liteparse

LiteParse — PDF Parsing & Visual Extraction

Fast, local PDF parser (LlamaIndex) for text extraction with bounding box coordinates, page screenshots, and visual element detection. Wraps a Node.js CLI (lit) via a Python wrapper.

Installation

uv pip install liteparse
# CLI available as `lit` (requires Node.js/npx)

Current version: 1.2.1. Docs: https://developers.llamaindex.ai/liteparse/

CLI Usage

# Parse PDF to text
lit parse document.pdf

# Parse to JSON with bounding boxes
lit parse document.pdf --format json -o output.json

# Parse specific pages, no OCR (native PDFs)
lit parse document.pdf --format json --no-ocr --target-pages "1-5,10"

# Screenshot pages at high DPI
lit screenshot document.pdf -o ./screenshots --dpi 200

# Batch parse a directory
lit batch-parse ./pdfs ./outputs --recursive --extension ".pdf"

CLI Flags (lit parse)

Flag	Default	Description
-o, --output	stdout	Output file path
--format	text	text or json
--no-ocr	false	Disable OCR (use for native/digital PDFs)
--ocr-language	en	Language code (is for Icelandic)
--target-pages	all	Page ranges: "1-5,10,15-20"
--dpi	150	Rendering resolution
--num-workers	CPU-1	Parallel OCR workers
--max-pages	10000	Page limit
--password		For encrypted PDFs
-q, --quiet	false	Suppress progress

Python API

from liteparse import LiteParse

parser = LiteParse()

# Parse — use ocr_enabled=False for native PDFs (much faster)
result = parser.parse(
    "document.pdf",
    ocr_enabled=False,       # Skip OCR for digital PDFs
    ocr_language="is",       # Icelandic OCR when needed
    target_pages="1-10",     # Optional page selection
    dpi=150,                 # Resolution for OCR
)

# Full text
print(result.text)

# Per-page access
for page in result.pages:
    print(f"Page {page.pageNum}: {page.width}x{page.height}")
    print(f"  Text items: {len(page.textItems)}")
    print(f"  Bounding boxes: {len(page.boundingBoxes)}")