AgentGraphNodeExtensions

Helpers & Utilities · AiDotNet.Agentic.Agents

Bridges the agents layer and the graph runtime: adds an IAgent as a node in a StateGraph. The node maps the graph state to the agent’s input, runs the agent, and folds the result back into the state — so agents (executors, supervisors, swarms) become first-class steps in a typed, checkpointable, resumable graph.

For Beginners

The graph is a flowchart with state flowing through it; these helpers let one box in the flowchart be an agent. You say how to read the agent’s question out of the state and how to write its answer back, and the graph handles the rest (routing, retries, saving progress).

How It Works

This realizes the unified “typed deterministic graph + multi-agent” story: a graph can route between agent nodes with conditional edges, cycles, human-in-the-loop interrupts, and durable checkpointing, while each node is a full agent (which may itself coordinate a team).

Methods

Method Summary
StateGraph<TState> AddAgentNode(StateGraph<TState> graph, string name, IAgent<T> agent, Func<TState, IReadOnlyList<ChatMessage>> messagesSelector, Func<TState, AgentRunResult, TState> applyResult) Adds a graph node that runs an agent against a full message list derived from the state (for richer control than a single user message).
StateGraph<TState> AddAgentNode(StateGraph<TState> graph, string name, IAgent<T> agent, Func<TState, string> inputSelector, Func<TState, AgentRunResult, TState> applyResult) Adds a graph node that runs an agent against a full message list derived from the state (for richer control than a single user message).