How I Built a Multilingual Developer News Hub in a Weekend

I built a news aggregator that automatically translates article summaries to 5 languages the moment they're published. Here's how I did it with RSS feeds, Claude AI, and my own translation API.

The Problem

I wanted a developer news feed on shipi18n.com - curated content about React, Next.js, AI, and i18n. But our users speak different languages. Building separate feeds for each language? That's a maintenance nightmare.

What if every article was automatically available in Spanish, French, German, Japanese, and Chinese?

The Architecture

RSS Feeds → Superfeedr Webhook → Lambda → AI Summary → Translation API → DynamoDB
                                                           ↓
                                              React Frontend (i18next)

Three key components:

Feed aggregation - Superfeedr pushes new articles via webhook
AI summarization - Claude generates concise summaries
Auto-translation - Every summary gets translated instantly

Step 1: Setting Up RSS Aggregation

Instead of polling dozens of feeds, I use Superfeedr - they monitor RSS feeds and POST new items to my webhook.

// Lambda webhook handler
export async function handleSuperfeedrWebhook(req, res) {
  const payload = JSON.parse(req.body);

  for (const item of payload.items) {
    // Store article
    const result = await storeArticle(item, feedUrl);

    // Translate immediately
    if (result.created) {
      await queueTranslation(result.articleId, result.article);
    }
  }
}

Feeds I'm aggregating:

Frontend: Smashing Magazine, CSS-Tricks, Vercel Blog
AI: OpenAI, Anthropic, Hugging Face
i18n: W3C Internationalization
SaaS: Indie Hackers, Stripe

Step 2: AI-Generated Summaries

Most RSS feeds have terrible summaries - often just the first paragraph with HTML junk. Claude Haiku fixes that:

async function generateAISummary(article) {
  const response = await fetch('https://api.anthropic.com/v1/messages', {
    method: 'POST',
    headers: {
      'x-api-key': process.env.ANTHROPIC_API_KEY,
      'anthropic-version': '2023-06-01',
    },
    body: JSON.stringify({
      model: 'claude-3-haiku-20240307',
      max_tokens: 150,
      messages: [{
        role: 'user',
        content: `Write a 1-2 sentence summary of this article for developers.
                  Be concise and informative.
                  Title: "${article.title}"
                  URL: ${article.url}`
      }]
    })
  });

  const data = await response.json();
  return data.content[0].text;
}

Haiku is perfect for this - fast, cheap ($0.25/M input tokens), and the summaries are actually good.

Step 3: The Translation Magic

Here's where it gets seamless. I built Shipi18n specifically for this use case - translating strings programmatically without managing translation files.

async function queueTranslation(articleId, article) {
  const targetLangs = ['es', 'fr', 'de', 'ja', 'zh'];
  const translations = {};

  for (const lang of targetLangs) {
    // Translate title
    const titleResult = await translateText(article.title, 'en', lang);

    // Translate summary
    const summaryResult = await translateText(article.summary, 'en', lang);

    translations[lang] = {
      title: titleResult.translatedText,
      summary: summaryResult.translatedText,
    };
  }

  await updateArticleTranslations(articleId, translations);
}

The translateText function calls the translation engine directly:

import { translateJSON } from '@shipi18n/core'

const { result } = await translateJSON({
  content: { text },
  from: 'en',
  to: targetLang,
  provider: 'anthropic',   // uses ANTHROPIC_API_KEY from the environment
})

const translated = result.text

One function call, no service in between. The Lambda holds the LLM key and talks to the provider directly.

Step 4: Frontend with i18next

The React frontend uses i18next for UI translations, and the article translations come from the API:

const { i18n } = useTranslation();
const currentLang = i18n.language?.split('-')[0] || 'en';

// Get translated content based on user's language
const getTranslatedContent = (article) => {
  if (currentLang === 'en' || !article.translations?.[currentLang]) {
    return { title: article.title, summary: article.summary };
  }
  return {
    title: article.translations[currentLang]?.title || article.title,
    summary: article.translations[currentLang]?.summary || article.summary,
  };
};

When a Japanese user visits, they see:

React 19's new compiler dramatically improves performance
Reactチームが新しいコンパイラを発表。手動のメモ化なしで...

When a Spanish user visits:

El nuevo compilador de React 19 mejora drásticamente el rendimiento
El equipo de React anunció un nuevo compilador que...

Same article, automatic localization.

The Result

What I got:

News hub with 5 categories (i18n, Frontend, AI, SaaS, Creative)
Every article translated to 5 languages instantly
Language switcher that respects user's browser locale
Globe icon to preview all translations

Cost breakdown:

Superfeedr: ~$5/month for 30 feeds
Claude Haiku: ~$2/month for summaries
Shipi18n: free and open source; I only pay Anthropic for the tokens, and incremental mode means repeat runs cost nothing

Time to build: One weekend

Key Takeaways

Don't poll RSS feeds - Use a service like Superfeedr that pushes updates
AI summaries > raw excerpts - Claude Haiku is cheap enough to run on every article
Translate at ingest, not render - Doing it server-side means instant page loads
Cache translations - Shipi18n's 90-day cache means I only pay once per string

Next Steps

Update (August 2026): the news hub itself has been retired — it ran on the hosted backend we shut down when Shipi18n went fully open source. The architecture below still stands on its own, and you can see the same translation pipeline at work on this site: use the language switcher in the top navigation and watch every page switch between English, Spanish, Japanese, Arabic and seven more, all translated by the CLI described here.

Want to add automatic translations to your own project?

Get started with Shipi18n — open source, Apache-2.0, bring your own LLM key
Check out the CLI integration for translating and checking JSON files
Use @shipi18n/core programmatically in Node.js

Ship your app in every language.

Get Started Free →