robbond/confidence-engine
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

docs: record closure metadata capture

Browse Source
This commit is contained in:
robbond
2026-08-14 09:30:55 +01:00
parent 50ae28b325
commit fa821a53dd
2 changed files with 90 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/current-handoff.md
+8
View File
@@ -242,6 +242,14 @@ Experiment 54N tested whether an interpretation disagreement can be judged for m
- Focused fallback case also passes: when the proposal-selected target becomes known but another real unresolved unknown remains, the terminal target is discarded and the real unresolved node is selected.
- Focused preservation run passed via `npx vitest run tests/graph/apply-proposal.test.js -t "60B.43|60B.11|replaces downstream pricing"`; live rerun still required to prove the same outcome end-to-end under live conditions.
### 60B.45 closure metadata capture in canonical harness
- Extended `scripts/reproduce-multi-turn-investigation.mjs` and its deterministic harness tests to expose raw accepted-update closure metadata explicitly.
- Harness now emits/stores `finalActiveUnknownNodeId` from the final updated graph and `finalSelectedQuestion` from the final update result.
- Explicit `null` is now preserved for both fields rather than being omitted, closing the apparatus gap that forced 60B.44 to infer closure from absence.
- Focused validation passed via `npx vitest run tests/reproduce-multi-turn-investigation.harness.test.js` (`67/67`).
- The exact 60B.44 live closure case can now be rerun once with direct evidence for active-target clearing and final-question clearing.
### When This Knowledge-Management Phase Is Complete
Provisional criteria for review (all confirmed met by Experiment 38 cold-start test):