fix: Build errors and tesseract removal

ocr/app/extractors/manual_extractor.py

@@ -5,9 +5,9 @@ import time
 from dataclasses import dataclass, field
 from typing import Callable, Optional
 
-import pytesseract
 from PIL import Image
 
+from app.engines import create_engine, OcrConfig
 from app.preprocessors.pdf_preprocessor import pdf_preprocessor, PdfInfo
 from app.table_extraction.detector import table_detector, DetectedTable
 from app.table_extraction.parser import table_parser, ParsedScheduleRow
@@ -243,8 +243,9 @@ class ManualExtractor:
 
         # OCR the full page
         try:
-            image = Image.open(io.BytesIO(image_bytes))
-            ocr_text = pytesseract.image_to_string(image)
+            engine = create_engine()
+            ocr_result = engine.recognize(image_bytes, OcrConfig())
+            ocr_text = ocr_result.text
 
             # Mark tables as maintenance if page contains maintenance keywords
             for table in detected_tables:
@@ -358,8 +359,9 @@ class ManualExtractor:
 
             if not text and first_page.image_bytes:
                 # OCR first page
-                image = Image.open(io.BytesIO(first_page.image_bytes))
-                text = pytesseract.image_to_string(image)
+                engine = create_engine()
+                ocr_result = engine.recognize(first_page.image_bytes, OcrConfig())
+                text = ocr_result.text
 
             if text:
                 return self._parse_vehicle_from_text(text)

ocr/app/extractors/vin_extractor.py

@@ -316,8 +316,8 @@ class VinExtractor(BaseExtractor):
             single-line - Treat as a single text line
             single-word - Treat as a single word
 
-        For PaddleOCR, angle classification handles rotated/angled text
-        inherently, replacing the need for Tesseract PSM mode fallbacks.
+        PaddleOCR angle classification handles rotated/angled text
+        inherently, so no PSM mode fallbacks are needed.
 
         Returns:
             List of VIN candidates