feat: PDF-Beleg-Parser blockbasiert (mehrzeilig) + Seiten als Bild
- Neuer Parser erkennt mehrzeilige Auszüge (N26 & Co.): Beschreibung → Value Date → Buchungsdatum → Betrag stehen auf eigenen Zeilen; Betrags-Zeile schließt eine Buchung ab, Kopf-/Fußzeilen (Name/Adresse/eigene IBAN/Seitenzahl/ Spaltenköpfe) werden gefiltert, Gegen-IBAN und Zweck werden übernommen - Alter Ein-Zeilen-Heuristik-Parser bleibt als Fallback - /documents/preview liefert zusätzlich die Seiten als PNG-Data-URLs (PyMuPDF get_pixmap, kein poppler nötig) - Tests: N26-artige mehrzeilige Fixture + Seiten-Rendering (74 grün); am echten 34-Seiten-Auszug 207 Buchungen erkannt Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
@@ -38,6 +38,68 @@ def test_parse_transactions_heuristic():
|
||||
assert rows[1]["amount_cents"] == 300000 # Schweizer Tausender-Apostroph
|
||||
|
||||
|
||||
N26_TEXT = "\n".join([
|
||||
"GUIDO OLIVIER SCHMIT",
|
||||
"Riedstrasse 18, 8908 Hedingen",
|
||||
"IBAN: DE54100110012138281295 • BIC: NTSBDEB1XXX",
|
||||
"Issued on",
|
||||
"20.08.2026",
|
||||
"1 / 2",
|
||||
"Description",
|
||||
"Booking Date",
|
||||
"Amount",
|
||||
"La Flamme",
|
||||
"Mastercard • Bars & Restaurants",
|
||||
"Value Date 31.01.2026",
|
||||
"01.02.2026",
|
||||
"-97,00€",
|
||||
"Guido Schmit",
|
||||
"Income",
|
||||
"IBAN: LT293250031382885526 • BIC: REVOLT21XXX",
|
||||
"Gesendet von Revolut",
|
||||
"Value Date 01.02.2026",
|
||||
"01.02.2026",
|
||||
"+43,00€",
|
||||
"Noemi Konig",
|
||||
"Outgoing Transfers",
|
||||
"IBAN: DE65120300001207628981 • BIC: BYLADEM1001",
|
||||
"WG Miete Guido Schmit",
|
||||
"Value Date 02.02.2026",
|
||||
"02.02.2026",
|
||||
"-288,00€",
|
||||
])
|
||||
|
||||
|
||||
def test_parse_transactions_n26_blocks():
|
||||
from app.services import ocr
|
||||
|
||||
rows = ocr.parse_transactions(N26_TEXT)
|
||||
assert len(rows) == 3
|
||||
|
||||
assert rows[0]["date"] == "2026-02-01"
|
||||
assert rows[0]["amount_cents"] == -9700
|
||||
assert rows[0]["payee"] == "La Flamme"
|
||||
|
||||
assert rows[1]["amount_cents"] == 4300 # Eingang (+)
|
||||
assert rows[1]["payee"] == "Guido Schmit"
|
||||
assert rows[1]["counterparty_iban"] == "LT293250031382885526"
|
||||
|
||||
assert rows[2]["amount_cents"] == -28800
|
||||
assert rows[2]["payee"] == "Noemi Konig"
|
||||
# Gegen-IBAN der Miete – NICHT die eigene Kopf-IBAN (DE54…).
|
||||
assert rows[2]["counterparty_iban"] == "DE65120300001207628981"
|
||||
|
||||
|
||||
def test_render_pages_pdf():
|
||||
from app.services import ocr
|
||||
|
||||
doc = fitz.open()
|
||||
doc.new_page()
|
||||
pages = ocr.render_pages(doc.tobytes(), "beleg.pdf", "application/pdf")
|
||||
assert len(pages) == 1
|
||||
assert pages[0].startswith("data:image/png;base64,")
|
||||
|
||||
|
||||
def test_pdf_document_preview(admin):
|
||||
acc = _account(admin)
|
||||
pdf = _make_pdf()
|
||||
|
||||
Reference in New Issue
Block a user