fix(scope-engine): Normalize UPPERCASE trigger docs to lowercase ScopeDocumentType

Critical bug fix: mandatoryDocuments in Hard-Trigger-Rules used UPPERCASE names (VVT, TOM, DSE) that never matched lowercase ScopeDocumentType keys (vvt, tom, dsi). This meant no trigger documents were ever recognized as mandatory in buildDocumentScope(). - Add normalizeDocType() mapping function with alias support (DSE→dsi, LOESCHKONZEPT→lf, DSR_PROZESS→betroffenenrechte, etc.) - Fix buildDocumentScope() to use normalized doc types - Fix estimateEffort() to use lowercase keys matching ScopeDocumentType - Add 2 tests for UPPERCASE normalization and alias resolution Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
fix(rag): Dedup check, BGB split, GewO timeout, arithmetic fix
2026-03-12 09:39:31 +01:00 · 2026-03-12 09:39:09 +01:00
4 changed files with 325 additions and 46 deletions
--- a/.gitea/workflows/rag-ingest.yaml
+++ b/.gitea/workflows/rag-ingest.yaml
@@ -78,7 +78,7 @@ jobs:
            -e "SDK_URL=http://bp-compliance-ai-sdk:8090" \
            alpine:3.19 \
            sh -c "
-              apk add --no-cache curl bash coreutils git python3 > /dev/null 2>&1
+              apk add --no-cache curl bash coreutils git python3 unzip > /dev/null 2>&1
              mkdir -p /tmp/rag-ingestion/{pdfs,repos,texts}
              cd /workspace
              if [ '${PHASE}' = 'all' ]; then
--- a/admin-compliance/lib/sdk/tests/compliance-scope-engine.test.ts
+++ b/admin-compliance/lib/sdk/tests/compliance-scope-engine.test.ts
@@ -265,6 +265,42 @@ describe('buildDocumentScope', () => {
    })
  })

+  it('normalizes UPPERCASE trigger doc names to lowercase ScopeDocumentType', () => {
+    const t = trigger('HT-test', 'L2', {
+      category: 'test',
+      mandatoryDocuments: ['VVT', 'TOM', 'DSFA'],
+    })
+    const docs = complianceScopeEngine.buildDocumentScope('L2', [t], [])
+    const vvt = docs.find((d: any) => d.documentType === 'vvt')
+    const tom = docs.find((d: any) => d.documentType === 'tom')
+    const dsfa = docs.find((d: any) => d.documentType === 'dsfa')
+    expect(vvt).toBeDefined()
+    expect(vvt!.requirement).toBe('mandatory')
+    expect(vvt!.triggeredBy).toContain('HT-test')
+    expect(tom).toBeDefined()
+    expect(tom!.requirement).toBe('mandatory')
+    expect(dsfa).toBeDefined()
+    expect(dsfa!.requirement).toBe('mandatory')
+  })
+
+  it('normalizes aliased doc names (DSE→dsi, LOESCHKONZEPT→lf)', () => {
+    const t = trigger('HT-alias', 'L2', {
+      category: 'test',
+      mandatoryDocuments: ['DSE', 'LOESCHKONZEPT', 'DSR_PROZESS'],
+    })
+    const docs = complianceScopeEngine.buildDocumentScope('L2', [t], [])
+    const dsi = docs.find((d: any) => d.documentType === 'dsi')
+    const lf = docs.find((d: any) => d.documentType === 'lf')
+    const betroffenenrechte = docs.find((d: any) => d.documentType === 'betroffenenrechte')
+    expect(dsi).toBeDefined()
+    expect(dsi!.requirement).toBe('mandatory')
+    expect(dsi!.triggeredBy).toContain('HT-alias')
+    expect(lf).toBeDefined()
+    expect(lf!.requirement).toBe('mandatory')
+    expect(betroffenenrechte).toBeDefined()
+    expect(betroffenenrechte!.requirement).toBe('mandatory')
+  })
+
  it('documents sorted: mandatory first', () => {
    const decision = complianceScopeEngine.evaluate([
      ans('data_art9', ['gesundheit']),
--- a/admin-compliance/lib/sdk/compliance-scope-engine.ts
+++ b/admin-compliance/lib/sdk/compliance-scope-engine.ts
@@ -1328,6 +1328,38 @@ export class ComplianceScopeEngine {
    return maxDepthLevel(levelFromScore, maxTriggerLevel)
  }

