NovelForge

LLM-assisted web novel writing application. Designed for long-form serial fiction (500+ chapters, ~1M words) with AI generation, a persistent knowledge graph, and RAG-powered context retrieval for consistency at scale.

Stack

• Frontend: Vite 8 + React SPA, TipTap v3 editor, Tailwind CSS
• Backend: Fastify, Drizzle ORM, BullMQ
• Database: PostgreSQL 16 (pgvector), Redis 7
• LLM: Anthropic API (Claude), Voyage AI (embeddings)
• Auth: Clerk

Quick Start

Prerequisites: Docker, Node.js 22+

# Install dependencies (npm workspaces — run from root)
npm install

# Copy environment configs
cp .env.example .env                       # Docker infra (Postgres, Redis)
cp backend/.env.example backend/.env       # Server secrets + runtime config
cp frontend/.env.example frontend/.env     # VITE_-prefixed public config
# Fill in API keys in backend/.env

# Start databases
docker compose up -d

# Run migrations
cd backend && npm run db:migrate

# Start dev servers (two terminals)
npm run dev:backend     # Fastify on :3001
npm run dev:frontend    # Vite on :5173

Project Layout

docs/               Developer documentation (architecture, index)
user_guide/         Operational docs (deployment, subscriptions, env config)
sessions/           Implementation session logs (per build phase)
shared/             Shared TypeScript types
frontend/           Vite + React SPA
backend/            Fastify API server + BullMQ workers
scripts/            Seed data, benchmarks, backups
tests/              Unit and integration tests

Documentation

See docs/index.md for a full map of all project documentation.

Spec

The full MVP specification is in MVP_REVISED_v4.md. Implementation follows the phased plan in Section 18.