minion-orchestrator

You can orchestrate LLM subagent workloads and shell jobs through a single routing interface that handles submission, monitoring, and lifecycle controls. It provides durable persistence using a Postgres-backed queue to ensure reliability and auditability.

Pausing and resuming long-running LLM agent tasks mid-flight requires a durable, restart-tolerant queue. This orchestrator supports mid-flight steering, pausing, resuming, and cancellation across subagent pathways to ensure restart durability.

Durable job queue routing for deterministic shell scripts works by backing the queue with Postgres to track progress and persist state. This ensures that submitted shell jobs remain observable and can survive restarts without losing execution context.

You can monitor and steer multiple subagent fan-out tasks through one unified interface. The system provides observable routing with progress tracking, allowing you to manage lifecycle states across both shell and subagent workloads simultaneously.

You need Postgres to manage durable orchestrations for LLM agents, as the system relies on a Postgres-backed queue. This database dependency provides the necessary progress tracking and restart durability for observable workload routing.

To cancel in-progress deterministic jobs without losing history, the orchestrator tracks lifecycle events and progress within its durable Postgres-backed queue. This ensures cancelled shell jobs and subagent workloads remain fully auditable after termination.