code wiki / _hdl_build / nx_vizsla_dedup_gate.nx
nx_vizsla_dedup_gate.nx source
↩ module page · 225 lines · 10287 B
1// nx_vizsla_dedup_gate.nx -- VIZSLA dedup/merge gate. Construction-known oracle + contact->dedup integration.
2// Ingest 5 fields via BLESSED nx_vizsla_contact (date 2026-01-01) -- TWO records of the same person:
3// momA email=mom@x.com name=Mom momB email=mom@x.com phone=5551212 dad email=dad@x.com
4// momA+momB share email mom@x.com => duplicate candidates. dad's email is unique => NOT a candidate.
5// merge survivor=momA loser=momB @2026-02-01 ; unified view of momA = email+name+phone (phone pulled from the
6// merged-in loser) = 3 fields. canonical(momB)=momA. After merge mommaster<-momA, canonical(momB)=mommaster.
7// Rows:
8// 1 ingest-fields contact load new=5 seg-4101 (integration precondition)
9// 2 detect candidate value=mom@x.com count=2 ; dad@x.com NOT a candidate ; candidates=1
10// 3 merge survivor=momA loser=momB new=1 (additive mrg: record, loser NOT destroyed)
11// 4 canonical-loser canonical momB -> momA
12// 5 unified-view view momA -> fields=3 AND key=phone value=5551212 (UNION pulled loser's field)
13// 6 view-via-loser view momB -> SUMMARY contact=momA fields=3 (query loser, get survivor's merged profile)
14// 7 transitive merge mommaster<-momA ; canonical momB -> mommaster (chain momB->momA->mommaster)
15// 8 idempotent re-merge momA<-momB same date -> new=0 (CID dedup)
16// 9 neg-self-merge merge momA<-momA -> fail loud exit 1
17// Evidence: VIZSLA-DEDUP-GATE -> stdout + knowledge/status/vizsla_gate.log; exit 0 iff 9/9.
18// license_tier: ORIGINAL
19import "nx_syscalls.nx"
20
21func og_slen(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } return n }
22func og_p(s: *u8) -> i64 { sys_write(1, s, og_slen(s)); return 0 }
23
24func og_cat(dst: *u8, off: i64, s: *u8) -> i64 {
25 var i: i64 = 0
26 while s[i] != (0 as u8) { dst[off + i] = s[i]; i = i + 1 }
27 return off + i
28}
29
30func og_catn(dst: *u8, off: i64, v: i64) -> i64 {
31 var o: i64 = off
32 var m: i64 = v
33 if m < 0 { dst[o] = 45 as u8; o = o + 1; m = 0 - m }
34 let t: *u8 = sys_mmap(28)
35 var k: i64 = 0
36 if m == 0 { t[0] = 48 as u8; k = 1 }
37 while m > 0 { t[k] = (48 + (m % 10)) as u8; m = m / 10; k = k + 1 }
38 var i: i64 = 0
39 while i < k { dst[o + i] = t[k - 1 - i]; i = i + 1 }
40 return o + k
41}
42
43func og_write(path: *u8, content: *u8) -> i64 {
44 let fd: i64 = sys_openat_wr(path, 0x1a4)
45 if fd < 0 { return 0 - 1 }
46 sys_write(fd, content, og_slen(content))
47 sys_close(fd)
48 return 0
49}
50
51func og_readall(path: *u8, szout: *i64) -> *u8 {
52 let fd: i64 = sys_openat_rd(path)
53 if fd < 0 { szout[0] = 0 - 1; return 0 as *u8 }
54 let sz: i64 = sys_lseek(fd, 0, 2)
55 sys_lseek(fd, 0, 0)
56 let buf: *u8 = sys_mmap(sz + 64)
57 var got: i64 = 0
58 var n: i64 = 1
59 while n > 0 { n = sys_read(fd, (buf as i64 + got) as *u8, 65536); if n > 0 { got = got + n } }
60 sys_close(fd)
61 szout[0] = got
62 return buf
63}
64
65func og_runv(elf: *u8, args: *i64, outpath: *u8) -> i64 {
66 let pid: i64 = sys_fork()
67 if pid == 0 {
68 if (outpath as i64) != 0 {
69 let ofd: i64 = sys_openat_wr(outpath, 0x1a4)
70 if ofd >= 0 { sys_dup3(ofd, 1, 0); sys_dup3(ofd, 2, 0) }
71 }
72 let argv: *i64 = sys_mmap(128) as *i64
73 argv[0] = elf as i64
74 var i: i64 = 0
75 var go: i64 = 1
76 while go == 1 { if args[i] == 0 { go = 0 } else { argv[i + 1] = args[i]; i = i + 1 } }
