feat(regulatory): 对齐附件4目录核查规则

review_agent/regulatory_review/services/structure_check.py

@@ -5,7 +5,27 @@ from review_agent.regulatory_review.schemas import Finding
 
 def run_structure_check(document_texts: dict[str, str], rule_set: dict) -> list[Finding]:
     findings: list[Finding] = []
+    combined_all_text = "\n".join(document_texts.values())
     for requirement in rule_set.get("requirements", []):
+        if requirement.get("structure_required") and not _contains_any(
+            combined_all_text,
+            [requirement.get("title", ""), *requirement.get("aliases", [])],
+        ):
+            findings.append(
+                Finding(
+                    rule_code=requirement["code"],
+                    category="structure",
+                    severity=requirement.get("severity", "medium"),
+                    title=f"申报资料目录缺少{requirement['title']}章节",
+                    detail=f"未在申报资料目录或章节标题候选中发现{requirement['title']}。",
+                    suggestion=requirement.get("suggestion", ""),
+                    evidence={
+                        "attachment4_code": requirement.get("attachment4_code"),
+                        "expected_title": requirement["title"],
+                        "aliases": requirement.get("aliases", []),
+                    },
+                )
+            )
         required_sections = requirement.get("required_sections") or []
         if not required_sections:
             continue
@@ -14,7 +34,7 @@ def run_structure_check(document_texts: dict[str, str], rule_set: dict) -> list[
             continue
         combined_text = "\n".join(matching_docs.values())
         for section in required_sections:
-            if section in combined_text:
+            if _contains_any(combined_text, [section]):
                 continue
             findings.append(
                 Finding(
@@ -39,3 +59,12 @@ def _matching_documents(document_texts: dict[str, str], keywords: list[str]) ->
         if any(str(keyword).lower() in haystack for keyword in keywords):
             result[name] = text
     return result
+
+
+def _contains_any(text: str, needles: list[str]) -> bool:
+    normalized = _normalize_title(text)
+    return any(_normalize_title(needle) in normalized for needle in needles if needle)
+
+
+def _normalize_title(value: str) -> str:
+    return "".join(str(value).lower().replace("/", "").replace("／", "").split())