nx_wiki_answer_index.nx

nx_wiki_answer_index.nx -- INDEX DRIVER for the instant-answer search over the 19-page live wiki corpus (real-tree deepened) (IMS Thrust B rung 1). Builds the "id<TAB>text" search-source TSV the reusable onsite ingestor (nx_onsite_index) consumes: one row per live wiki page (13 prior content pages + 6 batch-#2 pages; the search page itself is excluded by design), id = slug, text = the page's VISIBLE text with HTML scrubbed. REUSE not reinvent -- the HTML->text scrub is nx_html_to_text (strips tags, suppresses <script>/<style>, collapses whitespace, decodes entities) VERBATIM; the slug->web_assets file map mirrors nx_ims_corpus_refresh EXACTLY (start lives in ai_start.html, the /wiki/ econsim is econsim_wiki.html), so we index the SAME bytes vpub publishes (== what is served at https://nishifamily.com/wiki/<slug>.html). After this writes knowledge/index/wiki_answer_src.tsv, the build step runs the EXISTING ingestor: nx_onsite_index knowledge/index/wiki_answer_src.tsv https://nishifamily.com/wiki/ \ knowledge/index/wiki.idx knowledge/index/wiki.manifest which produces the durable inverted index + manifest the answer organ + onsite client read, and the committed knowledge/registry/onsite_sites.tsv row (site=wiki) points at exactly that pair. Hardcoded fixed 19-page job (no argv) so `nx_sov_build_run.elf nx_wiki_answer_index` builds AND runs it in one shot. ADDITIVE-ONLY: writes the search-source TSV; touches no live page. license_tier: ORIGINAL