Setup steps is the list of things that have to happen before translation is really running. There is one list, and you see it at two scopes: everything you can reach, and one project.
Who this is for
Whoever is wiring things up, usually a developer or admin. Anyone who can open a project can see that project's setup steps.
What happens next
The same five steps at both scopes, in the order the product forces on them:
Create your first project: a project holds one set of keys and the languages you ship it in.
Add a target language: pick the locales you want to ship.
Add your source strings: the source text everything else is built from. Nothing can be translated until it is here.
Run your first translation: let GLOSSA draft the missing strings in one pass.
Invite a teammate: add the translators and reviewers who will work on this project. Every plan includes seats, Free included, and a teammate pays nothing.
You see the card labelled Across your workspaces on the dashboard, before your first strings land, where it takes the page. It covers every project you can reach. Labelled This project, the same five steps narrowed to one project sit on that project's Get started page, where Create your first project counts as done the moment you are there. Steps the project already satisfies count from the start too: target languages you picked while creating it, and a teammate already in the workspace.
Two steps tick on something more than the thing existing:
Run your first translation counts a run that actually produced translations, either strings GLOSSA wrote or strings answered from translation memory. A run that failed, that was cancelled, that is still going, or that found nothing to do leaves the step open.
Invite a teammate counts a second person who reaches the workspace, an Owner of its account included, or an invitation that has not expired yet. A Billing member of the account reaches no workspace, so does not count. An invitation nobody accepted before it expired stops counting.
Progress is worked out from your real state every time a page loads. No progress is stored, so finishing a step anywhere updates every place the steps appear, and doing them out of order works exactly as well as doing them in order.
A row is not a link. The next step in order is highlighted, and an unfinished row carries a link named for what it does: Create a project, Add a language, Add strings, Open the editor, Invite someone. Three rows carry something else instead:
a step that is done offers nothing, because there is nowhere useful to go;
a step whose page you already have open says "You are on the page for this step." rather than linking to where you are;
a step your account does not have permission for names who does it, so you are not sent to a screen that would refuse you. A translator, for example, reads "Someone who can manage the project’s languages does this step." under Add a target language.
Beside the title the card counts N of M complete and draws a bar of the same figures. Under the rows, "The highlighted row is the next step in order. Anything below it can be done early, and the list follows." explains the highlight. On the dashboard the finished steps at the top are folded behind a line reading "N already done", so what is left comes first; open it to see them. Once every step is done the list says so and stays: "Every step on this list is done. It is a setup ladder rather than a queue of work, so it stays finished even while there are strings waiting to be translated."
On the project's Get started page, below the setup steps, Upload your language files takes a JSON, YAML, XLIFF, .strings, .arb or .po file straight from your machine. Nothing is written until you confirm what was detected. Under that, Push from your codebase gives you ready-made snippets: pick a framework preset and the snippets rewrite themselves for it, with your real project details filled in. If you can manage tokens, the strip above the snippets mints the credential the snippets need, and shows it once: "Deploy token minted. Copy it now, it is shown once." So copy it before you leave the page.
If it doesn't work
If the token strip is not there, one of two things is true. Either your plan does not include API access, in which case the page says so and you upload your files instead; or you do not have permission to mint deploy tokens in this project, and an admin can do it for you.
If a step you have already done still reads as not done, it is Run your first translation more often than not. A run that ended without writing a single string does not count, and neither does one that is still going. Open the editor, run the missing strings again, and the step ticks once something lands.
If you were granted specific projects rather than a whole workspace, and none of them has strings yet, you do not see the account-wide card at all. Its steps would all be work somebody else already did, and none of it is work you could do.
Seed a sample project fills a project with example content so you can see the whole flow before wiring up your own. It is a real project, so it asks the same two things creating one asks: that you are allowed to create projects here, and that your plan has room for another. On a project's Get started page the button is there when both are true. The dashboard offers the same thing as Explore a sample, and shows it locked rather than hiding it, with the reason on the lock: Needs a spare project when the workspace is at its limit, Admins and owners when creating projects is not something your role does.
If Seed a sample project does report a refusal after you press it, the workspace filled up between the page loading and the click, usually because a teammate or another tab created a project. The message appears under the page heading and says what the limit is. Free up a project or move up a plan, then press it again.
Related
Create a deploy token
The command line tool
Translate with AI
Your dashboard