Skip to main content
State stores fields in a DirectedAcyclicHypergraph (DAH). The DAH generalizes a directed acyclic graph by allowing a hyperedge with multiple source nodes and one target node.

​
Node indexes

Each node contains:
  • a generated UUID used for internal identity;
  • a string path used for public lookup;
  • an optional payload; and
  • incoming hyperedges plus reverse dependent links.
The graph maintains both a UUID-to-node map and a path-to-UUID secondary index. State.get_field() and other public state methods address nodes by path.

​
Hyperedge semantics

A target may have multiple incoming hyperedges. Readiness follows an OR of AND groups: a target is ready when any incoming hyperedge has all of its sources satisfied. Cycle checks treat each source-to-target relationship as a directed edge for ordering. Single-parent field hierarchy uses one-source hyperedges. Callers can reach the underlying structure with state.get_dah() when they need multi-source relationships:

​
State integration

set_field() automatically creates parent nodes and path-segment dependencies. copy() recreates nodes and hyperedges in a new state; interpretive state deep-copies its Pydantic field payloads. remove_field() removes the selected node and incident hyperedges. It does not promise a recursive subtree delete, so remove descendant paths explicitly if that is the desired application behavior.

​
Stability boundary

DAH classes are importable source modules and have direct unit coverage, but they are internal infrastructure within a source-only alpha. Prefer the State methods for application code unless the hypergraph operations are specifically required and tested by your integration.