Create Document Translation Job

from sarvamai import SarvamAI
import requests
client = SarvamAI(api_subscription_key="YOUR_SARVAM_API_KEY")
job = client.document_translation.create(
    job_name="chapter1",
    source_language_code="en-IN",
    target_language_codes=["hi-IN", "ta-IN"],
    original_filename="chapter1.pdf",
)
# SDK does not upload the file — PUT bytes to the signed URL, then start.
with open("chapter1.pdf", "rb") as f:
    requests.put(
        job.upload_url,
        data=f,
        headers={
            "Content-Type": "application/pdf",
            "x-ms-blob-type": "BlockBlob",
        },
    )
client.document_translation.start(job_id=job.job_id)

{
  "job_id": "550e8400-e29b-41d4-a716-446655440000",
  "upload_url": "https://api.sarvam.ai/translate/document/uploads/550e8400-e29b-41d4-a716-446655440000?sig=...",
  "expires_in_hours": 24,
  "source_language_code": "en-IN",
  "target_language_codes": [
    "hi-IN",
    "ta-IN"
  ]
}

Document Translation

Create a document translation job. Returns a job_id and a short-lived signed upload_url.

Flow:

POST /translate/document/jobs with source/target languages and options.
PUT the raw file bytes to upload_url with header Content-Type: <mime>.
POST /translate/document/jobs/{job_id}/start to begin the pipeline.
Poll GET /translate/document/jobs/{job_id}/live-status until translation completes.
POST /translate/document/jobs/{job_id}/export?lang=<code>&format=<fmt> to enqueue an async export for each target language.
Poll GET /translate/document/jobs/{job_id}/export/status until export_state is Completed, then download from download_url.

A single job translates into every language in target_language_codes. Exports run asynchronously — do not expect the download URL in the trigger response.

## Document Translation Create a document translation job. Returns a `job_id` and a short-lived signed `upload_url`. **Flow:** 1. `POST /translate/document/jobs` with source/target languages and options. 2. `PUT` the raw file bytes to `upload_url` with header `Content-Type: <mime>`. 3. `POST /translate/document/jobs/{job_id}/start` to begin the pipeline. 4. Poll `GET /translate/document/jobs/{job_id}/live-status` until translation completes. 5. `POST /translate/document/jobs/{job_id}/export?lang=<code>&format=<fmt>` to enqueue an async export for each target language. 6. Poll `GET /translate/document/jobs/{job_id}/export/status` until `export_state` is `Completed`, then download from `download_url`. A single job translates into every language in `target_language_codes`. Exports run asynchronously — do not expect the download URL in the trigger response.

Authentication

api-subscription-keystring

API Key authentication via header

Request

This endpoint expects an object.

job_namestringRequired1-255 characters

Human-readable job name.

source_language_codeenumRequired

Source language (BCP-47).

original_filenamestringRequired>=1 character

Original filename with extension. Supported: PDF, OOXML, legacy Office, HTML, ODF.

target_language_codeslist of enumsRequired

Target languages (BCP-47).

genreenum or nullOptional

Genre enum name (e.g. NON_FICTION). Use the enum constant name, not the human-readable label.

metadatamap from strings to any or nullOptional

Arbitrary key-value metadata attached to the job.

auto_processbooleanOptionalDefaults to true

When true (default), translation begins automatically after the document is parsed.

model_typeenum or nullOptional

Translation model quality tier: lite (faster, short-form content) or plus (higher quality, long-form content). Defaults to plus when omitted.

use_native_numeralsboolean or nullOptional

null (default) uses the application default. true uses native numerals; false uses international numerals.

style_guidelinesstring or nullOptional

Global tone, terminology, and formatting instructions (up to ~4000 characters). Applies to every target language.

language_specific_guidelinesmap from strings to strings or nullOptional

Per-language overrides (up to ~4000 characters each). Keys must match entries in target_language_codes.

Response

Job created.

job_idstring

Unique job identifier.

upload_urlstring

Signed URL for direct file upload.

expires_in_hoursinteger

Hours until the upload URL expires.

source_language_codestring

Echo of request source language.

target_language_codeslist of strings

Echo of target languages.

Errors

400

Bad Request Error

401

Unauthorized Error

402

Payment Required Error

500

Internal Server Error

Document Translation

Create a document translation job. Returns a job_id and a short-lived signed upload_url.

Flow:

POST /translate/document/jobs with source/target languages and options.
PUT the raw file bytes to upload_url with header Content-Type: <mime>.
POST /translate/document/jobs/{job_id}/start to begin the pipeline.
Poll GET /translate/document/jobs/{job_id}/live-status until translation completes.
POST /translate/document/jobs/{job_id}/export?lang=<code>&format=<fmt> to enqueue an async export for each target language.
Poll GET /translate/document/jobs/{job_id}/export/status until export_state is Completed, then download from download_url.

A single job translates into every language in target_language_codes. Exports run asynchronously — do not expect the download URL in the trigger response.