Skip to main content

Assignment: LangGraph Foundations & State Management

Assignment Metadata​

FieldDescription
Assignment NameLangGraph Foundations & State Management
CourseLangGraph and Agentic AI
Project Namelanggraph-chatbot-agent
Estimated Time90 minutes
FrameworkPython 3.10+, LangGraph, LangChain, OpenAI API

Learning Objectives​

By completing this assignment, you will be able to:

  • Understand LangGraph architecture and the role of Messages in State
  • Implement State Management with messages-centric pattern using TypedDict
  • Build Nodes and Edges with LangChain message types
  • Configure MemorySaver checkpointer for conversation persistence
  • Create a complete chatbot workflow with conditional routing

Problem Description​

You are tasked with building a conversational AI agent using LangGraph that can:

  1. Maintain conversation history across multiple turns
  2. Use proper message types (HumanMessage, AIMessage, SystemMessage)
  3. Implement conditional routing based on user intent
  4. Persist state using MemorySaver for multi-session conversations

Technical Requirements​

Environment Setup​

  • Python 3.10 or higher
  • Required packages:
    • langgraph >= 0.2.0
    • langchain >= 0.1.0
    • langchain-openai >= 0.1.0

API Requirements​

  • OpenAI API key configured as environment variable

Tasks​

Task 1: State Definition (20 points)​

  1. Define an AgentState using TypedDict that includes:

    • messages: Using Annotated[List[AnyMessage], add_messages] pattern
    • user_name: Context field for personalization
    • session_id: Context field for tracking

  2. Implement proper message handling that:

    • Uses add_messages reducer for automatic message merging
    • Handles all message types (HumanMessage, AIMessage, SystemMessage)

Task 2: Node Implementation (30 points)​

  1. Create a chatbot node that:

    • Reads messages from state
    • Adds a personalized system message if not present
    • Calls the LLM with message history
    • Returns new AIMessage to state

  2. Create a routing node that:

    • Analyzes the last user message
    • Routes to appropriate handler based on intent (help, general, goodbye)

Task 3: Graph Construction (30 points)​

  1. Build the StateGraph with:

    • Entry point set to chatbot node
    • Conditional edges based on routing logic
    • Proper END state configuration

  2. Configure MemorySaver checkpointer for state persistence

  3. Implement multi-turn conversation that maintains context across invocations

Task 4: Testing & Validation (20 points)​

  1. Create test conversations demonstrating:

    • Multi-turn dialogue with context retention
    • Different routing paths (help, general, goodbye)
    • Session persistence (same thread_id across calls)

  2. Debug output showing message history at each step

Submission Requirements​

Required Deliverables​

  • Source code in Python script or Jupyter notebook
  • README.md with setup instructions
  • Screenshots showing multi-turn conversation output
  • Graph visualization (using get_graph().draw_mermaid_png())

Submission Checklist​

  • State uses messages-centric pattern with add_messages
  • All node functions follow the correct signature (state) -> dict
  • Conditional routing works correctly
  • MemorySaver enables conversation persistence
  • Code runs without errors

Evaluation Criteria​

CriteriaPoints
State definition with messages20
Node implementation30
Graph construction & routing30
Testing & validation15
Code quality and documentation5
Total100

Hints​

tip
  • Use state["messages"][-1] to access the most recent message
  • The add_messages reducer automatically handles message deduplication
  • Use config = {"configurable": {"thread_id": "unique-id"}} for session tracking
  • Reference the knowledge document for node function patterns

References​