Translation with Mayura

Mayura is Sarvam AI's translation model optimized for Indian languages with support for script variations, numeral formats, and code-mixed text.

Installation

pip install sarvamai

Quick Start

from sarvamai import SarvamAI

client = SarvamAI()

response = client.translate.translate(
    input="Hello, how are you?",
    source_language_code="en-IN",
    target_language_code="hi-IN",
    model="mayura:v1"
)

print(response.translated_text)  # "नमस्ते, आप कैसे हैं?"

Supported Languages

Code	Language	Code	Language
hi-IN	Hindi	ta-IN	Tamil
bn-IN	Bengali	te-IN	Telugu
kn-IN	Kannada	ml-IN	Malayalam
mr-IN	Marathi	gu-IN	Gujarati
pa-IN	Punjabi	or-IN	Odia
en-IN	English	auto	Auto-detect

Translation Directions

• English → Indian Language: Translate English to any supported Indian language
• Indian Language → English: Translate any Indian language to English
• Indian → Indian: Translate between Indian languages (via English pivot)

Translation Modes

Formal Translation

response = client.translate.translate(
    input="Please submit the report by Friday",
    source_language_code="en-IN",
    target_language_code="hi-IN",
    model="mayura:v1",
    mode="formal"
)

Casual Translation

response = client.translate.translate(
    input="Hey, what's up?",
    source_language_code="en-IN",
    target_language_code="hi-IN",
    model="mayura:v1",
    mode="casual"
)

Script Control

Choose the output script for languages with multiple scripts:

# Hindi in Devanagari (default)
response = client.translate.translate(
    input="Hello",
    source_language_code="en-IN",
    target_language_code="hi-IN",
    output_script="devanagari"
)
# Output: "नमस्ते"

# Hindi in Latin (transliteration)
response = client.translate.translate(
    input="Hello",
    source_language_code="en-IN",
    target_language_code="hi-IN",
    output_script="latin"
)
# Output: "Namaste"

Numeral Format

Control numeral representation:

# International numerals (default)
response = client.translate.translate(
    input="The price is 500 rupees",
    source_language_code="en-IN",
    target_language_code="hi-IN",
    numeral_format="international"
)
# Output: "कीमत 500 रुपये है"

# Native numerals
response = client.translate.translate(
    input="The price is 500 rupees",
    source_language_code="en-IN",
    target_language_code="hi-IN",
    numeral_format="native"
)
# Output: "कीमत ५०० रुपये है"

Code-Mixed Input

Mayura handles code-mixed text (e.g., Hinglish):

response = client.translate.translate(
    input="Yaar, let's go for a movie tonight",
    source_language_code="auto",  # Auto-detect
    target_language_code="hi-IN",
    model="mayura:v1"
)
# Output: "यार, चलो आज रात फिल्म देखने चलते हैं"

Batch Translation

Translate multiple texts:

texts = [
    "Hello",
    "How are you?",
    "Thank you"
]

responses = []
for text in texts:
    response = client.translate.translate(
        input=text,
        source_language_code="en-IN",
        target_language_code="hi-IN",
        model="mayura:v1"
    )
    responses.append(response.translated_text)

print(responses)
# [
    "नमस्ते",
    "आप कैसे हैं?",
    "धन्यवाद"
]

JavaScript

import { SarvamAI
} from "sarvamai";

const client = new SarvamAI();

const response = await client.translate.translate({
  input: "Hello, how are you?",
  sourceLanguageCode: "en-IN",
  targetLanguageCode: "hi-IN",
  model: "mayura:v1"
});

console.log(response.translatedText);

cURL

curl -X POST "https://api.sarvam.ai/translate" \
  -H "api-subscription-key: $SARVAM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input": "Hello, how are you?",
    "source_language_code": "en-IN",
    "target_language_code": "hi-IN",
    "model": "mayura:v1"
}'

Parameters

Parameter	Type	Required	Description
input	string	Yes	Text to translate
source_language_code	string	Yes	Source language or auto
target_language_code	string	Yes	Target language code
model	string	Yes	mayura:v1
mode	string	No	formal or casual
output_script	string	No	devanagari, latin, etc.
numeral_format	string	No	international or native

Response

{
    "request_id": "abc123",
    "translated_text": "नमस्ते, आप कैसे हैं?",
    "source_language_code": "en-IN",
    "target_language_code": "hi-IN"
}

See references/languages.md for language-specific notes.