Translate Text and Documents with Azure SDKs

Wraps Azure's REST-style TypeScript clients for text translation (@azure-rest/ai-translation-text) and document translation (@azure-rest/ai-translation-document), each installed alongside @azure/identity. The text client authenticates with an API key plus region (TranslatorCredential) against a regional endpoint, or a bare credential against the global endpoint; the document client authenticates with either a DefaultAzureCredential token or an API key against a Cognitive Services endpoint.

const response = await client.path("/translate").post({
  body: {
    inputs: [
      {
        text: "Hello, how are you?",
        language: "en",
        targets: [
          { language: "es" },
          { language: "fr" },
        ],
      },
    ],
  },
});

Text translation covers translating one or more inputs to multiple target languages in one call, with per-target options (textType Plain/Html, profanityAction NoAction/Marked/Deleted, an LLM-specific tone); listing supported languages (code, name, and native name); transliterating text between scripts (e.g. Han characters to Latin, with source/target script parameters); and language detection with a confidence score. Every call checks isUnexpected(response) before reading the body, throwing the returned error object if the call failed.

Document translation covers a single-document translation posted as multipart form data and streamed back to a file (asNodeStream()), and batch document translation against Azure Blob Storage containers: generating time-limited SAS URLs for source (read/list) and target (read/write/list) containers, starting a batch job referencing those SAS URLs and target languages, and extracting the operation ID from the operation-location response header. Batch status is polled via the batch's status endpoint (overall status plus a success/total summary) and its documents endpoint, paginated with a paginate() helper to iterate every document's individual status. A supported-formats endpoint lists each format's name and file extensions.

Stated best practices: omit the language parameter to auto-detect the source language, batch multiple texts into one call for efficiency, use time-limited SAS tokens for document translation rather than long-lived credentials, always check isUnexpected(response) before accessing the body, and use regional endpoints for lower latency.

Use it for text or document translation needs in a TypeScript application: single or batch text translation, transliteration, language detection, single-document translation, or batch document translation against blob storage. As with any skill output, verify results against the current environment, and stop to ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.

Input is text or a document (or a container of documents) plus target language(s) and, for documents, storage container access. Output is translated text per target language, a transliterated string in the target script, a detected language with confidence, a translated document file, or a batch translation job's status and per-document results.

Uses @azure-rest/ai-translation-text and @azure-rest/ai-translation-document against the Azure Translator/Cognitive Services API, @azure/identity for token-based auth, and @azure/storage-blob to generate SAS URLs for batch document translation source and target containers.

TypeScript developers adding text or document translation, transliteration, or language detection to an application via Azure's Translator service.