Reference

`knowledge.document`

domain: knowledge

A registry record for an INGESTED document — a law, act, government order, notification, standard, guideline, court judgment, or general reference. The raw bytes live in the content-addressed filestore (`blob` = their sha2-256), so this record is metadata + a digest, never the payload. The document-variety axes are ORTHOGONAL, not one flat type: `doc_class` (what kind), `legal_force` (the axis that changes decisions — mandatory law vs advisory guideline vs adjudicative order vs informational reference), `issuer`/`jurisdiction`/temporal validity, and `citation_style` (how to cite it: section/clause vs table/row vs paragraph). Its versions/chunks (the Merkle-manifest corpus) are separate records (knowledge.version / knowledge.chunk) that reference this doc.

Field	Type	Required	Description
`schema`	`—`	✓
`id`	`string`	✓
`doc_id`	`string`	✓
`title`	`string`	✓
`doc_class`	`string`	✓	the kind of instrument (a governed vocabulary in the full model)
`legal_force`	`string`	✓	the axis that drives decision strength: a binding law ⇒ mandatory rules; a guideline ⇒ advisory; a court order ⇒ adjudicative; general knowledge ⇒ reference
`issuer`	`string`		the authority that issued it, e.g. CPCB, CAQM, Parliament, High Court — ties to ref.source
`jurisdiction`	`object`		where it applies, e.g. {level: national\|state\|local, code: ...}
`effective_from`	`string`
`effective_to`	`string/null`
`supersedes`	`string/null`		doc_id this replaces/amends
`citation_style`	`string`		how to cite it: 'section' \| 'clause' \| 'table' \| 'paragraph' \| 'page' — drives the chunk locator + citation format
`format`	`string`	✓	the raw bytes' format (drives the S3 extractor)
`blob`	`string`	✓	sha2-256 filestore id of the raw document bytes (content-addressed)
`checksum`	`string`		= blob (the digest is the integrity check)
`source`	`object`		provenance of acquisition: {url, retrieved_at} or {upload: true}
`current_version`	`string/null`		knowledge.version id of the latest chunked/embedded version (set by S3)
`status`	`string`		acquired \| chunked \| indexed
`tenant`	`string`