fix(dkv): extract invoice number/date from PDF, rename export files to RG-DKV format

- Parser now extracts Rechnungsnummer (DD/DDDDDDDDD/DDD) and Rechnungsdatum
  from PDF text, so filename doesn't rely on email subject
- Export filename changed from DKV_YYYY-MM_... to RG-DKV-{nr}-{YYMMDD}.xlsx
  e.g. RG-DKV-26-650869002-002-260331.xlsx
- Subject fallback now also matches slash-separated invoice numbers (26/NNN/NNN)
- writeAndPrune simplified to accept baseName instead of separate fields
- Validation regex and prune prefix updated to match new RG-DKV- pattern

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

apps/api/src/dkv/dkv-export.service.ts

@@ -13,7 +13,7 @@ const MAX_EXPORT_FILES = 10;
 /**
  * Glob pattern for DKV export files — used for prune selection.
  */
-const DKV_FILE_PREFIX = 'DKV_';
+const DKV_FILE_PREFIX = 'RG-DKV-';
 const DKV_FILE_SUFFIX = '.xlsx';
 
 /**
@@ -128,7 +128,7 @@ export class DkvExportService {
    * @param invoiceMonth - Month string "YYYY-MM", e.g. "2026-04"
    * @returns The filename that was written (relative to user-files/)
    */
-  writeAndPrune(buffer: Buffer, rechnungsnummer: string, invoiceMonth: string): string {
+  writeAndPrune(buffer: Buffer, baseName: string): string {
     // Ensure the directory exists
     if (!fs.existsSync(this.userFilesDir)) {
       fs.mkdirSync(this.userFilesDir, { recursive: true });
@@ -136,7 +136,7 @@ export class DkvExportService {
     }
 
     // Build filename server-side — NEVER from request input (T-07-09 path-traversal mitigation)
-    const filename = `DKV_${invoiceMonth}_${rechnungsnummer}.xlsx`;
+    const filename = `${baseName}.xlsx`;
     const filePath = path.join(this.userFilesDir, filename);
 
     // Write the file

apps/api/src/dkv/dkv-parser.service.ts

@@ -31,7 +31,11 @@ export class DkvParserService {
    * @returns Array of vehicle blocks with transactions
    * @throws Error with generic message if no vehicle blocks are found
    */
-  async parsePdf(buffer: Buffer): Promise<DkvVehicleBlock[]> {
+  async parsePdf(buffer: Buffer): Promise<{
+    vehicles: DkvVehicleBlock[];
+    rechnungsnummer: string | null;
+    rechnungsdatum: string | null;
+  }> {
     let text: string;
     try {
       text = await this.extractText(buffer);
@@ -52,7 +56,30 @@ export class DkvParserService {
       `DKV PDF parsed: ${vehicles.length} vehicle(s), ${totalTx} transaction(s)`,
     );
 
-    return vehicles;
+    return {
+      vehicles,
+      rechnungsnummer: this.extractRechnungsnummer(text),
+      rechnungsdatum: this.extractRechnungsdatum(text),
+    };
+  }
+
+  /** Extract first DKV invoice number (DD/DDDDDDDDD/DDD) from PDF text. */
+  private extractRechnungsnummer(text: string): string | null {
+    const m = text.match(/\b(\d{2}\/\d{9}\/\d{3})\b/);
+    return m ? m[1] : null;
+  }
+
+  /** Extract invoice date ("DD.MM.YYYY") appearing near "Rechnungsdatum" in PDF text. */
+  private extractRechnungsdatum(text: string): string | null {
+    // "Rechnungsdatum" is followed (within ~200 chars) by a date string
+    const m = text.match(/Rechnungsdatum[:\s\t]*(\d{2}\.\d{2}\.\d{4})/);
+    if (m) return m[1];
+    // Fallback: second line after "Rechnungsdatum:" marker
+    const idx = text.indexOf('Rechnungsdatum');
+    if (idx === -1) return null;
+    const after = text.slice(idx, idx + 200);
+    const dm = after.match(/(\d{2}\.\d{2}\.\d{4})/);
+    return dm ? dm[1] : null;
   }
 
   // ─── Private: PDF text extraction ──────────────────────────────────────────

apps/api/src/dkv/dkv.service.ts

@@ -342,12 +342,12 @@ export class DkvService {
     vehicleFormatString: string,
   ): Promise<void> {
     // D-10: Up to 3 parse retries
-    let vehicles: DkvVehicleBlock[] | null = null;
+    let parseResult: Awaited<ReturnType<typeof this.parser.parsePdf>> | null = null;
     let parseError: string | null = null;
 
     for (let attempt = 1; attempt <= 3; attempt++) {
       try {
-        vehicles = await this.parser.parsePdf(pdfBuffer);
+        parseResult = await this.parser.parsePdf(pdfBuffer);
         parseError = null;
         break;
       } catch (err) {
@@ -358,12 +358,12 @@ export class DkvService {
       }
     }
 