77 argv[i + 1] = 0
78 let envp: *i64 = sys_mmap(16) as *i64
79 envp[0] = 0
80 sys_execve(elf, argv, envp)
81 sys_exit(127)
82 }
83 let st: *i64 = sys_mmap(16) as *i64
84 sys_wait4(pid, st, 0)
85 let sig: i64 = st[0] & 0x7f
86 if sig != 0 { return 128 + sig }
87 return (st[0] >> 8) & 0xff
88}
89
90func og_idx(buf: *u8, sz: i64, needle: *u8) -> i64 {
91 let n: i64 = og_slen(needle)
92 if n == 0 { return 0 - 1 }
93 var i: i64 = 0
94 while i + n <= sz {
95 var ok: i64 = 1
96 var j: i64 = 0
97 while j < n { if buf[i + j] != needle[j] { ok = 0; j = n } else { j = j + 1 } }
98 if ok == 1 { return i }
99 i = i + 1
100 }
101 return 0 - 1
102}
103
104func og_has(path: *u8, needle: *u8) -> i64 {
105 let szp: *i64 = sys_mmap(16) as *i64
106 let b: *u8 = og_readall(path, szp)
107 if og_idx(b, szp[0], needle) >= 0 { return 1 }
108 return 0
109}
110
111func og_row(name: *u8, pass: i64) -> i64 {
112 og_p("ROW " as *u8); og_p(name)
113 if pass == 1 { og_p(" PASS\n" as *u8) } else { og_p(" FAIL\n" as *u8) }
114 return pass
115}
116
117func og_args(a1: *u8, a2: *u8, a3: *u8, a4: *u8, a5: *u8, a6: *u8) -> *i64 {
118 let a: *i64 = sys_mmap(64) as *i64
119 a[0] = a1 as i64; a[1] = a2 as i64; a[2] = a3 as i64; a[3] = a4 as i64; a[4] = a5 as i64; a[5] = a6 as i64; a[6] = 0
120 return a
121}
122
123func main(argc: i64, argv: *i64) -> i64 {
124 og_p("=== VIZSLA DEDUP GATE: detect/merge/canonical/view + contact->dedup ===\n" as *u8)
125 var ob: *u8 = "/tmp/nx_vizsla_dedup.sov.elf" as *u8
126 if argc > 1 { ob = argv[1] as *u8 }
127 let pr: i64 = sys_openat_rd(ob)
128 if pr >= 0 { sys_close(pr) }
129 else {
130 og_p(" instrument missing -> rebuilding via durable runner\n" as *u8)
131 og_runv("_offc/nx_sov_build_run.elf" as *u8, og_args("nx_vizsla_dedup" as *u8, 0 as *u8, 0 as *u8, 0 as *u8, 0 as *u8, 0 as *u8), "/tmp/vdd_rebuild.out" as *u8)
132 }
133
134 let pfx: *u8 = sys_mmap(128)
135 var po: i64 = 0
136 po = og_cat(pfx, po, "/tmp/vddG" as *u8); po = og_catn(pfx, po, sys_now_us()); po = og_cat(pfx, po, "-" as *u8)
137 pfx[po] = 0 as u8
138
139 og_write("/tmp/vdd_fld.txt" as *u8, "FIELD 2026-01-01 momA email mom@x.com\nFIELD 2026-01-01 momA name Mom\nFIELD 2026-01-01 momB email mom@x.com\nFIELD 2026-01-01 momB phone 5551212\nFIELD 2026-01-01 dad email dad@x.com\n" as *u8)
140
141 var pass: i64 = 0
142 var r: i64 = 0
143
144 // 1 ingest fields via blessed contact organ
145 let rc1: i64 = og_runv("_offc/nx_vizsla_contact.elf" as *u8, og_args("load" as *u8, "/tmp/vdd_fld.txt" as *u8, pfx, "4101" as *u8, 0 as *u8, 0 as *u8), "/tmp/vdd_ing.txt" as *u8)
146 r = 0
147 if rc1 == 0 { if og_has("/tmp/vdd_ing.txt" as *u8, "VIZSLA-CONTACT-LOAD scanned=5 new=5 dup_infile=0 dup_instore=0 segment=seg-4101" as *u8) == 1 { r = 1 } }
148 pass = pass + og_row("ingest-fields" as *u8, r)
149
150 // 2 detect by email
151 og_runv(ob, og_args("detect" as *u8, pfx, "email" as *u8, 0 as *u8, 0 as *u8, 0 as *u8), "/tmp/vdd_det.txt" as *u8)
152 r = 0
153 if og_has("/tmp/vdd_det.txt" as *u8, "VIZSLA-DEDUP-CANDIDATE key=email value=mom@x.com count=2" as *u8) == 1 {
154 if og_has("/tmp/vdd_det.txt" as *u8, "VIZSLA-DEDUP-DETECT key=email candidates=1" as *u8) == 1 {
155 if og_has("/tmp/vdd_det.txt" as *u8, "dad@x.com" as *u8) == 0 { r = 1 }
156 }
157 }
158 pass = pass + og_row("detect" as *u8, r)
159
160 // 3 merge survivor=momA loser=momB
