Register the source
Identity, edition, owner, date, rights, location, and cryptographic fingerprint.
Books, reports, laws, policies, and records are source material. TEHUTI converts them into small, traceable knowledge objects an AI can retrieve, compare, cite, and reason over—without losing the original meaning.
Identity, edition, owner, date, rights, location, and cryptographic fingerprint.
Sections, pages, tables, definitions, citations, dates, and named entities.
One idea per object. Quotes remain distinct from summaries and inference.
Concepts, dependencies, conflicts, jurisdictions, time ranges, and evidence links.
Human review, confidence, embeddings, access policy, versioning, and evaluation.
A source is not “trained into” the system and forgotten. Each useful statement keeps its origin, meaning, limits, and review state.
{
"object_id": "ko_example_001",
"source_id": "src_book_001",
"source_type": "book",
"title": "Source title",
"author_or_issuer": "Named author or institution",
"source_location": {
"edition": "1",
"page": 42
},
"passage": "Exact source passage is stored here.",
"concepts": [
"financial principle",
"risk"
],
"claims": [
{
"text": "One atomic, testable claim.",
"type": "descriptive",
"support": "direct"
}
],
"relationships": [
{
"predicate": "QUALIFIES",
"object_id": "ko_example_002"
}
],
"jurisdiction": "US-IL",
"effective_date": "YYYY-MM-DD",
"confidence": 0.95,
"verification_status": "human_review_required",
"content_hash": "sha256:…",
"citation": "Author, Title, edition, page 42"
}
This defines the language TEHUTI will use. Source registration, extraction, review, retrieval, and evaluation will be added as controlled API capabilities. No book or source is marked AI-ready until it passes this contract.