Localise an app with AI and glot
Extract interface strings, plan a translation run, check variables and record how each entry was produced or reviewed. An offline mock mode is available for testing.
To localise an app with AI and glot, start with a screen the team can verify end to end. The sign-in page is a manageable example: labels, button text and a short error message all have a visible home. Moving them into a source catalogue gives each string an identifier that future translations can share.
The workflow can estimate a translation run before submitting it to the selected provider. Afterwards, it checks interpolation and plural handling. It can also inspect a running route for text that overflows its control. A local provenance file records how translations originated and whether a reviewer has looked at them, which helps an editor pick the next entries to revisit.
What it does
- Move interface wording into the source catalogue.
- Add new string keys without duplicating existing text.
- Review a cost estimate before live translation.
- Find broken placeholders and text that no longer fits.
- Record and export translation provenance.
Install and use
- Install the skill and official glot CLI, or use Node and npx. Live translation needs an authorised Anthropic or OpenAI API key and may incur charges. Mock mode is available offline.
- Ask “Localise our sign-in screen into the second language. Preserve the product name and estimate the run before submitting translations.”
- Check the rendered controls and validation results, then retain the catalogues and review file with the project.
FAQ
Source
Made by MattJeff / glot contributors. Original skill: glot, published on GitHub under the MIT licence.
View on GitHub