gpu-crab 0.1.0

While a local-llm tool runs on your Ollama box, a crab on a treadmill shows the loaded model, VRAM, GPU/CPU split and elapsed time above the prompt (amber when a model spills to CPU), and /gpu keeps a 'kept on the LAN' tally of what local models read instead of the cloud.

David Coulson 0 on davidcoulson/crab-workshop

Install

Claude Code
/plugin marketplace add davidcoulson/crab-workshop
/plugin install gpu-crab@crab-workshop

Run both lines inside Claude Code.

Hooks into

  • Prompt
  • Tool calls
  • Slash commands

Also in davidcoulson/crab-workshop

More prompt mods

All prompt mods
Repo stars
0
Repo forks
0
License
MIT
Last commit
Checked
d964ccc,
Source on GitHub