DAG orchestration · retries · run visibility
Once you have enough scripts, nobody knows what depends on what or which step failed. People run them by hand, in order, and start over when something breaks.
Tasks become nodes and dependencies form a DAG. A scheduler walks it in topological order, retries per policy and resumes from the failing node. Every node records state and logs.
Four states (queued / running / ok / failed) with queue-based concurrency limits; one front-end page shows pass state and timing across the whole chain.
Download source ↓ (free with an account)