Simple Bot

Build a basic stateless agent that communicates with ChatOpenAI in LangGraph

In this section, we will see how to build a basic stateless AI agent using ChatOpenAI and LangGraph.

Objectives

1. Define the State schema for storing message history.
2. Initialize ChatOpenAI and invoke it from a graph node.
3. Build a stateless agent graph with a single processor node.
4. Compile and run the agent.

Agent I: Simple Agent Bot

Goal

Build a simple stateless agent using LangGraph that takes user messages and passes them directly to an LLM for processing.

Sample Input

{"messages": [HumanMessage(content="Tell me a quick 1-line joke.")]}

Sample Output

"Why don't skeletons fight each other? They don't have the guts."

Plan

1. Define the state schema (AgentState) containing messages (a list of messages).
2. Define the node function process that invokes the LLM with the messages.
3. Build the graph by adding the node and connecting START to it and it to END.
4. Compile and invoke the agent.

Code Implementation

1. Define the State Schema

We define a schema AgentState with a list of messages:

from typing import TypedDict
from langchain_core.messages import HumanMessage

class AgentState(TypedDict):
    messages: list[HumanMessage]

2. Define the Node Function

The node function receives the state, invokes the model, and prints the response:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o")

def process(state: AgentState) -> AgentState:
    response = llm.invoke(state["messages"])
    print(f"\nAI: {response.content}")
    return state

3. Build and Compile the Graph

We setup the StateGraph, register the node, and link entry and finish points:

from langgraph.graph import StateGraph, START, END

graph = StateGraph(AgentState)
graph.add_node("process", process)

graph.add_edge(START, "process")
graph.add_edge("process", END) 

agent = graph.compile()

4. Invoke the Agent

We invoke the compiled agent:

user_msg = [HumanMessage(content="Tell me a quick 1-line joke.")]
agent.invoke({"messages": user_msg})

Exercise: Pirate Speak Translator Agent 🏴‍☠️

Goal

Build a custom system prompt agent that translates whatever user inputs into dramatic Pirate Speak.

Sample Input

{"messages": [HumanMessage(content="Hello, how are you doing today?")]}

Sample Output

"Ahoy, matey! How be ye farin' this fine day?"

Plan

1. Reuse or create an AgentState containing the message list.
2. Write a node function pirate_node that inserts a SystemMessage specifying the pirate persona ahead of the user messages, invokes the LLM, and prints the pirate response.
3. Initialize the StateGraph, register the pirate node, hook it to START/END, and compile.
4. Test the compiled agent with a human message and verify the pirate speak output.

Solution

from typing import TypedDict
from langchain_core.messages import HumanMessage, SystemMessage
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, START, END

# 1. State Schema
class AgentState(TypedDict):
    messages: list[HumanMessage]

llm = ChatOpenAI(model="gpt-4o")

# 2. Node Function
def pirate_node(state: AgentState) -> AgentState:
    sys_msg = SystemMessage(content="You are a pirate. Translate the user's input into dramatic pirate speak.")
    response = llm.invoke([sys_msg] + state["messages"])
    print(f"\nPirate AI: {response.content}")
    return state

# 3. Create Graph
graph = StateGraph(AgentState)
graph.add_node("pirate_agent", pirate_node)
graph.add_edge(START, "pirate_agent")
graph.add_edge("pirate_agent", END)
app = graph.compile()

# 4. Invoke
app.invoke({"messages": [HumanMessage(content="Hello, how are you doing today?")]})

Practice & Exercises

To reinforce what you’ve learned in this section, practice with the interactive notebook:

Practice & Exercises

Practice defining simple agent structures, customizing system instructions, and compiling basic workflows.

💻 VS Code | 🚀 Colab | 📥 Download