Skip to main content
Not everything a contact reads comes from a flow. Button labels, command responses, and system messages are supplied by the platform, and this screen is where they are changed.

What is in here

A table of translation keys, one row each: the group it belongs to, the key itself, a description of when it is used, and a column per language the tenant has enabled. Keys are grouped by where they appear — buttons.* for the labels on generated buttons, commands.* for what built-in commands reply, and so on. The description column says exactly when each is used, which matters because the key name alone rarely does.

Defaults are already filled in

Every cell arrives with a sensible default in every enabled language. The screen is not a translation task waiting to be done — the assistant works without anyone opening it. Filling a cell overrides that default for that language. Leave a cell alone and the catalogue default stands; edit it and your text is used from then on.
Override selectively. Changing buttons.yes from “Yes” to “Sure” affects every confirmation the assistant ever offers, in every flow, forever. That is sometimes exactly right — a brand with a particular voice — and sometimes an accident nobody connects to the button they changed months earlier.

Where this sits among the languages

Three different things carry language, and they are edited in three places: A language must be enabled in Settings before a column for it appears here.

Per assistant

This screen belongs to one assistant. Two assistants can answer the same command differently — useful when one is formal and another is not, and worth knowing before wondering why an override did not take effect somewhere else.