nodes.py

from dotenv import load_dotenv
from langgraph.prebuilt import ToolNode

from react import react_agent_runnable, tools
from state import AgentState

load_dotenv()


def run_agent_reasoning_engine(state: AgentState):
    agent_outcome = react_agent_runnable.invoke(state)
    print(f"reasoning_agent_outcome: {agent_outcome}")
    return {"agent_outcome": agent_outcome}


tool_executor = ToolNode(tools)


def execute_tools(state: AgentState):
    """
    Execute the tool chosen by the agent
    Takes the agent's action from state and runs the appropriate tool
    Returns the tool's output as part of the agent's intermediate steps
    """
    agent_action = state["agent_outcome"]
    output = tool_executor.invoke(agent_action)
    return {"intermediate_steps": [(agent_action, str(output))]}