docs-summarizer

CLI tool to summarize documents and books with AI.

Features

• Auto-detect and skip front/back matter (covers, dedications, etc.)
• Two summary modes: concise (quick overview) or detailed (with examples)
• Interactive chapter selection with -i flag
• Modular provider architecture (Claude CLI by default, extensible for other APIs)
• Progress tracking with chapter-by-chapter logging
• Outputs both individual chapter summaries and complete book summaries

Installation

git clone --recursive git@github.com:fabiogaliano/docs-summarizer.git
cd docs-summarizer
bun install
bun run setup

Usage

# Summarize a single EPUB
bun run src/index.ts ./book.epub

# Summarize all EPUBs in a folder
bun run src/index.ts ./books/

# Interactive mode - select chapters manually
bun run src/index.ts ./book.epub -i

# Specify output directory
bun run src/index.ts ./book.epub -o ./output/

# Detailed mode with examples
bun run src/index.ts ./book.epub -m detailed

# Use a different model
bun run src/index.ts ./book.epub --model sonnet

# Skip if already summarized
bun run src/index.ts ./books/ --skip-existing

Options

-o, --output <dir>     Output directory (default: same as epub)
-m, --mode <mode>      Summary mode: concise (default) or detailed
--provider <provider>  AI provider: claude-cli (default)
--model <model>        Model to use (default: haiku)
--skip-existing        Skip if summaries already exist
-i, --interactive      Interactively select chapters to summarize
-h, --help             Show this help

Output Structure

{book-name}/
├── book.json                 # Manifest from epub-splitter
├── chapters/                 # Raw markdown chapters
├── summaries/
│   ├── concise/
│   │   ├── 01_introduction.md
│   │   ├── 02_chapter_1.md
│   │   └── ...
│   └── detailed/
│       ├── 01_introduction.md
│       └── ...
├── summary-concise.md        # Quick book overview
└── summary-detailed.md       # Rich book summary

Architecture

Layers

• EPUB Processing: Handles file splitting and chapter detection
• Providers: Pluggable AI backends (Claude CLI, future: OpenAI, Anthropic API)
• Summarizers: Chapter and book-level summarization logic
• UI: Interactive chapter selection with prompts library
• Output: File writing and formatting

Smart Chapter Detection

Automatically skips:

• Front matter (covers, dedications, tables of contents)
• Back matter (indexes, bibliography, acknowledgments)

Prefers starting from:

1. "Introduction" chapter
2. First preface/prologue/foreword
3. After last front matter section

Summary Modes

Concise (~200-300 words per chapter):

• Key points only
• Quick overview suitable for rapid reading

Detailed (~500-800 words per chapter):

• Includes notable examples and case studies
• Practical takeaways for non-fiction
• Better context preservation

Requirements

• Bun runtime
• Rust toolchain (to build epub-chapter-splitter)
• Claude CLI configured and available in PATH

Future

• PDF support
• DOCX support
• Additional AI providers (OpenAI, Anthropic API, Ollama)
• Batch processing with progress reporting
• Custom prompt templates
• Highlight extraction