PDF Parser Light

A lightweight desktop and CLI tool that converts PDFs into clean Markdown using the Google Gemini API. Formats math into LaTeX equations and preserves complex tables without requiring a paid subscription or third-party service.

---

Why PDF Parser Light?

Most PDF extraction tools fall into two categories:

1. Traditional OCR & PDF libraries (like Tesseract, PyPDF, or pdfplumber): Fast and local, but often mangle multi-column layouts, tables, and mathematical formulas into broken plain text.
2. Commercial OCR / Document AI APIs (like Mathpix or AWS Textract): High accuracy, but lock you into monthly subscriptions, proprietary dashboards, or per-page billing.

PDF Parser Light takes a different approach:

• Direct API access: Calls Gemini's vision models directly using your own API key. No middlemen, no tracking, and no added subscription fees.
• Works within Gemini's free tier: Handles daily document conversion within Google's free quota (20 requests/day on Flash, falling back to Flash Lite for higher volume).
• Multimodal extraction: Gemini understands full page layouts in one shot, converting formulas into clean inline/block LaTeX ($...$ / $$...$$) and multi-column data into Markdown/HTML tables.
• Both GUI & CLI: Use the drag-and-drop desktop app or automate large batches via terminal.

---

Features

• LaTeX math & table preservation: Extracts formulas into LaTeX and structured data into Markdown/HTML tables without summarizing.
• Desktop GUI & CLI: CustomTkinter desktop interface with live progress, plus a CLI for batch processing directories.
• Automatic page chunking: Automatically splits documents larger than 20 pages into chunks to prevent timeouts and context overflows.
• Quota tracking & model fallback: Tracks daily free tier usage locally and automatically falls back to Lite models if primary quotas run low.
• Resume support: Resumes processing from partial output files if a job is interrupted mid-way.
• Cross-platform: Pre-built binaries available for macOS, Windows, and Linux.

---

Installation

Option 1: Standalone App (No Python required)

Download the pre-built binary for your OS from the Releases page:

• macOS: Download pdf-parser-light-macos.zip, extract, and move PDF Parser Light.app to /Applications.
• Windows: Download pdf-parser-light-windows.zip, extract, and run PDF Parser Light.exe.
• Linux: Download pdf-parser-light-linux.tar.gz, extract, and run PDF Parser Light. (Requires Tk: sudo apt install python3-tk).

Opening Unsigned Binaries (First-Time Setup)

Because these builds are open-source releases built via GitHub Actions without commercial developer certificates, your OS will block them by default on first launch:

• macOS (Gatekeeper):
  • Finder / System Settings: Right-click (or Control-click) PDF Parser Light.app in Finder, click Open, and confirm Open. On macOS Sequoia (15+), if it doesn't open, go to System Settings → Privacy & Security, scroll down to the Security section, and click Open Anyway.
  • Terminal: Alternatively, remove the quarantine attribute directly:

    xattr -d com.apple.quarantine "/Applications/PDF Parser Light.app"
    

• Windows (SmartScreen):
  • Click More info on the popup, then click Run anyway.

---

Option 2: Install via pip

pip install pdf-parser-light

Launch the GUI or CLI:

pdf-parser-light-gui   # Launches Desktop GUI
pdf-parser-light       # Runs CLI

---

Usage

Desktop Application

1. Launch the app and enter your Gemini API Key (check Remember API Key to save locally).
2. Drag and drop a PDF file (or click Browse).
3. Click Process.
4. Copy the result or save it directly as .md or .txt.

Command Line (CLI)

Set your Gemini API key:

export GEMINI_API_KEY="your_api_key_here"

Examples

# Convert a single PDF
pdf-parser-light document.pdf --output output.md

# Batch process an entire directory of PDFs
pdf-parser-light ./pdf_folder/ --output ./markdown_output/

# Process a specific page range (e.g. pages 1 to 50)
pdf-parser-light document.pdf --pages 1-50

# Resume an interrupted parsing job
pdf-parser-light document.pdf --resume

# Check remaining free requests for today
pdf-parser-light --usage

# Reset local quota counter
pdf-parser-light --reset-quota

# Custom transcription prompt
pdf-parser-light document.pdf --prompt "Transcribe equations only into LaTeX."

Options

Option	Description
input_path	Path to a .pdf file or a directory containing PDFs.
--output, -o	Output file (single PDF) or output directory (batch mode).
--pages, -p	Page range to parse (e.g. 1-50, 40-120, or 10).
--resume, -r	Resume parsing from existing partial output.
--prompt	Custom instructions for the model.
--api-key	Pass API key explicitly (overrides GEMINI_API_KEY).
--usage	Print remaining daily free requests and exit.
--reset-quota	Reset local request counter to 0.
--force	Bypass local quota check and run anyway.

---

Daily Quota & Rate Limits

• Free Tier Limits: By default, the app tracks 20 free requests/day locally for primary Flash models (configurable via GEMINI_FREE_LIMIT).
• Model Fallbacks: When your primary quota runs out, requests cascade to Flash Lite models (gemini-3.5-flash-lite, gemini-3.1-flash-lite), which offer significantly higher capacity (up to 500 free requests/day per model on Google AI Studio).
• Retries: Automatically retries with exponential backoff on HTTP 429 rate limits or transient errors.

---

Privacy & Security

• Direct Connections: All requests and file uploads go directly to Google's Gemini API endpoints using the official google-genai SDK. No third-party servers, analytics, or telemetry proxies are involved.
• Data Sensitivity Notice: Do not upload confidential, sensitive, or restricted documents. Files are processed remotely on Google's servers in accordance with Google Gemini API terms.
• File Retention & Cleanup: Files are uploaded temporarily to Gemini API storage for transcription and deleted immediately upon completion. If a process terminates unexpectedly, files remain until Google's temporary storage retention limit expires.
• Key Storage: Saved API keys are stored locally on your machine in standard user configuration directories (~/Library/Application Support/pdf_parser_light/ on macOS, %LOCALAPPDATA%\pdf_parser_light\ on Windows, ~/.config/pdf_parser_light/ on Linux).

---

Development

git clone https://github.com/jtaroreh/pdf-parser-light.git
cd pdf-parser-light

python3 -m venv venv
source venv/bin/activate
pip install -e ".[test]"

pytest -v

Build App Bundle (macOS)

./build.sh

Output is created in dist/PDF Parser Light.app.