minion-orchestrator

To coordinate durable background tasks, you can use a Postgres-backed job queue that enables reliable routing, progress tracking, and lifecycle control for both deterministic shell scripts and LLM subagent workflows.

Yes, you can pause, resume, cancel, and replay background workflows. The orchestrator provides lifecycle controls with structured progress tracking and transcripts for full observability over shell jobs and subagent tasks.

Postgres-backed job routing works by distinguishing between deterministic shell jobs and LLM subagent workflows, applying precise routing logic to ensure reliable task processing and scalable fan-out parallel execution across agents.

The best way to track progress for parallel agent workflows is to use an orchestrator that provides end-to-end observability with structured progress updates and transcripts, ensuring reliable monitoring during fan-out execution.

Yes, you need Postgres to run this durable background task orchestration. The system relies on a Postgres-backed queue to guarantee reliable task processing, lifecycle management, and progress tracking for shell jobs and subagents.

You should not use a durable orchestrator for background work if your tasks are stateless, require no lifecycle controls like pause or replay, or do not need the reliable routing and observability provided by a Postgres-backed queue.