Feature

Review, Correction and Approval

Documents with a disputed, unreadable or missing field wait for a human — who can correct fields and approve.

What is Review, Correction and Approval in Mahad OCR?

Available now. Status is assigned automatically: could not be read → failed, at least one field disputed, unreadable or missing → review, every required field confirmed by evidence → verified. Input: Any processed document. Limitation: Multi-step approval chains, reviewer assignment and SLA timers are not implemented — review is a single-step action today.

Available now

What it does

Status is assigned automatically: could not be read → failed, at least one field disputed, unreadable or missing → review, every required field confirmed by evidence → verified. Anything not verified appears in the review queue where fields can be edited inline and approved or rejected.

How it works

1Document processed and each field checked against its evidence
2Non-verified documents enter the review queue
3Reviewer corrects fields inline
4Approve or reject — the action is audit-logged

Sample result

Illustrative response shape using synthetic data.

{ "status": "review", "result": { "decision": "review", "review_reasons": ["expiry_date: unreadable"] } }

Supported inputs

Any processed document.

Fields returned

status
field_review (✓ / ⚠ per field, with reason and both readings)
edited field values
approval record

API

POST /v1/documents/{id}/approve · /reject

Security & privacy

Every approval and rejection is written to the hash-chained audit log.

Current limitations

Multi-step approval chains, reviewer assignment and SLA timers are not implemented — review is a single-step action today.