nx_mvault_measure.nx

nx_mvault_measure.nx -- NishiCaption-grade MEASUREMENT atom for the vault's comprehensive first-pass tagging (the training-corpus label). WHY (operator 2026-07-17): "first tagging pass comprehensive ... joy caption like exercise ... classification for our models and loras both llm and image ... state of the art." The elder-ai NishiCaption design (95 axes, 3 legs, round-trip) is the SOTA schema; this is its capture atom on the sovereign store so the EXPENSIVE VLM/math ensemble pass runs ONCE and nothing is lost. A measurement is a triangulated axis reading: "<axis>=<value>@<leg>#<conf>" e.g. hip_waist_ratio=700/1000@math#920 hair_color=blonde@vlm#980 value = lossless (ivalue/scale) for regressors, a token for enum/classifier leg = which of MATH/VLM/HUMAN/FUSED produced it (triangulation) conf = permille confidence (training threshold + fusion weight) Round-trip variance (directive vs measured) = the self-improvement + LoRA training signal. Caption modes = the 3 NishiCaption emit forms. Extensible by construction (like nx_mvault_tag): NO axis list is hardcoded, so all 95 axes -- and future ones -- are just data. Composes nx_mvault_tag (mv_u_dec). license_tier: ORIGINAL