+  /**
+   * Normalisiert UPPERCASE Dokumenttyp-Bezeichner aus den Hard-Trigger-Rules
+   * auf die lowercase ScopeDocumentType-Schlüssel.
+   */
+  private normalizeDocType(raw: string): ScopeDocumentType | null {
+    const mapping: Record<string, ScopeDocumentType> = {
+      VVT: 'vvt',
+      TOM: 'tom',
+      DSFA: 'dsfa',
+      DSE: 'dsi',
+      AGB: 'vertragsmanagement',
+      AVV: 'av_vertrag',
+      COOKIE_BANNER: 'einwilligung',
+      EINWILLIGUNGEN: 'einwilligung',
+      TRANSFER_DOKU: 'daten_transfer',
+      AUDIT_CHECKLIST: 'audit_log',
+      VENDOR_MANAGEMENT: 'vertragsmanagement',
+      LOESCHKONZEPT: 'lf',
+      DSR_PROZESS: 'betroffenenrechte',
+      NOTFALLPLAN: 'notfallplan',
+      AI_ACT_DOKU: 'ai_act_doku',
+      WIDERRUFSBELEHRUNG: 'widerrufsbelehrung',
+      PREISANGABEN: 'preisangaben',
+      FERNABSATZ_INFO: 'fernabsatz_info',
+      STREITBEILEGUNG: 'streitbeilegung',
+      PRODUKTSICHERHEIT: 'produktsicherheit',
+    }
+    // Falls raw bereits ein gueltiger ScopeDocumentType ist
+    if (raw in DOCUMENT_SCOPE_MATRIX) return raw as ScopeDocumentType
+    return mapping[raw] ?? null
+  }
+
  /**
   * Baut den Dokumenten-Scope basierend auf Level und Triggers
   */
@@ -1338,11 +1370,18 @@ export class ComplianceScopeEngine {
  ): RequiredDocument[] {
    const requiredDocs: RequiredDocument[] = []
    const mandatoryFromTriggers = new Set<ScopeDocumentType>()
+    // Mapping: normalisierter DocType → original Rule-Strings (fuer triggeredBy Lookup)
+    const triggerDocOrigins = new Map<ScopeDocumentType, string[]>()

-    // Sammle mandatory docs aus Triggern
+    // Sammle mandatory docs aus Triggern (normalisiert)
    for (const trigger of triggers) {
      for (const doc of trigger.mandatoryDocuments) {
-        mandatoryFromTriggers.add(doc as ScopeDocumentType)
+        const normalized = this.normalizeDocType(doc)
+        if (normalized) {
+          mandatoryFromTriggers.add(normalized)
+          if (!triggerDocOrigins.has(normalized)) triggerDocOrigins.set(normalized, [])
+          triggerDocOrigins.get(normalized)!.push(doc)
+        }
      }
    }

@@ -1352,6 +1391,7 @@ export class ComplianceScopeEngine {
      const isMandatoryFromTrigger = mandatoryFromTriggers.has(docType)

