minion-orchestrator

You can orchestrate background jobs for shell commands and LLM subagents by submitting them to a unified queue interface. This allows you to manage deterministic scripts and LLM reasoning tasks together, tracking their progress and lifecycle in one place.

To persist background jobs and survive gateway restarts, you need to use Postgres. The orchestrator uses Postgres to store job state, ensuring that both shell and LLM subagent work resumes reliably across sessions and unexpected interruptions.

Yes, you can monitor and steer an LLM subagent during execution. The orchestrator provides job monitoring capabilities that allow you to track progress, send messages to steer agents dynamically, and manage the overall job lifecycle.

The best way to manage durable background work for data processing and LLM reasoning is using a unified queue interface. It combines deterministic shell job execution with LLM subagent orchestration, providing observable and steerable job management.

You submit a shell job to the background queue by executing a CLI command with specific parameters. Provide the command string and working directory in a JSON payload, such as passing the cmd and cwd arguments to the job submission script.