Arvin Xu 82cc885394 🐛 fix(llm-generation-tracing): backfill task_brief / task_brief_judge scenario (#15198) ...

🐛 fix(llm-generation-tracing): backfill task_brief/task_brief_judge scenario

Brief generation and judge call sites only set `metadata.trigger`, so the
tracing hook fell back to `scenario='unknown'` for every row. Surfaced via
the unknown-scenario cleanup pass: 433 task-brief + 26 task-brief-judge
rows landed in unknown, alongside 434 task-handoff rows that still used
the dashed trigger string.

- Add `task_brief` and `task_brief_judge` to `TRACING_SCENARIOS`
- Add `_PROMPT_VERSION` + `_SCHEMA_NAME` constants for both brief chains,
  matching the existing `TASK_TOPIC_HANDOFF_*` convention
- Wire explicit `tracing: { promptVersion, scenario, schemaName }` at all
  three task-lifecycle generateObject call sites
- Normalize `metadata.trigger` to underscored ids
  (`task_handoff` / `task_brief` / `task_brief_judge`) to match the
  `RequestTrigger` enum convention

2026-05-25 16:40:37 +08:00

..

src

🐛 fix(llm-generation-tracing): backfill task_brief / task_brief_judge scenario (#15198)

2026-05-25 16:40:37 +08:00

package.json

♻️ refactor(desktop): consolidate global shortcuts (LOBE-7181) (#13880)

2026-04-17 00:32:05 +08:00

vitest.config.mts

✨ feat(task-template): add home recommendation system with skill connect (#14214)

2026-04-28 18:11:00 +08:00