fix: Section splitter only splits at classified headings + LLM gets full text

Two critical fixes:

1. Section splitter: Only lines that classify as a known doc_type
   (cookie, social_media, dsfa, etc.) trigger section splits.
   Random short lines ("Typen", "Funktionale Cookies") no longer
   split sections — they all had blank lines before them in the
   extracted HTML text.

2. LLM verification: Sub-section checks now pass the full document
   text to the LLM, not just the section fragment. This lets the
   LLM find content that the section splitter missed.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

backend-compliance/compliance/api/agent_doc_check_routes.py

@@ -212,6 +212,7 @@ async def _check_single_document(entry: DocCheckEntry) -> list[DocCheckResult]:
             all_results.append(main_result)
 
             # Sub-section checks (auto-detected from headings)
+            # Pass full doc_text for LLM verification fallback
             for section in sections:
                 if section["word_count"] < 100:
                     continue
@@ -219,6 +220,7 @@ async def _check_single_document(entry: DocCheckEntry) -> list[DocCheckResult]:
                     section["text"], section["doc_type"],
                     section["title"], entry.url,
                     section["word_count"],
+                    full_text=doc_text,
                 )
                 all_results.append(sub_result)
 
@@ -232,8 +234,16 @@ async def _check_single_document(entry: DocCheckEntry) -> list[DocCheckResult]:
         )]
 
 
-async def _run_checklist(text: str, doc_type: str, label: str, url: str, word_count: int = 0) -> DocCheckResult:
-    """Run checklist against text, then LLM-verify failed checks."""
+async def _run_checklist(
+    text: str, doc_type: str, label: str, url: str,
+    word_count: int = 0, full_text: str = "",
+) -> DocCheckResult:
+    """Run checklist against text, then LLM-verify failed checks.
+
+    Args:
+        full_text: Optional full document text for LLM verification.
+                   If empty, uses `text` (the section fragment).
+    """
     findings = check_document_completeness(text, doc_type, label, url)
 
     all_checks: list[CheckItem] = []
@@ -259,7 +269,7 @@ async def _run_checklist(text: str, doc_type: str, label: str, url: str, word_co
         try:
             from compliance.services.doc_checks.llm_verify import verify_failed_checks
             overturns = await verify_failed_checks(
-                text,
+                full_text or text,
                 [{"id": c.id, "label": c.label, "hint": c.hint} for c in failed],
                 label,
             )
@@ -338,31 +348,30 @@ def _split_into_sections(text: str, parent_label: str, url: str) -> list[dict]:
                 "word_count": len(sec_text.split()),
             })
 
-    prev_blank = False
     for line in lines:
         stripped = line.strip()
+        # Only split at headings that classify as a known document type.
+        # This prevents table content ("Funktionale Cookies", "Typen")
+        # from triggering section splits.
         is_heading = (
             5 < len(stripped) < 80
             and not stripped.endswith(".")
             and not stripped.endswith(",")
             and stripped[0].isupper()
-            # Require preceding blank line to distinguish real headings
-            # from table content ("Funktionale Cookies", "Session Cookies")
-            and prev_blank
         )
-        is_skip = is_heading and stripped.lower().strip() in SKIP_HEADINGS
+        classified = _classify_section(stripped) if is_heading else None
+        is_real_heading = is_heading and classified is not None
+        is_skip = is_real_heading and stripped.lower().strip() in SKIP_HEADINGS
 
-        if is_heading and not is_skip and current_heading:
+        if is_real_heading and not is_skip and current_heading:
             _save_section(current_heading, current_text)
 
-        if is_heading and not is_skip:
+        if is_real_heading and not is_skip:
             current_heading = stripped
             current_text = []
         else:
             current_text.append(line)
 
-        prev_blank = len(stripped) == 0
-
     # Last section
     if current_heading:
         _save_section(current_heading, current_text)