Document Auto-Split

Document auto-splitting groups pages of a multi-page batch document into logical subsets (e.g., separating a 20-page PDF containing 5 different invoices into 5 logical documents).

Sync vs. Async Processing

To protect performance, document splitting branches automatically based on the page count of the uploaded file:

• Synchronous Path (Files under 20 pages): The API returns a 200 OK response with a full split manifest immediately.
• Asynchronous Path (Files 20 pages or more): The API returns a 202 Accepted response with a job_id. The client polls the status endpoint until the manifest becomes available.

Plan Limits

• Pro: Up to 25 pages per batch file.
• Business: Up to 100 pages per batch file.

Response Manifest

The split manifest contains:

• splits: Logical document splits, including page ranges (start, end, list of page indexes), type classification (doc_type), and review status.
• unassigned_pages: Pages that were skipped or could not be classified.
• warnings: Operational notifications (e.g. fallback heuristics used).
• metrics: Execution performance details (e.g. pages per second).

Dashboard

Auto-split is also available in the Parsurai dashboard's Split tab for Pro and Business accounts.

Split PDF download

For synchronous split responses, you can optionally request per-split PDF files as inline Base64 output alongside the manifest. This output is sync-only — it is not available on the asynchronous/polled path — and large packets may be too large to return inline. If the response is too large, request the manifest-only output instead.

Manual split review

Pro and Business accounts can review and adjust split results directly in the dashboard: merge or split page ranges, reclassify document types, ignore pages, and extract accepted segments. This is a dashboard review tool for accounts with auto-split access, not a separate pricing tier.