Novel Storyteller

The project focuses on the development of an AI-powered story generation pipeline capable of generating structured, age-appropriate narratives from high-level user inputs such as topic, genre, age group, fear level, and story length. The system combines Large Language Models (LLMs), graph-based workflow orchestration, deterministic validation, and state management to ensure coherent and contextually consistent storytelling. The primary objective of the project is to improve controllability, narrative consistency, and reliability in automated story generation systems. The architecture follows a modular multi-stage pipeline beginning with specification building and narrative pattern selection. Scene plans are generated using LLMs or deterministic fallback mechanisms and are organized into a Directed Acyclic Graph (DAG) where each scene represents a node with dependencies and contextual requirements. A topological execution engine processes scenes sequentially while maintaining a centralized narrative state. Each generated scene passes through state-diff extraction, safety validation, chronology checks, character consistency verification, and repair loops to ensure narrative quality and structural correctness before updating the global story state. The system demonstrates the integration of AI generation workflows with structured execution and validation pipelines for long-form content generation. Its modular and extensible design makes it suitable for applications such as educational storytelling platforms, AI-assisted creative writing tools, personalized children’s stories, and interactive fiction systems.