Knowledge Graph and Memory

The long-memory setup and reports point toward a graph-friendly archive: claims, sources, entities, events, topics, contradictions, and review states should be modeled separately.

Core Ideas

Entity pagesFuture records can track people, organizations, texts, doctrines, places, technologies, claims, and signals.

Claim provenanceEvery strong claim should keep source spans, review state, contradiction state, and last-reviewed dates.

Memory boundaryRaw sources stay in archive; hot memory keeps only current routing and reviewed summaries.