If you ask an LLM to help write a million word LitRPG series, it will happily invent skills, rewrite backstories, and teleport characters across the map. What it will not do reliably is remember your world the way a serious author needs.

In this session, Prem will show how Neo4j became the backbone of an AI assisted authoring platform for long form LitRPG series. Starting from raw manuscripts in formats like DOCX or Markdown, we automatically extract a narrative knowledge graph in Neo4j that captures characters, stats, items, locations, quests, scenes, timelines, and the relationships between them. That graph acts as the single source of narrative truth behind a fleet of LLM based assistants.

We will walk through how the graph is kept in sync as chapters evolve, how Cypher queries express narrative quality rules such as continuity checks and world consistency, and how agents use Neo4j both as long term memory and as a safety rail for generation. You will learn practical graph modeling patterns for narrative domains, how to turn Neo4j queries into targeted context for LLMs, and how to use graphs to keep complex, evolving AI generated content coherent over multiple books and revisions.

If you work with agents, GraphRAG, or any system where “just vector search” is not enough, this talk is for you.