      if (requirement === 'mandatory' || isMandatoryFromTrigger) {
+        const originDocs = triggerDocOrigins.get(docType) ?? []
        requiredDocs.push({
          documentType: docType,
          label: DOCUMENT_TYPE_LABELS[docType],
@@ -1361,7 +1401,7 @@ export class ComplianceScopeEngine {
          sdkStepUrl: DOCUMENT_SDK_STEP_MAP[docType],
          triggeredBy: isMandatoryFromTrigger
            ? triggers
-                .filter((t) => t.mandatoryDocuments.includes(docType as any))
+                .filter((t) => t.mandatoryDocuments.some((d) => originDocs.includes(d)))
                .map((t) => t.ruleId)
            : [],
        })
@@ -1410,29 +1450,33 @@ export class ComplianceScopeEngine {
   * Schätzt den Aufwand für ein Dokument (in Stunden)
   */
  private estimateEffort(docType: ScopeDocumentType): number {
-    const effortMap: Record<ScopeDocumentType, number> = {
-      VVT: 8,
-      TOM: 12,
-      DSFA: 16,
-      AVV: 4,
-      DSE: 6,
-      EINWILLIGUNGEN: 6,
-      LOESCHKONZEPT: 10,
-      TRANSFER_DOKU: 8,
-      DSR_PROZESS: 8,
-      NOTFALLPLAN: 12,
-      COOKIE_BANNER: 4,
-      AGB: 6,
-      WIDERRUFSBELEHRUNG: 3,
-      PREISANGABEN: 2,
-      FERNABSATZ_INFO: 4,
-      STREITBEILEGUNG: 1,
-      PRODUKTSICHERHEIT: 8,
-      AI_ACT_DOKU: 12,
-      AUDIT_CHECKLIST: 8,
-      VENDOR_MANAGEMENT: 10,
+    const effortMap: Partial<Record<ScopeDocumentType, number>> = {
+      vvt: 8,
+      tom: 12,
+      dsfa: 16,
+      av_vertrag: 4,
+      dsi: 6,
+      einwilligung: 6,
+      lf: 10,
+      daten_transfer: 8,
+      betroffenenrechte: 8,
+      notfallplan: 12,
+      vertragsmanagement: 10,
+      audit_log: 8,
+      risikoanalyse: 6,
+      schulung: 4,
+      datenpannen: 6,
+      zertifizierung: 8,
+      datenschutzmanagement: 12,
+      iace_ce_assessment: 8,
+      widerrufsbelehrung: 3,
+      preisangaben: 2,
+      fernabsatz_info: 4,
+      streitbeilegung: 1,
+      produktsicherheit: 8,
+      ai_act_doku: 12,
    }
-    return effortMap[docType] || 6
+    return effortMap[docType] ?? 6
  }

  /**
--- a/scripts/ingest-legal-corpus.sh
+++ b/scripts/ingest-legal-corpus.sh
@@ -64,6 +64,22 @@ upload_file() {
    return 0  # Don't abort script
  fi

+  # Dedup-Check: Prüfe ob regulation_id bereits in Qdrant vorhanden ist
+  local reg_id
+  reg_id=$(echo "$metadata_json" | python3 -c "import sys,json; print(json.load(sys.stdin).get('regulation_id',''))" 2>/dev/null || echo "")
+  if [[ -n "$reg_id" && -n "${QDRANT_URL:-}" ]]; then
+    local existing
+    existing=$(curl -sk --max-time 5 -X POST "${QDRANT_URL}/collections/${collection}/points/scroll" \
+      -H "Content-Type: application/json" \
+      -d "{\"filter\":{\"must\":[{\"key\":\"regulation_id\",\"match\":{\"value\":\"$reg_id\"}}]},\"limit\":1}" \
+      2>/dev/null | python3 -c "import sys,json; r=json.load(sys.stdin).get('result',{}); print(len(r.get('points',[])))" 2>/dev/null || echo "0")
+    if [[ "$existing" -gt 0 ]] 2>/dev/null; then
+      log "⏭ Skip (already in Qdrant): $label [regulation_id=$reg_id]"
+      SKIPPED=$((SKIPPED + 1))
+      return 0
+    fi
+  fi
+
  local filesize
  filesize=$(stat -f%z "$file" 2>/dev/null || stat -c%s "$file" 2>/dev/null || echo 0)
  if [[ "$filesize" -lt 100 ]]; then
@@ -76,7 +92,7 @@ upload_file() {

  # Use longer timeout for large files (>500KB)
  local curl_opts="$CURL_OPTS"
-  if [[ "$filesize" -gt 512000 ]]; then
+  if [[ "$filesize" -gt 384000 ]]; then
    curl_opts="$CURL_OPTS_LARGE"
    log "  (large file, using extended timeout)"
  fi
@@ -833,29 +849,210 @@ phase_verbraucherschutz() {
    fi
  done

