@celestoai/pi
v0.1.3
Published
Run Pi coding tools in a Celesto computer
Maintainers
Readme
Celesto for Pi
Run Pi's coding tools away from your laptop while keeping its interface and conversation history local. A Celesto computer is an isolated remote computer that holds your project and runs the tools. Model credentials are the secrets Pi uses to access an AI model. They stay on your machine.
Read the complete Pi cloud setup guide for synchronization, security, lifecycle, and troubleshooting details.
Quickstart
Before starting, install Pi and configure it with a model provider. You also need Node.js 22.19 or newer and a Celesto API key.
Install the extension:
pi install npm:@celestoai/piSign in once with the Celesto CLI:
pip install celesto
celesto auth loginAlternatively, export CELESTO_API_KEY or add it to the project's local .env file:
CELESTO_API_KEY="your-celesto-api-key"The bundled Celesto TypeScript SDK checks the shell environment first, then .env, then credentials saved by celesto auth login.
From the project you want to work on, start Pi:
pi --celestoThe extension creates a Celesto computer, copies the project to $HOME/workspace, and runs Pi's read, write, edit, and bash tools there.
Copy changes back
The Celesto workspace is the active copy while Pi is running. Copy changes between it and your local project explicitly:
/celesto syncThe sync command preserves independently changed files under .celesto-conflicts/ instead of overwriting local work.
Check the computer
/celesto statusThis shows the computer name, workspace, cleanup behavior, and current sync revision.
Keep the computer
Computers created by the extension are deleted only after a successful final sync. Keep one for later use with:
/celesto keepThe command prints the exact celesto computer delete command to use when you no longer need it.
Reuse an existing computer
Get a computer name from celesto computer create or celesto computer list, then pass it to Pi:
pi --celesto --celesto-computer curieThe extension never deletes a computer selected by the caller. If $HOME/workspace already contains files, those files remain authoritative until you run /celesto sync. A non-empty legacy /workspace is moved into $HOME/workspace automatically when the home workspace is empty.
Files that are not copied
The extension reads .gitignore and then applies .celestoignore overrides. It also excludes common secrets, dependency directories, build output, files larger than 25 MB, and symbolic links. .git remains available so Pi can inspect branches and diffs.
Add project-specific patterns to .celestoignore:
fixtures/private/
*.large-test-dataModel-provider credentials are never copied into the Celesto computer. CELESTO_API_KEY is used only by the local SDK to contact Celesto.
Current scope
This first version uses compressed archives and base64 file transfer. It intentionally does not override Pi's grep, find, or ls tools yet. Native workspace transfer and automatic synchronization require future Celesto backend support.
