Skip to content

ai-i18n

v0.6.0 Feature

This release adds 3 notable features for engineering teams evaluating rollout.

✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

ai angular angular-i18n anthropic crowdin i18n
+8 more
internationalization llm localization lokalise openai phrase react-intl vue-i18n

Summary

AI summary

Added post‑translation placeholder validation with retry using a custom user prompt.

Full changelog

What's New

Placeholder validation + retry for LLM translation output

After the preserveOrder parser fix reduced validation errors from 3,137 to 209, the remaining errors were caused by the LLM itself dropping placeholders during translation. This release adds post-translation validation that detects missing placeholders and retries with a targeted prompt.

Changes:

  • Added customUserPrompt to TranslationRequest for targeted retry prompts
  • Added findMissingPlaceholders() and buildPlaceholderRetryPrompt() helpers in context-builder
  • Added retryBrokenPlaceholders() step in TranslationOrchestrator.translate() — runs after existing missing-unit retry, validates each translation for placeholder completeness, and retries broken ones with a prompt that explicitly names the dropped placeholders
  • Updated all 3 providers (Anthropic, OpenAI, Ollama) to respect customUserPrompt
  • Retry accepts the fix only when re-validation passes (zero missing placeholders); otherwise keeps the original translation and logs warnings

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Track ai-i18n

Get notified when new releases ship.

Sign up free

About ai-i18n

GitHub Action that automatically translates i18n files (XLIFF, JSON) using LLM providers (Anthropic, OpenAI, Ollama) with glossary support, change detection, and reviewed translation preservation.

All releases →

Related context

Beta — feedback welcome: [email protected]