MLX 4-bit conversion of the
Wyrdsekai companion voice base,
for Apple Silicon hosts. Pulled automatically by
wyrd setup on macOS (arm64) when the MLX
voice runtime is bootstrapped; serves on
:8201 via
mlx_lm.server.
GGUF equivalent for llama.cpp hosts:
wyrdsekai/companion-3.5-4b-gguf.
Training data: synthetic corpora only (no user conversations).
