AI · Document & photo interpretation

Document reading

Extract page regions and reading order while preserving page and source identity.

Workflow

1. Request document readingSelect supported pages without altering the original.
2. Prepare page geometryNormalize page geometry and identify candidate reading regions.
3. Order readable regionsProduce ordered regions with page coordinates and abstain when structure is unsupported.
4. Return unavailableThe workflow returns an explicit unavailable or refused outcome without inventing content or mutating unrelated records.
5. Return reading resultAttach ordered regions or unreadable outcomes to the original document pages.

Failure boundaries

• Rotated order wrongPartial/incorrect: Do not infer meaning
• No text regionUnreadable page: Keep source
• Result detached from pageInvalid: Reject result

ai.read.document