LangGraph — Animate | AI Wisdom

← Back to topic
🎞 ANIMATE · Mode 3 of 5 · +100 XP— LangGraph
📦State
🔵Nodes
🔗Edges
⚙️Compile
▶️Invoke
🎯 Next: Test →
Step 1 / 6
Speed
Defining State
Define AgentState TypedDict with message accumulator💡 Analogy
Like a shared Google Doc where each contributor appends their section rather than overwriting.
AgentState is a TypedDict that flows through every node. The add_messages reducer appends new messages rather than replacing the list.
Console Output
> AgentState: {messages: [...], next_action: str}
> messages uses add_messages reducer (append semantics)
●langgraph_agent.py
Defining StateL8–10
1from typing import Annotated, TypedDict
2from langgraph.graph import StateGraph, START, END
3from langgraph.graph.message import add_messages
4from langchain_openai import ChatOpenAI
5from langchain_core.tools import tool
6from langgraph.prebuilt import ToolNode
7 
8llm = ChatOpenAI(model="gpt-4o")
9 
10class AgentState(TypedDict):
11    messages: Annotated[list, add_messages]
12    next_action: str
13 
14@tool
15def search(query: str) -> str:
16    """Search for information."""
17    return f"Results for: {query} — top 3 articles found"
18 
19@tool
20def calculator(expr: str) -> float:
21    """Evaluate a math expression."""
22    return eval(expr)
23 
24tools = [search, calculator]
25llm_with_tools = llm.bind_tools(tools)
26tool_node = ToolNode(tools)
27 
28def agent_node(state: AgentState) -> AgentState:
29    response = llm_with_tools.invoke(state["messages"])
30    return {"messages": [response]}
31 
32def router(state: AgentState) -> str:
33    last = state["messages"][-1]
34    if hasattr(last, "tool_calls") and last.tool_calls:
35        return "tools"
36    return END
37 
38graph = StateGraph(AgentState)
39graph.add_node("agent", agent_node)
40graph.add_node("tools", tool_node)
41graph.add_edge(START, "agent")
42graph.add_conditional_edges("agent", router, {"tools": "tools", END: END})
43graph.add_edge("tools", "agent")
44app = graph.compile()
45 
46result = app.invoke({"messages": [("user", "What is sqrt(144) + 10?")]})
47print(result["messages"][-1].content)
47 lines
Timeline — 6 Steps
Defining State
Define AgentState TypedDict with message accumulator💡 Analogy
Like a shared Google Doc where each contributor appends their section rather than overwriting.
AgentState is a TypedDict that flows through every node. The add_messages reducer appends new messages rather than replacing the list.
Console Output
> AgentState: {messages: [...], next_action: str}
> messages uses add_messages reducer (append semantics)

Was this page helpful?

Discussion

Define AgentState TypedDict with message accumulator

💡 Analogy

Like a shared Google Doc where each contributor appends their section rather than overwriting.

AgentState is a TypedDict that flows through every node. The add_messages reducer appends new messages rather than replacing the list.

Console Output

> AgentState: {messages: [...], next_action: str}
> messages uses add_messages reducer (append semantics)

Define AgentState TypedDict with message accumulator

💡 Analogy

Like a shared Google Doc where each contributor appends their section rather than overwriting.

AgentState is a TypedDict that flows through every node. The add_messages reducer appends new messages rather than replacing the list.

Console Output

> AgentState: {messages: [...], next_action: str}
> messages uses add_messages reducer (append semantics)