nx_onsite_frontier_fetch.nx
buildroot/runtime/nx_onsite_frontier_fetch.nx
about
nx_onsite_frontier_fetch.nx -- SOVEREIGN onsite-search S-CLASS FRONTIER research fetch.
Rung S-RESEARCH-2 of the onsite-search arc: the 7-axis exceed census proved we are PARITY on the lexical
core + EXCEEDS on sovereignty, but BEHIND on the neural/semantic rung -- and that rung is what blocks a
real, full S-class. The operator's directive: find out, HARDWARE-RUNG-UP, what we need to build to close it.
Global Rule 4 (no assumptions): every rung + every new census cell must cite a REAL fetch, not memory.
Reuses the team's OWN proven sovereign HTTPS stack (nx_tls13_client_session_run + nx_https_get_complete; no
browser/node/curl), validated against the real Mozilla CA store -- the IDENTICAL path proven live in
nx_onsite_research_fetch (the srch_* corpus, 2026-06-16). Fetches the fact-dense, canonical, parens/comma-free
open pages (en.wikipedia.org, static, NishiBot-allowed, no redirect) for the WHOLE hardware-up neural stack:
SILICON / VECTOR-MATH KERNEL (the lowest rung -- what an embedding similarity actually computes):
/wiki/Dot_product -> srch_dotproduct.raw
/wiki/Cosine_similarity -> srch_cosine.raw
/wiki/Basic_Linear_Algebra_Subprograms -> srch_blas.raw (GEMM = the matmul kernel layer)
VECTOR REPRESENTATION (turning text into vectors -- the model rung):
/wiki/Word_embedding -> srch_embedding.raw
/wiki/Word2vec -> srch_word2vec.raw
/wiki/Large_language_model -> srch_llm.raw (embedding + cross-encoder rerank model)
ANN INDEX / SCALE (searching millions of vectors sub-linearly):
/wiki/Nearest_neighbor_search -> srch_ann.raw (approximate NN frame)
/wiki/Hierarchical_navigable_small_world -> srch_hnsw.raw (the dominant ANN graph index)
/wiki/Locality-sensitive_hashing -> srch_lsh.raw
/wiki/Vector_quantization -> srch_vq.raw (PQ = compress vectors for RAM budget)
/wiki/Vector_database -> srch_vectordb.raw
RETRIEVAL FRAME (hybrid fusion + evaluation):
/wiki/Information_retrieval -> srch_ir.raw
expect_exit: 0
license_tier: ORIGINAL (clone of nx_onsite_research_fetch; only the URL list differs)
dependencies 11 imports · 0 importers
diagram shows first 10 each side; +1 more imports, +0 more importers in the complete lists below.
imports: nx_syscalls.nxnx_csprng.nxnx_x509_trust_store.nxnx_trust_store_load_from_certdata.nxnx_tls13_client_validate_certificate.nxnx_tls13_client_session_run.nxnx_https_url_for_fetch.nxnx_https_url_connect.nxnx_https_get.nxnx_https_get_complete.nxnx_http_response_parse.nx
imported by: nobody (leaf or entry point)
call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown
structs
| none |
consts
| 44 | const K_MAGIC_4194304: i64 = 4194304 |
functions
| 46 | func sf_puts(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } sys_write(1, s, n); return 0 } |
| 47 | func sf_putn(v: i64) -> i64 |
| 55 | func sf_strlen(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } return n } called by 1: one |
| 58 | func fetch_page(store: *TrustStore, full_url: *u8, path: *u8, path_len: i64, out_path: *u8) -> i64 called by 1: one calls 14: sf_putssys_mmapnx_csprng_fillnx_url_newnx_https_url_for_fetchnx_https_url_connect+8 |
| 115 | func one(store: *TrustStore, full: *u8, path: *u8, out: *u8) -> i64 |
| 120 | func main() -> i64 |