THRPY · EMC²

The undercurrent.
Not the sentence.

MMRY · this hostRemember

Title and content. Cite or abstain. Feeling is never the index.

EMC² · THRPY edgeResonate

Live silhouette versus historic feeling vectors. Theme echo. Thresholded cosine.

RealSeriesThen accommodate

High resonance may qualify a subtle Action — UI, pace, voice — inside bounds already set.

Mathematics ↓ Document library ↗

01 / THE MATH

A silhouette,
not a keyword.

The live turn is embedded locally (MiniLM-class, typically 384-d). That vector is the emotional silhouette of this sitting. Historic moments in the eKG are scored with cosine similarity. Resonance is \([-1, 1]\). Matches below 0.4 are dropped. High resonance (> 0.85) means the same theme may be back, even if “anxiety” was never typed.

  1. 01
    High > 0.85Undercurrent aligns with a past vector. Words may be completely different.
  2. 02
    Medium 0.65–0.85An echo. May inform background VAD or ADF confidence. Not a caption.
  3. 03
    Floor 0.4Below this, detectUndercurrent does not return the row. Treated as a fresher trajectory.
  4. 04
    Epiphany is separateBreakthrough is language + VAD + optional hypothesis test. Cosine is supporting context only.

Expected result Ask “what did we agree about Thursday?” — cite MMRY or abstain. Ask “have we been in this feeling before?” — that is an EMC² question on the THRPY engine, labelled, not a rewrite of Thursday.

02 / WHERE IT RUNS

Live on the edge.
Stub in some Python.

  1. 01
    LivedetectUndercurrent in @thrpy/core-engine and SensAffect ekg-core. Cosine vs stored vectors. Full EMC² is the thrpy.ai tier in THRPY platform docs.
  2. 02
    Toolssynthesize_multi_modal_hypothesis may pass emc2_resonance_query (“Grief”, “Anxiety”) against the subconscious cloud.
  3. 03
    StubOlder rflow_emc2.py returns example resonances. Do not treat that file as the live engine.
  4. 04
    This hostMMRY Cloud does not run embeddings for recall. Charter: full-text, cite or abstain. Optional feeling rows stay derived.

03 / READ MORE