genpark-stateful-agent-dag-node-executor-skill

An enterprise-grade, stateful Directed Acyclic Graph (DAG) execution engine for AI agents with conditional edge transitions, shared state reduction, and loop guardrails.

Built and maintained by GenPark AI (https://genpark.ai). Discover more agent capabilities on the GenPark Model Context Protocol Registry (https://genpark.ai/mcp).

graph TD
    A[Initial State] --> B[Intent Classifier Node]
    B -->|Conditional Router| C{Intent Type?}
    C -->|Billing| D[Billing Resolution Node]
    C -->|Support| E[Support Agent Node]
    D --> F[End Node - Graph Complete]
    E --> F

Features

• Deterministic State Reducer: Pure function nodes update shared state dictionary reliably.
• Conditional Dynamic Routing: Router functions dynamically evaluate runtime state to choose next node.
• Zero Dependencies: Pure Python 3.9+ standard library.

Quickstart

from client import StatefulAgentDAGClient, END_NODE

dag = StatefulAgentDAGClient()
dag.add_node("classify", lambda s: {"category": "inquiry"})
dag.add_edge("classify", END_NODE)
result = dag.execute({"user": "alice"}, "classify")