homie: a Mac notch voice assistant running Qwen 3.5 4B locally, with rules bypassing the LLM
KungFuSaifooo · reddit · 2026-11-22
homie is an open-source (MIT) voice assistant living in the Mac notch: Qwen 3.5 4B (Q4KM) via llama.cpp on Metal, Kokoro-82M TTS on the Neural Engine, Apple on-device STT. Since the 4B model only hits 15-30 tok/s, 600 lines of pattern rules handle everyday commands (214/214, zero false fires), skipping the LLM entirely; prompt state caches to disk. On an M5 16GB it reaches 0.43s p50 to first sound for commands, 1.17s for LLM turns. The README documents failed experiments (spec decoding, MLX vs llama.cpp, quant bake-off).
Related event: Open-source homie runs a fully local voice assistant in the Mac notch(2 posts)→
More from Infra
- Gemini Ultra's $100 plan includes 1000 monthly Colab credits—about 55 H100 hours — iskander · 2026-11-22
- Hy team pushes back on ultra-sparse MoE: sparsity recovers efficiency, not quality — douglovefishing12 · 2026-11-22
- SpaceX lenders push back on $40B of AI chips launched into space — flngr · 2026-11-22
- Homie: an offline Mac notch voice assistant running Qwen 3.5 4B locally — KungFuSaifooo · 2026-11-22
- AMD and PyTorch ship a playbook to run GPT-OSS-20B inference locally with ROCm — PyTorch · 2026-11-22
- Microsoft's full Outlook outage PIR confirms AI tooling wasn't to blame — aronchick · 2026-11-22