golem-update-running-agents

To update running agents without losing state, you can trigger an agent update that either auto-replays the operation log against the new component version or performs manual snapshot-based state migration using save-snapshot and load-snapshot exports.

You should use snapshot migration instead of automatic operation log replay when the component interface or state changes break compatibility, requiring manual state transfer via save-snapshot and load-snapshot exports.

Yes, you can update all affected running agents in bulk during a deploy, or target them explicitly via component and application commands, allowing them to switch to the newly deployed revision without losing behavioral continuity.

During component version migration, operational controls allow you to wait for completion and optionally prevent wakeups for suspended agents, ensuring they remain suspended until the update process is fully finalized.

Deployed agents stay on the old code path because deploying a new component version does not automatically migrate them; you must explicitly trigger an agent update to switch their execution to the newly deployed revision.

Yes, the Skill supports individual agent updates, bulk application updates, and component-wide updates, providing multiple paths to migrate already-running agents to a new component revision while preserving behavioral continuity.