@magnitudedev/pi-extension

Run local AI models in Pi with Magnitude

Packages

Package details

extension

Install @magnitudedev/pi-extension from npm and Pi will load the resources declared by the package manifest.

npm repohome report
$ pi install npm:@magnitudedev/pi-extension
Package
@magnitudedev/pi-extension
Version
0.0.3
Published
Sep 16, 2026
Downloads
508/mo · 508/wk
Author
anerli
License
MIT
Types
extension
Size
273.9 KB
Dependencies
4 dependencies · 3 peers
Pi manifest JSON
{
  "image": "https://raw.githubusercontent.com/magnitudedev/magnitude/main/assets/readme/pi-preview.png",
  "extensions": [
    "./dist/magnitude.js"
  ]
}

Security note

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

README

Magnitude for Pi

Run Pi on local models. Free, private, and offline.

Magnitude runs the best local models for your hardware, plugged into the agent you already use. It profiles your machine, recommends the models that fit, then downloads, tunes, and runs them.

Get started

Install the extension:

pi install npm:@magnitudedev/pi-extension

Open Pi and accept the setup offer. Magnitude walks you through the best models for your machine, downloads your selection, and connects it to Pi. You don't need Magnitude installed beforehand.

You can also start setup anytime with /magnitude-setup.

Requires Pi 0.83.0 or newer. Magnitude supports macOS and Linux, and Windows through WSL. An internet connection is needed for installation and model downloads; after that, you can work offline.

Why Magnitude?

  • Free to run: no token costs, API keys, or rate limits
  • Private and offline: local model requests stay on your machine
  • Recommends what fits: the best models for your hardware, with estimated tok/s
  • Tuned end to end: inference settings chosen for your machine
  • Models on demand: loaded when Pi needs them, unloaded when idle or memory gets tight
  • Live performance stats: see model loading, prompt processing, and time spent working

After each completed run, a summary shows the model, time worked, time to first token, and tokens per second. It stays in your chat history so you can refer back to it.

Using your models

Choose an installed Magnitude model from Pi's /model selector. To discover or install more models, ask your agent:

Use Magnitude to recommend the best local models for my hardware and help me install one.

Use /magnitude-setup to run guided setup again, or /stop-model to unload the active model and free its memory.

Already using Magnitude?

Connect your installed models to Pi:

magnitude connections add pi

Then restart Pi or run /reload.

Learn more