foxctl-flow-orchestration

DAG workflow orchestration routes structured envelopes between execution nodes using typed edges with transforms like jq_filter and regex_extract. It reshapes payloads and maintains dependency order so agent outputs correctly feed into downstream steps.

You build a fan-out DAG by defining a workspace-scoped named flow graph with agent, skill, and transform nodes. Configure parallel implementers to run simultaneously before a final reviewer node, using edge triggers to manage dependency execution order.

Yes, this workflow engine supports human-in-the-loop orchestration through durable workflow state. It maintains named flows in a workspace-scoped database with status tracking, logs, stop capabilities, and output injection for interactive pipeline designs.

Yes, pipeline automation supports HTTP and PTY node types alongside agent, skill, and transform nodes. These nodes execute within the directed acyclic graph, allowing external commands and web requests to integrate into the typed envelope dataflow.

DAG workflows require acyclic dependencies, meaning circular execution loops between nodes are not supported. Additionally, you must pre-define a workspace-scoped named flow graph with explicit edge triggers and typed transforms before execution can begin.