lenses

A lens is a saved way of looking at a recording: a prompt for the language model, a declared shape for the answer, and a view that renders it. "What was decided?", "Who owes whom what by when?", "Summarise this for someone who missed it" — each is a lens.

Run a lens on one recording, or flip it always-on ("Run on every Recording") and every future recording arrives with that insight already extracted, by the insight step of the pipeline. Always-on lenses share the local model's time, so the app is honest about the cost when you enable one.

your library

Settings → lenses holds your collection in two tabs: installed (with run, edit, enable, delete per lens) and discover (the marketplace). yapyap preinstalls a starter set during onboarding — all editable, all removable.

making them yours

Every lens is editable, and you can build your own from scratch:

• the prompt — the actual instruction to the model. Edit it freely; Reset to default brings an installed lens back to factory.
• the schema — the shape of the answer (a checklist, a table of owners and due dates, prose…). The builder gives you a full editor with live preview.
• the view — how results render. Regenerate view rebuilds it from the schema if you've changed the shape.

Drafts autosave while you edit; Discard draft abandons them explicitly.

when a lens stays quiet

Lenses fail politely, in plain language, on the recording itself:

• "The Lens output did not match its declared schema." — the model's answer didn't fit the declared shape; run it again or loosen the schema.
• "Lens needs pipeline inputs … which have not run yet" — it depends on a pipeline step (say, speaker recognition) that hasn't completed or failed.
• "Deferred … will pick up on the next Recording." — the insight engine was busy or unavailable; nothing is lost.