refactor: make PDF parser dependencies optional and lazy-load parser

loko/stock/pdf_parser.py

@@ -12,8 +12,15 @@ import logging
 from io import BytesIO
 from typing import Dict, List, Any, Optional, Tuple
 
+try:
     import pypdfium2 as pdfium
+except ImportError:
+    pdfium = None
+
+try:
     import numpy as np
+except ImportError:
+    np = None
 
 from django.db.models import Q
 from .models import Supplier, Product, WarehouseLocation
@@ -91,8 +98,13 @@ class PdfQuoteParser:
         self._temp_path = None
         self.is_ocr = False
 
-    def _get_document(self) -> Tuple[pdfium.PdfDocument, str]:
+    def _get_document(self) -> Tuple[Any, str]:
         """Ouvre le document PDF via pypdfium2."""
+        if pdfium is None:
+            raise RuntimeError(
+                "Le module 'pypdfium2' n'est pas installé sur le serveur. "
+                "Veuillez exécuter 'pip install -r requirements/base.txt' pour activer l'analyse PDF locale."
+            )
         if isinstance(self.file_or_path, str):
             return pdfium.PdfDocument(self.file_or_path), self.file_or_path
         elif hasattr(self.file_or_path, 'temporary_file_path'):
@@ -146,7 +158,7 @@ class PdfQuoteParser:
                     pass
             self.close()
 
-    def _parse_vector(self, doc: pdfium.PdfDocument) -> Dict[str, Any]:
+    def _parse_vector(self, doc: Any) -> Dict[str, Any]:
         """Analyse un PDF avec couche texte vectorielle (recherche spatiale précise)."""
         full_text = ""
         for page in doc:
@@ -167,7 +179,7 @@ class PdfQuoteParser:
             'total_items': len(enriched_items),
         }
 
-    def _parse_with_ocr(self, doc: pdfium.PdfDocument) -> Dict[str, Any]:
+    def _parse_with_ocr(self, doc: Any) -> Dict[str, Any]:
         """Analyse un PDF scanné via RapidOCR."""
         try:
             from rapidocr_onnxruntime import RapidOCR

loko/stock/views.py

@@ -64,7 +64,7 @@ from .models import (
     SerializedItemEvent,
     notify_supervisors_pending_product,
 )
-from .pdf_parser import PdfQuoteParser, normalize_unit
+from .pdf_parser import normalize_unit
 from .permissions import (
     warehouse_view_required,
     warehouse_manage_required,
@@ -1649,6 +1649,7 @@ def purchase_order_from_pdf(request):
                 orig_pdf_name = pdf_file.name
 
                 try:
+                    from .pdf_parser import PdfQuoteParser
                     parser = PdfQuoteParser(temp_pdf_path)
                     parse_result = parser.parse()
                     step = 'review'

requirements/base.txt

@@ -44,5 +44,6 @@ boto3>=1.34.0
 markdown
 onnxruntime>=1.19.0
 rapidocr-onnxruntime>=1.2.0
+pypdfium2>=5.0.0
 pywebpush>=2.0.0
 py-vapid>=1.9.0