pdfalive 0.5.0

A Python library and CLI tool that uses LLMs to enhance PDF files

---

pdfalive: A Python library and set of CLI tools to bring PDF files alive with the magic of LLMs.

Features:

• Automatically generate a Table of Contents via PDF Bookmarks for PDF file using LLMs. Supports arbitrarily large files with intelligent batching.
• Automatically detect if OCR is needed to parse text from raster data. If needed, performs OCR via Tesseract OCR library.
• Choose which LLM to use from any vendor. Supports using local models via ollama as well. Retry logic included for handling rate limits.

Installation

the tesseract library is required for OCR. This is used for PDFs where text is not parsed. On MacOS, can install via Homebrew:

brew install tesseract

You can then install the pdfalive package via pip for example:

pip install pdfalive

Usage

To use the CLIs described below, you can install the python package (pip install pdfalive), or run the cli directly using uvx:

uvx pdfalive generate-toc input.pdf output.pdf

More detailed examples of the CLI sub-commands are provided below. You can also use --help on the main command-line and any of the sub-commands to get an idea of the different options supported.

generate-toc

Automatically generate clickable Table of Contents (using PDF bookmarks) for a PDF file. The tool extracts font and text features from the PDF and uses an LLM to intelligently identify chapter and section headings.

Basic usage:

pdfalive generate-toc input.pdf output.pdf

Choosing an LLM: By default we use the latest OpenAI model, but you can use any LLM supported by LangChain:

pdfalive generate-toc --model-identifier 'claude-sonnet-4-5' input.pdf output.pdf

Set the appropriate API key for your provider (e.g., OPENAI_API_KEY, ANTHROPIC_API_KEY).

Scanned PDFs: OCR is enabled by default. If your PDF is a scanned document without extractable text, OCR will be performed automatically to extract text before TOC generation.

By default, the OCR text layer is discarded after TOC generation (preserving original file size). To include the OCR text layer in the output (making it searchable):

pdfalive generate-toc --ocr-output scanned.pdf output.pdf

To disable automatic OCR detection entirely:

pdfalive generate-toc --no-ocr input.pdf output.pdf

Improving TOC quality with postprocessing:

For documents with a printed table of contents page, you can enable LLM postprocessing to refine the generated TOC:

pdfalive generate-toc --postprocess input.pdf output.pdf

Postprocessing uses an additional LLM call to:

• Remove duplicate entries
• Fix typos in section titles
• Cross-reference against any printed TOC found in the document
• Add missing entries and correct page numbers

Other useful options:

• --force - Overwrite existing TOC if the PDF already has bookmarks
• --ocr-language - Set OCR language (default: eng). Use Tesseract language codes like deu, fra, etc.

extract-text

Extract text from scanned PDFs using OCR and save to a new PDF with an embedded text layer:

pdfalive extract-text input.pdf output.pdf

This is useful when you want a searchable/selectable text layer without generating a TOC.

Development

We use uv to manage the library. To install locally can run e.g. with:

uv sync
uv pip install -e .

We use ruff for formatting and linting, mypy for static type checking, and pytest for running unit-tests. We also use pre-commit for ensuring high-quality commits.