lokalise-reference-architecture
Implement a production-ready architecture for translation management using Lokalise.
Install
mkdir -p .claude/skills/lokalise-reference-architecture && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/5941" && unzip -o skill.zip -d .claude/skills/lokalise-reference-architecture && rm skill.zipInstalls to .claude/skills/lokalise-reference-architecture
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Implement Lokalise reference architecture with best-practice projectKey capabilities
- →Define translation lifecycle from code to deployment
- →Implement build-time and OTA translation delivery paths
- →Structure i18n code with separate loaders and middleware
- →Generate TypeScript types from source JSON files
- →Validate translation coverage and format
How it works
The architecture separates translation concerns into specific modules for initialization, loading, and middleware, while providing scripts to sync strings between the codebase and Lokalise via CI/CD.
Inputs & outputs
When to use lokalise-reference-architecture
- →Design new Lokalise integration
- →Standardize project translation structure
- →Review architecture best practices
About this skill
Lokalise Reference Architecture
Overview
A production-ready architecture for integrating Lokalise into web applications. Covers the end-to-end translation flow from source code through CI/CD and Lokalise to deployed translations, recommended project structure for i18n, file organization conventions, multi-app translation sharing, and the tradeoffs between OTA (over-the-air) and build-time translation loading.
Prerequisites
- Node.js 18+ with TypeScript
@lokalise/node-apiSDK installed (npm install @lokalise/node-api)- An i18n framework selected (i18next, react-intl, vue-i18n, or equivalent)
- Lokalise project created with at least one source language configured
- Basic understanding of CI/CD pipelines (GitHub Actions, GitLab CI, or equivalent)
Instructions
Step 1: Architecture Diagram
The translation lifecycle follows this flow:
┌─────────────┐ ┌──────────┐ ┌─────────────┐
│ Source Code │────▶│ CI/CD │────▶│ Lokalise │
│ │ │ (upload) │ │ (TMS) │
│ en.json │ └──────────┘ │ │
│ t('key') │ │ ┌────────┐ │
└─────────────┘ │ │Transla-│ │
│ │ tors │ │
│ └────────┘ │
┌─────────────┐ ┌──────────┐ │ │
│ Deploy │◀────│ CI/CD │◀────│ Download │
│ │ │ (build) │ │ │
│ CDN/Server │ └──────────┘ └─────────────┘
└──────┬──────┘ ┌─────────────┐
│ │ Lokalise │
│ (OTA path) │ OTA CDN │
│◀────────────────────────────│ │
│ └─────────────┘
┌──────▼──────┐
│ Users │
│ (browser/ │
│ mobile) │
└─────────────┘
Two delivery paths:
- Build-time (solid arrows): Translations downloaded during CI build, bundled into the application. Changes require a new deployment.
- OTA (dashed arrow): Translations fetched from Lokalise CDN at runtime. Changes appear without redeployment. Adds a network dependency.
Step 2: Project Structure for i18n
Organize your codebase to separate translation concerns from business logic:
project-root/
├── src/
│ ├── i18n/
│ │ ├── index.ts # i18n initialization and configuration
│ │ ├── client.ts # Lokalise API client wrapper
│ │ ├── loader.ts # Translation loader (build-time or OTA)
│ │ ├── fallback.ts # Fallback translation logic
│ │ ├── types.ts # TypeScript types for translation keys
│ │ └── middleware.ts # Express/Next.js locale detection middleware
│ │
│ ├── locales/
│ │ ├── en.json # Source language (committed to git)
│ │ ├── de.json # Downloaded from Lokalise (gitignored or committed)
│ │ ├── fr.json
│ │ ├── es.json
│ │ └── ja.json
│ │
│ ├── locales-fallback/ # Static fallback copy (always committed)
│ │ ├── en.json
│ │ ├── de.json
│ │ └── ...
