frontseat-action

The official GitHub Action for Frontseat. Installs the Frontseat CLI plus declared plugins, restores the REAPI build cache between runs, and drives any lifecycle command (build, ship, lint, test, conform, …) on the workspace.

Quick start

jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - uses: frontseat-dev/frontseat-action@v1
        with:
          mode: build

That single step replaces a manual curl mise.run | sh + tool install + REAPI cache restoration block — frontseat build // runs in the same job, and a hashed cache of .frontseat/cache/ survives across runs.

Shipping a release

- uses: frontseat-dev/frontseat-action@v1
  with:
    mode: ship
    args: --prerelease alpha
  env:
    GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

mode: ship runs the full Frontseat ship lifecycle (assemble → changelog → catalog → sign → deploy → upload → release → prepare → publish → announce). GitHub release creation needs GITHUB_TOKEN exposed to the step.

Inputs

Name	Default	Description
mode	build	Lifecycle command to run (build, ship, lint, test, package, verify, conform, sync). Empty string installs Frontseat and exits.
entities	//	Entity selector passed to the command.
version	latest	Frontseat CLI version. Resolved against the frontseat releases via the frontseat-mise plugin.
plugins	""	Comma-separated plugin names (e.g. go,maven,docker). When empty, the action defers to whatever the workspace's mise.toml declares.
args	""	Extra arguments appended to the Frontseat command.
exec-mode	local	Passed as --exec-mode.
cache-mode	offline	Passed as --cache-mode.
cache	"true"	Enable REAPI local-cache persistence via actions/cache.
working-directory	.	Directory containing frontseat.yaml.
mise-version	""	Override the mise CLI version (passed to jdx/mise-action's version input).

What it does, in order

1. Configures frontseat-mise if your workspace doesn't already declare it. Drops a small mise.toml overlay into .mise/ so this action works in workspaces that haven't pinned Frontseat themselves.
2. Installs mise and all declared tools via jdx/mise-action with caching enabled. This downloads the Frontseat CLI, every plugin tarball for the runner's OS/arch, and the language toolchains declared in your workspace's mise.toml. On a warm cache this is seconds.
3. Restores the REAPI cache from actions/cache keyed on the hash of your workspace manifests (go.sum, go.work.sum, pom.xml, package-lock.json, Cargo.lock). Subsequent runs that don't change dependencies skip already-completed actions.
4. Runs the Frontseat command with the configured --exec-mode and --cache-mode.

Why not just call frontseat directly

You can — and for a quick experiment that's the right answer. The action exists to encapsulate three things that are tedious to get right by hand:

• Auth wiring for gh (used by frontseat-mise for version discovery) and the GitHub release step in mode: ship.
• REAPI cache persistence between ephemeral runner instances. Without it, every frontseat build rebuilds everything from scratch.
• Plugin installation through frontseat-mise's plugins option, which needs a token configured correctly to talk to the GitHub API.

License

Apache 2.0. See LICENSE.