Ingestion
Scan/OCR Ingestion
Scanned PDFs, images, and signed documents are preserved as raw originals, OCR-normalized, and routed into legal review before any canonical clause JSON is eligible for Merkle hashing.
Raw uploads
Verified2
Original files preserved
OCR jobs
Mock2
Mock provider only
Clause candidates
Review3
OCR-derived source material
Merkle eligible
Approved1
2 blocked pending review
OCR Jobs
External OCR disabled| Document | Provider | Extraction | Candidates | Confidence |
|---|---|---|---|---|
| facility-agreement-demo-scan.pdf Facility Agreement | mock | 1 page 3 blocks / 3 lines | 2 candidates 1 approved / 1 review | 93% |
| demand-letter-demo-signed.pdf Demand Letter | mock | 1 page 3 blocks / 2 lines | 1 candidates 0 approved / 1 review | 94% |
Review Queue
Legal review requiredFacility Agreement
Page 21 / Clause 21.1
A payment default candidate should be routed for legal review before becoming canonical.
Confidence 93% / Extracted 02 May, 11:10 UTC
Demand Letter
Page 1 / Paragraph 2
A demand package may be staged for review but not issued by the application.
Confidence 94% / Extracted 02 May, 11:00 UTC
Facility Agreement
Page 21 / Payment status table
Payment status table follows. Metric Status Payment overdue Review
Confidence 88% / Extracted 02 May, 11:10 UTC
Controls
Evidence boundaryRaw original preserved
Uploads retain raw storage URIs and source hashes before OCR processing.
Human legal review gate
Candidates stay evidential until a reviewer approves canonical clause JSON.
Merkle exclusion by default
Unreviewed OCR output is blocked from canonical clause Merkle roots.