lsm / HyperNeo / 31149186378

40%

push

github

fix(space): recover preset worker agents orphaned from template tracking (#2363)

* fix(space): recover preset worker agents orphaned from template tracking

Preset agents that lost template tracking (template_name/template_hash NULL)
were invisible to drift detection, so their prompts silently went stale (e.g.
the NeoKai→HyperNeo rebrand never propagated to seeded spaces) and there was
no UI reset path.

- getAgentDriftReport now includes preset-named rows with null templateName as
  `orphaned` entries; syncFromTemplate/getTemplateSyncPreview re-attach by name
  lookup so the existing Apply flow fixes them.
- Editor gains an edit-mode "Reset to <Preset> default" button that reloads the
  preset's description/tools/customPrompt and re-stamps tracking on save.
- Migration 170 re-backfills tracking on orphaned rows (M106 is one-shot/marked
  and can't catch rows re-orphaned by later edits), stamping the row hash so
  divergent rows surface as updateAvailable rather than silently in-sync.

Tests: daemon unit (manager drift/sync/preview re-attach + migration-170) and
web (editor reset button).

* fix(space): align getTemplateSyncPreview.customized with orphaned drift semantics

For an orphaned (re-attached) row storedHash is null, so the row-vs-stored
comparison was unconditionally true. Mirror getAgentDriftReport's orphaned
branch: customized means "the row diverges from the current preset"
(rowHash !== liveHash), not the meaningless row-vs-null check. Harmless today
(no agent UI consumes preview.customized) but removes the misleading value.
Adds a preview test mirroring the drift-report orphaned cases.

* fix(space): leave divergent preset rows as orphans in M170 backfill

M170 stamped template_hash to each orphaned preset row's own fingerprint,
which made the drift report's `customized` signal always false for those
rows. The UI then offered a direct "Apply" (generic confirm only) that
could overwrite a user's local edits with no field-level diff revie... (continued)

19319 of 55389 branches covered (34.88%)

Branch coverage included in aggregate %.

27 of 67 new or added lines in 5 files covered. (40.3%)

28977 of 65800 relevant lines covered (44.04%)

276.53 hits per line