azure-ai-translation-text-py

OverviewHistoryStatsSecurity

npx skills add microsoft/skills --skill azure-ai-translation-text-py

Documentation

SKILL.md

npx skills add ...

Documentation

SKILL.md

Azure AI Text Translation SDK for real-time text translation, transliteration, language detection, and dictionary lookup. Use for translating text content in applications. Triggers: "text translation", "translator", "translate text", "transliterate", "TextTranslationClient".

npx skills add microsoft/skills --skill azure-ai-translation-text-py

Documentation

SKILL.mdView source

Azure AI Text Translation SDK for Python

Client library for Azure AI Translator text translation service for real-time text translation, transliteration, and language operations.

Installation

pip install azure-ai-translation-text

Environment Variables

AZURE_TRANSLATOR_ENDPOINT=https://<resource>.cognitiveservices.azure.com  # Required for Entra ID auth (must be a custom subdomain endpoint)
AZURE_TOKEN_CREDENTIALS=prod # Required only if DefaultAzureCredential is used in production
# Only required for the legacy API-key auth path below:
AZURE_TRANSLATOR_KEY=<your-api-key>
AZURE_TRANSLATOR_REGION=<your-region>  # e.g., eastus, westus2; required when authenticating with a key against the global endpoint

Authentication & Lifecycle

🔑 Two rules apply to every code sample below:

Prefer DefaultAzureCredential. It works locally (Azure CLI / VS Code / Developer CLI) and in Azure (managed identity, workload identity) with no code change. Avoid connection strings, account/API keys — they bypass Entra audit and rotation.

Local dev: DefaultAzureCredential works as-is.

Production: set AZURE_TOKEN_CREDENTIALS=prod (or AZURE_TOKEN_CREDENTIALS=<specific_credential>) to constrain the credential chain to production-safe credentials.

Wrap every client in a context manager so HTTP transports, sockets, and token caches are released deterministically:

Sync: with <Client>(...) as client:

Async: async with <Client>(...) as client: and async with DefaultAzureCredential() as credential: (from azure.identity.aio)

Snippets may abbreviate this setup, but production code should always follow both rules.

import os
from azure.identity import DefaultAzureCredential, ManagedIdentityCredential
from azure.ai.translation.text import TextTranslationClient
 
# Local dev: DefaultAzureCredential. Production: set AZURE_TOKEN_CREDENTIALS=prod or AZURE_TOKEN_CREDENTIALS=<specific_credential>
credential = DefaultAzureCredential(require_envvar=True)
# Or use a specific credential directly in production:
# See https://learn.microsoft.com/python/api/overview/azure/identity-readme?view=azure-python#credential-classes
# credential = ManagedIdentityCredential()
 
with TextTranslationClient(
    endpoint=os.environ["AZURE_TRANSLATOR_ENDPOINT"],
    credential=credential,
) as client:
    result = client.translate(body=["Hello, world!"], to=["es"])

Legacy: API Key (existing keyed deployments)

New code should use DefaultAzureCredential above. The Translator service has two specifics that make API-key auth still common in existing deployments:

Token-credential auth requires a custom subdomain endpoint (https://<resource>.cognitiveservices.azure.com). If you only have the global endpoint (https://api.cognitive.microsofttranslator.com), you must either provision a custom subdomain or stay on the key-based path until you do.
Key + region is the canonical setup against the global endpoint. The region is sent as the Ocp-Apim-Subscription-Region header and is required whenever you use a multi-service or global Translator key.

import os
from azure.core.credentials import AzureKeyCredential
from azure.ai.translation.text import TextTranslationClient
 
# Key + region against the global endpoint (most common keyed setup)
with TextTranslationClient(
    credential=AzureKeyCredential(os.environ["AZURE_TRANSLATOR_KEY"]),
    region=os.environ["AZURE_TRANSLATOR_REGION"],
) as client:
    result = client.translate(body=["Hello, world!"], to=["es"])
 
