Скачать или смотреть Lecture 25 : LangGraph Workflows: State, Nodes, Edges, Conditional Routing & Agentic AI Explained

In Lecture 25 of the Gen AI Series, we dive deep into one of the most important foundations of Agentic AI development — LangGraph Workflows. This class focuses on how state, nodes, edges, conditional routing, and workflow design come together to build powerful, controllable AI systems.

This lecture is a complete, step-by-step walkthrough starting from the basics of state containers, creating nodes as functions, defining edges, and designing multiple workflow patterns used in modern agentic AI.

🔥 What You Will Learn in Lecture 25

✔️ What is State in LangGraph

✔️ State as a shared memory container

✔️ Using TypedDict / Pydantic to create schema

✔️ How nodes are created as functions

✔️ Node input/output mechanism

✔️ How state is updated partially or fully

✔️ What nodes can and cannot modify

✔️ Creating different types of Workflows:

Sequential Workflow

Parallel Workflow

Mixed (Sequential + Parallel) Workflow

Conditional Workflow (with dotted edges)

✔️ Understanding conditional edges and routing logic

✔️ Role of Start and End nodes

✔️ How LangGraph compiles and visualizes workflows

✔️ When and why workflows are used in Agentic AI

✔️ How LangGraph compares to frameworks like CrewAI, AutoGen, N8N, ADK

✔️ Why LangGraph is more powerful for real production-grade agent systems

📌 Highlights from the Practical Demo

Creating a state schema (PersonalInfo, RoutingState)

Writing node functions (nodeA, nodeB, categorize, handle_math, etc.)

Updating state inside nodes

Adding nodes to the graph

Connecting nodes with edges

Implementing conditional routing using:

builder.add_conditional_edges(
"categorize",
route_question,
{
"math": "math_handler",
"science": "science_handler",
"general": "general_handler"
}
)


Visualizing the final workflow

Running examples like:

“What is Newton’s third law?” → routed to Science Handler

📚 Who Should Watch This Lecture?

This lecture is perfect for:

AI/ML engineers

Students learning GenAI workflow design

Developers building Agentic AI, RAG systems, or multi-step reasoning pipelines

Anyone preparing for interviews on LangChain, LangGraph, CrewAI, Autonomous Agents, etc.