-  # BGB komplett (Fernabsatz, Digitale Inhalte, Kaufrecht, AGB-Recht)
-  download_pdf \
-    "https://www.gesetze-im-internet.de/bgb/BGB.pdf" \
-    "$WORK_DIR/pdfs/BGB_full.pdf"
-  if [[ -f "$WORK_DIR/pdfs/BGB_full.pdf" ]]; then
-    upload_file "$WORK_DIR/pdfs/BGB_full.pdf" "$col" "compliance" "legal_reference" "2025" \
-      '{"regulation_id":"bgb_komplett","regulation_name_de":"BGB (komplett: AGB-Recht, Fernabsatz, Digitale Inhalte, Kaufrecht)","category":"vertragsrecht","license":"public_domain_§5_UrhG","source":"gesetze-im-internet.de"}' \
-      "BGB komplett"
+  # BGB in Teilen statt komplett (2.7MB PDF ist zu gross fuer CPU-Embeddings)
+  # gesetze-im-internet.de bietet XML-Download pro Gesetz
+  local bgb_xml="$WORK_DIR/pdfs/bgb_xml.zip"
+  curl -sL "https://www.gesetze-im-internet.de/bgb/xml.zip" -o "$bgb_xml" 2>/dev/null
+  if [[ -f "$bgb_xml" && $(stat -f%z "$bgb_xml" 2>/dev/null || stat -c%s "$bgb_xml" 2>/dev/null || echo 0) -gt 1000 ]]; then
+    local bgb_extract="$WORK_DIR/pdfs/bgb_xml"
+    mkdir -p "$bgb_extract"
+    unzip -qo "$bgb_xml" -d "$bgb_extract" 2>/dev/null || true
+
+    # Relevante BGB-Abschnitte als Text extrahieren und einzeln uploaden
+    # Die XML-Datei hat <norm> Elemente mit <metadaten><enbez>§ 305</enbez>
+    local bgb_xmlfile
+    bgb_xmlfile=$(find "$bgb_extract" -name "*.xml" | head -1)
+    if [[ -n "$bgb_xmlfile" ]]; then
+      # BGB Teil 1: AGB-Recht §§ 305-310
+      python3 -c "
+import xml.etree.ElementTree as ET, sys, re
+tree = ET.parse('$bgb_xmlfile')
+root = tree.getroot()
+ns = {'': root.tag.split('}')[0].lstrip('{') if '}' in root.tag else ''}
+text_parts = []
+capture = False
+for norm in root.iter():
+    if norm.tag.endswith('norm'):
+        enbez = norm.find('.//' + ('{' + ns[''] + '}' if ns[''] else '') + 'enbez')
+        if enbez is not None and enbez.text:
+            num = re.search(r'§\s*(\d+)', enbez.text)
+            if num:
+                n = int(num.group(1))
+                capture = 305 <= n <= 310
+        else:
+            capture = False
+        if capture:
+            for t in norm.itertext():
+                text_parts.append(t.strip())
+with open('$WORK_DIR/pdfs/BGB_AGB_305_310.txt', 'w') as f:
+    f.write('BGB AGB-Recht §§ 305-310\n\n' + '\n'.join(p for p in text_parts if p))
+" 2>/dev/null
+      if [[ -f "$WORK_DIR/pdfs/BGB_AGB_305_310.txt" && $(wc -c < "$WORK_DIR/pdfs/BGB_AGB_305_310.txt") -gt 100 ]]; then
+        upload_file "$WORK_DIR/pdfs/BGB_AGB_305_310.txt" "$col" "compliance" "legal_reference" "2025" \
+          '{"regulation_id":"bgb_agb","regulation_name_de":"BGB AGB-Recht (§§ 305-310)","category":"vertragsrecht","license":"public_domain_§5_UrhG","source":"gesetze-im-internet.de"}' \
+          "BGB AGB-Recht §§ 305-310"
+      fi
+
+      # BGB Teil 2: Fernabsatzrecht §§ 312-312k
+      python3 -c "
+import xml.etree.ElementTree as ET, sys, re
+tree = ET.parse('$bgb_xmlfile')
+root = tree.getroot()
+text_parts = []
+capture = False
+for norm in root.iter():
+    if norm.tag.endswith('norm'):
+        enbez = norm.find('.//' + ('{' + root.tag.split('}')[0].lstrip('{') + '}' if '}' in root.tag else '') + 'enbez')
+        if enbez is not None and enbez.text:
+            if re.search(r'§\s*312', enbez.text):
+                capture = True
+            elif re.search(r'§\s*31[3-9]|§\s*32', enbez.text):
+                capture = False
+        else:
+            if capture and not any(norm.itertext()):
+                capture = False
+        if capture:
+            for t in norm.itertext():
+                text_parts.append(t.strip())
+with open('$WORK_DIR/pdfs/BGB_Fernabsatz_312.txt', 'w') as f:
+    f.write('BGB Fernabsatzrecht §§ 312-312k\n\n' + '\n'.join(p for p in text_parts if p))
+" 2>/dev/null
+      if [[ -f "$WORK_DIR/pdfs/BGB_Fernabsatz_312.txt" && $(wc -c < "$WORK_DIR/pdfs/BGB_Fernabsatz_312.txt") -gt 100 ]]; then
+        upload_file "$WORK_DIR/pdfs/BGB_Fernabsatz_312.txt" "$col" "compliance" "legal_reference" "2025" \
+          '{"regulation_id":"bgb_fernabsatz","regulation_name_de":"BGB Fernabsatzrecht (§§ 312-312k)","category":"verbraucherschutz","license":"public_domain_§5_UrhG","source":"gesetze-im-internet.de"}' \
+          "BGB Fernabsatzrecht §§ 312-312k"
+      fi
+
