DEV Community

xidioda xidioda
xidioda xidioda

Posted on • Originally published at khaleejiapi.dev

How We Built an AI Translation API with Google Gemini

#webdev #ai #machinelearning #api

Why AI Translation?

Traditional translation APIs (Google Translate, DeepL) work well for general text, but they struggle with:

  1. Arabic dialects — Gulf Arabic vs Egyptian vs Levantine
  2. Formality levels — Formal business Arabic vs casual chat
  3. Context-aware translation — Technical terms, cultural nuances

We decided to use Google Gemini 2.5 Flash as our translation engine. It's fast (< 500ms), understands context, and supports dialect control.

Architecture 

Client → /api/v1/translate → Cache Check (Valkey)
                                  ↓ miss
                             Gemini 2.5 Flash
                                  ↓
                             Cache Store (24h TTL)
                                  ↓
                             Response
Enter fullscreen mode Exit fullscreen mode

The Prompt

The key to good AI translation is the prompt. Here's our approach:

Translate the following text from {source} to {target}.
Formality: {formality}
Dialect: {dialect}

Rules:
- Maintain the original meaning and tone
- Use the specified dialect if provided
- Return ONLY the translated text, no explanations
Enter fullscreen mode Exit fullscreen mode

Caching Strategy

Translation is expensive (API credits + latency). We cache aggressively:

  • Cache key: Hash of text + source + target + formality + dialect
  • TTL: 24 hours in Valkey
  • Fallback: In-memory Map if Valkey is unavailable

This gives us 0ms latency on cache hits and reduces our Gemini API costs by ~80%.

SDK Usage

TypeScript 

import { KhaleejiAPI } from '@khaleejiapi/sdk';

const client = new KhaleejiAPI({ apiKey: 'YOUR_KEY' });

const result = await client.communication.translateAI({
  text: 'Hello, how are you?',
  target: 'ar',
  formality: 'formal',
  dialect: 'gulf'
});

console.log(result.translated);
// "السلام عليكم، كيف حالكم؟"
Enter fullscreen mode Exit fullscreen mode

Python 

from khaleejiapi import KhaleejiAPI

client = KhaleejiAPI(api_key="YOUR_KEY")

result = client.translate(
    text="Hello, how are you?",
    target="ar",
    formality="formal",
    dialect="gulf"
)

print(result["translated"])
# "السلام عليكم، كيف حالكم؟"
Enter fullscreen mode Exit fullscreen mode

Supported Languages

We support 22+ languages, including all major MENA languages:

Code Language Dialect Support
ar Arabic gulf, egyptian, levantine, maghrebi
fa Persian —
ur Urdu —
tr Turkish —
en English —
fr French —
hi Hindi —

Performance

Metric Value
Cache hit latency <5ms
Cache miss latency 300-500ms
Cache hit rate ~75%
Supported languages 22+
Max text length 5,000 chars

Try It 

curl -X POST https://khaleejiapi.dev/api/v1/translate \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Good morning", "target": "ar", "dialect": "gulf"}'
Enter fullscreen mode Exit fullscreen mode

Get your free API key at khaleejiapi.dev/signup.

Originally published on KhaleejiAPI Blog. Sign up for a free API key at khaleejiapi.dev/signup — 1,000 requests/month, no credit card required.

Top comments (0)