pi-set-model

Pi extension that remembers model and thinking level per project folder

Packages

Package details

extension

Install pi-set-model from npm and Pi will load the resources declared by the package manifest.

$ pi install npm:pi-set-model
Package
pi-set-model
Version
0.1.2
Published
Jul 22, 2026
Downloads
212/mo · 212/wk
Author
hk_net
License
EUPL-1.2
Types
extension
Size
25.5 KB
Dependencies
0 dependencies · 2 peers
Pi manifest JSON
{
  "extensions": [
    "./set-model.ts"
  ]
}

Security note

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

README

pi-set-model

Remembers Pi's selected model and thinking level for each project folder. Requires pi >=0.81.1.

After selecting a model and thinking level, explicitly save that combination for the current working directory with /set-model set. New Pi sessions opened in that directory restore it, rather than retaining the model selected in another project. Later /model or thinking changes do not modify the saved preference unless you run /set-model set again.

Preferences are stored in the project at .pi/set-model.json; they are never shared with another project. Pi must trust the project before the extension reads or writes this project-local setting.

Install

pi install npm:pi-set-model

Or install the full collection:

pi install git:git@github.com:hknet/pi-extensions@main

Commands

/set-model          # view the saved preference for this folder
/set-model view     # view the saved preference for this folder
/set-model set      # save the active model and thinking level
/set-model clear    # stop restoring a preference for this folder

Autocomplete completes /setm... to /set-model without adding a trailing space, then offers view, set, and clear.

The folder is Pi's current working directory. Run Pi from the project root when you want one preference for the whole project. The setting is saved at <cwd>/.pi/set-model.json.

If the saved model is unavailable or has no configured API key, Pi keeps the active model and shows a warning. Thinking levels are restored after the model and are clamped to that model's supported levels by Pi. When the session ends, the model and thinking level that were active before the project preference was applied are restored.

Issues and feedback

Found a bug or have a feature request? Please report it on GitHub Issues.

For security vulnerabilities, please use GitHub's private vulnerability reporting instead of opening a public issue.