+      # BGB Teil 3: Kaufrecht + Gewährleistung §§ 433-480
+      python3 -c "
+import xml.etree.ElementTree as ET, sys, re
+tree = ET.parse('$bgb_xmlfile')
+root = tree.getroot()
+text_parts = []
+capture = False
+for norm in root.iter():
+    if norm.tag.endswith('norm'):
+        enbez = norm.find('.//' + ('{' + root.tag.split('}')[0].lstrip('{') + '}' if '}' in root.tag else '') + 'enbez')
+        if enbez is not None and enbez.text:
+            num = re.search(r'§\s*(\d+)', enbez.text)
+            if num:
+                n = int(num.group(1))
+                capture = 433 <= n <= 480
+        else:
+            capture = False
+        if capture:
+            for t in norm.itertext():
+                text_parts.append(t.strip())
+with open('$WORK_DIR/pdfs/BGB_Kaufrecht_433_480.txt', 'w') as f:
+    f.write('BGB Kaufrecht §§ 433-480\n\n' + '\n'.join(p for p in text_parts if p))
+" 2>/dev/null
+      if [[ -f "$WORK_DIR/pdfs/BGB_Kaufrecht_433_480.txt" && $(wc -c < "$WORK_DIR/pdfs/BGB_Kaufrecht_433_480.txt") -gt 100 ]]; then
+        upload_file "$WORK_DIR/pdfs/BGB_Kaufrecht_433_480.txt" "$col" "compliance" "legal_reference" "2025" \
+          '{"regulation_id":"bgb_kaufrecht","regulation_name_de":"BGB Kaufrecht + Gewaehrleistung (§§ 433-480)","category":"verbraucherschutz","license":"public_domain_§5_UrhG","source":"gesetze-im-internet.de"}' \
+          "BGB Kaufrecht §§ 433-480"
+      fi
+
+      # BGB Teil 4: Widerrufsrecht §§ 355-361
+      python3 -c "
+import xml.etree.ElementTree as ET, sys, re
+tree = ET.parse('$bgb_xmlfile')
+root = tree.getroot()
+text_parts = []
+capture = False
+for norm in root.iter():
+    if norm.tag.endswith('norm'):
+        enbez = norm.find('.//' + ('{' + root.tag.split('}')[0].lstrip('{') + '}' if '}' in root.tag else '') + 'enbez')
+        if enbez is not None and enbez.text:
+            num = re.search(r'§\s*(\d+)', enbez.text)
+            if num:
+                n = int(num.group(1))
+                capture = 355 <= n <= 361
+        else:
+            capture = False
+        if capture:
+            for t in norm.itertext():
+                text_parts.append(t.strip())
+with open('$WORK_DIR/pdfs/BGB_Widerruf_355_361.txt', 'w') as f:
+    f.write('BGB Widerrufsrecht §§ 355-361\n\n' + '\n'.join(p for p in text_parts if p))
+" 2>/dev/null
+      if [[ -f "$WORK_DIR/pdfs/BGB_Widerruf_355_361.txt" && $(wc -c < "$WORK_DIR/pdfs/BGB_Widerruf_355_361.txt") -gt 100 ]]; then
+        upload_file "$WORK_DIR/pdfs/BGB_Widerruf_355_361.txt" "$col" "compliance" "legal_reference" "2025" \
+          '{"regulation_id":"bgb_widerruf","regulation_name_de":"BGB Widerrufsrecht (§§ 355-361)","category":"verbraucherschutz","license":"public_domain_§5_UrhG","source":"gesetze-im-internet.de"}' \
+          "BGB Widerrufsrecht §§ 355-361"
+      fi
+
+      # BGB Teil 5: Digitale Produkte §§ 327-327u
+      python3 -c "
+import xml.etree.ElementTree as ET, sys, re
+tree = ET.parse('$bgb_xmlfile')
+root = tree.getroot()
+text_parts = []
+capture = False
+for norm in root.iter():
+    if norm.tag.endswith('norm'):
+        enbez = norm.find('.//' + ('{' + root.tag.split('}')[0].lstrip('{') + '}' if '}' in root.tag else '') + 'enbez')
+        if enbez is not None and enbez.text:
+            if re.search(r'§\s*327', enbez.text):
+                capture = True
+            elif re.search(r'§\s*328', enbez.text):
+                capture = False
+        if capture:
+            for t in norm.itertext():
+                text_parts.append(t.strip())
+with open('$WORK_DIR/pdfs/BGB_Digital_327.txt', 'w') as f:
+    f.write('BGB Digitale Produkte §§ 327-327u\n\n' + '\n'.join(p for p in text_parts if p))
+" 2>/dev/null
+      if [[ -f "$WORK_DIR/pdfs/BGB_Digital_327.txt" && $(wc -c < "$WORK_DIR/pdfs/BGB_Digital_327.txt") -gt 100 ]]; then
+        upload_file "$WORK_DIR/pdfs/BGB_Digital_327.txt" "$col" "compliance" "legal_reference" "2025" \
+          '{"regulation_id":"bgb_digital","regulation_name_de":"BGB Digitale Produkte (§§ 327-327u)","category":"verbraucherschutz","license":"public_domain_§5_UrhG","source":"gesetze-im-internet.de"}' \
+          "BGB Digitale Produkte §§ 327-327u"
+      fi
+    else
+      warn "BGB XML file not found in archive"
+    fi
+  else
+    warn "BGB XML download failed"
  fi

