evidence_summary", [])
@@ -183,6 +186,7 @@ class ScoringAgent:
                 }
             )
         data["dimension_scores"] = normalized_dimensions or self._fallback_score("percentage", guideline_refs)["dimension_scores"]
+        data["score_details"] = self._normalize_score_details(data.get("score_details"), data["dimension_scores"])
         data["errors"] = self._normalize_errors(data.get("errors"))
         data["improvement_plan"] = self._ensure_list(data.get("improvement_plan"))
         data["evidence_summary"] = self._ensure_list(data.get("evidence_summary"))
@@ -195,6 +199,29 @@ class ScoringAgent:
             data["score_type"] = "percentage"
         return data
 
+    def _normalize_score_details(self, raw_details: object, dimension_scores: list[dict]) -> list[dict]:
+        """评分明细归一化：生成可落库的 training_score_detail 数据。"""
+        source = raw_details if isinstance(raw_details, list) and raw_details else dimension_scores
+        details = []
+        for item in source:
+            if not isinstance(item, dict):
+                continue
+            deducted_reason = item.get("deducted_reason")
+            if not deducted_reason:
+                deducted_reason = "；".join(str(value) for value in item.get("deductions", []) if value)
+            details.append(
+                {
+                    "rule_id": item.get("rule_id"),
+                    "dimension": str(item.get("dimension") or "综合表现"),
+                    "score": float(item.get("score") or 0),
+                    "deducted_reason": self._truncate(deducted_reason or "", 260),
+                    "evidence_message_ids": self._ensure_list(item.get("evidence_message_ids") or item.get("evidence")),
+                    "ai_confidence": float(item.get("ai_confidence") or 0.85),
+                    "comment": self._truncate(item.get("comment") or item.get("improvement") or "", 220),
+                }
+            )
+        return details
+
     def _normalize_errors(self, errors: object) -> list[dict]:
         """错误项归一化：转为报告可渲染的扣分项。"""
         normalized = []
@@ -320,6 +347,35 @@ class ScoringAgent:
                     "improvement": "用 SOAP 结构归纳病情，把证据、判断和计划串联起来。",
                 },
             ],
+            "score_details": [
+                {
+                    "rule_id": None,
+                    "dimension": "信息获取",
+                    "score": 20,
+                    "deducted_reason": "既往喘息史、过敏史、疫苗接种史、家属照护能力等信息不够完整。",
+                    "evidence_message_ids": ["围绕发热、咳嗽、喘息等核心症状展开问诊。"],
+                    "ai_confidence": 0.85,
+                    "comment": "完成主要症状追问，但儿科专科病史仍需补充。",
+                },
+                {
+                    "rule_id": None,
+                    "dimension": "分析推理",
+                    "score": 16,
+                    "deducted_reason": "鉴别诊断和严重程度判断未充分引用血氧、胸片和炎症指标。",
+                    "evidence_message_ids": ["主要诊断指向支气管肺炎。"],
+                    "ai_confidence": 0.84,
+                    "comment": "诊断方向基本正确，但严重程度分层需要更清晰。",
+                },
+                {
+                    "rule_id": None,
+                    "dimension": "检查利用",
+                    "score": 12,
+                    "deducted_reason": "对 SpO2、胸片异常和炎症指标的临床意义解释不够具体。",
+                    "evidence_message_ids": ["胸片、血氧或炎症指标可支持肺炎诊断和严重程度判断。"],
+                    "ai_confidence": 0.84,
+                    "comment": "关键检查申请较完整，但检查结果解释仍可细化。",
+                },
+            ],
             "errors": [
                 {
                     "title": "信息采集不够系统",
@@ -359,6 +415,13 @@ class ScoringAgent:
             }
             for item in data.get("dimension_scores", [])
         ]
+        converted["score_details"] = [
+            {
+                **item,
+                "score": round(float(item.get("score", 0)) / 20, 1),
+            }
+            for item in data.get("score_details", [])
+        ]
         return converted
 
     def _truncate(self, value: Any, limit: int) -> str: