nx_txtscan.nx

nx_txtscan.nx -- CLI over nx_txtscan_lib: sovereign, BYTE-ORIENTED text search. The estate already had two sovereign greps (nx_codegrep, nx_shelltool grep) and both are line-oriented, because both reimplemented grep's worldview rather than only its independence. On knowledge/library/write_nsfwbot_flowgpt_2601_14324.txt -- 112,755 bytes, ONE line, which is what scraped text normally looks like -- asking for a pattern returns the whole file as a single "match". This organ reports OFFSETS, which every byte stream has, instead of lines, which only some text has. nx_txtscan find <path> <pattern> [ctx] every hit: offset + a clamped context window nx_txtscan count <path> <pattern> occurrences (overlapping) nx_txtscan ctl <path> non-tab/LF/CR control bytes + run lengths `ctl` is not a bonus mode: it is the capability that found 211 corrupted sources in the local tree, each carrying one 119-byte run of 0x0E where a migration should have rewritten a number formatter. nx_cc accepts 0x0E as whitespace, so every one of those files compiles clean. u2605HONEST TRUNCATION IS PART OF THE CONTRACT. This organ prints `truncated=1` with the counts whenever it did not see everything -- because a capped listing that reports absence is a silent wrong answer, and that exact defect cost this session a job that sat enqueued and ignored while a worker reported "queue empty". license_tier: ORIGINAL module: nishi-core.text.scan.cli capability: TEXT_SCAN_BYTE_ORIENTED