independent-background-agents

You can run durable Hermes background tasks by separating them from the gateway process and starting them as transient systemd-run units. This ensures your long-running jobs persist independently across any gateway interruptions or restarts.

Per-task pinning locks a specific provider and model to an individual Hermes background job. This mechanism ensures reproducible results by maintaining consistent configurations throughout the task lifecycle, even when running multiple jobs simultaneously.

You manage auditable logs by directing the output of each Hermes background task to a stable file path. This centralized logging approach allows you to retrieve unit status and historical logs for restoration and review.

Yes, you can manage multiple long-running Hermes tasks using systemd-run. It starts transient systemd units for each job, enabling a restart-safe lifecycle with per-task model and provider decisions.

The best way to ensure reproducible results is to use a restart-safe lifecycle with per-task model and provider pinning. This approach locks configurations per job and centralizes auditable logs for each independent task.

Your Hermes jobs stop because they are likely running within the gateway process. You must separate them and start transient systemd units via systemd-run to achieve a restart-safe lifecycle that survives interruptions.