# Key against a custom subdomain endpoint (no region required)
with TextTranslationClient(
    endpoint=os.environ["AZURE_TRANSLATOR_ENDPOINT"],
    credential=AzureKeyCredential(os.environ["AZURE_TRANSLATOR_KEY"]),
) as client:
    result = client.translate(body=["Hello, world!"], to=["es"])

Basic Translation

# Translate to a single language
result = client.translate(
    body=["Hello, how are you?", "Welcome to Azure!"],
    to=["es"]  # Spanish
)
 
for item in result:
    for translation in item.translations:
        print(f"Translated: {translation.text}")
        print(f"Target language: {translation.to}")

Translate to Multiple Languages

result = client.translate(
    body=["Hello, world!"],
    to=["es", "fr", "de", "ja"]  # Spanish, French, German, Japanese
)
 
for item in result:
    print(f"Source: {item.detected_language.language if item.detected_language else 'unknown'}")
    for translation in item.translations:
        print(f"  {translation.to}: {translation.text}")

Specify Source Language

result = client.translate(
    body=["Bonjour le monde"],
    from_parameter="fr",  # Source is French
    to=["en", "es"]
)

Language Detection

result = client.translate(
    body=["Hola, como estas?"],
    to=["en"]
)
 
for item in result:
    if item.detected_language:
        print(f"Detected language: {item.detected_language.language}")
        print(f"Confidence: {item.detected_language.score:.2f}")

Transliteration

Convert text from one script to another:

result = client.transliterate(
    body=["konnichiwa"],
    language="ja",
    from_script="Latn",  # From Latin script
    to_script="Jpan"      # To Japanese script
)
 
for item in result:
    print(f"Transliterated: {item.text}")
    print(f"Script: {item.script}")

Dictionary Lookup

Find alternate translations and definitions:

result = client.lookup_dictionary_entries(
    body=["fly"],
    from_parameter="en",
    to="es"
)
 
for item in result:
    print(f"Source: {item.normalized_source} ({item.display_source})")
    for translation in item.translations:
        print(f"  Translation: {translation.normalized_target}")
        print(f"  Part of speech: {translation.pos_tag}")
        print(f"  Confidence: {translation.confidence:.2f}")

Dictionary Examples

Get usage examples for translations:

from azure.ai.translation.text.models import DictionaryExampleTextItem
 
result = client.lookup_dictionary_examples(
    body=[DictionaryExampleTextItem(text="fly", translation="volar")],
    from_parameter="en",
    to="es"
)
 
for item in result:
    for example in item.examples:
        print(f"Source: {example.source_prefix}{example.source_term}{example.source_suffix}")
        print(f"Target: {example.target_prefix}{example.target_term}{example.target_suffix}")

Get Supported Languages

# Get all supported languages
languages = client.get_supported_languages()
 
# Translation languages
print("Translation languages:")
for code, lang in languages.translation.items():
    print(f"  {code}: {lang.name} ({lang.native_name})")
 
# Transliteration languages
print("\nTransliteration languages:")
for code, lang in languages.transliteration.items():
    print(f"  {code}: {lang.name}")
    for script in lang.scripts:
        print(f"    {script.code} -> {[t.code for t in script.to_scripts]}")
 
# Dictionary languages
print("\nDictionary languages:")
for code, lang in languages.dictionary.items():
    print(f"  {code}: {lang.name}")

Break Sentence

Identify sentence boundaries:

result = client.find_sentence_boundaries(
    body=["Hello! How are you? I hope you are well."],
    language="en"
)
 
for item in result:
    print(f"Sentence lengths: {item.sent_len}")