-    if (!vehicles) {
+    if (!parseResult) {
       // Record parse failure in history (D-10)
       await this.prisma.dkvInvoiceHistory.create({
         data: {
           tenantId,
-          rechnungsnummer: this._extractInvoiceNumber(email.subject, email.uid),
+          rechnungsnummer: this._buildRechnungsnummer(null, email.subject, email.uid),
           anzahlFahrzeuge: 0,
           anzahlTransaktionen: 0,
           status: 'Fehler',
@@ -373,9 +373,13 @@ export class DkvService {
       return;
     }
 
-    // Extract invoice metadata for filename + history record
+    const { vehicles, rechnungsnummer: rgNr, rechnungsdatum: rgDat } = parseResult;
-    const rechnungsnummer = this._extractInvoiceNumber(email.subject, email.uid);
+
-    const invoiceMonth = this._extractInvoiceMonth(vehicles);
+    // Build filename: RG-DKV-{nr}-{date}.xlsx
+    const rechnungsnummer = this._buildRechnungsnummer(rgNr, email.subject, email.uid);
+    const datePart = rgDat ? _formatDateYYMMDD(rgDat) : this._extractInvoiceMonth(vehicles).replace('-', '');
+    const exportBaseName = `RG-DKV-${rechnungsnummer}-${datePart}`;
+
     const anzahlTransaktionen = vehicles.reduce((s, v) => s + v.transactions.length, 0);
 
     // Build export rows — look up drivers from vehicle master
@@ -383,7 +387,7 @@ export class DkvService {
 
     // Generate xlsx buffer and write to user-files/ (DkvExportService)
     const xlsxBuffer = this.exporter.buildExcelBuffer(exportRows);
-    const exportFilename = this.exporter.writeAndPrune(xlsxBuffer, rechnungsnummer, invoiceMonth);
+    const exportFilename = this.exporter.writeAndPrune(xlsxBuffer, exportBaseName);
 
     // D-16: SMTP send with 3-retry exponential backoff
     let smtpStatus: 'Verarbeitet' | 'Versand fehlgeschlagen' = 'Verarbeitet';
@@ -552,7 +556,7 @@ export class DkvService {
       filename.includes('/') ||
       filename.includes('\\') ||
       filename.includes('..') ||
-      !/^DKV_[\w\-]+\.xlsx$/.test(filename)
+      !/^(RG-DKV-|DKV_)[\w\-]+\.xlsx$/.test(filename)
     ) {
       throw new BadRequestException('Invalid export filename');
     }
@@ -624,17 +628,19 @@ export class DkvService {
   }
 
   /**
-   * Extract a DKV invoice number from the email subject line.
+   * Build a safe rechnungsnummer string for use in filenames.
-   * Pattern: DD-DDDDDDDDD-DDD (e.g. "26-651566449-001")
+   * Priority: PDF-extracted number → email subject → uid fallback.
-   * Falls back to "email-{uid}" when no match is found.
+   * All slashes and non-safe chars are replaced so the value is path-safe.
    */
-  private _extractInvoiceNumber(subject: string, uid: number | string): string {
+  private _buildRechnungsnummer(
-    const match = subject?.match(/(\d{2}-\d{9}-\d{3})/);
+    fromPdf: string | null,
-    if (match?.[1]) return match[1];
+    subject: string,
-    // Fallback when invoice number cannot be parsed from subject.
+    uid: number | string,
-    // Exchange UniqueIds are base64 and can contain '+', '/', '=' which would
+  ): string {
-    // introduce path separators into the generated filename (WR-04).
+    if (fromPdf) return fromPdf.replace(/\//g, '-');
-    // Sanitise to [a-zA-Z0-9-] before the value reaches the filesystem.
+    // DKV email subjects carry the invoice number with slashes: "26/650869002/002"
+    const m = subject?.match(/(\d{2}[\/\-]\d{9}[\/\-]\d{3})/);
+    if (m?.[1]) return m[1].replace(/\//g, '-');
     return `email-${String(uid).replace(/[^a-zA-Z0-9\-]/g, '_')}`;
   }
 
@@ -694,6 +700,13 @@ function _delay(ms: number): Promise<void> {
   return new Promise((resolve) => setTimeout(resolve, ms));
 }
 
+/** Format "DD.MM.YYYY" → "YYMMDD" for compact filename date segment. */
+function _formatDateYYMMDD(date: string): string {
+  const [d, m, y] = date.split('.');
+  if (!d || !m || !y) return date.replace(/\./g, '');
+  return `${y.slice(2)}${m}${d}`;
+}
+
 /**
  * Normalize a Kennzeichen for fuzzy vehicle lookup.
  * DKV PDF extraction may omit hyphens or use spaces as separators.