Skip to main content
Super Turtle orchestration now uses a durable conductor state model. The dashboard reads that state directly so restart/recovery visibility does not depend on chat context.

​
Canonical State Paths

Conductor data is persisted under:
  • .superturtle/state/events.jsonl
  • .superturtle/state/workers/
  • .superturtle/state/wakeups/
  • .superturtle/state/inbox/
handoff.md and legacy summaries are rendered from this canonical state.

​
/api/conductor

The conductor API exposes a machine-readable snapshot:
  • workers — active + recent worker lifecycle records (run_id, lifecycle_state, checkpoint metadata)
  • wakeups — pending delivery items (completion, failure, milestone, timeout)
  • inbox — pending meta-agent background context items
This is the same data source used by dashboard panels and worker detail pages.

​
Why This Matters

  • Worker completion/failure survives bot restarts
  • Wakeups can be replayed if delivery was interrupted
  • Multi-worker supervision is deterministic and auditable
  • Driver/model switches do not lose background worker updates