Sync primary diagnosis from discharge table

患者首页处理/数据可视化网页端/app/main.py

@@ -146,9 +146,6 @@ FIELD_GROUPS: list[dict[str, Any]] = [
         "fields": [
             ("outpatient_diagnosis", "门急诊诊断", "text", None),
             ("outpatient_diagnosis_code", "门急诊诊断编码", "text", None),
-            ("primary_diagnosis", "主要诊断", "text", None),
-            ("primary_diagnosis_code", "主要诊断编码", "text", None),
-            ("primary_admission_condition", "主要诊断入院病情", "text", None),
             ("discharge_diagnoses", "出院诊断", "json", None),
             ("injury_poisoning_external_cause", "损伤中毒外部原因", "text", None),
             ("injury_poisoning_code", "损伤中毒疾病编码", "text", None),
@@ -385,9 +382,6 @@ AI_REDACT_PATTERNS = (
 AI_UPLOAD_FIELDS = [
     "outpatient_diagnosis",
     "outpatient_diagnosis_code",
-    "primary_diagnosis",
-    "primary_diagnosis_code",
-    "primary_admission_condition",
     "discharge_diagnoses",
     "operations",
     "pathology_diagnosis",
@@ -1808,6 +1802,35 @@ def json_changed_fields(field: str, old_value: Any, new_value: Any) -> list[dict
     ]
 
 
+def main_diagnosis_row_index(rows: list[Any]) -> int:
+    for index, row in enumerate(rows):
+        if isinstance(row, dict) and str(row.get("诊断类别") or "").strip() == "主要诊断":
+            return index
+    return 0
+
+
+def main_diagnosis_from_discharge_diagnoses(value: Any) -> dict[str, str]:
+    rows = value if isinstance(value, list) else []
+    if not rows:
+        return {"primary_diagnosis": "", "primary_diagnosis_code": "", "primary_admission_condition": ""}
+    index = main_diagnosis_row_index(rows)
+    row = rows[index] if index < len(rows) and isinstance(rows[index], dict) else {}
+    return {
+        "primary_diagnosis": str(row.get("出院诊断") or row.get("诊断名称") or "").strip(),
+        "primary_diagnosis_code": str(row.get("疾病编码") or row.get("诊断编码") or "").strip(),
+        "primary_admission_condition": str(row.get("入院病情") or "").strip(),
+    }
+
+
+def sync_primary_diagnosis_updates(updates: dict[str, Any], before: dict[str, Any]) -> None:
+    if "discharge_diagnoses" not in updates:
+        return
+    derived = main_diagnosis_from_discharge_diagnoses(updates.get("discharge_diagnoses"))
+    for field, value in derived.items():
+        if comparable(before.get(field)) != comparable(value):
+            updates[field] = value
+
+
 def build_record_updates(
     record_id: int,
     fields: dict[str, Any],
@@ -1838,6 +1861,8 @@ def build_record_updates(
     if "inpatient_no" in updates and not str(updates["inpatient_no"] or "").strip():
         raise HTTPException(status_code=400, detail="患者号不能为空")
 
+    sync_primary_diagnosis_updates(updates, before)
+
     manual_note = manual_note.strip()
     if manual_note:
         current = before.get("review_notes") or []
@@ -2343,7 +2368,7 @@ def build_ai_prompt(record: dict[str, Any], pdf_context: dict[str, Any], privacy
 5. 只有仍需要复核或纠正的问题，classification 返回 "problem"，并写入 remaining_issues。
 6. 如果手术表格中能看到“手术及操作编码”列但对应单元格为空，写“编码栏可见但为空白”，不要写“编码区域未在首页显示”；单元格本来为空且无需补录时不要因此判为 problem。
 7. 手术操作名称可能因为换行被结构化解析截断；如果 PDF定位文本或局部截图中显示完整多行名称，请把完整名称放入 suggested_updates。
-8. 门急诊诊断编码只能用于 outpatient_diagnosis_code；不要把门急诊诊断编码复制到 primary_diagnosis_code 或 discharge_diagnoses[].疾病编码，除非出院诊断表格对应“疾病编码”单元格本身清楚显示该编码。
+8. 门急诊诊断编码只能用于 outpatient_diagnosis_code；主要诊断请修正 discharge_diagnoses 中“诊断类别=主要诊断”的行，不要把门急诊诊断编码复制到 discharge_diagnoses[].疾病编码，除非出院诊断表格对应“疾病编码”单元格本身清楚显示该编码。
 9. remaining_issues 只写当前文档复核人应该特别注意的内容；不要写如何修改，不要重复 suggested_updates，不要写“无需补录/无需处理/首页原貌”这类已判定无问题的说明。
 10. 不要编造 PDF 中看不见的内容，不要输出置信度。
 
@@ -2353,7 +2378,7 @@ def build_ai_prompt(record: dict[str, Any], pdf_context: dict[str, Any], privacy
   "summary": "一句话结论，60字以内",
   "method": "AI视觉核验：PDF文本定位+局部截图，对照复核定位和结构化字段",
   "suggested_updates": [
-    {{"field": "字段名或路径，例如 primary_diagnosis_code 或 operations[0].麻醉方式", "value": "PDF中应写入的值", "reason": "20字以内"}}
+    {{"field": "字段名或路径，例如 discharge_diagnoses[0].疾病编码 或 operations[0].麻醉方式", "value": "PDF中应写入的值", "reason": "20字以内"}}
   ],
   "remaining_issues": ["AI觉得有必要复核的内容，最多3条；无则返回空数组"],
   "evidence": [
@@ -2660,8 +2685,22 @@ def ai_outpatient_code_leak(path: tuple[str, int | None, str | None], value: Any
     return bool(target_is_diagnosis_code)
 
 
+PRIMARY_DIAGNOSIS_AI_TARGETS = {
+    "primary_diagnosis": "出院诊断",
+    "主要诊断": "出院诊断",
+    "主要诊断名称": "出院诊断",
+    "primary_diagnosis_code": "疾病编码",
+    "主要诊断编码": "疾病编码",
+    "primary_admission_condition": "入院病情",
+    "主要诊断入院病情": "入院病情",
+}
+
+
 def ai_update_path(field_text: str, item: dict[str, Any]) -> tuple[str, int | None, str | None] | None:
     text = str(field_text or "").strip()
+    for keyword, column in sorted(PRIMARY_DIAGNOSIS_AI_TARGETS.items(), key=lambda entry: len(entry[0]), reverse=True):
+        if keyword in text:
+            return ("discharge_diagnoses", -1, column)
     if text in EDITABLE_FIELDS:
         return (text, None, None)
     for name, meta in FIELD_META.items():
@@ -2726,6 +2765,10 @@ def ai_suggested_updates(result: dict[str, Any], before: dict[str, Any]) -> tupl
                 continue
             rows = old_value if isinstance(old_value, list) else []
             rows = [dict(row) if isinstance(row, dict) else {} for row in rows]
+            if index == -1 and field == "discharge_diagnoses":
+                if not rows:
+                    rows = [{"诊断类别": "主要诊断", "出院诊断": "", "疾病编码": "", "入院病情": ""}]
+                index = main_diagnosis_row_index(rows)
             if index < 0 or index >= len(rows):
                 continue
             compare_old = rows[index].get(key)
@@ -2750,6 +2793,7 @@ def ai_suggested_updates(result: dict[str, Any], before: dict[str, Any]) -> tupl
             }
         )
 
+    sync_primary_diagnosis_updates(updates, before)
     return updates, changed_fields