kiokuko 1.0.44 → 1.0.46

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.
Files changed (58) hide show
  1. package/dist/embedding/model-download.d.ts.map +1 -1
  2. package/dist/embedding/model-download.js +52 -9
  3. package/dist/embedding/model-download.js.map +1 -1
  4. package/dist/embedding/model-installation.d.ts.map +1 -1
  5. package/dist/embedding/model-installation.js +11 -0
  6. package/dist/embedding/model-installation.js.map +1 -1
  7. package/dist/mcp/chatgpt-contract.d.ts +251 -2
  8. package/dist/mcp/chatgpt-contract.d.ts.map +1 -1
  9. package/dist/mcp/chatgpt-contract.js +21 -1
  10. package/dist/mcp/chatgpt-contract.js.map +1 -1
  11. package/dist/mcp/chatgpt-server.d.ts +4 -1
  12. package/dist/mcp/chatgpt-server.d.ts.map +1 -1
  13. package/dist/mcp/chatgpt-server.js +33 -3
  14. package/dist/mcp/chatgpt-server.js.map +1 -1
  15. package/dist/mcp/correction-review.d.ts +85 -0
  16. package/dist/mcp/correction-review.d.ts.map +1 -0
  17. package/dist/mcp/correction-review.js +81 -0
  18. package/dist/mcp/correction-review.js.map +1 -0
  19. package/dist/memory/correction-classification.d.ts +47 -0
  20. package/dist/memory/correction-classification.d.ts.map +1 -0
  21. package/dist/memory/correction-classification.js +143 -0
  22. package/dist/memory/correction-classification.js.map +1 -0
  23. package/dist/memory/correction-contract.d.ts +136 -0
  24. package/dist/memory/correction-contract.d.ts.map +1 -0
  25. package/dist/memory/correction-contract.js +42 -0
  26. package/dist/memory/correction-contract.js.map +1 -0
  27. package/dist/memory/git-provenance.d.ts.map +1 -1
  28. package/dist/memory/git-provenance.js +28 -0
  29. package/dist/memory/git-provenance.js.map +1 -1
  30. package/dist/reranker/runtime.d.ts.map +1 -1
  31. package/dist/reranker/runtime.js +121 -124
  32. package/dist/reranker/runtime.js.map +1 -1
  33. package/docs/chatgpt.md +12 -0
  34. package/docs/cli-contract.md +10 -2
  35. package/docs/client-compatibility.md +1 -1
  36. package/docs/correction-classification.md +122 -0
  37. package/docs/evaluations/correction-classification-2026-10-21/assessment.json +717 -0
  38. package/docs/evaluations/correction-classification-2026-10-21/built-both-product-pair.log +14 -0
  39. package/docs/evaluations/correction-classification-2026-10-21/development.json +877 -0
  40. package/docs/evaluations/correction-classification-2026-10-21/dsh-build.log +9 -0
  41. package/docs/evaluations/correction-classification-2026-10-21/dsh-final-2-typecheck.log +0 -0
  42. package/docs/evaluations/correction-classification-2026-10-21/dsh-final-2-unit.log +43 -0
  43. package/docs/evaluations/correction-classification-2026-10-21/dsh-final-3-typecheck.log +0 -0
  44. package/docs/evaluations/correction-classification-2026-10-21/held-out.json +2671 -0
  45. package/docs/evaluations/correction-classification-2026-10-21/live-mcp-final-2.json +32 -0
  46. package/docs/evaluations/correction-classification-2026-10-21/live-mcp-final-2.log +1 -0
  47. package/docs/evaluations/correction-classification-2026-10-21/live-mcp-final.json +32 -0
  48. package/docs/evaluations/correction-classification-2026-10-21/main-build.log +17 -0
  49. package/docs/evaluations/correction-classification-2026-10-21/main-final-2-typecheck.log +0 -0
  50. package/docs/evaluations/correction-classification-2026-10-21/main-final-2-unit.log +64 -0
  51. package/docs/evaluations/correction-classification-2026-10-21/metrics-final.log +11 -0
  52. package/docs/evaluations/correction-classification-2026-10-21/native-final.log +57 -0
  53. package/docs/evaluations/correction-classification-2026-10-21/package-evidence.json +51 -0
  54. package/docs/evaluations/correction-classification-2026-10-21/timing-final.json +53 -0
  55. package/docs/evaluations/correction-classification-2026-10-21/verification.json +199 -0
  56. package/migrations/015_correction_classification.sql +11 -0
  57. package/package.json +1 -1
  58. package/templates/chatgpt/memory-policy.json +1 -1
@@ -0,0 +1,9 @@
1
+
2
+ > kiokuko-dsh@0.2.17 build
3
+ > npm run clean:dist && tsc -p tsconfig.build.json && node scripts/build-model-transports.mjs && node scripts/compile-skill-prompts.mjs && node scripts/build-dsh-client.mjs && node scripts/build-history-codecs.mjs && node scripts/build-plugin-entry.mjs && node scripts/guardrail-artifacts.mjs --verify-bundled --quiet && node scripts/verify-micro-judge.mjs
