Microsoft Azure AI Translation Document Py

Azure AI Document Translation SDK for Python

Client library for Azure AI Translator document translation service for batch document translation with format preservation.

Installation

pip install azure-ai-translation-document

Environment Variables

AZURE_DOCUMENT_TRANSLATION_ENDPOINT=https://<resource>.cognitiveservices.azure.com  # Required for all auth methods
# Storage for source and target documents
AZURE_SOURCE_CONTAINER_URL=https://<storage>.blob.core.windows.net/<container>?<sas>  # Required for all auth methods
AZURE_TARGET_CONTAINER_URL=https://<storage>.blob.core.windows.net/<container>?<sas>  # Required for all auth methods
AZURE_TOKEN_CREDENTIALS=prod # Required only if DefaultAzureCredential is used in production
AZURE_DOCUMENT_TRANSLATION_KEY=<your-api-key>  # Only required for the legacy API-key auth path below

Authentication & Lifecycle

🔑 Two rules apply to every code sample below:

1. 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.
2. 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.document import DocumentTranslationClient

# 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 DocumentTranslationClient(
    endpoint=os.environ["AZURE_DOCUMENT_TRANSLATION_ENDPOINT"],
    credential=credential,
) as client:
    statuses = list(client.list_translation_statuses())

Legacy: API Key (existing keyed deployments)

New code should use DefaultAzureCredential above. Use AzureKeyCredential only if you have an existing keyed deployment that hasn't been migrated to Entra ID yet — for example, regulated environments still completing their Entra rollout.

import os
from azure.core.credentials import AzureKeyCredential
from azure.ai.translation.document import DocumentTranslationClient, SingleDocumentTranslationClient

with DocumentTranslationClient(
    endpoint=os.environ["AZURE_DOCUMENT_TRANSLATION_ENDPOINT"],
    credential=AzureKeyCredential(os.environ["AZURE_DOCUMENT_TRANSLATION_KEY"]),
) as client:
    statuses = list(client.list_translation_statuses())

# SingleDocumentTranslationClient accepts the same key-based credential.

Basic Document Translation

from azure.ai.translation.document import DocumentTranslationInput, TranslationTarget

source_url = os.environ["AZURE_SOURCE_CONTAINER_URL"]
target_url = os.environ["AZURE_TARGET_CONTAINER_URL"]

# Start translation job
poller = client.begin_translation(
    inputs=[
        DocumentTranslationInput(
            source_url=source_url,
            targets=[
                TranslationTarget(
                    target_url=target_url,
                    language="es"  # Translate to Spanish
                )
            ]
        )
    ]
)

# Wait for completion
result = poller.result()

print(f"Status: {poller.status()}")
print(f"Documents translated: {poller.details.documents_succeeded_count}")
print(f"Documents failed: {poller.details.documents_failed_count}")

Multiple Target Languages

poller = client.begin_translation(
    inputs=[
        DocumentTranslationInput(
            source_url=source_url,
            targets=[
                TranslationTarget(target_url=target_url_es, language="es"),
                TranslationTarget(target_url=target_url_fr, language="fr"),
                TranslationTarget(target_url=target_url_de, language="de")
            ]
        )
    ]
)

Translate Single Document

from azure.ai.translation.document import SingleDocumentTranslationClient
from azure.identity import DefaultAzureCredential

with open("document.docx", "rb") as f:
    document_content = f.read()

with SingleDocumentTranslationClient(endpoint, DefaultAzureCredential()) as single_client:
    result = single_client.translate(
        body=document_content,
        target_language="es",
        content_type="application/vnd.openxmlformats-officedocument.wordprocessingml.document"
    )

# Save translated document
with open("document_es.docx", "wb") as f:
    f.write(result)

Check Translation Status

# Get all translation operations
operations = client.list_translation_statuses()

for op in operations:
    print(f"Operation ID: {op.id}")
    print(f"Status: {op.status}")
    print(f"Created: {op.created_on}")
    print(f"Total documents: {op.documents_total_count}")
    print(f"Succeeded: {op.documents_succeeded_count}")
    print(f"Failed: {op.documents_failed_count}")

List Document Statuses

# Get status of individual documents in a job
operation_id = poller.id
document_statuses = client.list_document_statuses(operation_id)

for doc in document_statuses:
    print(f"Document: {doc.source_document_url}")
    print(f"  Status: {doc.status}")
    print(f"  Translated to: {doc.translated_to}")
    if doc.error:
        print(f"  Error: {doc.error.message}")

Cancel Translation

# Cancel a running translation
client.cancel_translation(operation_id)

Using Glossary

from azure.ai.translation.document import TranslationGlossary

poller = client.begin_translation(
    inputs=[
        DocumentTranslationInput(
            source_url=source_url,
            targets=[
                TranslationTarget(
                    target_url=target_url,
                    language="es",
                    glossaries=[
                        TranslationGlossary(
                            glossary_url="https://<storage>.blob.core.windows.net/glossary/terms.csv?<sas>",
                            file_format="csv"
                        )
                    ]
                )
            ]
        )
    ]
)

Supported Document Formats

# Get supported formats
formats = client.get_supported_document_formats()

for fmt in formats:
    print(f"Format: {fmt.format}")
    print(f"  Extensions: {fmt.file_extensions}")
    print(f"  Content types: {fmt.content_types}")

Supported Languages

# Get supported languages
languages = client.get_supported_languages()

for lang in languages:
    print(f"Language: {lang.name} ({lang.code})")

Async Client

from azure.ai.translation.document.aio import DocumentTranslationClient
from azure.identity.aio import DefaultAzureCredential

async def translate_documents():
    async with DefaultAzureCredential() as credential:
        async with DocumentTranslationClient(
            endpoint=endpoint,
            credential=credential,
        ) as client:
            poller = await client.begin_translation(inputs=[...])
            result = await poller.result()

Supported Formats

Storage Requirements

• Source and target containers must be Azure Blob Storage
• Use SAS tokens with appropriate permissions:
  • Source: Read, List
  • Target: Write, List

Best Practices

1. 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.
2. 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.
3. Use SAS tokens with minimal required permissions
4. Monitor long-running operations with poller.status()
5. Handle document-level errors by iterating document statuses
6. Use glossaries for domain-specific terminology
7. Separate target containers for each language
8. Use async client for multiple concurrent jobs
9. Check supported formats before submitting documents