Split

Split finds where each document starts and ends inside one combined file, for example a loan packet with an application, bank statements and tax returns scanned as one PDF. Split is a server feature.

curl -s -X POST http://localhost:8000/split -H "content-type: application/json" -d '{
  "input": "reducto://packet.pdf",
  "split_description": [
    {"name": "Application", "description": "The signed loan application form"},
    {"name": "Bank statement", "description": "Monthly bank statements", "partition_key": "account number"},
    {"name": "Tax return", "description": "Federal tax returns"}
  ],
  "settings": {"allow_page_overlap": true}
}'

With Reducto's SDK: client.split.run(input=..., split_description=[...]).

Field	Meaning
split_description[]	name, description, optional partition_key to split one type further (one part per account, per year)
split_rules	free-text rules, default: sections may only overlap at their first and last page
settings.allow_page_overlap	default true
settings.deep_split	a slower, more careful pass
settings.table_cutoff	truncate or preserve

The response lists each section with its pages and a confidence (conf). A section whose boundary was changed by a rule rather than by evidence is reported with low confidence.