openfitz 🚀

A Fast, Lightweight Python Library for PDF Document Parsing, Spatial Text Extraction, and Raster Rendering Powered by Google PDFium.

Features

• Permissive MIT License: 100% free for commercial use, private SaaS backends, and microservices.
• Powered by Google PDFium: Utilizes the battle-tested, high-performance C++ engine that powers Google Chrome's PDF viewer via pypdfium2.
• Top-Left Coordinate System: Native (x0, y0, x1, y1) bounding box coordinate mapping with zero coordinate inversion drift.
• Spatial Text Extraction: Extract text bounded by exact clipping rectangles or inspect full document block, line, and span hierarchies.
• Raster Rendering & Pixel Inspection: Render pages or clipped areas directly to memory bitmaps with raw packed RGB .samples for checkbox/darkness ratio analysis.
• Bookmark / TOC Traversal: Fast hierarchical Table of Contents and bookmark extraction.
• Flexible Namespaces: Usable as either import fitz or import openfitz.

Installation

pip install openfitz

Quickstart

import fitz

# 1. Open a PDF document (from file path or bytes)
doc = fitz.open("sample.pdf")
print("Total pages:", doc.page_count)

# 2. Extract full page text or bounded text
page = doc[0]
full_text = page.get_text("text")

# Clipped extraction (top-left origin coordinates)
field_rect = fitz.Rect(50, 100, 300, 150)
field_text = page.get_text("text", clip=field_rect)
print("Extracted field:", field_text.strip())

# 3. Spatial text hierarchy (blocks, lines, spans)
page_dict = page.get_text("dict")
for block in page_dict["blocks"]:
    for line in block["lines"]:
        for span in line["spans"]:
            print(f"Text: {span['text']} | Box: {span['bbox']} | Font: {span['font']}")

# 4. Render to Pixmap (for image export or darkness/checkbox detection)
mat = fitz.Matrix(2, 2)  # 2x zoom
pix = page.get_pixmap(matrix=mat, clip=field_rect)
print("Pixel buffer length:", len(pix.samples))
pix.save("field_preview.png")

# 5. Extract Bookmarks / Table of Contents
toc = doc.get_toc(simple=True)
for level, title, page_num in toc:
    print(f"Level {level}: {title} (page {page_num})")

doc.close()

Architecture

openfitz wraps Google PDFium's C++ rasterization and layout engine into clean, Pythonic classes:

Running Tests

git clone https://github.com/jagadeesh88705/openfitz.git
cd openfitz
pip install -e .[dev]
pytest -v

License

MIT License. Copyright (c) 2026 Jagadeesh Murugesan.