cookbook/03_teams/24_regenerate/README.md
Redo the last response of a team run. regenerate=True drops only the trailing
assistant answer, keeps intermediate tool exchanges (member delegations are not
re-run), and re-generates a fresh final answer.
Regenerate always forks — a new team run_id with fresh metrics; the source
is always retained. replace_original controls only history visibility of the
source:
replace_original=True) — source marked REGENERATED and hidden; new run replaces itreplace_original=False — source stays COMPLETED and visible, so both attempts showMember rows the original team produced stay attached to the original team — fork
replays the team's own model loop with a new run_id; delegated work that
already happened is a fact of history, not state to copy.
| Example | What it shows |
|---|---|
01_regenerate.py | team.continue_run(regenerate=True) — new run_id, fresh metrics, original team and members untouched. |
.venvs/demo/bin/python cookbook/03_teams/24_regenerate/01_regenerate.py
Foundation: ../23_checkpointing/.