translation-health

Run a translation health check to compare each target language file against the source language. It reports translated, missing, extra, TODO, and empty keys per language along with a coverage percentage, so you can see exactly which locales need work.

Use the deep analysis mode, which greps source files for key references in t(), $t(), formatMessage(), and similar calls. It accounts for dynamically constructed keys via configurable prefixes and auto-inferred template-literal patterns, so runtime-built key families are not falsely flagged as dead.

Yes, it extracts interpolation variables from both source and target strings for every translated key. It flags keys where the target is missing a source variable or contains an extra variable not present in the source, which usually indicates a typo.

No, the audit is strictly read-only. It analyzes translation files and source code, then outputs a structured report with severity classifications, leaving all fixes to a separate sync step or manual editing.

Naive grep-based detection misses keys built at runtime, such as t(`Api.${endpoint}`). This check loads dynamicKeyPrefixes and dynamicKeyPatterns from config and auto-infers prefixes from template literals and concatenation in code, suppressing those families from the dead-key report.