Document intake model
ProSe should treat documents as evidence-bearing records, not disposable uploads. The platform should preserve the source file, metadata context where appropriate, review state, and links between extracted information and the source.
What good ingest supports
Readable text
OCR and extraction can make documents searchable, but extracted text should remain tied to original files.
Human review
People should be able to correct fields, reject bad extraction, and confirm important information before relying on it.
Source traceability
Facts, events, tasks, and evidence entries should trace back to uploaded material when possible.
Safe packets
Packet preparation should include only what belongs in the packet and should make missing material visible.
Public expectation
Document handling should be clear enough for ordinary people and rigorous enough for serious legal-operation workflows.