Core concepts

Understand sites, captures, screens, tutorials, storyboards, renders, videos, and personas.

Shadow is built on a few core entities. Understanding them up front makes the rest of the docs read naturally. Tutorials are the first product line built on this substrate; Insights, Demos, and Assist reuse the same entities.

Site

A site is one product. It owns a public key, an API secret, and a set of branding overrides. You can install the SDK on multiple domains, but each one should have its own site so captures are scoped correctly.

Capture

A capture is a single user session recorded by the SDK. It contains a stream of events (navigation, clicks, inputs) and an anonymized DOM snapshot for each distinct screen.

Screen

A screen is a canonical page within a site — the AI clusters all captures of /projects/new into one screen. You can label screens with a human title ("New project form") and a purpose ("Capture name, plan, and billing email") to give the AI better grounding.

Tutorial

A tutorial is your authored intent — “Show how to invite a teammate.” It references a goal, an optional persona, and becomes a storyboard after composition.

Storyboard

A storyboard is an ordered list of scenes. Each scene picks one screen, a highlight target, a narration script, and a target duration. The dashboard lets you reorder, edit, and regenerate individual scenes.

Render

A render is one pass of Shadow’s video pipeline that produces an MP4 from a storyboard. Renders are queued and processed asynchronously; the dashboard reports status until the video is ready.

Video

A video is the published artifact of a successful render. It has a unique publicSlug, a presigned playback URL, an embed-enable flag, and optional marketing demo flag.

Persona

A persona describes an example user (e.g. "Jane, a solo designer, plan: starter, sample data: [...]"). Shadow uses personas to keep narration grounded in realistic data without leaking real PII.