DyFlow

Overview

DyFlow dynamically constructs and revises workflows from intermediate feedback during execution instead of fixing one plan upfront. It treats workflow repair as part of runtime behavior rather than a separate offline ceremony.

Why it matters

It matters because many tasks do not reveal the right procedure in advance. Dynamic revision is often more realistic than pretending the first graph will be perfect.

Distinctive trait

Its distinctive trait is in-flight workflow revision: intermediate evidence can trigger restructuring before the run is over.

Relationships

Read DyFlow with self-evolving-workflows, runtime error-recovery patterns, and JudgeFlow. It is a useful contrast to the more static search style of AFlow and MermaidFlow.