Figeometrica — executable theory infrastructure for rhetorical figures

A theory-compilation framework that turns classical rhetoric (≈456 named figures of speech, historically defined only in prose) into structured, machine-checkable specifications. Includes: per-figure JSON entries with FigureSignature contract blocks and an evidence ladder enforced by CI; a deterministic matcher producing byte-exact evidence spans; a witness protocol with inverse reconstruction; versioned vocabulary manifests recording how the theory itself evolves. Multilingual evidence layer (English/Indonesian).