Skip to content

Checkpoint Recovery

Save and restore workflow blackboard state for time-travel debugging and failure recovery.

When to use

  • Long-running multi-agent workflows
  • Resume after worker crash or deliberate pause
  • Audit and replay via Studio Replay page

Minimal code

See examples/04_checkpoint.py:

from hiveflow import HiveFlow, HiveFlowConfig, CheckpointManager

hf = HiveFlow(HiveFlowConfig())
await hf.start()

manager = CheckpointManager(hf.blackboard)
checkpoint_id = await manager.create_checkpoint(
    workflow_id="my-workflow",
    metadata={"step": "after_ingest"},
)

# ... later ...
await manager.restore_checkpoint(checkpoint_id)

Studio

  • Checkpoints page lists snapshots for the active workflow.
  • Replay (/replay) correlates intent_id with blackboard audit entries.