GuidoandClaude Opus 4.8
7461fddb7a
Backend: - services/ocr: PDF-Text via PyMuPDF, OCR (Tesseract) für Scans/Bilder, heuristischer Parser (Datum + Betrag je Zeile, CH/DE/US-Beträge). - routers/documents: /documents/preview (Upload -> Text -> Kandidaten -> gemeinsame Prüf-/Buchungslogik des CSV-Imports). csv_import refaktoriert: detect_rows() von CSV und Beleg gemeinsam genutzt. - requirements: PyMuPDF, pytesseract, pillow; install.sh installiert tesseract-ocr (+ deu). pytest: Heuristik-Parser und PDF-Vorschau. Frontend: - Import-Seite: CSV/PDF-Umschalter; PDF-Modus mit Beleg-Auslesen, extrahiertem Text zur Kontrolle und Best-Effort-Hinweis; gleiche Prüfansicht + Buchen. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
128 lines
4.0 KiB
Python
128 lines
4.0 KiB
Python
"""PDF-/Bild-Belege lokal auslesen (kein Cloud-Dienst).
|
||
|
||
- Digitale PDFs: Text direkt via PyMuPDF.
|
||
- Gescannte PDFs / Bilder: OCR via Tesseract (pytesseract).
|
||
Danach ein heuristischer Parser, der Zeilen mit Datum + Betrag findet. Das ist
|
||
bewusst Best-Effort – das Ergebnis landet in derselben Prüfansicht wie der
|
||
CSV-Import und wird dort korrigiert.
|
||
"""
|
||
|
||
import io
|
||
import re
|
||
from datetime import datetime
|
||
from decimal import Decimal, InvalidOperation
|
||
|
||
try: # PyMuPDF (Text + Rendern)
|
||
import fitz # type: ignore
|
||
except Exception: # noqa: BLE001
|
||
fitz = None
|
||
|
||
try: # OCR
|
||
import pytesseract # type: ignore
|
||
from PIL import Image # type: ignore
|
||
except Exception: # noqa: BLE001
|
||
pytesseract = None
|
||
Image = None
|
||
|
||
|
||
class ExtractionError(RuntimeError):
|
||
pass
|
||
|
||
|
||
def _image_to_text(data: bytes) -> str:
|
||
if pytesseract is None or Image is None:
|
||
raise ExtractionError("OCR nicht verfügbar (pytesseract/Pillow/Tesseract fehlen).")
|
||
img = Image.open(io.BytesIO(data))
|
||
return pytesseract.image_to_string(img, lang="deu+eng")
|
||
|
||
|
||
def _pdf_to_text(data: bytes) -> str:
|
||
if fitz is None:
|
||
raise ExtractionError("PDF-Verarbeitung nicht verfügbar (PyMuPDF fehlt).")
|
||
parts: list[str] = []
|
||
with fitz.open(stream=data, filetype="pdf") as doc:
|
||
for page in doc:
|
||
txt = page.get_text()
|
||
if txt and txt.strip():
|
||
parts.append(txt)
|
||
elif pytesseract is not None and Image is not None:
|
||
pix = page.get_pixmap(dpi=200)
|
||
img = Image.open(io.BytesIO(pix.tobytes("png")))
|
||
parts.append(pytesseract.image_to_string(img, lang="deu+eng"))
|
||
return "\n".join(parts)
|
||
|
||
|
||
def extract_text(data: bytes, filename: str, content_type: str) -> str:
|
||
name = (filename or "").lower()
|
||
if name.endswith(".pdf") or content_type == "application/pdf":
|
||
return _pdf_to_text(data)
|
||
return _image_to_text(data)
|
||
|
||
|
||
# --- Heuristischer Zeilen-Parser -------------------------------------------
|
||
|
||
DATE_RE = re.compile(r"\b(\d{4}-\d{2}-\d{2}|\d{1,2}[./]\d{1,2}[./]\d{2,4})\b")
|
||
# Betrag: entweder mit Tausender-Gruppen (1'234.56 / 1.234,56) oder schlicht (3000.00).
|
||
AMOUNT_RE = re.compile(r"[-+]?(?:\d{1,3}(?:['.,\s]\d{3})+|\d+)[.,]\d{2}(?!\d)")
|
||
_DATE_FORMATS = ["%Y-%m-%d", "%d.%m.%Y", "%d.%m.%y", "%d/%m/%Y", "%d/%m/%y"]
|
||
|
||
|
||
def _norm_date(token: str):
|
||
for fmt in _DATE_FORMATS:
|
||
try:
|
||
return datetime.strptime(token, fmt).date()
|
||
except ValueError:
|
||
continue
|
||
return None
|
||
|
||
|
||
def _parse_amount_auto(s: str) -> int | None:
|
||
"""Betrag mit automatisch erkanntem Dezimaltrenner (CH/DE/US) -> Cent."""
|
||
s = s.strip()
|
||
neg = s.startswith("-") or (s.startswith("(") and s.endswith(")"))
|
||
m = re.search(r"[.,](\d{2})$", s)
|
||
if not m:
|
||
return None
|
||
int_digits = re.sub(r"\D", "", s[: m.start()])
|
||
frac = m.group(1)
|
||
try:
|
||
val = Decimal(f"{int_digits or '0'}.{frac}")
|
||
except InvalidOperation:
|
||
return None
|
||
cents = int(val * 100)
|
||
return -cents if neg else cents
|
||
|
||
|
||
def parse_transactions(text: str) -> list[dict]:
|
||
"""Findet Zeilen mit Datum + Betrag und macht daraus Kandidaten-Buchungen."""
|
||
rows: list[dict] = []
|
||
for line in text.splitlines():
|
||
line = line.strip()
|
||
if len(line) < 6:
|
||
continue
|
||
dm = DATE_RE.search(line)
|
||
if not dm:
|
||
continue
|
||
d = _norm_date(dm.group(0))
|
||
if d is None:
|
||
continue
|
||
rest = f"{line[: dm.start()]} {line[dm.end():]}"
|
||
amounts = AMOUNT_RE.findall(rest)
|
||
if not amounts:
|
||
continue
|
||
amount = _parse_amount_auto(amounts[0])
|
||
if amount is None or amount == 0:
|
||
continue
|
||
desc = rest
|
||
for a in amounts:
|
||
desc = desc.replace(a, " ")
|
||
desc = re.sub(r"\s+", " ", desc).strip(" -–|;,.")
|
||
rows.append({
|
||
"date": d.isoformat(),
|
||
"amount_cents": amount,
|
||
"payee": desc[:120] or None,
|
||
"purpose": None,
|
||
"counterparty_iban": None,
|
||
})
|
||
return rows
|