fix: avoid RapidOCR torch fallback in default OCR setup

docling/models/stages/ocr/auto_ocr_model.py

@@ -95,27 +95,6 @@ def __init__(
                 except ImportError:
                     _log.info("easyocr cannot be used because it is not installed.")
 
-            if self._engine is None:
-                try:
-                    import torch
-                    from rapidocr import EngineType, RapidOCR  # type: ignore
-
-                    self._engine = RapidOcrModel(
-                        enabled=self.enabled,
-                        artifacts_path=artifacts_path,
-                        options=RapidOcrOptions(
-                            backend="torch",
-                            bitmap_area_threshold=self.options.bitmap_area_threshold,
-                            force_full_page_ocr=self.options.force_full_page_ocr,
-                        ),
-                        accelerator_options=accelerator_options,
-                    )
-                    _log.info("Auto OCR model selected rapidocr with torch.")
-                except ImportError:
-                    _log.info(
-                        "rapidocr cannot be used because rapidocr or torch is not installed."
-                    )
-
             if self._engine is None:
                 _log.warning("No OCR engine found. Please review the install details.")
 

pyproject.toml

@@ -243,7 +243,7 @@ cli = [
 # CONVENIENCE BUNDLES
 # ============================================================================
 standard = [
-  'docling-slim[format-pdf,models-local,feat-ocr-rapidocr,format-office,format-web,format-latex,format-email,feat-chunking,extract-core,service-client,cli]',
+  'docling-slim[format-pdf,models-local,feat-ocr-rapidocr-onnx,format-office,format-web,format-latex,format-email,feat-chunking,extract-core,service-client,cli]',
 ]
 
 all = [

tests/test_options.py

@@ -123,6 +123,13 @@ def test_kserve_v2_binary_data_deprecated_alias():
     assert options.use_binary_data is True
 
 
+def test_standard_extra_installs_rapidocr_onnx():
+    pyproject = Path("pyproject.toml").read_text()
+
+    assert "feat-ocr-rapidocr-onnx" in pyproject
+    assert "models-local,feat-ocr-rapidocr," not in pyproject
+
+
 def test_e2e_conversions(test_doc_path):
     for converter in get_converters_with_table_options():
         print(f"converting {test_doc_path}")