01
kettle / technical_overview

Technical overview.

Kettle processes collection material, writes extracted records to a graph, and keeps citations on claims. Object IDs and signing tools run internally. The public verification specification is still in development.

SHA-256 Object IDsClaim CitationsSensitivity GatingGraph RetrievalPublic Verifier · Developing
02
kettle / public_architecture

System map

These are the three names used on the public site.

Kettle RefineryCurrent

Reads collection material and turns it into clean records.

  • OCR and handwriting recognition
  • Layout reconstruction and segmentation
  • Field normalization and record extraction
  • Document, person, event, and claim models
  • Collection-specific processing rules
GemyndCurrent

Stores the records and supports search and research.

  • Multi-stage graph ingestion
  • Claims with source references
  • Sensitivity filters
  • Authenticated research access
  • Exports covered by the collection agreement
Verification Standard

Defines record IDs, signatures, manifests, and change events. Kettle has an internal reference implementation. The public specification and outside verifier are unfinished.

03
kettle / access_boundaries

Access boundaries

Source files, extracted records, and research results have separate controls.

Zone A - source and ingest
  • ●Authorized source access
  • ●Path and input validation
  • ●Sensitivity and quarantine controls
  • ●Record anchoring integration
Zone B - extracted records
  • ●Authenticated access
  • ●Role-based retrieval
  • ●Quarantine exclusion
  • ●PII output handling
Zone C - research and delivery
  • ●Supporting claim IDs
  • ●Page and document citations
  • ●Collection use rules
  • ●Outside verification tools
A verified state matches an ID or signed commitment created earlier. It does not prove that every historical statement is true, and it does not grant a right to use the record.
04
kettle / status

Current status

The labels below refer to the capability named in each row.

CapabilityStatusScope
Document and graph ingestionDEPLOYEDCollection workflows produce clean text, records, citations, and graph data.
Research retrievalDEPLOYEDSearch results carry supporting claims and source references.
OCR and segmentationFUNCTIONALHistorical material is working; difficult layouts and reading order still need improvement.
Object IDs and signingFUNCTIONALAn internal reference implementation creates IDs and tests signatures.
Public verification specificationROADMAPVersioned specification, test vectors, and outside verifier.
Institution-held source pilotROADMAPMove signed IDs and approved data between systems while the institution keeps the master files.
05
kettle / internal_names

Internal names

Engineering notes may refer to Wreon, Aletheia, Kettle Core, and ALBC. Those are internal components, not additional products.

Read the verification model or see deployments and status.