161 let rc3: i64 = og_runv(ob, og_args("merge" as *u8, pfx, "momA" as *u8, "momB" as *u8, "2026-02-01" as *u8, 0 as *u8), "/tmp/vdd_m1.txt" as *u8)
162 r = 0
163 if rc3 == 0 { if og_has("/tmp/vdd_m1.txt" as *u8, "VIZSLA-DEDUP-MERGE survivor=momA loser=momB new=1" as *u8) == 1 { r = 1 } }
164 pass = pass + og_row("merge" as *u8, r)
165
166 // 4 canonical of the loser
167 og_runv(ob, og_args("canonical" as *u8, pfx, "momB" as *u8, 0 as *u8, 0 as *u8, 0 as *u8), "/tmp/vdd_c1.txt" as *u8)
168 r = og_has("/tmp/vdd_c1.txt" as *u8, "VIZSLA-DEDUP-CANONICAL id=momB canonical=momA" as *u8)
169 pass = pass + og_row("canonical-loser" as *u8, r)
170
171 // 5 unified view of survivor: 3 fields incl phone pulled from loser
172 og_runv(ob, og_args("view" as *u8, pfx, "momA" as *u8, 0 as *u8, 0 as *u8, 0 as *u8), "/tmp/vdd_v1.txt" as *u8)
173 r = 0
174 if og_has("/tmp/vdd_v1.txt" as *u8, "VIZSLA-DEDUP-VIEW-SUMMARY contact=momA fields=3" as *u8) == 1 {
175 if og_has("/tmp/vdd_v1.txt" as *u8, "VIZSLA-DEDUP-VIEW contact=momA key=phone value=5551212" as *u8) == 1 { r = 1 }
176 }
177 pass = pass + og_row("unified-view" as *u8, r)
178
179 // 6 view via the loser id -> resolves to survivor
180 og_runv(ob, og_args("view" as *u8, pfx, "momB" as *u8, 0 as *u8, 0 as *u8, 0 as *u8), "/tmp/vdd_v2.txt" as *u8)
181 r = og_has("/tmp/vdd_v2.txt" as *u8, "VIZSLA-DEDUP-VIEW-SUMMARY contact=momA fields=3" as *u8)
182 pass = pass + og_row("view-via-loser" as *u8, r)
183
184 // 7 transitive: merge mommaster<-momA then canonical momB -> mommaster
185 og_runv(ob, og_args("merge" as *u8, pfx, "mommaster" as *u8, "momA" as *u8, "2026-03-01" as *u8, 0 as *u8), "/tmp/vdd_m2.txt" as *u8)
186 og_runv(ob, og_args("canonical" as *u8, pfx, "momB" as *u8, 0 as *u8, 0 as *u8, 0 as *u8), "/tmp/vdd_c2.txt" as *u8)
187 r = og_has("/tmp/vdd_c2.txt" as *u8, "VIZSLA-DEDUP-CANONICAL id=momB canonical=mommaster" as *u8)
188 pass = pass + og_row("transitive" as *u8, r)
189
190 // 8 idempotent re-merge momA<-momB
191 let rc8: i64 = og_runv(ob, og_args("merge" as *u8, pfx, "momA" as *u8, "momB" as *u8, "2026-02-01" as *u8, 0 as *u8), "/tmp/vdd_m3.txt" as *u8)
192 r = 0
193 if rc8 == 0 { if og_has("/tmp/vdd_m3.txt" as *u8, "VIZSLA-DEDUP-MERGE survivor=momA loser=momB new=0" as *u8) == 1 { r = 1 } }
194 pass = pass + og_row("idempotent" as *u8, r)
195
196 // 9 neg self-merge
197 let rc9: i64 = og_runv(ob, og_args("merge" as *u8, pfx, "momA" as *u8, "momA" as *u8, "2026-01-01" as *u8, 0 as *u8), "/tmp/vdd_neg.txt" as *u8)
198 r = 0
199 if rc9 == 1 { r = 1 }
200 pass = pass + og_row("neg-self-merge" as *u8, r)
201
202 let permil: i64 = (pass * 1000) / 9
203 let logfd: i64 = sys_openat_append("knowledge/status/vizsla_gate.log" as *u8, 0x1a4)
204 var fdi: i64 = 0
205 while fdi < 2 {
206 var fd: i64 = 1
207 if fdi == 1 { fd = logfd }
208 if fd > 0 {
209 let line: *u8 = sys_mmap(256)
210 var o: i64 = 0
211 o = og_cat(line, o, "VIZSLA-DEDUP-GATE epoch=" as *u8)
212 o = og_catn(line, o, sys_now_realtime_sec())
213 o = og_cat(line, o, " rows=9 pass=" as *u8)
214 o = og_catn(line, o, pass)
215 o = og_cat(line, o, " permil=" as *u8)
216 o = og_catn(line, o, permil)
217 if pass == 9 { o = og_cat(line, o, " verdict=GREEN\n" as *u8) } else { o = og_cat(line, o, " verdict=RED\n" as *u8) }
218 sys_write(fd, line, o)
219 }
220 fdi = fdi + 1
221 }
222 if logfd > 0 { sys_close(logfd) }
223 if pass == 9 { return 0 }
224 return 1
225}