Docs, template assistant

Just describe it.

The naming language is powerful and it is also, honestly, a small programming language. Run /templateassistant, say what you want in your own words, and it writes the template for you. Free on every server, same as everything else.

How it works

  1. Run /templateassistant in the room you want to name, or run it anywhere and pick a channel from the list. You need Manage Channels, the same permission /template needs.
  2. Describe the name you want. Plain words, your own language, no syntax.
  3. It comes back with a template, a short explanation, and a preview of how the name renders in four different situations: one person with nothing playing, three people in a game, the owner streaming, and a game that reports party info.
  4. Hit Apply, or Refine to adjust it, or Cancel. Nothing changes until you apply.

The previews are the part worth reading. A template is only correct in the situations it will actually meet, and the one situation you can see while setting it up, your own room right now, is the one least likely to break. A name that looks perfect while you are playing something can render blank the moment nobody is.

What language it answers in

The templates themselves are always the same English tokens, because that is what the engine reads. The explanation is written for you, and it picks a language like this:

  1. If you ask for a language, it uses that one. Put "reply in English" or "responde en español" anywhere in your request and it will.
  2. Otherwise it uses your Discord app's language setting, the one under Discord's User Settings, Language. Not the language you typed in.
  3. If you clearly wrote in a different language from that, it will usually match what you wrote instead.

So if it answers you in a language you did not expect, it is following your Discord language setting. That is the usual cause. Either change it in Discord, which fixes it for good, or just ask for the language you want in your next message, which fixes it for that request.

Why it works this way, since it is a fair question: guessing the language from what you typed sounds obvious and turned out to be unreliable in testing. It would drift, and answer in Spanish to an English question, more often on longer requests. Your Discord setting is a real signal rather than a guess, so that is what it leans on, with the two escape hatches above for when the guess would have been better.

What it can and cannot do

It knows every token the engine knows, and it will tell you when you have asked for something that does not exist rather than inventing a token that quietly does nothing. Ask it to change the name based on whether the room is locked, or on the time of day, or on how many people are in the room, and it will say so and offer the closest thing that does work. There is no condition for room headcount, for instance, though it can happily show the count.

Every template it writes is checked before you see it: rendered against those four situations, and re-checked for the mistakes that look fine but are not, like a name that renders empty or a condition that can never match. If a template fails those checks it never reaches you, it goes back for another attempt.

The monthly limit

Each server gets 200 builds a month, and the count resets on the 1st. You will almost certainly never see this: naming is something you set up once and then forget, and 200 is far past any real use. It exists so a stuck client cannot run up a bill on our side.

It is not a paid limit. Every plan gets the same 200 and upgrading does not raise it. Nothing is told to you until a server passes 100, so normal use never even learns there is a limit. If you do hit it, /template still works by hand and the full token reference has everything the assistant knows.

What gets sent, and where

When you ask it something, what you typed goes to OpenAI, along with your server's current name and status templates, its "no game" label, and your Discord language setting. Nothing else. No message content, no member names, no presence, no IDs. The full detail is in the privacy policy, section 7.

If you self-host, the assistant is off until you point it at a model yourself, and it works with any OpenAI-compatible endpoint, including one running on your own machine. Nothing reaches us or OpenAI unless you configure it to.

Something not behaving? Ask in the support server, or start with the token reference if you would rather write templates by hand.