PDF_X_tract / app.py
ShankarC's picture
Update app.py
c079558 verified
Raw History Blame Contribute Delete
5.47 kB
import gradio as gr
from docling.document_converter import DocumentConverter, PdfFormatOption
from docling.datamodel.pipeline_options import PdfPipelineOptions
from docling_core.types.doc import ImageRefMode
from docx import Document as WordDocument
from docx.shared import Inches
import re
import io
import base64
import tempfile
def write_markdown_to_word(md_text: str, word_doc: WordDocument, include_images=False):
image_pattern = re.compile(r'!\[.*?\]\((data:image\/png;base64,[^\)]+)\)')
lines = md_text.splitlines()
i = 0
while i < len(lines):
line = lines[i].rstrip()
# -- NEW: REMOVE image placeholder tags if not including images <!-- image --> or <!-- image ... --> or embed image
if not include_images and (line.strip() == "<!-- image -->" or line.strip().startswith("<!-- image")):
i += 1
continue
# Images (base64-encoded)
match = image_pattern.search(line)
if match:
if include_images:
image_data = match.group(1)
header, b64data = image_data.split(',', 1)
img_stream = io.BytesIO(base64.b64decode(b64data))
word_doc.add_picture(img_stream, width=Inches(4))
i += 1
continue
if not line:
word_doc.add_paragraph("")
i += 1
continue
if line.startswith("#"):
level = min(line.count("#"), 4)
heading_text = line.lstrip("#").strip()
word_doc.add_heading(heading_text, level=level)
i += 1
continue
if line.startswith("```"):
code_lines = []
i += 1
while i < len(lines) and not lines[i].startswith("```"):
code_lines.append(lines[i])
i += 1
word_doc.add_paragraph("\n".join(code_lines))
if i < len(lines) and lines[i].startswith("```"):
i += 1
continue
if line.lstrip().startswith(("- ", "* ")):
while i < len(lines) and lines[i].lstrip().startswith(("- ", "* ")):
item = lines[i].lstrip()[2:].strip()
word_doc.add_paragraph(f" - {item}")
i += 1
continue
if "|" in line:
if i + 1 < len(lines) and set(lines[i + 1].replace("|", "").replace(" ", "")) <= set("-:"):
header_cells = [c.strip() for c in line.strip("|").split("|")]
i += 2
data_rows = []
while i < len(lines) and "|" in lines[i]:
row_cells = [c.strip() for c in lines[i].strip("|").split("|")]
data_rows.append(row_cells)
i += 1
n_cols = len(header_cells)
n_rows = len(data_rows)
tbl = word_doc.add_table(rows=n_rows + 1, cols=n_cols)
tbl.style = "Table Grid"
for c_idx, col_name in enumerate(header_cells):
tbl.cell(0, c_idx).text = str(col_name)
for r_idx, row in enumerate(data_rows):
for c_idx in range(n_cols):
val = row[c_idx] if c_idx < len(row) else ""
tbl.cell(r_idx + 1, c_idx).text = str(val)
word_doc.add_paragraph("")
continue
word_doc.add_paragraph(line)
i += 1
def pdf_to_word(pdf_file, include_images_radio):
try:
include_images = (include_images_radio == "Yes")
if include_images:
pipeline_options = PdfPipelineOptions()
pipeline_options.images_scale = 2.0
pipeline_options.generate_page_images = True
pipeline_options.generate_picture_images = True
converter = DocumentConverter(
format_options={
"pdf": PdfFormatOption(pipeline_options=pipeline_options)
}
)
else:
converter = DocumentConverter()
with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp_pdf:
tmp_pdf.write(pdf_file)
pdf_path = tmp_pdf.name
result = converter.convert(pdf_path)
doc = result.document
if include_images:
md = doc.export_to_markdown(image_mode=ImageRefMode.EMBEDDED)
else:
md = doc.export_to_markdown()
word_doc = WordDocument()
write_markdown_to_word(md, word_doc, include_images=include_images)
# Write Word doc to temp file and return file path
with tempfile.NamedTemporaryFile(delete=False, suffix=".docx") as tmp_docx:
word_doc.save(tmp_docx)
docx_path = tmp_docx.name
return docx_path, ""
except Exception as e:
# Return 'None' for the file, and the error message for the error component
return (None, f"Error: {str(e)}")
with gr.Blocks() as demo:
gr.Markdown("## PDF to Word (Text, Tables, and Optional Images) via Docling")
pdf_input = gr.File(label="Upload PDF", type="binary")
include_images = gr.Radio(["Yes", "No"], label="Include Images", value="No")
btn = gr.Button("Convert and Download Word (.docx)")
output = gr.File(label="Download Word File (.docx)")
error_box = gr.Textbox(label="Error (if any)", interactive=False)
btn.click(fn=pdf_to_word, inputs=[pdf_input, include_images], outputs=[output, error_box])
demo.launch()