vanishDOCS

Working trees and results

A Run executes one command against your live working tree, so you do not need to commit first. Vanish always leaves credentials, private keys and .env files out of the stored snapshot. By default it also skips dependency and cache directories such as node_modules, .venv, .git and .cache. To send one of them anyway, add a negated pattern such as !node_modules/ to a .vanishignore file at the root of your repository.

A Run token identifies the filesystem changes a Run captured. Vanish stages them beside your working tree instead of writing them into it: vanish results lists staged results, and vanish results apply TOKEN --dry-run previews applying one. Pass --apply, or put apply = true in .vanish.toml, to write a Run's results into your tree as it finishes. If you changed a file during the Run that the Run also wrote, your version stays and the whole result stays staged.

Exit codes, signals and output come back as they would locally: exit 42 returns 42, a command killed by SIGTERM returns 143, a command writing to a closed pipe returns 141, Ctrl-C returns 130, and each output stream arrives byte for byte. Three things differ: workers run Linux, each Run adds startup time, and stdout and stderr are not interleaved in the order they were written.

Processes do not survive between Runs. Files and captured environment changes carry over, but a database server started by an earlier command is not still running. Start the server and its client inside one Run.