Make your strings translatable

Every string you add ends up in front of translators of 64 languages. A little care on your side saves them a lot of guessing.

• In an integration, text lives in strings.json, next to the code. It covers the setup flow, entity names and states, actions, errors, and repair issues.
• In the frontend, text lives in src/translations/en.json.

Never put user facing text directly in code. Text in code cannot be translated.

Many strings already exist, like "Device is already configured" or "Failed to connect". Common strings live in homeassistant/strings.json in core. Point to them with a reference instead of writing your own:

{
  "config": {
    "abort": {
      "already_configured": "[%key:common::config_flow::abort::already_configured_device%]"
    }
  }
}

Every reused string is one string less to translate, in every language.

• Use placeholders for values, never glue sentences together. "Found {count} devices" can be translated. "Found " + count + " devices" cannot, because word order differs per language.
• Use plurals properly in the frontend. Write {count, plural, one {# device} other {# devices}} instead of "device(s)". Integration strings in strings.json don't use this syntax, so phrase those so they work for any number, like "Devices found: {count}".
• Name placeholders after what they hold. {device_name} tells a translator more than {name}.
• Keep sentences whole. Do not split one sentence over two strings.

In core, compile the strings of your integration to see them in the interface. Replace wled with the domain of your integration, or use --all for every integration:

python3 -m script.translations develop --integration wled

Once your pull request is merged, the new English strings are uploaded to Lokalise automatically. Translators pick them up from there, and the translations arrive with the next build of the project, see how translations get back. You do not need to do anything else.