Mostlysane — Local AI Config Generator

Get working on real work with local AI. Don't go crazy over AI. Stay Mostlysane.

Failed to load model database. Make sure the page is served from a local server (not file://).

Recommended Model

Run this from your llama.cpp build directory. Models go in ~/AI/models/ by default — put them there or edit the paths above.

One-time calibration per model. The profile is quantization-invariant — same file works for Q8, Q5, Q4 quants.

No manual steps. Just run this in your terminal.

Installs dependencies, clones the Mostlysane fork, builds llama.cpp, and optionally downloads a model. Use the manual guide below for custom setups.