nx_sov_ledger.nx

nx_sov_ledger.nx -- CANONICAL generic sovereign ledger on nx_seg_store. The DRY primitive that makes "sovereign is as easy as a TSV" and kills the FRICTION root-cause of the flat-file habit. WHY THIS EXISTS (census 2026-07-14, runtime+_hdl_build/16029 .nx): ~599 organs write .tsv, ~1407 .log, ~965 .txt for INTERNAL state vs only 324 seg_store adopters -- a ~10:1 flat-file default. Root cause is not a missing store (nx_seg_store has time-travel history + compaction + postings = SOTA) but (a) no enforcement sensor and (b) writing a TSV is 3 lines while seg_store means begin/add/commit + segid discipline + a reader. This lib collapses that to one call. Dependency-light (seg_store + syscalls ONLY) so ANY organ can adopt without dragging a domain stack. SOVEREIGN RECORD FORMAT (not burdened by others' non-SOTA decisions): typed BINARY records -- n i64 stored as raw 8n little-endian bytes, NO tab/comma delimiters, NO stringly-typed parsing. (Even seg_store adopters like nx_seedlib still pack tab bytes into the value; this is the fix for that too.) sov_put(prefix,key,val,vlen) KV put, idempotent latest-wins, segid=sys_now_us() (per-op-in-loop safe -- ms collides & silently drops a record, [[reference-segstore-segid-use-now-us]]) sov_has(prefix,key) 1/0 existence (replaces "does <file>.tsv exist" resume checks) sov_get(prefix,key,ptrout,lenout) newest value (zero-copy ptr into the store map) sov_get_copy(prefix,key,out,cap) newest value copied into out (aligned); returns len or -1 sov_put_ints(prefix,key,vals,n) typed record: n i64 -> raw 8n bytes sov_get_ints(prefix,key,out,maxn) read them back -> count, or -1 if absent (never fabricated) license_tier: ORIGINAL