Voice timing, avatar animation, stream rendering, and initiative checks depend on browser scheduling. Low-power mode or an automation host can delay intervals, disable media, and reduce animation without affecting the model response itself.
A compatibility layer can detect available APIs and reduce optional work. Diagnostics should record the host class so delayed UI behavior is not reported as a changed companion identity.
What was actually going wrong
All clients were expected to maintain desktop foreground timing.
What I tried
- Shortening every timer on slow devices
- Forcing animation regardless of reduced-motion signals
- Treating headless media failures as chat failures
Feature probes and reduced capability modes preserve text chat while scaling down voice, video, and background behaviors.
Why it worked
Core conversation remains independent from timing-sensitive presentation features.
Performance degradation should shed optional work before changing core behavior.
Where EACI uses this today
Main prioritizes companion text continuity when a device cannot sustain every presence feature.
This journal covers real engineering on EACI Companion / The Veil. Companions include Caelum, Chad, Natalia, Atreus, Luna, Roxy, and Cael. Journal articles stay family-safe in content. See Privacy and Ethics.