Technical Documentation

How Writan works under the hood. This section explains the architecture, data model, and AI generation pipeline.

Architecture

• The Node Tree Architecture — The hierarchical data model that structures every story
• Data, Generation & Output Layers — The three layers that exist on every node
• Database Schema — The PostgreSQL schema (projects, nodes, pips)
• Store Architecture — Zustand stores, local-first sync, and data flow patterns

AI & Generation

• How AI Generation Works — The generation endpoint and prompt composition (partial)
• What Data Feeds Into Generation — How context is composited from the node tree (partial)

Feedback & Analysis

• The Pip System — The universal feedback surface (craft pips implemented)
• The Writing Rules Engine — 27 mechanical rules with client-side detectors