4
+
5
+
6
+ > kiokuko-dsh@0.2.17 clean:dist
7
+ > node --input-type=module --eval "import { rmSync } from 'node:fs'; rmSync('dist', { recursive: true, force: true });"
8
+
9
+ Skill prompts: 44 validated resources compiled
@@ -0,0 +1,43 @@
1
+ ✔ ordinary execution intake retrieves a stored correction without saving memory or asking another intake question (3595.100208ms)
2
+ ✔ first use persists only host-owned metadata; body survives a lossy summary and unchanged reuse survives restart (477.942209ms)
3
+ ✔ legacy/caller metadata cannot activate a rule; shadow, unknown and reference never trigger the compatibility gate (246.901458ms)
4
+ ✔ title, body, summary, revision and provider configuration invalidate cached meaning without rewriting old entries (307.440125ms)
5
+ ✔ concurrent updates cannot receive a classification from an earlier revision, and concurrent identical lookups share inference (183.599834ms)
6
+ ✔ outage remains explicit, is cached and cannot become reference (149.436292ms)
7
+ ✔ malformed remains explicit, is cached and cannot become reference (141.796125ms)
8
+ ✔ abstain remains explicit, is cached and cannot become reference (153.135833ms)
9
+ ✔ timeout remains explicit, is cached and cannot become reference (144.18075ms)
10
+ ✔ parent cancellation never persists an asynchronous result (137.726875ms)
11
+ ✔ cached meaning cannot survive entry deletion or a foreign workspace identity (133.785584ms)
12
+ ✔ corrupt persisted classification fails closed without repeating inference or repairing memory content (139.587ms)
13
+ ✔ the semantic request carries the full body, explicit abstention and no execution effects (0.207959ms)
14
+ ✔ the additive cache migration upgrades a real prior schema without changing legacy memory (140.538958ms)
15
+ ✔ retrieval failure, no match, missing delivery, source tampering and version-only decisions remain distinct (3431.499625ms)
16
+ ✔ literal write detector leaves CI pins, lower bounds, test fixtures, prose and reads alone (0.991666ms)
17
+ ✔ public decision check rejects version-only incomplete claims only after actual source-bound verification (301.434334ms)
18
+ ✔ native confirmation binds one operation and optional observer failures do not hold other work (196.079666ms)
19
+ ✔ evaluation denominators include positive abstentions/outages as misses and all gold non-special cases (2.214375ms)
20
+ ✔ empty denominators are unavailable and cannot pass release gates (0.239958ms)
21
+ ✔ held-out corpus is disjoint, balanced and aligned across the five declared languages (2.186125ms)
22
+ ✔ legacy Choice normalizes without changing its serialized shape (4.989042ms)
23
+ ✔ Noul and Score retain measured values and reject type, rubric and probability corruption (123.092292ms)
24
+ ✔ only memory reuse Noul accepts 300 questions and answers (127.301875ms)
25
+ ✔ correction meaning uses the same Choice contract through typesafe (96.055291ms)
26
+ ✔ correction meaning uses the same Choice contract through nimble (10.997166ms)
27
+ ✔ correction meaning uses the same Choice contract through laya-strict (38.531917ms)
28
+ ✔ correction meaning uses the same Choice contract through laya-v1 (4.819584ms)
29
+ ✔ strict truncation admission fails before correction inference; decision-off performs no inference (5.200792ms)
30
+ ✔ Laya config remains optional, old normalized JSON/digest stable, socket resolves once before binding (52.686167ms)
31
+ ✔ Laya serializes complete canonical state and ordered numeric IDs, verifies worker before evidence, uses strict prediction (29.4645ms)
32
+ ✔ Laya acceptance uses probabilities, margin and conservative rounding; entropy/action are irrelevant (17.674583ms)
33
+ ✔ Laya rejects corrupt identity, question/choice sets, probabilities and usage (63.061208ms)
34
+ ✔ compaction forwards per-part preflight through wrapper before any batch inference, and overflow preserves readiness (26.107625ms)
35
+ ✔ declared module assets contain every authoritative migration exactly once (1.394417ms)
36
+ ℹ tests 35
37
+ ℹ suites 0
38
+ ℹ pass 35
39
+ ℹ fail 0
40
+ ℹ cancelled 0
41
+ ℹ skipped 0
42
+ ℹ todo 0
43
+ ℹ duration_ms 5324.294333