│ │
│ └── components/
│ └── ... # Components use t('key') from i18n
│
├── scripts/
│ ├── lokalise-pull.sh # Download translations from Lokalise
│ ├── lokalise-push.sh # Upload source strings to Lokalise
│ ├── validate-translations.ts # Check coverage, placeholders, format
│ └── generate-types.ts # Generate TypeScript types from en.json
│
├── .github/workflows/
│ ├── lokalise-upload.yml # Upload on push to main
│ └── lokalise-download.yml # Download during build
│
└── lokalise.config.ts # Lokalise project configuration
Step 3: Core Configuration
// src/i18n/index.ts
import i18next from 'i18next';
import { initReactI18next } from 'react-i18next'; // or vue-i18n, svelte-i18n, etc.
import en from '../locales/en.json';
export const SUPPORTED_LOCALES = ['en', 'de', 'fr', 'es', 'ja'] as const;
export type SupportedLocale = typeof SUPPORTED_LOCALES[number];
export const DEFAULT_LOCALE: SupportedLocale = 'en';
i18next
.use(initReactI18next)
.init({
resources: { en: { translation: en } },
lng: DEFAULT_LOCALE,
fallbackLng: DEFAULT_LOCALE,
supportedLngs: [...SUPPORTED_LOCALES],
load: 'languageOnly', // 'de' not 'de-DE'
returnEmptyString: false, // Treat '' as missing → use fallback
interpolation: { escapeValue: false },
detection: {
order: ['cookie', 'navigator', 'htmlTag'],
caches: ['cookie'],
},
});
export default i18next;
Step 4: Lokalise API Client Wrapper
// src/i18n/client.ts
import { LokaliseApi } from '@lokalise/node-api';
interface LokaliseClientConfig {
apiToken: string;
projectId: string;
rateLimitPerSec?: number;
}
export class LokaliseClient {
private api: LokaliseApi;
private projectId: string;
private requestTimestamps: number[] = [];
private maxRequestsPerSec: number;
constructor(config: LokaliseClientConfig) {
this.api = new LokaliseApi({ apiKey: config.apiToken });
this.projectId = config.projectId;
this.maxRequestsPerSec = config.rateLimitPerSec ?? 6;
}
/**
* Download all translation files as a zip bundle URL.
*/
async downloadTranslations(options?: {
format?: string;
branch?: string;
}): Promise<string> {
await this.rateLimit();
const projectId = options?.branch
? `${this.projectId}:${options.branch}`
: this.projectId;
const response = await this.api.files().download(projectId, {
format: options?.format ?? 'json',
original_filenames: true,
directory_prefix: '',
export_empty_as: 'base',
export_sort: 'first_added',
});
return response.bundle_url;
}
// Additional methods: listKeys(), getStatistics(), uploadFile()
// follow the same pattern — call this.rateLimit() before each API call.
/**
* Simple rate limiter: 6 requests per second max.
*/
private async rateLimit(): Promise<void> {
const now = Date.now();
this.requestTimestamps = this.requestTimestamps.filter(t => now - t < 1000);
if (this.requestTimestamps.length >= this.maxRequestsPerSec) {
const oldestInWindow = this.requestTimestamps[0];
const waitMs = 1000 - (now - oldestInWindow);
if (waitMs > 0) {
await new Promise(resolve => setTimeout(resolve, waitMs));
}
}
this.requestTimestamps.push(Date.now());
}
}
Step 5: File Organization Conventions
Follow these conventions for translation file organization:
Flat keys (recommended for most projects — simpler grep, no nesting ambiguity):
{
"homepage.hero.title": "Welcome to MyApp",
"homepage.hero.subtitle": "The best app ever",
"settings.profile.name_label": "Full Name",
"errors.not_found": "Page not found"
}
Nested keys work better for large projects with clear module boundaries, where each top-level key maps to a feature area. Both formats are supported by Lokalise and i18next.
Key naming conventions:
- Use dot notation:
module.section.element - Use snake_case for key segments:
user_profile, notuserProfile - Prefix by feature area:
checkout.payment.card_label - Use consistent suffixes:
_title,_label,_button,_error,_placeholder - Keep keys under 100 characters (Lokalise hard limit is 1024 chars per key name)
File naming:
- One file per locale:
en.json,de.json,fr.json - For large apps, split by namespace:
common.json,auth.json,dashboard.json - Namespace files go in subdirectories:
locales/en/common.json,locales/en/auth.json
Step 6: Multi-App Translation Sharing
When multiple applications share translations (e.g., web app + mobile app + marketing site):
Lokalise Project: "MyCompany Shared"
├── Tags: shared, web-only, mobile-only, marketing-only
│
├── Shared keys (tag: shared)
│ ├── common.button.ok
│ ├── common.button.cancel
│ └── common.error.generic
│
├── Web-only keys (tag: web-only)
│ ├── web.nav.dashboard
│ └── web.nav.settings
│
└── Mobile-only keys (tag: mobile-only)
├── mobile.nav.home
└── mobile.permissions.camera
Download by tag to get only the keys each app needs:
# Web app — download shared + web-only
lokalise2 file download \
--token "$LOKALISE_API_TOKEN" \
--project-id "$LOKALISE_PROJECT_ID" \
--format json \
--filter-tags "shared,web-only" \
--original-filenames=false \
--bundle-structure "locales/%LANG_ISO%.json" \
--unzip-to "./"
# Mobile app — download shared + mobile-only
lokalise2 file download \
--token "$LOKALISE_API_TOKEN" \
--project-id "$LOKALISE_PROJECT_ID" \
--format json \
--filter-tags "shared,mobile-only" \
--original-filenames=false \
--bundle-structure "src/translations/%LANG_ISO%.json" \
--unzip-to "./"
Alternative: Separate projects with key linking. Lokalise does not natively share keys across projects, so tag-based filtering within a single project is the recommended approach for shared translations.
