Problem
Test Manager's Execute runs the last generated code snapshot for a case — and that snapshot is not regenerated or invalidated when the test is re-authored. kane-cli testmd run uploads the new recording, TMS shows the updated steps, but Execute keeps running the frozen old code until someone clicks re-export in the UI.
This produced two rounds of convincing-but-false failures for us:
- 2026-07-12 — suite re-pointed from
localhost to a deployed URL ({{base_url}} variable). TMS execution kept navigating to http://localhost:3000 → ERR_CONNECTION_REFUSED, days after every recording was updated. Root cause: stale snapshot.
- 2026-07-22 16:11 — tests gained self-seeding API preconditions (reset + seed + verify). TMS execution ran without them and failed on state. Minutes later, after a manual re-export, the same case passed all 31 steps (16:15 run) and re-executed green again.
In both cases the failure read like a test/app defect, and the auto-RCA reinforced the misread ("Platform Failure — KaneAI read the displayed balance incorrectly"). The actual rule — "re-export after every re-author" — is currently tribal knowledge, enforced by one manual UI click per test case (9 cases in our servicing suite today, 25 suite-wide).
Ask (in preference order)
- Auto-invalidate/regenerate the execution snapshot when a test's
commit_id advances — or at minimum surface a "code snapshot is older than the latest recording" warning on the case and at Execute time.
- A CLI command to regenerate the cloud snapshot for one test, e.g.
kane-cli testmd export <path> --cloud, or extend testmd sync (which already pushes the replay bundle) to also refresh the execution snapshot.
- A batch form: regenerate all stale snapshots in a folder/project.
Any of these turns the CLI re-author → TMS execute pipeline into something CI can drive end-to-end; today it has a mandatory human click in the middle.
Env: kane-cli 0.6.4 · TMS cloud executor (Auteur code export / testmu runtime) · Windows 11 · account toddl
Problem
Test Manager's Execute runs the last generated code snapshot for a case — and that snapshot is not regenerated or invalidated when the test is re-authored.
kane-cli testmd runuploads the new recording, TMS shows the updated steps, but Execute keeps running the frozen old code until someone clicks re-export in the UI.This produced two rounds of convincing-but-false failures for us:
localhostto a deployed URL ({{base_url}}variable). TMS execution kept navigating tohttp://localhost:3000→ERR_CONNECTION_REFUSED, days after every recording was updated. Root cause: stale snapshot.In both cases the failure read like a test/app defect, and the auto-RCA reinforced the misread ("Platform Failure — KaneAI read the displayed balance incorrectly"). The actual rule — "re-export after every re-author" — is currently tribal knowledge, enforced by one manual UI click per test case (9 cases in our servicing suite today, 25 suite-wide).
Ask (in preference order)
commit_idadvances — or at minimum surface a "code snapshot is older than the latest recording" warning on the case and at Execute time.kane-cli testmd export <path> --cloud, or extendtestmd sync(which already pushes the replay bundle) to also refresh the execution snapshot.Any of these turns the CLI re-author → TMS execute pipeline into something CI can drive end-to-end; today it has a mandatory human click in the middle.
Env: kane-cli 0.6.4 · TMS cloud executor (Auteur code export / testmu runtime) · Windows 11 · account toddl