Document primitive

Split API

Divide multi-page documents into typed subdocuments, then route each section into the right extraction, review, or automation path.

Try in playground Read docs

Mixed vendor packet

12 pages

invoicepages 1, 2, 5, 6

receiptpages 3, 4

contractpages 7-12

Upload a packet

Send a mixed PDF that contains invoices, receipts, contracts, or custom section types.

Define subdocuments

Describe the page types you want the API to identify and separate.

Route every section

Use returned page groups to run extraction, review, or downstream workflow branches.

Request

Send the document and describe the sections you expect.

`POST /v1/splits` returns page assignments for each subdocument. Add consensus when the split controls a production workflow.

documentRequired file or URL

The multi-page PDF or image document to split.

subdocumentsRequired array

Typed sections to identify, each with a name and description.

modelOptional string

Use retab-micro, retab-small, or retab-large.

contextOptional string

Business context that helps classify ambiguous pages.

n_consensusOptional integer

Run multiple voters and return confidence data.

Python

Starter call

from retab import Retab

client = Retab()

result = client.splits.create(
    document="batch.pdf",
    model="retab-small",
    subdocuments=[
        {
            "name": "invoice",
            "description": "Invoice pages with billing details and line items",
        },
        {
            "name": "receipt",
            "description": "Receipt or proof-of-payment pages",
        },
        {
            "name": "contract",
            "description": "Agreement pages with legal terms",
        },
    ],
    context="Q4 vendor packet",
)

for split in result.output:
    print(split.name, split.pages)

Response

Page groups

{
  "output": [
    { "name": "invoice", "pages": [1, 2, 5, 6] },
    { "name": "receipt", "pages": [3, 4] }
  ],
  "consensus": {
    "likelihoods": [
      { "name": 0.95, "pages": [0.95, 0.95, 0.94, 0.94] },
      { "name": 1.0, "pages": [1.0, 1.0] }
    ]
  }
}

Starter call

from retab import Retab client = Retab() result = client.splits.create( document="batch.pdf", model="retab-small", subdocuments=[ { "name": "invoice", "description": "Invoice pages with billing details and line items", }, { "name": "receipt", "description": "Receipt or proof-of-payment pages", }, { "name": "contract", "description": "Agreement pages with legal terms", }, ], context="Q4 vendor packet", ) for split in result.output: print(split.name, split.pages)

{ "output": [ { "name": "invoice", "pages": [1, 2, 5, 6] }, { "name": "receipt", "pages": [3, 4] } ], "consensus": { "likelihoods": [ { "name": 0.95, "pages": [0.95, 0.95, 0.94, 0.94] }, { "name": 1.0, "pages": [1.0, 1.0] } ] } }