-  # EGBGB (Muster-Widerrufsbelehrung Anlage 1+2)
-  download_pdf \
-    "https://www.gesetze-im-internet.de/bgbeg/BGBEG.pdf" \
-    "$WORK_DIR/pdfs/BGBEG.pdf"
-  if [[ -f "$WORK_DIR/pdfs/BGBEG.pdf" ]]; then
-    upload_file "$WORK_DIR/pdfs/BGBEG.pdf" "$col" "compliance" "legal_reference" "2025" \
-      '{"regulation_id":"egbgb","regulation_name_de":"EGBGB (Muster-Widerrufsbelehrung, Informationspflichten)","category":"verbraucherschutz","license":"public_domain_§5_UrhG","source":"gesetze-im-internet.de"}' \
-      "EGBGB (Muster-Widerrufsbelehrung)"
+  # EGBGB — XML statt PDF (BGBEG.pdf war leer)
+  local egbgb_xml="$WORK_DIR/pdfs/bgbeg_xml.zip"
+  curl -sL "https://www.gesetze-im-internet.de/bgbeg/xml.zip" -o "$egbgb_xml" 2>/dev/null
+  if [[ -f "$egbgb_xml" && $(stat -f%z "$egbgb_xml" 2>/dev/null || stat -c%s "$egbgb_xml" 2>/dev/null || echo 0) -gt 1000 ]]; then
+    local egbgb_extract="$WORK_DIR/pdfs/egbgb_xml"
+    mkdir -p "$egbgb_extract"
+    unzip -qo "$egbgb_xml" -d "$egbgb_extract" 2>/dev/null || true
+    local egbgb_xmlfile
+    egbgb_xmlfile=$(find "$egbgb_extract" -name "*.xml" | head -1)
+    if [[ -n "$egbgb_xmlfile" ]]; then
+      # Art. 246a EGBGB (Informationspflichten Fernabsatz) + Anlage 1+2 (Widerrufsbelehrung)
+      python3 -c "
+import xml.etree.ElementTree as ET
+tree = ET.parse('$egbgb_xmlfile')
+root = tree.getroot()
+text_parts = []
+for norm in root.iter():
+    if norm.tag.endswith('norm'):
+        # Capture all text — EGBGB is not too large
+        parts = [t.strip() for t in norm.itertext() if t.strip()]
+        if any('246' in p or 'Anlage' in p or 'Widerruf' in p or 'Muster' in p for p in parts):
+            text_parts.extend(parts)
+with open('$WORK_DIR/pdfs/EGBGB_Widerruf.txt', 'w') as f:
+    f.write('EGBGB - Informationspflichten und Muster-Widerrufsbelehrung\n\n' + '\n'.join(text_parts))
+" 2>/dev/null
+      if [[ -f "$WORK_DIR/pdfs/EGBGB_Widerruf.txt" && $(wc -c < "$WORK_DIR/pdfs/EGBGB_Widerruf.txt") -gt 100 ]]; then
+        upload_file "$WORK_DIR/pdfs/EGBGB_Widerruf.txt" "$col" "compliance" "legal_reference" "2025" \
+          '{"regulation_id":"egbgb","regulation_name_de":"EGBGB (Muster-Widerrufsbelehrung, Art. 246a + Anlage 1+2)","category":"verbraucherschutz","license":"public_domain_§5_UrhG","source":"gesetze-im-internet.de"}' \
+          "EGBGB Muster-Widerrufsbelehrung"
+      fi
+    fi
+  else
+    warn "EGBGB XML download failed"
  fi

