fix(extract): backfill booking code/total and harden the reference match

Apply the deterministic confirmation-code and total fill to vendor-template results too (not just model output), and require the captured reference to contain a digit so a bare 'Confirmation'/'Reference' label no longer grabs the next prose word.
2026-06-30 18:46:00 +00:00 · 2026-06-26 09:08:36 +02:00
parent 9aaf313d59
commit b175ef4626
2 changed files with 191 additions and 13 deletions
@@ -68,10 +68,18 @@ export function detectFlightNumbers(text: string): string[] {
  return out;
 }

-/** The booking/confirmation code, pulled once for the whole document. */
+/**
+ * The booking/confirmation code, pulled once for the whole document. Covers the German
+ * "Bestätigungs-Code" (Airbnb) and "Reservation No." (rental brokers) on top of the PNR /
+ * Buchungsnummer / Confirmation forms. The match is left-most in the text, so a customer
+ * "Reservation No." that precedes a vendor "Supplier Reference" wins.
+ */
 export function extractBookingRef(text: string): string | undefined {
+  // The captured code must contain a digit: real PNRs/booking codes effectively always
+  // do, while the case-insensitive [A-Z0-9] class would otherwise grab a following prose
+  // word ("Confirmation\nThank you…" → "Thank") after a bare label.
  const m = text.match(
-    /(?:PNR|Buchungs(?:code|nummer|referenz)|Booking\s*(?:reference|code|number)|Confirmation(?:\s*number)?|Reservierungsnummer|Best(?:ä|ae)tigungsnummer|Reference)\s*:?\s*([A-Z0-9]{5,})/i,
+    /(?:PNR|Buchungs(?:code|nummer|referenz)|Booking\s*(?:reference|code|number)|Confirmation\s*(?:number|code)?|Reservierungsnummer|Reservation\s*(?:No\.?|Number|Nr\.?)|Best(?:ä|ae)tigungs[-\s]?(?:nummer|code)|Reference)\s*:?\s*((?=[A-Z0-9]*\d)[A-Z0-9]{5,})/i,
  );
  return m?.[1];
 }
@@ -173,12 +181,38 @@ async function extractSingle(text: string, ctx: RouterContext): Promise<FlatLike
 * Run the router on extracted document text and return schema.org KiReservation nodes.
 * Returns `[]` (never throws for content reasons) so the caller degrades gracefully.
 */
+/**
+ * Schicht 2 — fill the booking-wide fields the per-reservation extraction doesn't carry:
+ * the confirmation/PNR and the booking total. Applied to BOTH the deterministic vendor
+ * results AND the model output, so a vendor template that read the structured fields but
+ * whose narrow ref/price regex missed still gets the broad doc-wide deterministic value.
+ * Never overrides a value the source already provided.
+ */
+function fillBookingWideFields(flats: Array<Record<string, unknown>>, text: string): void {
+  const ref = extractBookingRef(text);
+  const total = extractTotalPrice(text);
+  // A small model sometimes emits an empty string for a price it didn't find, which is
+  // not `null` — treat blank/whitespace as "no price" so the deterministic total still wins.
+  const priceMissing = (v: unknown) => v == null || (typeof v === 'string' && v.trim() === '');
+  flats.forEach((f, i) => {
+    if (!f.booking_reference && ref) f.booking_reference = ref;
+    // The total belongs to the booking, so attach it once (the first item).
+    if (i === 0 && total && priceMissing(f.price)) {
+      f.price = total.price;
+      if (f.currency == null) f.currency = total.currency;
+    }
+  });
+}
+
 export async function routeExtraction(text: string, ctx: RouterContext): Promise<{ kiItems: KiReservation[]; warnings: string[] }> {
  const warnings: string[] = [];

