Publishing

Publishing is the moment your draft becomes the thing your community actually experiences. Juno front-loads the risk: everything that can be checked is checked before the switch flips, so "it published" and "it works" are nearly the same statement.

The toolbar: Validate, Save, Publish

  • Save stores your draft. Do it freely; it affects nobody.
  • Validate checks the whole workflow and lists every problem it finds. Click any item and the canvas jumps to the offending block or wire.
  • Publish runs the same checks, and ships only if they pass.

Reading validation results

Issues come in two severities, and the distinction is practical:

  • Errors block publishing. A required input with nothing wired or typed in, a branch that leads nowhere, a reference to a setting or stored value that no longer exists, a block that can't run on the current runtime. Fix these; the workflow cannot ship around them.
  • Warnings ship, but read them. The classic example is a best-effort block: it works, but its effect isn't guaranteed a retry if Discord fails at the worst moment. A warning is the editor saying "you should know this before your users do."

Validation covers the entire workflow — main graph, every function, every setting reference — in one pass. There is no "it broke in the part I didn't have open."

The pre-publish review

Alongside issue checking, publishing shows you the workflow's footprint:

  • Permissions — the combined list of everything the Juno bot needs on your server for this workflow. If something in the list surprises you, a block you forgot about is asking for it.
  • Settings — whether required settings for this server are still unanswered.

Thirty seconds here saves the classic failure mode of shipping a workflow that silently can't act.

When publish is refused

Two refusals are worth recognizing:

  • Outstanding errors — the list is shown; fix and go again.
  • The draft moved — someone else saved or published while your screen was open on older changes. Juno refuses to ship a version of the workflow nobody has actually seen assembled; reload, confirm the graph is what you intend, and publish again.

After the switch

The new version takes over from the old one cleanly — runs in flight finish on the version they started on, new triggers start on the new one. The old version stays in history, so rolling back is a click, not a rebuild. Then head to Monitoring and watch the first few runs land — the last step of publishing is seeing it work.