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) correlatesintent_idwith blackboard audit entries.