Expand description
Knowledge corpus handlers — atoms, domains, TF-IDF search, fold, index.
Atoms and domains are stored in dedicated knowledge_atoms /
knowledge_domains tables (V19 migration) separate from the notes/entities
substrate. This lets the knowledge corpus scale to hundreds of thousands of
items without polluting the general-purpose store.
Verbs implemented here:
upsert_atoms— bulk insert/update atoms by slugupsert_domains— bulk insert/update domains (named atom groups)knowledge.get— fetch one atom or domain by ID or slugknowledge.list— paginated listingdelete_atoms— soft-delete by slugstats— corpus statisticsindex— backfill embeddings + FTS for atomsfold— budget-constrained knapsack selectionknowledge.search— TF-IDF + embedding rerank (default; opt out with rerank=false)