TMR System

TRION Meaning Representation.

Meaning is normalised before routing decides. Same intent. Different text. Same route.

status: core model implemented — full-coverage rollout in progress

TMR is the missing link between what a user writes and what the pipeline processes. Instead of applying routing directly to raw text, TMR normalises the meaning first — independent of phrasing, language, or context. The typed representation model is implemented and already drives occurrence-mapped projections in production for container and memory-recall routing. Meanings outside that mapped scope still fall back to a sanitised shadow trace while coverage keeps expanding.

raw_text → TMR → routing_frame

Normalisation Before Routing.

The RoutingFrame operates on a normalised meaning representation — not on the raw text. What the user writes is input. What TMR produces from it already drives routing for its mapped occurrences today, with broader coverage rolling out.

speech_act · predicate · theme · roles · confidence · provenance

One Typed Representation.

TMR captures meaning as a typed, immutable structure — speech act, predicate, roles, temporality, polarity, modality, and more — with confidence and provenance attached to every field, so routing never has to guess.