feat: add pypdf as a method

app.py

@@ -24,6 +24,7 @@ from backends import (  # convert_zerox,
     convert_img2table,
     convert_marker,
     convert_mineru,
+    convert_pypdf,
     convert_sycamore,
     convert_unstructured,
 )
@@ -79,6 +80,8 @@ def convert_document(path, method, start_page=0, enabled=True):
         text, debug_image_paths = convert_img2table(path, file_name)
     elif method == "GMFT (table-only)":
         text, debug_image_paths = convert_gmft(path, file_name)
+    elif method == "PyPDF":
+        text, debug_image_paths = convert_pypdf(path, file_name)
     else:
         raise ValueError(f"Unsupported method: {method}")
 

backends/__init__.py

@@ -5,6 +5,7 @@ from .gmft import convert_gmft
 from .img2table import convert_img2table
 from .marker import convert_marker
 from .mineru import convert_mineru
+from .pypdf import convert_pypdf
 from .syca import convert_sycamore
 from .unstructured import convert_unstructured
 
@@ -20,6 +21,8 @@ __all__ = [
     "convert_img2table",
     "convert_gmft",
     "convert_sycamore",
+    "convert_pypdf",
+    # "convert_zerox",
 ]
 
 SUPPORTED_METHODS = [
@@ -32,6 +35,7 @@ SUPPORTED_METHODS = [
     "Gemini (API)",
     "Img2Table (table-only)",
     "GMFT (table-only)",
+    "PyPDF",
     # "Zerox"
 ]
 SUPPORTED_METHODS_METADATA = {
@@ -89,4 +93,10 @@ SUPPORTED_METHODS_METADATA = {
         "url": "https://github.com/aryn-ai/sycamore",
         "documentation": "https://sycamore.readthedocs.io/en/stable/",
     },
+    "PyPDF": {
+        "name": "PyPDF",
+        "description": "PyPDF is a pure-Python PDF toolkit that can help you read, write, and manipulate PDF documents.",
+        "url": "https://github.com/py-pdf/pypdf",
+        "documentation": "https://pypdf.readthedocs.io/en/stable",
+    },
 }

backends/pypdf.py

@@ -0,0 +1,10 @@
+from pypdf import PdfReader
+
+
+def convert_pypdf(path: str, file_name: str):
+    pdf = PdfReader(path)
+    pages = pdf.pages
+
+    text = "\n\n".join([page.extract_text(0) for page in pages])
+
+    return text, []

requirements.txt

@@ -18,6 +18,7 @@ openai
 sycamore-ai[local-inference]
 img2table
 gmft
-opencv-contrib-python
 unimernet==0.2.3
 transformers<5.0.0,>=4.45.2
+pypdf
+opencv-contrib-python