Language Detection: Dynamic Multilingual Support

This article explains how we automatically detect and serve content in the user's preferred language.

The Problem: Serving Global Users

Users come from different countries speaking different languages:

  • India: Hindi, Bengali, Tamil, Telugu, Gujarati, Marathi, Kannada, Malayalam, Punjabi, Urdu

  • Europe: German, French, Spanish, Italian, Portuguese, Russian

  • Asia: Japanese, Korean, Chinese

We need to detect their language preference and serve translated content automatically.

Detection Priority

We check multiple sources in order:

1. URL parameter (?lang=hi)

2. Session (stored after first detection)

3. Cookie (translation_lang or lang)

4. Browser Accept-Language header (future)

5. Default (English)

Why This Order?

URL parameter first: Allows explicit language switching

Session second: Persists during browsing session

Cookie third: Persists across sessions (30 days)

Default last: Fallback for new users

Implementation

def get_locale():
    # 1. Check URL parameter
    lang_param = request.args.get("lang")
    if lang_param in valid_langs:
        return lang_param

    # 2. Check session
    if "translation_lang" in session:
        return session["translation_lang"]

    # 3. Check cookie
    # ... (implementation details omitted)

Language Switching

Users can switch language via URL:

/p/Treo-N100-8-256?lang=hi

This triggers:

1. Set session: session["translation_lang"] = "hi"

2. Set cookie: translation_lang=hi; Max-Age=2592000 (30 days)

3. Redirect: Remove lang parameter from URL

Benefit: Clean URLs after language selection

Supported Languages

We support 20 languages:

Indian languages (10):

  • Hindi (hi)

  • Bengali (bn)

  • Tamil (ta)

  • Telugu (te)

  • Gujarati (gu)

  • Marathi (mr)

  • Kannada (kn)

  • Malayalam (ml)

  • Punjabi (pa)

  • Urdu (ur)

European languages (7):

  • German (de)

  • French (fr)

  • Spanish (es)

  • Italian (it)

  • Portuguese (pt)

  • Russian (ru)

Asian languages (3):

  • Japanese (ja)

  • Korean (ko)

  • Chinese (zh)

Content Translation

Different content types use different translation methods:

Static Templates (Flask-Babel)

Navigation, buttons, labels translated via .po files:

{{ _('Add to Cart') }}  → "कार्ट में जोड़ें" (Hindi)

Dynamic Content (TranslationManager)

Product names, features translated via phrase tables:

"Mini PC" → "मिनी पीसी" (Hindi)

Long Content (DeepSeek AI)

Product descriptions, articles translated via AI:

"Compact office computer..." → "कॉम्पैक्ट ऑफिस कंप्यूटर..." (Hindi)

See: Translation System for details

Regional Formatting

Language selection also affects formatting:

Number Format

English: 1,234.56

European: 1.234,56 (reversed separators)

Indian: 1,23,456 (lakhs/crores grouping)

Date Format

English: Feb 18, 2026

European: 18.02.2026

Indian: 18/02/2026

Currency

English: ₹1,234

European: €1.234

International: $1,234

Language Selector UI

Users can switch language via dropdown:

Location: Top navigation bar

Display: Flag icon + language name

Behavior: Reload page with ?lang=<code>

SEO Considerations

URL Structure

We use single URL for all languages (no /hi/, /de/ prefixes):

Benefit: Simpler architecture, no duplicate content issues

Trade-off: Less SEO value per language

Hreflang Tags

We add alternate language links:

<link rel="alternate" hreflang="en" href="https://www.thinvent.in/p/Treo-N100">
<link rel="alternate" hreflang="hi" href="https://www.thinvent.in/p/Treo-N100?lang=hi">
<link rel="alternate" hreflang="de" href="https://www.thinvent.in/p/Treo-N100?lang=de">

Benefit: Google knows which language version to show

Fallback Strategy

If translation missing for selected language:

1. Try English: Most content has English version

2. Show original: Better than nothing

3. Queue for translation: Add to translation queue for batch processing

Performance

Caching

Translations are cached in memory:

Phrase tables: Loaded once at startup

AI translations: Cached in database

Benefit: No translation overhead per request

Lazy Loading

Translations loaded only when needed:

Product descriptions: Loaded on product page view

Query page content: Loaded on query page view

Benefit: Faster initial page load

References

Technical Concepts

Libraries

Related Articles

Summary

Language detection uses a priority system:

Detection order:

  • ✅ URL parameter (?lang=hi)

  • ✅ Session (during browsing)

  • ✅ Cookie (30 days)

  • ✅ Default (English)

Supported:

  • ✅ 20 languages (10 Indian, 7 European, 3 Asian)

  • ✅ Regional formatting (numbers, dates, currency)

  • ✅ SEO hreflang tags

Translation methods:

  • ✅ Flask-Babel (static templates)

  • ✅ TranslationManager (product names, features)

  • ✅ DeepSeek AI (descriptions, articles)

Performance:

  • ✅ In-memory caching

  • ✅ Lazy loading

  • ✅ Database caching for AI translations

This enables automatic language detection and seamless multilingual experience.

← Back to Documentation Index