Translation Options

result = client.translate(
    body=["Hello, world!"],
    to=["de"],
    text_type="html",           # "plain" or "html"
    profanity_action="Marked",  # "NoAction", "Deleted", "Marked"
    profanity_marker="Asterisk", # "Asterisk", "Tag"
    include_alignment=True,      # Include word alignment
    include_sentence_length=True # Include sentence boundaries
)
 
for item in result:
    translation = item.translations[0]
    print(f"Translated: {translation.text}")
    if translation.alignment:
        print(f"Alignment: {translation.alignment.proj}")
    if translation.sent_len:
        print(f"Sentence lengths: {translation.sent_len.src_sent_len}")

Async Client

from azure.ai.translation.text.aio import TextTranslationClient
from azure.identity.aio import DefaultAzureCredential
 
async def translate_text():
    async with DefaultAzureCredential() as credential:
        async with TextTranslationClient(
            credential=credential,
            endpoint=endpoint,
        ) as client:
            result = await client.translate(
                body=["Hello, world!"],
                to=["es"]
            )
            print(result[0].translations[0].text)

Client Methods

Method	Description
`translate`	Translate text to one or more languages
`transliterate`	Convert text between scripts
`detect`	Detect language of text
`find_sentence_boundaries`	Identify sentence boundaries
`lookup_dictionary_entries`	Dictionary lookup for translations
`lookup_dictionary_examples`	Get usage examples
`get_supported_languages`	List supported languages

Best Practices

Pick sync OR async and stay consistent. Do not mix azure.xxx sync clients with azure.xxx.aio async clients in the same call path. Choose one mode per module.
Always use context managers for clients and async credentials. Wrap every client in with Client(...) as client: (sync) or async with Client(...) as client: (async). For async DefaultAzureCredential from azure.identity.aio, also use async with credential: so tokens and transports are cleaned up.
Batch translations — Send multiple texts in one request (up to 100)
Specify source language when known to improve accuracy
Use async client for high-throughput scenarios
Cache language list — Supported languages don't change frequently
Handle profanity appropriately for your application
Use html text_type when translating HTML content
Include alignment for applications needing word mapping

Reference Files

File	Contents
references/capabilities.md	Additional non-hero capabilities, operation-group coverage, and production checklists.
references/non-hero-scenarios.md	Dedicated non-hero examples for secondary/advanced scenarios.

npx skills add microsoft/skills --skill azure-ai-translation-text-py

Documentation

SKILL.md

npx skills add microsoft/skills --skill azure-ai-translation-text-py

Documentation

SKILL.mdView source

Azure AI Text Translation SDK for Python

Client library for Azure AI Translator text translation service for real-time text translation, transliteration, and language operations.

Installation

pip install azure-ai-translation-text

Environment Variables

AZURE_TRANSLATOR_ENDPOINT=https://<resource>.cognitiveservices.azure.com  # Required for Entra ID auth (must be a custom subdomain endpoint)
AZURE_TOKEN_CREDENTIALS=prod # Required only if DefaultAzureCredential is used in production
# Only required for the legacy API-key auth path below:
AZURE_TRANSLATOR_KEY=<your-api-key>
AZURE_TRANSLATOR_REGION=<your-region>  # e.g., eastus, westus2; required when authenticating with a key against the global endpoint

Authentication & Lifecycle

🔑 Two rules apply to every code sample below:

Prefer DefaultAzureCredential. It works locally (Azure CLI / VS Code / Developer CLI) and in Azure (managed identity, workload identity) with no code change. Avoid connection strings, account/API keys — they bypass Entra audit and rotation.

Local dev: DefaultAzureCredential works as-is.

Production: set AZURE_TOKEN_CREDENTIALS=prod (or AZURE_TOKEN_CREDENTIALS=<specific_credential>) to constrain the credential chain to production-safe credentials.

