Symbolic Representations, Constraints, and Reasoning

Documentation status: guide — see Maturity and evidence.

Symbolic structures make generated domain knowledge inspectable and testable.

A candidate concept should have explicit identity, ontology level, dependencies, relations, and constraints. A candidate relation should identify its roles and participant types. A candidate rule should state the proposition it protects rather than remain an informal comment.

Example constraint

Foundational package
  MUST NOT depend on a sector package

Primary role
  MUST NOT overlap another primary role
  of the same type and context
  over the same validity interval

Cooperation with generation

Neural proposal
  -> symbolic structure
  -> constraint evaluation
  -> contradiction / missing-data detection
  -> revised proposal
  -> validation

This loop allows the system to distinguish a fluent proposal from a structurally admissible model.