fix the classficiations retrieval

scrap.py

@@ -75,21 +75,24 @@ async def scrap_patent_async(client: AsyncClient, patent_url: str) -> PatentScra
     # pub_num = soup.select_one("h2#pubnum").get_text(strip=True)
     # get the h2 with id ="pubnum" and extract the text
 
-    # Classification codes (CPC + IPC, flat list, deduplicated by code)
-    code_pattern = re.compile(r'[A-Z]\d{2}[A-Z]\d+/\d+')
+    # Classification codes (CPC + IPC, flat list, deduplicated by code).
+    # Google Patents renders each code in a <span> inside a <li>. Leaf entries
+    # (no nested <ul>) carry exactly one code; parent entries are breadcrumbs.
+    leaf_code_re = re.compile(r'^[A-Z]\d{2}[A-Z]\d+/\d+$')
     classifications = []
     seen_codes: set[str] = set()
-    cls_section = soup.find("section", id="classifications")
-    if cls_section:
-        for item in cls_section.find_all("li"):
-            text = item.get_text(separator=" ", strip=True)
-            match = code_pattern.search(text)
-            if match:
-                code = match.group()
-                if code not in seen_codes:
-                    seen_codes.add(code)
-                    desc = text.replace(code, "", 1).strip()
-                    classifications.append(ClassificationCode(code=code, description=desc))
+    for li in soup.find_all("li"):
+        if li.find("ul"):
+            continue
+        span = li.find("span")
+        if not span:
+            continue
+        code = span.get_text(strip=True)
+        if leaf_code_re.match(code) and code not in seen_codes:
+            seen_codes.add(code)
+            full_text = li.get_text(separator=" ", strip=True)
+            desc = full_text.replace(code, "", 1).strip().lstrip("—").lstrip("-").strip()
+            classifications.append(ClassificationCode(code=code, description=desc))
 
     return PatentScrapResult(
         # publication_number=pub_num,