code wiki / (root) / nx_research_fetch.nx

nx_research_fetch.nx

buildroot/runtime/nx_research_fetch.nx

21259 B350 linesdepth 21pulls 146 transitivereach 0 importersview sourcekind tooltopic research
docsdependenciesstructsconstsfunctions

about

nx_research_fetch.nx -- the Nishi RESEARCHER's sovereign web-fetch for the SPEED / BEAT-CUDA arc. Reuses the proven sovereign TLS-1.3 stack (nx_https_fetch_follow: url->connect->TLS1.3 handshake validated vs the Mozilla CA store->GET->redirect-follow->dechunk) -- 100% sovereign (own TLS, nx_cc->nxasm, no curl/wget). Fetches a CITED source on efficient LLM inference + cross-vendor (CUDA-alternative) GPU compute, saves the raw body to knowledge/fetched/ for analysis. The only non-Nishi inputs are the CA-root DATA + the fetched page (exactly the researcher's contract). license_tier: ORIGINAL

dependencies 4 imports · 0 importers

nx_syscalls.nx nx_x509_trust_store.nx nx_trust_store_load_from_certdata. nx_https_fetch_follow.nx nx_research_fetch.nx

imports: nx_syscalls.nxnx_x509_trust_store.nxnx_trust_store_load_from_certdata.nxnx_https_fetch_follow.nx

imported by: nobody (leaf or entry point)

call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown

main rf_puts sys_write nx_trust_store_load_from_c sys_mmap nxa_die sys_write ↻ sys_exit nxa_lock_take nxa_lock_addr sys_write ↻ nxa_lock_give nxa_lock_addr ↻ nxa_report_overrun sys_write ↻ nxa_dump_printable sys_write ↻ nxa_dump_sizes sys_write ↻ sys_read_file sys_openat_rd sys_lseek sys_mmap ↻ sys_read sys_munmap sys_close nx_nss_certdata_parse sys_mmap ↻ _pat_class_cert _pat_value_octal _pat_end _find_newline _starts_with _parse_octal_line sys_mmap ↻ _is_space _decode_octal_escape _is_octal rf_putn sys_write ↻

structs

none

consts

11const K_MAGIC_5381: i64 = 5381
12const RF_CERTDATA_CAP: i64 = 4194304 // Mozilla certdata.txt measures ~1MB; this is ITS reserve alone
19const RF_BODY_CAP: i64 = 536870912 // RAISED 2026-09-14 (deliberately, per the refusal's own remedy) from 128 MiB for the BRIGHT leetcode documents parquet, which filled the 128 MiB reserve and was refused as maybe-cut; address space only, pages commit lazily
20const K_MAGIC_2000: i64 = 2000
21const K_MAGIC_2048: i64 = 2048
22const RF_PIPE: i64 = 124
23const RF_NL: i64 = 10
24const RF_PROV_PATH: *u8 = "knowledge/status/fetch_provenance.jrnl"
29const RF_PROV_ROWCAP: i64 = 2048
179const RF_TRUNC_NAMES: *u8 = "knowledge/status/fetch_truncation.names"
180const RF_TRUNC_ARM: *u8 = "knowledge/status/fetch_truncation_arm.conf"

functions

31func rf_puts(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } sys_write(1, s, n); return 0 }
32func rf_putn(v: i64) -> i64
called by 1: main calls 2: sys_writesys_mmap
45func rf_has(hay: *u8, n: i64, needle: *u8) -> i64
called by 2: rf_trunc_armedmain
62func rf_hexnib(v: i64) -> i64 { if v < 10 { return 48 + v } return 87 + v } // 0-9 then a-f
63func rf_build_path(url: *u8, argc: i64, argv: *i64, out: *u8) -> i64
called by 1: main calls 1: rf_hexnib
116func rf_hex32(dst: *u8, o: i64, digest: *u8) -> i64
called by 1: rf_provenance calls 1: rf_hexnib
128func rf_prov_cat(dst: *u8, o: i64, s: *u8) -> i64
134func rf_prov_num(dst: *u8, o: i64, v: i64) -> i64
called by 2: rf_provenancerf_trunc_name calls 1: sys_mmap
145func rf_provenance(url: *u8, mirror: *u8, body: *u8, n: i64) -> i64
192func rf_trunc_armed() -> i64
called by 1: main calls 3: sys_mmapsys_read_filerf_has
207func rf_trunc_name(mirror: *u8, expected: i64, got: i64) -> i64
225func main(argc: i64, argv: *i64) -> i64