ai-language-detect Edge Function — Ai

ai-language-detect — Detect the language of a text string

Identifies the natural language of text using character trigram frequency analysis. Supports 13 languages with no external API calls.

---

Supported Languages

English, Spanish, French, German, Portuguese, Italian, Dutch, Polish, Russian, Chinese, Japanese, Korean, Arabic.

---

API

POST /api/ai-language-detect

Request body

Field	Type	Required	Default	Description
text	string	✅	—	Text to detect the language of
topN	number	❌	1	Number of language candidates to return, ranked by confidence

Success response (200)

{
  "success": true,
  "data": {
    "detected": [
      { "language": "English", "code": "en", "confidence": 0.82 }
    ],
    "primary": { "language": "English", "code": "en", "confidence": 0.82 }
  }
}

Error responses

Code	HTTP	When
INVALID_INPUT	400	Missing or empty text
INTERNAL_ERROR	500	Unexpected error

---

Usage

cURL

curl -X POST "$FUNCTION_URL" \
  -H "Content-Type: application/json" \
  -d '{"text": "Bonjour le monde, comment allez-vous?", "topN": 3}'

TypeScript / JavaScript (HTTP)

const response = await fetch(FUNCTION_URL, {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ text: userMessage }),
});
const { data } = await response.json();
if (data.primary.code !== 'en') {
  // Route to multilingual prompt
}

Direct import (Node / Bun / Deno)

import { aiLanguageDetect } from '@aerostack/functions/ai-language-detect';

const { primary } = aiLanguageDetect({ text: 'Hola mundo' });
console.log(primary.language); // "Spanish"

---

Use Cases

• Auto-routing user messages to language-specific system prompts before calling an LLM.
• Tagging ingested documents with their language for multilingual search and filtering.
• Validating that user-submitted content matches the expected language for a service.

---

Notes

• Detection uses character trigram frequency analysis — no API keys or network calls required.
• Confidence is reduced for texts shorter than 20 characters.
• For very short inputs, increase text length for reliable detection.
• Confidence values are normalised to [0, 1].