Description
The system needs to know how many concurrent sessions a body can handle based on experience type and GPU specs.
- Flat experiences: depends on GPU VRAM and experience VRAM requirement (e.g. RTX 4080 SUPER 16GB can run 3x mercator-talks at ~5GB each)
- VR experiences: always 1 per body (headset exclusive)
- Desktop (screen share): 1 per body (DXGI captures one desktop)
Capacity should be computed from experience library VRAM data + hydrabody GPU reporting (now working via nvidia-smi fix in v1.8.10). Currently hardcoded to 1 via the body occupancy guard in hydraneckwebrtc v1.10.23.
Depends on multi-session support actually being implemented (see related investigation issue).
Nebula's reasoning: This is a new capability — computing dynamic session capacity from GPU specs and experience VRAM requirements. Medium priority because the current hardcoded limit of 1 works correctly for all existing use cases (VR and desktop are always 1 anyway). This only becomes actionable after issue #67 (multi-session investigation) determines whether parallel flat-experience streams are feasible. Description is clear with good breakdown by experience type and explicit dependency noted.