Wrap every client in a context manager so HTTP transports, sockets, and token caches are released deterministically:

Sync: with <Client>(...) as client:

Async: async with <Client>(...) as client: and async with DefaultAzureCredential() as credential: (from azure.identity.aio)

Snippets may abbreviate this setup, but production code should always follow both rules.

import os
from azure.identity import DefaultAzureCredential, ManagedIdentityCredential
from azure.ai.translation.text import TextTranslationClient
 
# Local dev: DefaultAzureCredential. Production: set AZURE_TOKEN_CREDENTIALS=prod or AZURE_TOKEN_CREDENTIALS=<specific_credential>
credential = DefaultAzureCredential(require_envvar=True)
# Or use a specific credential directly in production:
# See https://learn.microsoft.com/python/api/overview/azure/identity-readme?view=azure-python#credential-classes
# credential = ManagedIdentityCredential()
 
with TextTranslationClient(
    endpoint=os.environ["AZURE_TRANSLATOR_ENDPOINT"],
    credential=credential,
) as client:
    result = client.translate(body=["Hello, world!"], to=["es"])

Legacy: API Key (existing keyed deployments)

New code should use DefaultAzureCredential above. The Translator service has two specifics that make API-key auth still common in existing deployments:

Token-credential auth requires a custom subdomain endpoint (https://<resource>.cognitiveservices.azure.com). If you only have the global endpoint (https://api.cognitive.microsofttranslator.com), you must either provision a custom subdomain or stay on the key-based path until you do.
Key + region is the canonical setup against the global endpoint. The region is sent as the Ocp-Apim-Subscription-Region header and is required whenever you use a multi-service or global Translator key.

import os
from azure.core.credentials import AzureKeyCredential
from azure.ai.translation.text import TextTranslationClient
 
# Key + region against the global endpoint (most common keyed setup)
with TextTranslationClient(
    credential=AzureKeyCredential(os.environ["AZURE_TRANSLATOR_KEY"]),
    region=os.environ["AZURE_TRANSLATOR_REGION"],
) as client:
    result = client.translate(body=["Hello, world!"], to=["es"])
 
# Key against a custom subdomain endpoint (no region required)
with TextTranslationClient(
    endpoint=os.environ["AZURE_TRANSLATOR_ENDPOINT"],
    credential=AzureKeyCredential(os.environ["AZURE_TRANSLATOR_KEY"]),
) as client:
    result = client.translate(body=["Hello, world!"], to=["es"])

Basic Translation

# Translate to a single language
result = client.translate(
    body=["Hello, how are you?", "Welcome to Azure!"],
    to=["es"]  # Spanish
)
 
for item in result:
    for translation in item.translations:
        print(f"Translated: {translation.text}")
        print(f"Target language: {translation.to}")

Translate to Multiple Languages

result = client.translate(
    body=["Hello, world!"],
    to=["es", "fr", "de", "ja"]  # Spanish, French, German, Japanese
)
 
for item in result:
    print(f"Source: {item.detected_language.language if item.detected_language else 'unknown'}")
    for translation in item.translations:
        print(f"  {translation.to}: {translation.text}")

Specify Source Language

result = client.translate(
    body=["Bonjour le monde"],
    from_parameter="fr",  # Source is French
    to=["en", "es"]
)

Language Detection

result = client.translate(
    body=["Hola, como estas?"],
    to=["en"]
)
 
for item in result:
    if item.detected_language:
        print(f"Detected language: {item.detected_language.language}")
        print(f"Confidence: {item.detected_language.score:.2f}")

Transliteration

Convert text from one script to another:

result = client.transliterate(
    body=["konnichiwa"],
    language="ja",
    from_script="Latn",  # From Latin script
    to_script="Jpan"      # To Japanese script
)
 
for item in result:
    print(f"Transliterated: {item.text}")
    print(f"Script: {item.script}")

Dictionary Lookup

Find alternate translations and definitions:

result = client.lookup_dictionary_entries(
    body=["fly"],
    from_parameter="en",
    to="es"
)
 
for item in result:
    print(f"Source: {item.normalized_source} ({item.display_source})")
    for translation in item.translations:
        print(f"  Translation: {translation.normalized_target}")
        print(f"  Part of speech: {translation.pos_tag}")
        print(f"  Confidence: {translation.confidence:.2f}")

