[major] Forcing fresh authoring (deleting output-*) mints a new testcase_id → duplicate TMS cases; need a re-author-in-place flag

[tlemmonds]

Environment: kane-cli 0.4.10 · Node 24.14.1 · Windows 11

Summary
The replayable action cache and the TMS case identity live in the same directory. output-<stem>/.internal/meta.json holds the testcase_id that --name reads to update a case in place:

{"testcase_id":"01KWTZ0VYX…","session_name":"Orig-Approve-AtCap","test_id":"cbfe0d13-…","project_id":"…","folder_id":"…"}

Because the only reliable way to avoid brittle replay (companion issue) is to delete output-<stem>/, doing so also deletes the testcase_id. The next testmd run --name <same> can't find the existing case and creates a new one. Iterating a test to green therefore litters the folder with duplicates — we accumulated ~25 TMS cases for 11 tests this way.

Steps to reproduce

1. kane-cli testmd run t_test.md --name Foo → creates TMS case A (id stored in output-t/.internal/meta.json).
2. rm -rf output-t/ (needed to dodge brittle replay), then kane-cli testmd run t_test.md --name Foo again.
3. A second TMS case is created instead of updating case A.

Impact
You must choose one of:

• reliable runs (clear cache → fresh authoring → duplicate TMS cases), or
• stable case identity (keep cache → brittle replay).

There is also no CLI to delete a TMS case, so cleanup is manual in the dashboard.

Ask

• Decouple the action cache from the TMS-case mapping — persist testcase_id (and project/folder) separately from the replayable step-actions, so clearing replay state doesn't orphan the case; and/or
• Add a --reauthor / --fresh flag that re-authors from scratch but keeps the same testcase_id (updates in place); and
• Add a testmd delete --remote (or test-cases delete) for dashboard-free cleanup.

Notes
Root cause of the duplicate-case symptom is the coupling; brittle replay is what forces the cache-clear that triggers it. Fixing the replay reliability removes the need for the workaround entirely.

[tlemmonds]

Cross-ref: the rm -rf output-* workaround this issue describes is forced by the replay-unreliability tracked in #85 (and #72, closed). Fixing #85 so replay is reliable would remove the need for the cache-clear that orphans the testcase_id and creates the duplicate cases described here.

[siddhant573]

Hi @tlemmonds , we have a flag --author to re-author everything on the same testcase_id. Do check this out