Currently, run-local-model-lab is restricted to Apple Silicon and MLX. There's no equivalent skill for Windows, Linux, or Mac users who prefer Ollama as their local model runner.
I've built a cross-platform sibling skill, run-ollama-lab, to bridge this gap. What's currently done:
SKILL.md and reference.md, mirroring the structure of the existing MLX lab section-for-section.- The three registration points (
README.md, skills/README.md, skills/understudy/SKILL.md) updated.
- Passes
npm run skills:validate.
Testing status:
The change passes lint and all internal skill links resolve. I haven't yet completed a live end-to-end run (pull → serve → curl the OpenAI-compatible endpoint) — my machine has 8GB RAM, and pulling/running the flagship gemma4:e2b model caused a crash before the run finished. The documented commands match Ollama's published model library and docs, but aren't yet confirmed by a completed local run on my end.
Is a cross-platform Ollama skill something the project wants? If so, would you prefer a draft PR now for early structural feedback, or should I wait until I can validate the full flow on higher-RAM hardware?
Currently,
run-local-model-labis restricted to Apple Silicon and MLX. There's no equivalent skill for Windows, Linux, or Mac users who prefer Ollama as their local model runner.I've built a cross-platform sibling skill,
run-ollama-lab, to bridge this gap. What's currently done:SKILL.mdandreference.md, mirroring the structure of the existing MLX lab section-for-section.README.md,skills/README.md,skills/understudy/SKILL.md) updated.npm run skills:validate.Testing status:
The change passes lint and all internal skill links resolve. I haven't yet completed a live end-to-end run (pull → serve → curl the OpenAI-compatible endpoint) — my machine has 8GB RAM, and pulling/running the flagship
gemma4:e2bmodel caused a crash before the run finished. The documented commands match Ollama's published model library and docs, but aren't yet confirmed by a completed local run on my end.Is a cross-platform Ollama skill something the project wants? If so, would you prefer a draft PR now for early structural feedback, or should I wait until I can validate the full flow on higher-RAM hardware?