Systems
Wavefront
Audio devices, streaming, sample timing, immutable DSP graphs, mixing, voice boundaries, and acoustics direction.
Wavefront owns audio capture, devices, streaming decode, sample timing, DSP, mixing, music behavior, spatial playback, and optional acoustics. Collective owns optional online-service and voice-session policy; networking owns transport.
The intended signal path
- 01Source and stream
Bounded decode and buffering prepare audio without blocking the real-time callback.
- 02Immutable DSP graph
Prepared graph state crosses into the audio thread without shared mutable control.
- 03Spatial mix
Listener, source, music, and optional acoustics contribute to the output.
- 04Device
A recoverable backend feeds the selected hardware under a real-time-safe contract.
Current foundation
The architecture, ownership boundaries, and a meridian-audio scaffold crate exist. A mixer, decoder, device backend, music runtime, voice stack, and acoustic solver do not.
Why the boundary matters
Audio callbacks cannot wait for the world, allocate without bounds, or inherit arbitrary engine locks. Wavefront’s design prepares work away from the callback and crosses the boundary through bounded, inspectable state.
Optional voice and acoustics packs must also disappear cleanly when disabled. They should not leave tasks, listeners, allocations, or package data behind.