nx_frame_codec.nx

nx_frame_codec.nx -- TUTOR-AUTHORED SCAFFOLD (Claude), NOT team-emitter output. 2-D LOSSLESS image/video-FRAME codec: JPEG-LS/FFV1-class MED (median) predictor over causal neighbors + zigzag + Rice residuals. FOUNDED ON nx_nv1_lpc: it REUSES that organ's proven Rice bit I/O (nv1l_rice_put/get), best-k (nv1l_best_k) and the predict->zigzag->Rice machinery, lifted from 1-D audio samples to 2-D pixels. Bit-exact (lossless): decode(encode(f)) == f. Tamper-evident: an out-of-range reconstruction returns -7 (the nv1_lpc doctrine). This is the lossless-VIDEO fidelity anchor (the bird's video pixel-exact); JPEG/lossy demotes to a measured fallback tier. main() is the unforgeable self-test: round-trip bit-exact on gradient/checker/edge/extremes/noise + compression-actual (gradient must shrink) + tamper. NOT FLOATING (operator law 2026-06-13 "don't build things that hang in the air"): it sits on nx_nv1_lpc (a GREEN organ). BACK-FILL: the team RE-AUTHORS this from a DATA spec via the emitter-of-emitters (X-AUT-006c/e/f, spec 2026-06-13-codec-dual-spec-and-kat.md) -- this scaffold is the sanctioned one-time bootstrap, NOT credited as team self-authoring (meter-integrity). v1 = single 8-bit plane, single MED predictor; the adaptive predictor library + NFC1 container are the named follow-ons. No verbatim escape yet (buffers sized 2n so noise cannot overrun). license_tier: ORIGINAL