Dictionary Examples

Get usage examples for translations:

from azure.ai.translation.text.models import DictionaryExampleTextItem
 
result = client.lookup_dictionary_examples(
    body=[DictionaryExampleTextItem(text="fly", translation="volar")],
    from_parameter="en",
    to="es"
)
 
for item in result:
    for example in item.examples:
        print(f"Source: {example.source_prefix}{example.source_term}{example.source_suffix}")
        print(f"Target: {example.target_prefix}{example.target_term}{example.target_suffix}")

Get Supported Languages

# Get all supported languages
languages = client.get_supported_languages()
 
# Translation languages
print("Translation languages:")
for code, lang in languages.translation.items():
    print(f"  {code}: {lang.name} ({lang.native_name})")
 
# Transliteration languages
print("\nTransliteration languages:")
for code, lang in languages.transliteration.items():
    print(f"  {code}: {lang.name}")
    for script in lang.scripts:
        print(f"    {script.code} -> {[t.code for t in script.to_scripts]}")
 
# Dictionary languages
print("\nDictionary languages:")
for code, lang in languages.dictionary.items():
    print(f"  {code}: {lang.name}")

Break Sentence

Identify sentence boundaries:

result = client.find_sentence_boundaries(
    body=["Hello! How are you? I hope you are well."],
    language="en"
)
 
for item in result:
    print(f"Sentence lengths: {item.sent_len}")

Translation Options

result = client.translate(
    body=["Hello, world!"],
    to=["de"],
    text_type="html",           # "plain" or "html"
    profanity_action="Marked",  # "NoAction", "Deleted", "Marked"
    profanity_marker="Asterisk", # "Asterisk", "Tag"
    include_alignment=True,      # Include word alignment
    include_sentence_length=True # Include sentence boundaries
)
 
for item in result:
    translation = item.translations[0]
    print(f"Translated: {translation.text}")
    if translation.alignment:
        print(f"Alignment: {translation.alignment.proj}")
    if translation.sent_len:
        print(f"Sentence lengths: {translation.sent_len.src_sent_len}")

Async Client

from azure.ai.translation.text.aio import TextTranslationClient
from azure.identity.aio import DefaultAzureCredential
 
async def translate_text():
    async with DefaultAzureCredential() as credential:
        async with TextTranslationClient(
            credential=credential,
            endpoint=endpoint,
        ) as client:
            result = await client.translate(
                body=["Hello, world!"],
                to=["es"]
            )
            print(result[0].translations[0].text)

Client Methods

Method	Description
`translate`	Translate text to one or more languages
`transliterate`	Convert text between scripts
`detect`	Detect language of text
`find_sentence_boundaries`	Identify sentence boundaries
`lookup_dictionary_entries`	Dictionary lookup for translations
`lookup_dictionary_examples`	Get usage examples
`get_supported_languages`	List supported languages

Best Practices

Pick sync OR async and stay consistent. Do not mix azure.xxx sync clients with azure.xxx.aio async clients in the same call path. Choose one mode per module.
Always use context managers for clients and async credentials. Wrap every client in with Client(...) as client: (sync) or async with Client(...) as client: (async). For async DefaultAzureCredential from azure.identity.aio, also use async with credential: so tokens and transports are cleaned up.
Batch translations — Send multiple texts in one request (up to 100)
Specify source language when known to improve accuracy
Use async client for high-throughput scenarios
Cache language list — Supported languages don't change frequently
Handle profanity appropriately for your application
Use html text_type when translating HTML content
Include alignment for applications needing word mapping

Reference Files

File	Contents
references/capabilities.md	Additional non-hero capabilities, operation-group coverage, and production checklists.
references/non-hero-scenarios.md	Dedicated non-hero examples for secondary/advanced scenarios.