  local after
  after=$(collection_count "$col")
-  log "Collection $col: $before → $after chunks (+$((after - before)))"
+  local diff=$(( ${after:-0} - ${before:-0} )) 2>/dev/null || diff="?"
+  log "Collection $col: ${before:-?} → ${after:-?} chunks (+${diff})"

  # =========================================================================
  # H2: EU-Recht → bp_compliance_ce
@@ -915,7 +1112,8 @@ phase_verbraucherschutz() {
  done

  after=$(collection_count "$col")
-  log "Collection $col: $before → $after chunks (+$((after - before)))"
+  local diff=$(( ${after:-0} - ${before:-0} )) 2>/dev/null || diff="?"
+  log "Collection $col: ${before:-?} → ${after:-?} chunks (+${diff})"

  # =========================================================================
  # H3: NIST Security Frameworks → bp_compliance_security
@@ -956,7 +1154,8 @@ phase_verbraucherschutz() {
  fi

  after=$(collection_count "$col")
-  log "Collection $col: $before → $after chunks (+$((after - before)))"
+  local diff=$(( ${after:-0} - ${before:-0} )) 2>/dev/null || diff="?"
+  log "Collection $col: ${before:-?} → ${after:-?} chunks (+${diff})"

  # =========================================================================
  # Summary