@conarium-ai/core 0.2.45 → 0.2.46

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -1,5 +1,48 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.2.46 - three numbers that named something they were not
4
+
5
+ Every item here is a figure this project printed, published, or wrote down,
6
+ which turned out to measure something other than what it said. Two were
7
+ declared on the SCITT list before they were fixed; this is that debt.
8
+
9
+ **`exit_contract.mjs` reported its own work wrong.** The success line said
10
+ `27 invocation(s)`. It runs 25 processes; 27 was the pass counter, 23 cases
11
+ plus four assertions over two of them. A count that names something it is not,
12
+ printed by the file added to stop codes naming events they are not, in the
13
+ release that was about exactly that. Both numbers are now derived from the run
14
+ through a single `invoke()` wrapper and both are printed:
15
+ `25 invocation(s), 27 assertion(s)`. Add a case and the label moves on its own.
16
+
17
+ **`pack_locale.mjs` carried a figure its own subject forbids.** Its comment
18
+ said the letter pass finds 1041 where the changelog said 1034 — two
19
+ hand-written declarations of one number, inside the check whose whole purpose
20
+ is catching hand-written declarations drifting apart. Neither was reproducible:
21
+ on the 0.2.42 packed set the letter pass is 971 across 26, the word pass 310
22
+ across 20, and their union 1034 across 29. So 1034 was the union and never the
23
+ letter pass. The comment now describes the mechanism and states no figure at
24
+ all; the live number lives in `docs/claims/locale-residue.json`, where a run
25
+ compares it.
26
+
27
+ **And the number underneath both of them was a measurement of one disk.** See
28
+ the 0.2.43 entry, corrected in place: `324 / 22` came from a working tree
29
+ holding two gitignored files under `docs/audit/` that are in no commit. It
30
+ reached a changelog, and from there a public mailing list, where nobody can
31
+ reproduce it.
32
+
33
+ **`test/pack_tracked.mjs`** is the mechanism for that class, and it is not
34
+ about Turkish. Every path npm would ship must be tracked by git or fall under
35
+ one declared generated prefix, `dist/`, which `npm run build` reproduces from
36
+ source. Anything else is a file that exists on the packer's machine and in no
37
+ release, which means the package built here is not the package CI builds, and
38
+ every measurement taken over it is a measurement of somebody's disk.
39
+
40
+ The same mechanism once shipped a private key: `npm publish` from a machine
41
+ that had run the demo included `examples/_keys/audit-ed25519.pem`, which is why
42
+ package.json carries the sentence about .gitignore not binding npm and why
43
+ `pack_artefakt.mjs` locks that one path. This locks the class instead of the
44
+ path. It fails closed: if git cannot answer, the check does not pass.
45
+
3
46
  ## 0.2.45 - the rest of an outside report, and a check over the descriptions
4
47
 
5
48
  **Behaviour change in `conarium-verify` and `conarium-coverage`.** Read the two
@@ -118,11 +161,26 @@ plainly rather than implied by an untested green.
118
161
 
119
162
  `test/pack_locale.mjs` has always ended with a line naming what it does *not*
120
163
  fix: Turkish left under `docs/`, out of interface scope, printed so it stays
121
- visible. The line read `324 occurrences across 22 files`. The letter pass finds
122
- **1034 lines across 29 files**. The remainder was counted by `wordProbe` alone —
123
- the pass written for Turkish that carries *no* Turkish letter, and therefore the
124
- wrong instrument for Turkish prose. Both passes now run and the union is
125
- reported.
164
+ visible. The 0.2.43 note wrote that line down as `324 occurrences across 22
165
+ files`, and called **1034 lines across 29 files** the letter pass. Remeasured
166
+ on 22 August 2026 against both bases that could have produced those figures:
167
+ the published 0.2.42 tarball (508 extracted paths) and `npm pack --dry-run` at
168
+ `v0.2.42` (256 paths). The lists differ only by `dist/` build artefacts, which
169
+ `docsRemainder` already excludes (`bin/`, `scripts/` and `public/` too). On
170
+ that packed-doc set the 0.2.42 instrument (`wordProbe` only) is **310 lines
171
+ across 20 files**. `324 / 22` reproduces on neither base, and the reason
172
+ was found rather than left open: two files under `docs/audit/` that are
173
+ gitignored, present on the machine the figure was taken on, and in no commit.
174
+ Copy them into a clean v0.2.42 worktree and the same guard prints 324 across
175
+ 22; remove them and it prints 310 across 20. The number was never a fact about
176
+ the package. `.gitignore` does not bind npm, so a tree with extra files packs
177
+ a different package, and 0.2.46 adds `test/pack_tracked.mjs` to refuse that.
178
+ The `308 / 19` handover note below is the same class with the cause still
179
+ unknown. The **union** of letter and word on that set is **1034 / 29**; the
180
+ letter pass alone is 971 / 26. 1034 is the union, not the letter pass. The
181
+ remainder was counted by `wordProbe` alone — the pass written for Turkish that
182
+ carries *no* Turkish letter, and therefore the wrong instrument for Turkish
183
+ prose. Both passes now run and the union is reported.
126
184
 
