book-condenser 0.1.2

Read the essential book: create extractive abridgements that preserve the author's original passages.

Book Condenser

Create a shorter reading edition of a nonfiction book from the author’s original passages.

Book Condenser transforms an EPUB, PDF, DOCX, TXT, or Markdown book into a clean, tablet-friendly PDF abridgement. An AI model identifies the passages that carry the book’s central argument, evidence, concepts, chronology, and conclusions. The program then retrieves those passages from the source and assembles them into a shorter reading edition.

The result is shorter than the source, richer than a summary, and faithful to the author’s voice.

Preserve the author. Remove the excess.

How It Works

1. Recover structure
   The program identifies chapters, reading order, and back matter, while cleaning common extraction artifacts.

2. Validate the source
   A local structure report checks whether the recovered text is reliable before model-based selection begins.

3. Select essential passages
   The model determines the nonfiction form and selects coherent original passages that carry the book’s intellectual or narrative arc.

4. Balance the abridgement
   The program reduces redundancy, protects broad chapter coverage, limits overrepresentation of individual sections, and meets the requested target length.

5. Produce the reading edition
   The retained source passages are rendered as a professionally formatted, large-type PDF for tablet reading.

The AI acts as an editorial selector. The final edition remains grounded in the author’s original text.

Features

• Supports EPUB, PDF, DOCX, TXT, and Markdown input.
• Recovers structure from EPUB 2, EPUB 3, and text-based PDFs, including imperfect source files.
• Detects and excludes notes, bibliography, acknowledgments, indexes, and other non-reading matter.
• Stops before API calls when the parsed structure is unreliable or the source is likely image-only.
• Adapts passage selection to argumentative, historical, technical, biographical, and mixed nonfiction.
• Produces a tablet-optimized PDF as the primary output.
• Generates parsing and selection reports for traceability.

Requirements

You need:

• Python 3.10 or newer.
• An OpenAI API key for full condensation runs.
• A source book you are legally allowed to process and store.

Use Book Condenser with public-domain works, your own material, or works for which you have appropriate permission. Generated editions contain substantial source text.

EPUB is preferred when available because it usually provides cleaner chapter structure and text than PDF.

Installation

From PyPI, once released:

pip install book-condenser

From a local checkout:

python -m venv .venv
source .venv/bin/activate        # Windows: .venv\Scripts\activate
pip install -e .

Set your OpenAI API key:

export OPENAI_API_KEY="your-api-key-here"

Windows PowerShell:

$env:OPENAI_API_KEY="your-api-key-here"

Optionally choose the model:

export OPENAI_MODEL="gpt-5-mini"

When OPENAI_MODEL is not set, the CLI uses gpt-5-mini.

Quick Start

1. Check the source structure locally

Start with --parse-only. It validates the input and creates a report without sending book text to the API.

book-condenser path/to/book.epub \
  --output-dir out \
  --parse-only

Open the generated:

parsed_structure_report.md

Proceed when chapters are detected correctly, back matter is excluded appropriately, and the report indicates that extraction may proceed.

2. Generate the condensed reading edition

book-condenser path/to/book.epub \
  --output-dir out \
  --target-ratio 0.25 \
  --coverage-mode all \
  --chapter-max-share 0.08 \
  --pdf-font-size 14 \
  --apply-qc

A target ratio of 0.25 aims to retain approximately one quarter of the original book’s words.

The primary output is:

reading_abridgement.pdf

PDF Reading Edition

The default PDF is designed for comfortable reading on a small tablet:

• 7 × 10 inch portrait pages.
• Large 14 pt serif body text.
• Generous line spacing and clean chapter openings.
• Discreet markers between separated retained passages.
• Restrained running headers and page numbers.

Useful controls:

Option	Purpose	Default
--pdf-page-size	small-tablet, a5, or large-tablet	small-tablet
--pdf-font-size	Body font size from 11 to 20 pt	14
--pdf-font	auto, georgia, dejavu serif, or times	auto
--no-docx	Skip optional DOCX output	off

For larger text on a small screen:

book-condenser path/to/book.epub \
  --output-dir out \
  --pdf-font-size 15 \
  --no-docx \
  --apply-qc

Source Format Guidance

EPUB

EPUB is the recommended input. Book Condenser supports:

• EPUB 2 toc.ncx navigation.
• EPUB 3 navigation documents.
• Visible-heading recovery when navigation metadata is missing.
• Anchored subsections and common imperfect EPUB structures.

PDF

Text-based PDFs are supported. The program uses bookmarks when available and can attempt to recover sections from visible headings.

For a scanned or image-only PDF, run OCR first. When chapter boundaries are unreliable, provide a manual chapter map.

book-condenser path/to/book.pdf \
  --chapter-map examples/chapter_map.json \
  --output-dir out \
  --parse-only

Example chapter map:

[
  {"title": "Introduction", "start_page": 1, "end_page": 8},
  {"title": "Chapter One", "start_page": 9},
  {"title": "Chapter Two", "start_page": 28},
  {"title": "Bibliography", "start_page": 410}
]

Back matter remains visible in the structure report but is excluded from passage selection and word budgeting.

Main Controls

Argument	Meaning	Default
--target-ratio	Approximate share of source words retained	0.25
--candidate-ratio	Candidate passage pool before global pruning	0.42
--coverage-mode	Section coverage rule: all, major, or none	all
--chapter-max-share	Nominal maximum share from one chapter	0.08
--parse-only	Validate parsing without API calls	off
--apply-qc	Apply final model-based quality review	off
--chapter-map	Manual page map for difficult PDFs	none
--output-dir	Parent directory for generated run folders	abridgement_output
--reuse-output-dir	Replace prior generated artifacts in that folder	off

Outputs

A full run creates a folder such as:

out/book-<timestamp>/
    reading_abridgement.pdf
    parsed_structure_report.md
    selection_audit.md
    reading_abridgement.md
    reading_abridgement.docx
    book_metadata.json
    book_paragraphs.jsonl
    structural_overview.json
    chapter_candidates/
    scored_candidates.json
    global_selection.json
    quality_control.json

Files most users need:

File	Purpose
reading_abridgement.pdf	Final tablet-friendly reading edition
parsed_structure_report.md	Verification that the source was parsed correctly
selection_audit.md	Record of coverage and passage-selection decisions
reading_abridgement.docx	Optional editable copy

Keep output folders private by default. They may contain verbatim passages, local paths, and model-generated selection analysis.

Cost and Privacy

--parse-only runs locally and does not require API calls.

A full run sends structural context and source excerpts to the configured OpenAI model. API usage increases with book length, candidate-pool size, and use of final quality-control review.

Do not process confidential or restricted material unless your rights and API/provider settings permit it.

Development

Install development dependencies:

pip install -e ".[dev]"

Run checks:

ruff check .
pytest
python -m build
twine check dist/*

The package exposes:

book-condenser

and:

python -m book_condenser

License

Book Condenser is licensed under the PolyForm Noncommercial License 1.0.0. Commercial use requires a separate commercial license from the licensor.

Disclaimer

Book Condenser is provided as-is and does not provide legal advice. You are responsible for ensuring that source material, API use, and generated outputs comply with applicable copyright law, contract terms, platform policies, and other obligations.