Intro

Building an AI-based novel setting management platform that automatically extracts and organizes characters, locations, and items from stories while detecting consistency issues.

• Enables writers to view and manage novel settings (characters, items, places, relationships) in a structured way.
• Incorporates AI-powered extraction to detect setting conflicts (e.g., timeline or attribute mismatches).
• Provides a UI for exploring setting data, making it easier to maintain narrative consistency in long-form writing.
• Designed with extensibility for LLM-based reasoning and retrieval-augmented workflows.

Tech Stack

• Languages: Python, TypeScript, SQL
• Backend / API: FastAPI (Python), PostgreSQL with SQLAlchemy + Alembic for ORM/migrations
• Frontend: React (Next.js), Tailwind CSS
• AI/NLP: OpenAI API (LLM-based entity & relation extraction), custom prompt pipelines for tagging and conflict detection
• Database Design:
  • EntityAttributeAppearance – track character traits, items, and places across scenes
  • CharacterScene, CharacterItem, PlaceScene – link entities to narrative contexts
• Infra: Docker (local development), Vercel (frontend deployment), Render/Railway (backend hosting)

Key Contribution

• Designed a relational database schema in PostgreSQL to capture narrative entities and their relationships, including EntityAttributeAppearance, CharacterScene, CharacterItem, and PlaceScene.
• Implemented the backend architecture with FastAPI and PostgreSQL, exposing REST APIs for managing characters, items, places, and scenes, while ensuring consistent cross-entity retrieval.
• Developed an LLM-powered entity extraction pipeline using the OpenAI API, capable of identifying characters, items, places, and relationships directly from novel text and mapping them into the database.

Problem & Solution

Problem 1: Structuring complex relationships among entities

• Problem: Novels involve diverse entities (skills, items, places, characters) that must maintain multiple relationship types. For example, places require hierarchical relations (country → city → district).