Core concepts

NVS reads your manuscript scene by scene and maintains a few ledgers. Understanding them is most of understanding the app.

Scenes

The atomic unit. Each scene is one Markdown file (or section) with a stable scene_id — the anchor everything else hangs off, so renames and reorders never break the analysis.

Threads

A thread is a through-line — a plot, a relationship, a question the reader is holding. The thread ledger tracks where each one advances, stalls, or resolves, so a subplot can't quietly vanish for ten chapters.

Cast presence

Who is on stage in each scene. Presence makes rotating POV and ensemble casts legible — and catches a character being in two places at once.

Reveals

Information the reader learns, and when. The engine pairs a reveal with its setup so payoffs land in the right order.

Coherence findings

The output you act on: stale scope, dropped threads, contradictory state, duplicated reveals. Findings are surfaced as you write — an editing pass that runs continuously rather than at the end.

Durability

Your .md scenes are the source of truth. The analysis database is derived and can be rebuilt from source at any time — so the app's schema can evolve without putting your manuscript at risk.