127
185
  What the package was shipping under that number: `docs/superpowers/`,
128
186
  `docs/plans/`, `docs/specs/` and `docs/audit/` — dated internal design and
@@ -0,0 +1,67 @@
1
+ # Claims review — 0.2.46
2
+
3
+ - base: v0.2.45
4
+ - head: 696c694
5
+ - surface: sha256:e9d01e4e6f99670242922cf7c7d960d85adb9ca786d238d8a5f4279bc1c3eb43
6
+ - verdict: pass
7
+ - reviewer: maintainer, after a figure this file's process approved reached a
8
+ public archive and turned out to describe one machine
9
+
10
+ The surface hash is unchanged from 0.2.45. No file in `docs/claims/surfaces.json`
11
+ was touched: this release changes two guards, adds a third, and corrects a
12
+ changelog. That is exactly why the substantive finding below is not a finding
13
+ against the diff, and recording it here anyway is the point.
14
+
15
+ ## Findings
16
+
17
+ none, against the claim surfaces.
18
+
19
+ ## The reason this record is not empty
20
+
21
+ **A number that no claim-surface check could see was published to a mailing
22
+ list, and the process that let it through is this one.**
23
+
24
+ `324 Turkish lines across 22 packed doc files` was written into the 0.2.43
25
+ changelog and, on 22 August, into a message on the SCITT list. It is not
26
+ reproducible. Run v0.2.42's own `pack_locale.mjs` in a clean worktree and it
27
+ prints **310 across 20**. Copy in two gitignored files under `docs/audit/` that
28
+ sit on one machine and in no commit, and it prints **324 across 22**. The
29
+ figure measured a working directory.
30
+
31
+ Every gate held and none of them was pointed here:
32
+
33
+ - `claim_discipline` scans surfaces for retracted phrasings. A number is not a
34
+ phrasing.
35
+ - `denetci` compares the surface hash. `CHANGELOG.md` is not a surface, on
36
+ purpose, because a changelog is a record rather than a claim about the
37
+ present.
38
+ - `pack_locale` printed the figure. It had no reason to doubt its own output,
39
+ and it was correct about the tree it was given.
40
+ - `version_claim` compares what npm shipped against the tag. It compares
41
+ contents, not the provenance of the files that produced them.
42
+
43
+ So the defect was not that a check failed. There was no check whose subject was
44
+ *"is the tree this measurement was taken over the tree this repository
45
+ describes"*. `test/pack_tracked.mjs` is that check, and it is the whole
46
+ deliverable of this release. The two label repairs are consequences.
47
+
48
+ **What it does not close.** It establishes that the packed set matches the
49
+ repository plus `dist/`. It says nothing about figures measured over anything
50
+ else — a corpus, a database, a run on another machine — and nothing about
51
+ whether a number in prose was copied correctly from a run that did happen. A
52
+ figure written by hand into a document is still a hand-written figure; the only
53
+ one now mechanically pinned is the locale remainder, in
54
+ `docs/claims/locale-residue.json`.
55
+
56
+ ## Not a finding, recorded so the next reader does not re-derive it
57
+
58
+ The first attempt to show `pack_tracked` red used the two `docs/audit/` files
59
+ that caused the original defect. It stayed **green**, and correctly: 0.2.43
60
+ excluded `docs/audit` from the package, so those files are no longer packed
61
+ from any tree. The red was produced instead with a stray file under `docs/`,
62
+ which is where the class still bites.
63
+
64
+ Worth stating because the near-miss is instructive: a red test built from the
65
+ historical culprit can pass for a reason that has nothing to do with the check
66
+ being wrong, and a reviewer reading only the green would have concluded the
67
+ guard worked. It did work. The demonstration did not.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@conarium-ai/core",
3
- "version": "0.2.45",
3
+ "version": "0.2.46",
4
4
  "mcpName": "io.github.dogrucanemek-alt/conarium",
5
5
  "//files-artefakt": "Generated files under examples/ (private key, receipt/audit JSONL) are EXCLUDED. .gitignore does not bind npm: because the allow-list contained 'examples', `npm publish` on a machine that had run the demo was shipping _keys/audit-ed25519.pem (PRIVATE KEY) and the receipts. test/pack_artefakt.mjs locks this. The exclusions are LIMITED to examples/: receipts.jsonl and .pub.pem under test-vectors/ MUST ship, or a third party cannot run the conformance vectors.",
6
6
  "//files": "Allow-list of what enters the package. Without it npm sent everything: 296 files / 1.5 MB — CI workflows, gitleaks config, a 137 KB marketing image, the site's api/chat.js. A bank security team opens the tarball and sees those. scripts/ is here ON PURPOSE: conarium-reconcile references scripts/pg-snapshot.sql at runtime; without it the user files a support mail.",