Batch operations
Process ordered items and handle partial failures explicitly.
Use POST /v1/batch/syllabify or POST /v1/batch/transliterate.
Transliteration also requires the top-level scheme: "ala-lc-2011".
{
"items": [
{ "id": "first", "text": "က" },
{ "id": "invalid", "text": "\ud800" }
]
}The escaped lone surrogate deliberately demonstrates invalid_unicode. The
response is HTTP 200 with the first item successful and the second failed.
Always inspect each item's success, result or error. Results retain input
order and IDs. Do not retry successful items when a batch partially fails.
Limits
| Limit | Default |
|---|---|
| Text length | 100,000 Unicode code points per item |
| Batch size | 1–1,000 items |
| Combined text size | 1,000,000 UTF-8 bytes |
| HTTP body | 1 MiB, including JSON keys and escaping |
Any limit can reject a request. IDs must be unique, non-empty strings. Invalid envelopes fail the whole request; per-item text validation failures in a valid envelope are data. Unexpected internal failures are not disguised as successful item results.
Quotas
A batch reserves one request and the sum of submitted code points. Reservations occur before validation/processing, so later errors are not refunded. See Errors and quotas before designing retries.