code wiki / _hdl_build / nx_research_relation.nx
nx_research_relation.nx
buildroot/runtime/_hdl_build/nx_research_relation.nx
about
nx_research_relation.nx -- Researcher EXTRACT, rung 1 of the GROWN-ORGAN
semantic pipeline (permanent intelligence = sovereign Nishi organs, hardware
rung up; NO trained weights -- those are benchmark/initial only).
The number-only extractor (nx_research_extract) pulls "53" out of "53 percent"
-- but misses QUALITATIVE claims (concepts, tradeoffs, architecture), which is
the real depth gap vs deep-research. This organ extracts (subject, relation,
object) triples by an INSPECTABLE, data-driven cue table -- every decision is
visible (which cue matched, the exact spans), the opposite of a black box.
API: nx_rel_extract(text, n, subj_out, obj_out) -> relation index (>=0) | -1
rel_name(i) -> the relation type string for index i.
Reuses re_find / re_strlen from nx_research_extract (DRY). license_tier: ORIGINAL.
dependencies 2 imports · 5 importers
imports: nx_syscalls.nxnx_research_extract.nx
imported by: nx_rel_measure.nxnx_research_landscape.nxnx_research_relation_test.nxnx_research_relations.nxnx_research_relations_test.nx
structs
| none |
consts
| none |
functions
| 19 | func rel_n_cues() -> i64 { return 10 } called by 1: rel_find_cue |
| 21 | func rel_cue(i: i64) -> *u8 called by 1: rel_find_cue |
| 35 | func rel_name(i: i64) -> *u8 |
| 50 | func rel_find_cue(text: *u8, n: i64, cpos: *i64, clen: *i64) -> i64 |
| 70 | func rel_is_stop(c: u8) -> i64 called by 1: nx_rel_extract |
| 79 | func rel_slice(text: *u8, a0: i64, b0: i64, out: *u8) -> i64 called by 1: nx_rel_extract |
| 94 | func nx_rel_extract(text: *u8, n: i64, subj: *u8, obj: *u8) -> i64 |