How many review → fix rounds the AI runs when creating or rebuilding a slide — in deck generation, build-a-slide, and the editor's assistant. Each round renders the slide, runs the vision reviewer (same rules everywhere) and fixes any blockers. More rounds = better layouts but slower and more AI credits.

NVISO Slide Library slides carry a short AI content summary the generator uses to find relevant slides to reuse. New and edited slides get one automatically; this backfills any that are still missing (processed in batches until done). Admins only.

These instructions steer the deck generation agents (Outline, Designer, Reviewer) and the brand rules shared with the editor's AI assistant. Changes take effect on the next generation job (and within a minute for the editor assistant) — no redeployment needed. Reset returns a prompt to the built-in default shipped with the app.

Background templates offered in the deck editor (◧ Background in the Insert tab, and the "New slide" gallery). Each entry has a key, title, a CSS swatch for the preview, optional section cls, Reveal attrs (data-background-*), an optional css snippet and onDark (flips headings to white). {ASSETS} in values resolves to the deck's assets folder.

The slide-type categories layout templates are filed under in the library, and that the deck generator routes on — an agenda slide reaches for an agenda layout, a comparison for a comparison one. Each type has a stable key (used by the generator + stored on layouts) and an editable title. Add or rename freely; removing a type leaves any layouts filed under it as Uncategorised until re-filed. Keep section-divider — it also drives the per-deck divider rotation.

Curated NVISO facts the deck generator can draw on when writing slide content — service one-liners, methodology steps, framework blurbs (OWASP/MITRE/NIST), compliance facts, standard stats/definitions. Keep each item a short, self-contained, accurate fact; avoid client-specific or time-sensitive detail unless you keep it current. Keywords help the system find it (optional — the title and text are matched too). Facts are pulled into the outline at generation time, beneath the deck's own brief and uploaded files.

generation = the deck pipeline (designer + reviewer + wireframer); chat / edit = the editor assistant (free-form + layout-locked slot-fill); outline, review, describe-*, knowledge, picklayout, interview = the supporting AI features. Every model call is attributed here, so this is the FULL AI spend, not just generation.

Refreshes automatically every 10 seconds while this tab is open.

Token and cost figures per generation exist for decks generated since worker v22; older ones show 0.