-  // Schicht 0 — deterministic vendor templates (no LLM).
+  // Schicht 0 — deterministic vendor templates (no LLM). Still top-up the booking-wide
+  // fields so a template misses on the ref/price doesn't drop them when the doc-wide
+  // deterministic extractor would have found them.
  const vendor = matchVendorTemplate(text);
  if (vendor && vendor.length > 0) {
+    fillBookingWideFields(vendor as unknown as Array<Record<string, unknown>>, text);
    return { kiItems: nuExtractToKiReservations(vendor) as unknown as KiReservation[], warnings };
  }

@@ -191,16 +225,7 @@ export async function routeExtraction(text: string, ctx: RouterContext): Promise
  }

  // Schicht 2 — deterministic booking-wide fields the per-call schema doesn't carry.
-  const ref = extractBookingRef(text);
-  const total = extractTotalPrice(text);
-  flats.forEach((f, i) => {
-    if (!f.booking_reference && ref) f.booking_reference = ref;
-    // The total belongs to the booking, so attach it once (the first item).
-    if (i === 0 && total && f.price == null) {
-      f.price = total.price;
-      if (f.currency == null) f.currency = total.currency;
-    }
-  });
+  fillBookingWideFields(flats as unknown as Array<Record<string, unknown>>, text);

  const kiItems = nuExtractToKiReservations(flats as unknown as Record<string, unknown>[]) as unknown as KiReservation[];
  return { kiItems, warnings };
@@ -0,0 +1,153 @@
+import { describe, it, expect } from 'vitest';
+import { matchVendorTemplate } from '../../../../src/nest/llm-parse/router/vendor-templates';
+import { extractBookingRef, extractTotalPrice } from '../../../../src/nest/llm-parse/router/extraction-router';
+
+// The snippets below mirror the pdf-parse text layer of real confirmation PDFs
+// (Expedia hotel receipt, Airbnb booking, a broker rental-car voucher).
+
+const EXPEDIA_HOTEL = `Beleg
+Expedia-Reiseplan: 73222406755286
+Buchungsdatum: 27. Aug. 2025
+Buchungsdetails
+Mercure Tokyo Haneda Airport
+1 Chome-2-11 Haneda, Ota City, Tokyo, 144-0043 Japan
+Anreise: 3. Mai 2026
+Abreise: 22. Mai 2026
+1 Zimmer x 19 Nächte
+Zahlungsdetails
+Steuern und Gebühren 1.195,07 €
+Gesamtpreis 3.516,13 €
+Bezahlt`;
+
+const AIRBNB = `Zwei-Zimmer-Wohnung zwischen Venedig und
+Treviso!
+Check-in
+15:00
+Sa., 23. Aug.
+Check-out
+10:00
+Sa., 30. Aug.
+Bestätigungs-Code
+HMHJ9RTEEK
+Adresse
+Via Aldo Moro, 47 n. 15, Quarto d'Altino, Venetien 30020, Italien
+Bezahlter Betrag
+651,86 €`;
+
+const BROKER_RENTAL = `Reservation No.: G72820729
+MAIN DRIVER'S NAME: Felix Pakulat
+SUPPLIER DETAILS
+SICILY BY CAR (V2) Supplier Reference: IT587200464
+PICK-UP DETAILS
+Venice Marco Polo Airport
+Aug 23 2025 13:30
+DROP-OFF DETAILS
+Venice Marco Polo Airport
+Aug 30 2025 12:30
+Payment Details
+Amount Payable to
+Supplier:
+(Payable at Pick-up)
+EUR 300.21`;
+
+describe('expedia-hotel vendor template', () => {
+  it('extracts hotel name, address, stay dates, price and Reiseplan number', () => {
+    const out = matchVendorTemplate(EXPEDIA_HOTEL);
+    expect(out).toEqual([
+      {
+        type: 'hotel',
+        name: 'Mercure Tokyo Haneda Airport',
+        booking_reference: '73222406755286',
+        address: '1 Chome-2-11 Haneda, Ota City, Tokyo, 144-0043 Japan',
+        checkin_time: '2026-05-03',
