nx_asr_research_fetch.nx source
↩ module page · 72 lines · 4882 B
1// nx_asr_research_fetch.nx -- SOVEREIGN research round for THE DEEP: automatic speech recognition (operator
2// 2026-06-28 "...then move on to the deep"). The live-translation long-pole = audio->text. Mirrors
3// nx_video_feature_research_fetch: own TLS-1.3 + Mozilla CA, idempotent have-skip, nx_cc->nxasm. Banks the ASR
4// frontier (acoustic features, HMM/CTC, language models, decoding) -> knowledge/fetched/asr_*.raw so the deep
5// arc is grounded in cited sources before a line of acoustic code is written. expect_exit: 0 license_tier: ORIGINAL
6import "nx_syscalls.nx"
7import "nx_x509_trust_store.nx"
8import "nx_trust_store_load_from_certdata.nx"
9import "nx_https_fetch_follow.nx"
10const K_MAGIC_4194304: i64 = 4194304
11const K_MAGIC_8388608: i64 = 8388608
12
13func df_puts(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } sys_write(1, s, n); return 0 }
14func df_putn(v: i64) -> i64 {
15 if v == 0 { sys_write(1, "0" as *u8, 1); return 0 }
16 var m: i64 = v
17 if m < 0 { sys_write(1, "-" as *u8, 1); m = 0 - m }
18 let d: *u8 = sys_mmap(24); var k: i64 = 0
19 while m > 0 { d[k] = (48 + (m % 10)) as u8; m = m / 10; k = k + 1 }
20 var j: i64 = k - 1
21 while j >= 0 { sys_write(1, ((d as i64)+j) as *u8, 1); j = j - 1 }
22 return 0
23}
24func have_file(path: *u8) -> i64 { let fd: i64 = sys_openat_rd(path); if fd < 0 { return 0 } sys_close(fd); return 1 }
25func fetch_save(url: *u8, opath: *u8, store: *TrustStore, out: *u8, cap: i64) -> i64 {
26 if have_file(opath) == 1 { df_puts(opath); df_puts(" [have-skip]\n" as *u8); return 1 }
27 let status: *i64 = sys_mmap(8) as *i64
28 let n: i64 = nx_https_fetch_follow(url, store, out, cap, 6, status)
29 df_puts(url); df_puts(" status=" as *u8); df_putn(status[0]); df_puts(" bytes=" as *u8); df_putn(n)
30 if n <= 0 { df_puts(" FETCH-FAIL\n" as *u8); return 0 }
31 var gz: i64 = 0
32 if n >= 2 { if out[0] == 0x1f as u8 { if out[1] == 0x8b as u8 { gz = 1 } } }
33 if gz == 1 { df_puts(" [GZIP-skip]\n" as *u8); return 0 }
34 let fd: i64 = sys_openat_wr(opath, 0x1a4)
35 if fd < 0 { df_puts(" SAVE-FAIL\n" as *u8); return 0 }
36 sys_write(fd, out, n); sys_close(fd)
37 df_puts(" SAVED\n" as *u8)
38 return 1
39}
40
41func main() -> i64 {
42 let r: i64 = nx_trust_store_load_from_certdata("data/mozilla_certdata.txt" as *u8, 512, K_MAGIC_4194304)
43 if r <= 0 { df_puts("ASR: certdata load failed\n" as *u8); return 1 }
44 let store: *TrustStore = r as *TrustStore
45 df_puts("CA roots=" as *u8); df_putn(trust_store_count(store)); df_puts("\n" as *u8)
46 let cap: i64 = K_MAGIC_8388608
47 let out: *u8 = sys_mmap(cap)
48 var ok: i64 = 0
49
50 df_puts("== ACOUSTIC FRONT-END (PCM -> features) ==\n" as *u8)
51 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Mel-frequency_cepstrum" as *u8, "knowledge/fetched/asr_mfcc.raw" as *u8, store, out, cap)
52 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Mel_scale" as *u8, "knowledge/fetched/asr_melscale.raw" as *u8, store, out, cap)
53 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Spectrogram" as *u8, "knowledge/fetched/asr_spectrogram.raw" as *u8, store, out, cap)
54 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Window_function" as *u8, "knowledge/fetched/asr_window.raw" as *u8, store, out, cap)
55
56 df_puts("== ACOUSTIC MODEL + ALIGNMENT ==\n" as *u8)
57 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Acoustic_model" as *u8, "knowledge/fetched/asr_acoustic.raw" as *u8, store, out, cap)
58 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Hidden_Markov_model" as *u8, "knowledge/fetched/asr_hmm.raw" as *u8, store, out, cap)
59 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Connectionist_temporal_classification" as *u8, "knowledge/fetched/asr_ctc.raw" as *u8, store, out, cap)
60 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Dynamic_time_warping" as *u8, "knowledge/fetched/asr_dtw.raw" as *u8, store, out, cap)
61 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Phoneme" as *u8, "knowledge/fetched/asr_phoneme.raw" as *u8, store, out, cap)
62
63 df_puts("== LANGUAGE MODEL + DECODING ==\n" as *u8)
64 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Language_model" as *u8, "knowledge/fetched/asr_lm.raw" as *u8, store, out, cap)
65 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Beam_search" as *u8, "knowledge/fetched/asr_beam.raw" as *u8, store, out, cap)
66 ok = ok + fetch_save("https://en.wikipedia.org/wiki/Finite-state_transducer" as *u8, "knowledge/fetched/asr_fst.raw" as *u8, store, out, cap)
67
68 df_puts("asr research: fetched/have=" as *u8); df_putn(ok); df_puts(" / 12\n" as *u8)
69 if ok >= 10 { df_puts("verdict=GREEN (ASR corpus on disk; grep knowledge/fetched/asr_*.raw to ground the deep rung ladder)\n" as *u8); sys_exit(0); return 0 }
70 df_puts("verdict=PARTIAL (re-run to resume -- idempotent have-skip)\n" as *u8)
71 sys_exit(1); return 1
72}