ytb-books: Book Content Automation with Neo4j GraphRAG

Creating educational video content from books has always been a time-consuming process — from extracting key concepts, generating engaging scripts, creating visual representations of complex ideas like mathematical formulas, to assembling everything into polished videos. The ytb-books skill pack transforms this entire workflow into an automated, end-to-end pipeline powered by modern AI and graph database technologies.

What is ytb-books?

The ytb-books skill pack is a complete solution for book-based content automation that leverages Neo4j GraphRAG (Graph Retrieval-Augmented Generation) to intelligently generate episode sequences from book content. Unlike traditional RAG systems that rely on vector similarity alone, GraphRAG uses the semantic relationships between concepts stored in a Neo4j graph database to create more coherent, contextually rich educational episodes.

This open-source MIT-licensed tool is part of the ClickAround Creator Pro ecosystem and provides developers and educators with a production-ready framework for transforming static book content into dynamic, multi-format educational materials.

Core Capabilities

Neo4j GraphRAG Integration

The foundation of ytb-books is its tight integration with Neo4j, a native graph database. Instead of treating book content as isolated text chunks, ytb-books structures knowledge as interconnected nodes:

• Document nodes represent entire books or sections
• Episode nodes contain generated video scripts with metadata like episode number, title, and summary
• Scene nodes break down episodes into individual visual segments
• BookChunk nodes store the raw text content with optional embeddings

This graph structure enables intelligent navigation through book content — finding related concepts, identifying prerequisite knowledge, and generating logical episode sequences that follow the natural flow of learning.

LaTeX Math Formula Rendering

For STEM education content, ytb-books includes sophisticated LaTeX formula handling through its math-formula service. This component renders mathematical expressions as high-quality PNG images suitable for video overlays, while also providing Unicode text alternatives for subtitles and TTS (text-to-speech) generation.

The implementation uses MathJax v4 for TeX parsing and Sharp for SVG-to-PNG conversion, with careful attention to rendering quality and performance considerations. The system automatically handles color adjustments for different backgrounds and removes unnecessary wrapper elements to ensure clean output.

End-to-End Video Generation Pipeline

The complete automation pipeline follows this sequence:

1. Extract book content and store in Neo4j graph database
2. Use GraphRAG to identify optimal next episode based on current progress
3. Generate script using ContentPlanner service
4. Create scene images using NanoBanana API
5. Render mathematical formulas as PNG overlays
6. Synthesize speech using Gemini TTS
7. Compose final video using FFmpeg
8. Upload to YouTube platform

Technical Architecture

The ytb-books architecture is built around TypeScript services that work together in a modular fashion:

• BooksVideoService.ts serves as the main orchestrator, coordinating the entire pipeline
• ContentPlannerService.ts handles AI-driven script generation
• GhibliImageService.ts manages image generation for scenes
• MathFormulaService.ts handles LaTeX rendering and Unicode conversion
• Neo4jService.ts provides full CRUD operations for the graph database

The system supports multiple visual styles including math_character (primary), dark_academia, watercolor, minimalist, and retro_comic — allowing creators to match their brand identity or educational goals.

Getting Started

Installation is straightforward with the ClickAround plugin system:

/plugin marketplace add https://github.com/clickaround/marketplace
/plugin install ytb-books@clickaround

After installation, you'll need to configure your Neo4j instance connection details and set up the necessary APIs for image generation and text-to-speech. The system is designed to be extensible, making it easy to swap out components like the image generator or TTS service with alternatives that better fit your infrastructure.