Backfill & Scale
Reprocessing history and running pipelines at scale are simple in LeastAction because the date is a dimension of a task (logical_date), not baked into the run. The same operator/SQL/Python runs for any date you assign.
Backfill methods
- Set the date at import —
LeastActionGitToTasktakeslogical_datein the import payload; loop over a date range to create one instance per date. See CI/CD: Git-to-Task. - Bulk reschedule (UI) — select tasks in the table view and trigger a reschedule action to set a new
logical_dateon many tasks at once. - Ask the AI — "backfill daily sales from 2024-01-01 to today."
Catch-up is automatic (see Scheduling); dependencies hold per date.
Partitions — parallel & sharded runs
A partition is part of a task's primary key. Same name + same project + different partition = an independent instance with its own state, run date, and dependency chain. Use it for sharding (one partition per region), multi-tenant pipelines (one per client), or parallel report variants. A child depending on partition: NORTH_AMERICA waits only for that partition.
Run it — the worked example
The leastaction-pipelines-orchestration usecase (in ai/usecases) is the runnable, AI-implementable version of this guide: backfill methods, dependency wiring, partitions, and catch-up. Deploy it, or ask the AI to apply the pattern to your tasks.
Test before you use
Bulk actions (reschedule, skip-subtree, rerun-subtree) have broad effects — test on a few tasks first; keep action code in Git so changes are reversible.