Commands

Prints, for each target locale, the total key count and how many are translated, missing, or outdated. No credentials, no writes — the read-only snapshot. Add --json for a machine-readable form.

Translates the missing and outdated keys for one or all target locales and merges the results into config/locales/. Determines the work from the source-hash state, so unchanged keys are never re-sent.

• --locale de limits to one locale.
• --force re-translates everything (respecting protected manual edits).
• --force-key a.b.c re-translates specific keys — including manual-flagged ones, whose manual flag survives the rewrite.
• --dry-run shows the plan without writing.

On success it runs the configured after_translate hooks. Requires provider credentials.

Runs the configured validators and prints the violations. --strict and --strict-all make it exit non-zero on their respective violation types — this is the CI command. See Validators for the checks and Continuous integration for wiring.

quality lints a locale's text (defaults to the source locale) with the static rules, terminology list, and — with --ai — an LLM review pass. fix-quality rewrites the auto-fixable suggestions (universal fixes and British spellings) back into the locale files, preserving case; add --dry-run to preview. See Quality linting.

When the manual_edits validator is enabled, accept-edits records hand-edited target values as intentional so they are protected from being overwritten by the next translate and no longer flagged as hand-edited.

• Unscoped, it accepts exactly the keys the manual_edits validator flags — nothing else changes.
• --key a.b.c (repeatable) accepts specific keys, drifted or not.
• --all marks every translated key as manual — the blanket form, for adopting locallingo on an app whose translations were all hand-made.
• --locale de limits any of the above to one locale.

See Drift & state.

Refreshes the .i18n-state/ drift state from the current locale files — for initial setup on an existing app, or after editing source strings, so validate doesn't report spurious "outdated" keys. It updates each key's source_hash, backfills a target_hash baseline for entries that lack one (hand-added translations), and prunes entries whose keys were removed. An existing target_hash and all manual flags are always preserved, so hand-edit protection never depends on when sync last ran.

Prints a single CRC32 fingerprint of the current source translations — handy as a cache key (e.g. to skip a translate step in CI when the source hasn't changed). --json wraps it in an object.