Migrate translations from Transifex to AI-based workflow

[m3nu]

Summary

Move translation management from Transifex to an AI-based workflow that keeps translations fully within our repository. This will reduce external dependencies, enable faster iteration on translations, and provide better context-aware translations.

Background

Currently, Vorta uses Transifex for community translation management:

• Translatable strings are extracted from Python/UI files using pylupdate5
• English source file (vorta.en.ts) is pushed to Transifex
• Community translators work on Transifex's web interface
• Translated .ts files are pulled back and compiled to .qm format

While this has worked well, it introduces:

• External service dependency
• Friction for quick translation updates
• Inconsistent translation quality across languages
• Difficulty maintaining context for technical strings

Current State

Translation files already in repo:

• Source files: src/vorta/i18n/ts/vorta.{lang}.ts (13 languages)
• Compiled files: src/vorta/i18n/qm/vorta.{lang}.qm

Supported languages: Arabic, Czech, German, English, Spanish, Finnish, French, Galician, Italian, Dutch, Russian, Slovak, Swedish

Transifex integration to remove:

• .tx/config - Transifex configuration
• Makefile targets: translations-push, translations-pull, translations-update
• Reference in com.borgbase.Vorta.appdata.xml (translate URL)
• Contributor references in CONTRIBUTORS.md
• Documentation at https://vorta.borgbase.com/contributing/translations/

Proposed Changes

1. Confirm translations remain in repository

• Keep existing .ts and .qm files structure
• Continue using Qt Linguist format for compatibility
• Maintain translations-from-source and translations-to-qm Makefile targets

2. Remove Transifex integration

• Delete .tx/config
• Remove translations-push and translations-pull from Makefile
• Remove translations-update target (or repurpose for AI workflow)
• Update com.borgbase.Vorta.appdata.xml to remove Transifex URL
• Update documentation to reflect new workflow

3. Set up AI translation skill

Create a Claude Code skill (or similar) that can:

• Parse .ts files and identify untranslated/fuzzy strings
• Generate translations with full application context
• Review existing translations for consistency and accuracy
• Update .ts files with new translations
• Compile to .qm format after updates

4. Ensure context-aware translations

The AI translation system should have access to:

• String context: The <context> element in .ts files (e.g., "RepoTab", "ArchiveTab")
• UI location: Where the string appears (button, label, menu, tooltip)
• Application domain: Backup software terminology (repository, archive, prune, etc.)
• Style guide: Title case for buttons, colon suffixes for labels, etc.
• Existing translations: Maintain consistency with already-translated strings

5. Model selection considerations

Choose a model that:

• Handles technical terminology well
• Supports all 13 target languages
• Can maintain consistent tone/style across translations
• Is cost-effective for periodic batch updates

Implementation Tasks

1. Phase 1: Cleanup

   • Remove Transifex configuration and Makefile targets
   • Update documentation

2. Phase 2: AI Skill Development

   • Create skill to parse Qt .ts files
   • Implement translation generation with context
   • Add review/update capabilities
   • Integrate with existing translations-to-qm workflow

3. Phase 3: Validation

   • Test translations in running application
   • Verify RTL support (Arabic) works correctly
   • Ensure no regressions in existing translations

Files to Modify

• .tx/config - DELETE
• Makefile - Remove Transifex targets, add AI workflow targets
• src/vorta/assets/metadata/com.borgbase.Vorta.appdata.xml - Remove translate URL
• CONTRIBUTORS.md - Update translation contributor acknowledgments
• Documentation site - Update translation workflow docs

Open Questions

1. Should we maintain a way for community members to contribute translations manually (PRs with .ts file changes)?
2. What review process should AI-generated translations go through before merging?
3. Should we expand to additional languages now that the barrier to entry is lower?

Related Links

• Current translation docs: https://vorta.borgbase.com/contributing/translations/
• Qt Linguist format: https://doc.qt.io/qt-6/linguist-ts-file-format.html

Notes for Implementation

The skill should be able to run commands like:

# Review and update German translations
/translate-review de

# Generate missing translations for all languages
/translate-missing

# Full translation update cycle
/translate-update