vhotseat lets a local LLM agent drive ComfyUI on one GPU overnight without OOM
wandermyz · reddit · 2026-10-19
The author built open-source vhotseat (MIT) so a local LLM agent and ComfyUI can share one consumer GPU: it unloads the LLM before renders, frees ComfyUI models before LLM loads, never interrupts running jobs, and restarts ComfyUI only when idle to clear leftover RAM. On an RTX 5070 Ti 16GB, a pi agent serving Qwen3.8-Flash-Next (128K ctx) drove MiniMax H3 int8 video renders overnight — two render→review rounds, vision-based take selection, ffmpeg concatenation, four GPU handovers, zero OOM. Also ships a comfy-render CLI for batching and a generic vhotseat hold NAME -- COMMAND wrapper for any GPU job. Linux/NVIDIA only for now; install via uv tool install vhotseat.
More from coding & agent
- Bot-Interface: one /.well-known/bi.md file tells AI agents your site's APIs — natesiggard · 2026-10-19
- 'Vibe Coding' Has Lost Its Meaning: From Midnight Hacks to 500-Hour Agent Projects — chrisalbon · 2026-10-19
- Next Token Podcast #006: Grok Bot & Muse Hands-On, AI Reverse Engineering, Agent Pitfalls — vista8 · 2026-10-19
- Image Lab 2 rebuilds AI image workflow: AI acts as creative director before generating — woulatte · 2026-10-19
- Biscotti: a self-hosted Instinct-like iMessage assistant in ~850 LOC — lucasgelfond · 2026-10-19
- CS Freshman Money Guide: Ship a Real Tool First, Upgrade Gear Later — sven_ai · 2026-10-19