Get a download link

For high volumes, up to 100 million rows per file. Every row is checked for format and catalogs while the batch validates, so `errors` lists what to fix before anything is stamped. The receiver is not verified against the SAT: the file is expected to contain data already checked with fiscal validations, and a document the stamping authority refuses ends `failed` in the report with its reason. Each party is identified by its RFC or its heru id, and with neither the row is invalid. Create the batch with its `type` and, optionally, `defaults` for the columns you would repeat on every row; PUT the CSV, plain or gzip, to `upload_url` within 24 hours; wait for `ready` and process it with `POST /batches/{id}/process`. Each type has its own columns, listed in the [batches guide](/guides/batches). The same documents can go through the JSON endpoints one by one. A batch lives 20 days at most from this call: one still unfinished at that point is cancelled as `POST /batches/{id}/cancel` would. Needs the `batches:write` scope, plus the scope that creating each document needs one by one: `invoices:write` or `retention_certificates:write`.