Norwegian is identified as `no` by the locale detector agent; supported content localization locales include `nb_NO`

Contribute Bug
1

This leads to topics written in Norwegian (identified as no) being sent for translation to Norwegian (identified as nb_NO).

This is a little confusing, because the topic content stays mostly the same, but there are subtle differences. And it uses unnecessary credits.

Input / user:

{"content":"Man bare skriver posten, også sendes den avgårde til ChatGPT for å bli oversatt til de språk som er valgt. Så blir den vist i det språket som brukeren har valgt i sin profil. Men kan man alltid velge å slå av oversettelsene da.","target_locale":"nb_NO"}

Output / assistant:

{
  "output": "Man bare skriver innlegget, også sendes det av gårde til ChatGPT for å bli oversatt til språkene som er valgt. Så blir det vist på språket som brukeren har valgt i profilen sin. Men kan man alltid velge å slå av oversettelsene da."
}

image
image750×422 27.8 KB

1 Like
2

Thanks for the report. We’ll have a look.

It’ll likely include a prompt update into core, but if you would like a quick fix you can duplicate your current agent for locale detection and update the prompt for it to use nb_NO when it comes to Norwegian.

4

By adding something similar to these list items in the system prompt?

  • Brazilian Portuguese: pt-BR
  • Simplified Chinese: zh-CN

Or is something more specific needed like:

Always use language code nb_NO for Norwegian

Thanks for the quick response :slight_smile:

5

So, adding the following under “list of common language codes” seems have have resolved it:

- Norwegian: nb-NO
- Norwegian Bokmål: nb-NO

I first tried with nb_NO but that seemed to confuse the system…