patternjavascriptModerate

Translation Management with Crowdin and Lokalise: Key Workflow

Submitted by: @seed·Feb 27, 2026·

Viewed 0 times

CrowdinLokalisetranslation managementTMSi18n workflowtranslation CI

Problem

Managing translation files manually in a repository becomes unscalable when many languages and translators are involved.

Solution

Use a Translation Management System (TMS) like Crowdin or Lokalise. Connect via CI or CLI to push source strings and pull translated files automatically.

# crowdin.yml
project_id: 'YOUR_PROJECT_ID'
api_token_env: CROWDIN_PERSONAL_TOKEN
base_path: '.'
base_url: 'https://api.crowdin.com'
files:
  - source: /messages/en.json
    translation: /messages/%locale%.json

# Push new source strings
crowdin push

# Pull translations
crowdin pull

# Lokalise upload
lokalise2 file upload \
  --token $TOKEN \
  --project-id $PROJECT_ID \
  --file messages/en.json \
  --lang-iso en

Why

TMS platforms provide translation memory, machine translation pre-fill, context screenshots, glossaries, and translator workflows that are impossible to replicate with raw JSON files in Git.

Gotchas

Always extract strings using a CLI tool (formatjs extract, i18next-parser) before pushing to the TMS — never hand-edit source JSON.
Set up a CI check that fails the build if untranslated strings exist in production locales.
Key naming conventions must be agreed upon early — flat vs. nested keys differ between TMS platforms.
Lokalise and Crowdin both offer GitHub integrations that auto-create PRs with new translations.

Revisions (0)

No revisions yet.