Skip to main content
The reactor CLI is how you build and run a model. It is a single Go binary, and it is the only thing you install. There is no Python, no runtime package, and no media libraries on your host — all of it ships inside the image the CLI builds for your workspace. You need two things: the CLI, and a running Docker daemon.

macOS

Install with Homebrew:
Upgrade the same way:

Linux

Homebrew works here too, if you already run it:
Otherwise, download the release tarball into a directory on your PATH. The RELEASE variable controls which build you get, and latest always resolves to the newest release. Use the sudo tab to install system-wide, or the no-sudo tab to install into your home directory.
To see which release latest currently points at, read the VERSION file.

Pin to a specific release

CI pipelines that need a reproducible install should pin a release instead of tracking latest. Set RELEASE to a version from releases.reactor.inc, where versioned tarballs are kept indefinitely.

Windows

Not supported. Use WSL2, or the Linux download above.

Verify

Sample output:
The runtime version is informational: reactor is a Go binary with no Python dependency of its own. It tells you what a fresh reactor init will pin, which you can override per workspace with reactor init --runtime-version. Once a workspace exists, the pin lives in its requirements.txt and you bump it there. The reactor command with no arguments prints a welcome screen. To print it again later, run reactor welcome.

Check your setup

reactor version only proves the binary runs. To check the rest, run reactor doctor. It checks Docker, the workspace config files, your API key, and the connection to the Reactor API. Each check prints one line, a failed check tells you how to fix it, and the command exits non-zero when any check fails.
Docker and the config files are what building and running a model locally depend on. The API key and connection checks only matter once you deploy, so a fresh install that fails those two can still build and run.

Next

Build your own model

Write a ReactorModel, run it, and connect a client.

Model Anatomy

A working model read line by line.