KuberwastakenKuberwastakenCommunityยท1 Agent Skills Included

litmus

Autonomous overnight ML experiments with parallel research workers

Runs parallel machine learning experiments on your GPU overnight, with each worker on its own git branch. Eliminates manual experiment tracking, lost results, and repeated trial-and-error through a shared skills library and structured attempt records. A Director steers workers every two hours, a Synthesizer distills findings nightly, and a morning digest delivers the research narrative to your chat.
npx skills add Kuberwastaken/litmus --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install Litmus?โ–ผ

Run `npx skills add Kuberwastaken/litmus --all -g -y` in your terminal to install it globally, then run the setup script once to download training data.

What does Litmus do overnight?โ–ผ

It spawns multiple research workers that run ML experiments on your GPU, each on its own git branch, while a Director steers them and a Synthesizer distills findings into a reusable skills library.

Do I need a GPU to use Litmus?โ–ผ

Yes. Litmus requires an NVIDIA GPU with CUDA on Linux or macOS, plus git, python3, and the uv package manager.

How do I check experiment results?โ–ผ

Run the included status and results scripts to see per-agent experiment counts, leaderboards, and the full git history of every experiment commit.

Can I use Litmus without ML expertise?โ–ผ

Mostly yes. Setup is conversational and your agent handles scheduling and worker management, though basic ML knowledge helps you interpret the morning research digest.

Related Repositories in Data & Analytics

View All in Data & Analyticsโ†’