Knowledge lifecycle

CORE CONCEPT · KNOWLEDGE CHANGES WITH ITS SOURCES

Knowledge is not static after ingestion. lorebit treats source material, derived retrieval units, and answer citations as a lifecycle that must survive updates, withdrawals, and rebuilds.

Lifecycle stages

  1. Register a source and its owner, origin, and update expectations.
  2. Capture a version that can be traced back to the original material.
  3. Derive retrieval-ready units while retaining source and version links.
  4. Serve evidence for a query through the appropriate retrieval capabilities.
  5. Reconcile derived material when a source changes, disappears, or is corrected.
SOURCEVersioned material

Ownership and provenance remain attached to the original.

DERIVEDRetrieval units

Chunks, indexes, and relations are rebuildable artifacts, not the source of truth.

ANSWERVerifiable citations

Every response can show which evidence and version informed it.

Why adapters matter here

Different stores can own different derived artifacts. The lifecycle therefore asks for capabilities—such as preserving a source, querying full text, or rebuilding a semantic index—rather than making one database a permanent architectural dependency.

See database and index adapters.