Translate Persian to English as a separate layer.

Request

Send one text field and no unknown fields. The input may contain mixed Persian and English, but it must include at least one Persian-script letter and may contain at most 6,000 Unicode code points. The complete serialized JSON body must be at most 32 KiB (32,768 bytes).

const response = await fetch("https://api.vowelmarks.com/v1/translate", {
  method: "POST",
  headers: {
    authorization: "Bearer " + process.env.VOWELMARKS_API_KEY,
    "content-type": "application/json",
    "idempotency-key": jobId,
  },
  body: JSON.stringify({ text: "من فارسی می‌خوانم" }),
});

Response

{
  "request_id": "95b…",
  "engine_version": "vowelmarks-api-v1.2026-08-10",
  "original": "من فارسی می‌خوانم",
  "translation": "I read Persian.",
  "source_language": "fa",
  "target_language": "en",
  "usage": {
    "meter": "text", "input_code_points": 17,
    "multiplier": 2, "units": 34,
    "additional_units": 0, "cache_hit": false
  }
}

Metering and failure behavior

Translation charges two text units for each submitted Unicode code point. The returned English text and response metadata are unmetered. A successful organization-isolated cache hit still charges the normal 2× amount; validation, provider, and failed-generation responses charge nothing.

• Keep original visible if your product is meant to support Persian reading.
• Do not parse undocumented provider or model details; they are deliberately absent.
• Use request_id for support and Idempotency-Key for retries.
• A retryable 503 means no configured path produced a valid result, not that a lower-quality translation was substituted.