nx_vec_hybrid_gate.nx

nx_vec_hybrid_gate.nx -- INTEGRATION proof that R-VEC-0 (kernel) + R-VEC-1 (embeddings) + R-VEC-5 (RRF fusion) COMPOSE into working sovereign hybrid semantic search -- the felt S-class moment, measured end-to-end. Scenario (the classic lexical-recall failure): user query = "marriage". Only doc0 literally contains the token "marriage"; doc1 is the OTHER family-law doc (custody/children/parents/divorce) and never says "marriage", so pure keyword/BM25 search CANNOT find it. Sovereign distributional embeddings know marriage ~ divorce/custody (same context), so the vector ranker surfaces doc1; RRF fusion then RECOVERS doc1 into the results -- without losing the exact match (doc0 stays #1) and without false positives (the estate/injury docs stay out). This is the whole point of hybrid search: semantic recall on top of lexical precision. All sovereign, no-float, no external weights, no third party. expect_exit: 0 license_tier: ORIGINAL