hominfer
Self-hosted AI, minus the yak shaving.
hominfer runs local AI models like home services. Paste a Hugging Face repo and it tells you which variants fit your GPU, installs the missing engine, and gives every model a card to start, stop and profile with live VRAM, RAM, CPU and GPU load.
It is engine-agnostic: llama.cpp, vLLM, Ninfer are configuration, not code. It targets Linux / WSL2 with NVIDIA or AMD, including APUs.
Highlights
- Start/stop with profiles from any browser on the LAN
- Live usage: VRAM, RAM, CPU, GPU load
- Hugging Face “does it run here?” verdict per variant
- Engine catalog installs without sudo
- Engine metrics: requests, KV cache, prefix cache
- Usage stats over 24h to 90 days
Gallery
