Amanda Hanson / Product & Systems
← Back to Barnaby

SANITIZED EVIDENCE REGISTER / OCTOBER 15, 2026

Barnaby Test Evidence.

What was observed, what was reported, and what still needs a real execution.

This is a summary of recorded development evidence, not a new test run. No raw execution traces, private groceries, or identifiers are published. Node-level behavior does not establish end-to-end reliability.

Development observations

CheckEvidence classRecorded result
Pending state insertObserved at node levelAn unmatched quantity update produced a saved add request with item and quantity.
Select and route confirmationObserved at node levelA “yes add it” test returned a pending object and routed to Confirm.
Rebuild add requestObserved at node levelThe prepared request recovered the stored item and quantity.
Shared duplicate checkerObserved at node levelThe confirmed fixture reached the checker as a new item.
Cleanup branchReported component behaviorThe IF and deletion nodes were reported to fire; final row deletion and downstream persistence were not independently verified.
Existing quantity editReported development testA quantity-update path was reported tested; retained execution evidence is still needed.
Cancellation languageKnown defect“No don’t add it” failed the original exact-match regex. Broader wording was proposed, not verified end to end.

Acceptance cases still open

All cases below require documented execution evidence before being marked passed.

CaseScenarioExpected result — not yet verified
GL-01List after router changesActive items only; correct grouping.
GL-02New add with quantityCorrect Notion values; acknowledge after success.
GL-03 / 04Duplicate and reactivationNo extra page; reuse inactive matching page.
GL-05Remove and repeat removeCorrect Purchased state; harmless repeat.
GL-06Existing / missing updatePersist quantity; ask before creating a missing item.
GL-07Live confirmationCorrect create/reactivate/no-op and exact state cleanup.
GL-08Negative variants“No”, “no thanks”, and “no don’t add it”: no grocery write; clear pending state.
GL-09Unrelated and no-pending messagesPreserve chat and /today, /dinner, /groceries routing.
GL-10Authorization and ambiguityUnauthorized input cannot read/write; ambiguous match cannot alter wrong record.
ReliabilityRetries, offline, restartNo duplicate writes or false success.
State lifecycleExpired / multiple pending rowsPredictable selection, isolation, and cleanup.
Scope edgesMulti-item and combined editsValidate quantities/categories; combined-property updates not verified.

How to record the next test

Case ID · date · fictional input · expected Notion change · actual persisted change · Telegram reply · sanitized execution reference · pass/fail · defect reference.

Release boundary: Partial validation only. No production-readiness, security-validation, full-coverage, or quantified savings claim.

Technical Specification ↗