Status: Phase 1 placeholder. Substantive content lands in Phase 5 per §8.1.

Statement

Calls are 1:1 voice and video over WebRTC, with signalling carried on NIP-17-gift-wrapped kind:25050 events (NIP-100 alignment). The UI prioritises clarity of call state — ringing, connecting, in-call, ended — and respects that the protocol's live-call window is short: signal replays older than the window are dropped, and only call_ended history persists on the conversation timeline.

Why

Calls UI is the surface where the gap between "what the protocol does" and "what the user sees" is largest. ICE candidate exchange, TURN credential refresh, exporter-secret derivation, the 30s-no-answer / 15s-ICE-timeout watchdogs, the WakeLock during the call — none of this should be visible to the user, but all of it determines what state the UI is in.

Treating calls as a generic phone metaphor breaks at the seams: there is no "voicemail" because there is no centralised inbox, the call ends when either party drops because there is no server holding the channel open, and rejoining a call is conceptually a new call rather than reconnecting to the same one.

How

This stub is intentionally thin. Substantive C-1 content matures in Phase 5. Expected coverage at maturity:

Outgoing call ringing surface, with a clear cancel affordance
Incoming call surface, with accept / decline at A-3 hit-target sizes
In-call surface: minimal UI, audio/video streams, mute / camera-off / hangup; quality indicator only when degraded
Call-ended surface as a brief acknowledgment, not a modal — the conversation timeline shows the call_ended history message
Permission-denied surfaces (mic, camera) per §7.4
Reduced-motion (A-2) honoured through the entire call lifecycle

Phase 5 fills these in with audit citations and per-surface specifications.