+        checkout_time: '2026-05-22',
+        price: '3.516,13',
+        currency: 'EUR',
+      },
+    ]);
+  });
+
+  it('parses German abbreviated months (e.g. "4. Feb. 2026")', () => {
+    const bnb = EXPEDIA_HOTEL.replace('Anreise: 3. Mai 2026', 'Anreise: 4. Feb. 2026').replace(
+      'Abreise: 22. Mai 2026',
+      'Abreise: 6. Feb. 2026',
+    );
+    const out = matchVendorTemplate(bnb);
+    expect(out?.[0]).toMatchObject({ checkin_time: '2026-02-04', checkout_time: '2026-02-06' });
+  });
+});
+
+describe('broker-rental-voucher vendor template', () => {
+  it('extracts pickup/return depots, English date-times, price and the customer reservation no.', () => {
+    const out = matchVendorTemplate(BROKER_RENTAL);
+    expect(out).toEqual([
+      {
+        type: 'car',
+        operator: 'SICILY BY CAR', // the "(V2)" supplier-version tag is stripped
+        booking_reference: 'G72820729', // the customer ref, not the supplier reference
+        from_name: 'Venice Marco Polo Airport',
+        to_name: 'Venice Marco Polo Airport',
+        departure_time: '2025-08-23T13:30:00',
+        arrival_time: '2025-08-30T12:30:00',
+        price: '300.21',
+        currency: 'EUR',
+      },
+    ]);
+  });
+});
+
+describe('non-matching documents', () => {
+  it('returns null when no template applies', () => {
+    expect(matchVendorTemplate(AIRBNB)).toBeNull();
+    expect(matchVendorTemplate('just some unrelated text')).toBeNull();
+  });
+});
+
+describe('broker template — date & price variants', () => {
+  const VARIANT = `Reservation No.: AB123456
+SUPPLIER DETAILS
+GREEN MOTION Supplier Reference: XYZ
+PICK-UP DETAILS
+London Heathrow
+Aug 5, 2025 09:00 AM
+DROP-OFF DETAILS
+London Heathrow
+Aug 12, 2025 05:30 PM
+Payment Details
+Total to pay
+150.00 GBP`;
+
+  it('handles a comma date, a 12-hour clock and a trailing non-EUR currency', () => {
+    const out = matchVendorTemplate(VARIANT);
+    expect(out?.[0]).toMatchObject({
+      booking_reference: 'AB123456',
+      departure_time: '2025-08-05T09:00:00', // 09:00 AM
+      arrival_time: '2025-08-12T17:30:00', // 05:30 PM → 17:30
+      price: '150.00',
+      currency: 'GBP', // derived, not hard-coded EUR
+    });
+  });
+});
+
+describe('extractBookingRef', () => {
+  it('reads an Airbnb "Bestätigungs-Code"', () => {
+    expect(extractBookingRef(AIRBNB)).toBe('HMHJ9RTEEK');
+  });
+  it('prefers the customer "Reservation No." over a later "Supplier Reference"', () => {
+    expect(extractBookingRef(BROKER_RENTAL)).toBe('G72820729');
+  });
+  it('still reads a classic "Buchungsnummer" / "PNR"', () => {
+    expect(extractBookingRef('Buchungsnummer: ABC123')).toBe('ABC123');
+    expect(extractBookingRef('PNR XY7Q9Z')).toBe('XY7Q9Z');
+  });
+  it('does not capture a prose word after a bare "Confirmation"/"reference"', () => {
+    expect(extractBookingRef('Booking Confirmation\n\nThank you for choosing us')).toBeUndefined();
+    expect(extractBookingRef('For future reference please retain this email')).toBeUndefined();
+  });
+});
+
+describe('extractTotalPrice', () => {
+  it('reads an Airbnb "Bezahlter Betrag"', () => {
+    expect(extractTotalPrice(AIRBNB)).toEqual({ price: '651,86', currency: 'EUR' });
+  });
+});