Models
millfolio runs a language model on your Mac — it reads your documents and does the on-device reasoning. The model picker (in the status bar at the bottom) is a small catalog: download the models you want, and switch between the ones you have.
Download or use
Each model in the catalog is either:
- Downloaded — shown with Use (switch to it) or In use (the current one). Switching reloads the engine onto that model.
- Not downloaded — shown with Download. Fetching a model is a background job with a progress bar; when it finishes the model flips to Use.
Downloads happen through the app, not at install time — so getting the app is fast, and you only pull the weights you actually want.
“Won't fit in memory”
Larger models need more RAM than smaller ones. millfolio checks your Mac's memory and greys out any model that's unlikely to fit — its button reads Won't fit in memory, with a tooltip showing what it needs versus what you have — so you don't download something that can't run. The model you're currently running is never flagged.
Getting started is automatic. On first run millfolio provisions a sensible default chat model and the embedding model it needs to index your vault, in the background — so a new install works out of the box, and the catalog is there when you want to add or switch.