Data Model
What an engine stores: source records, the edge properties that join them into a hypergraph, connections and groups, the authority each record carries, and corrections and adjustments. The rules that govern them are in the governance model.
The model is designed to be as simple as possible while providing maximum flexibility in what it can represent.
The parts
Source record. Data — any content an agent or a program stores: a sentence, a document, a large JSON object, a part of a file — kept as it was written, with its source type, authority, the system that wrote it and when. See source records.
Edge properties. The core of the data model: label:value pairs on a record, like person:Sarah or topic:database. When two or more records share one, an edge is imputed — one N-ary edge touching every record that carries it. The records and their edge properties form the Edge Property Graph, a hypergraph. Agentic annotation, adjustments (a linked list), connections (1-to-1) and groups (membership) are idioms on that one mechanism. See edge properties.
Connections and groups. The asserted kind of edge: a named relation or membership, written by the engine as one reserved edge property value shared by every member. See connections and groups.
Authority. How much weight a record carries, set by its source type and returned by e_read. See authority.
Corrections. The truth, attached to the record it corrects. The record's content is not changed. See corrections.
Adjustments. A newer record linked to an earlier one that was right and has since changed. See updates and adjustments.
Governance. Policies, agent profiles, lanes and source types decide who reads and writes what, and what a record may be. See the governance model.
Durability and the mesh. The archive keeps a copy from which an engine is rebuilt; the mesh exchanges records between engines.