Step 7: OTA vs Build-Time Translation Loading
Choose the right delivery strategy based on your requirements:
| Factor | Build-Time | OTA |
|---|---|---|
| Latency | Zero (bundled) | Network request on first load |
| Update speed | Requires deployment | Instant (CDN cache) |
| Offline support | Full | Needs initial fetch + local cache |
| Bundle size | Increases with locales | Minimal (loaded on demand) |
| Reliability | No external dependency | Depends on Lokalise CDN |
| Best for | Server-rendered apps, SPAs with CI/CD | Mobile apps, rapid copy changes |
Build-time implementation (recommended for most web apps):
// src/i18n/loader-buildtime.ts
// Translations are imported statically — bundled at build time
import en from '../locales/en.json';
import de from '../
---
*Content truncated.*
When not to use it
- →When translation bundles must be strictly static and never updated via CDN
- →When the application does not use a supported i18n framework
Prerequisites
Limitations
- →OTA translations may flash on load if not merged with bundled translations
- →Circular imports can occur if translations are imported before i18n initialization
- →Nested keys with array values can cause type generation failures
How it compares
Unlike manual file management, this approach automates the synchronization of translation keys and supports both build-time bundling and runtime OTA updates.
Compared to similar skills
lokalise-reference-architecture side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| lokalise-reference-architecture (this skill) | 1 | 27d | Review | Intermediate |
| svelte-expert | 11 | 9mo | No flags | Intermediate |
| markdown-to-html | 16 | 6mo | Review | Beginner |
| openai-knowledge | 5 | 4mo | No flags | Intermediate |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by jeremylongshore
View all by jeremylongshore →You might also like
svelte-expert
Raudbjorn
Expert Svelte/SvelteKit development assistant for building components, utilities, and applications. Use when creating Svelte components, SvelteKit applications, implementing reactive patterns, handling state management, working with stores, transitions, animations, or any Svelte/SvelteKit development task. Includes comprehensive documentation access, code validation with svelte-autofixer, and playground link generation.
markdown-to-html
github
Convert Markdown files to HTML similar to `marked.js`, `pandoc`, `gomarkdown/markdown`, or similar tools; or writing custom script to convert markdown to html and/or working on web template systems like `jekyll/jekyll`, `gohugoio/hugo`, or similar web templating systems that utilize markdown documents, converting them to html. Use when asked to "convert markdown to html", "transform md to html", "render markdown", "generate html from markdown", or when working with .md files and/or web a templating system that converts markdown to HTML output. Supports CLI and Node.js workflows with GFM, CommonMark, and standard Markdown flavors.
openai-knowledge
openai
Use when working with the OpenAI API (Responses API) or OpenAI platform features (tools, streaming, Realtime API, auth, models, rate limits, MCP) and you need authoritative, up-to-date documentation (schemas, examples, limits, edge cases). Prefer the OpenAI Developer Documentation MCP server tools when available; otherwise guide the user to enable `openaiDeveloperDocs`.
http-generate
spring-ai-alibaba
Generates HTTP request examples for Spring Boot Web interfaces according to task specification and saves them as .http files in module-generate.md directories
ai-sdk-documentation
malob
This skill should be used when working with Vercel AI SDK, AI Gateway, streamText, generateText, generateObject, streamObject, tool calling, or AI SDK providers. Also relevant for "ai-sdk", "@ai-sdk/*" packages, or questions about AI SDK patterns, configuration, and best practices.
generating-api-contracts
jeremylongshore
Generate API contracts and OpenAPI specifications from code or design documents. Use when documenting API contracts and specifications. Trigger with phrases like "generate API contract", "create OpenAPI spec", or "document API contract".