Week 05 — LangGraph | SoftBricks Academy

What You'll Learn

LangGraph is where serious agent orchestration happens. You'll understand state machines, node and edge composition, conditional branching, and how to build multi-agent networks that coordinate without getting into infinite loops.

Session Schedule

Day	Time	Focus
Saturday	8:00 - 11:00 PM WAT	LangGraph Architecture & State Machines
Sunday	8:00 - 11:00 PM WAT	Multi-Agent Build

Pre-Requisites

Weeks 01-04 completed
CORTEX M1 started
Understanding of agent patterns from Week 02

Topics Covered

LangGraph State Machines

StateGraph, TypedDict state, node functions, compile & invoke. The foundation of building stateful, composable agent workflows.

StateGraph TypedDict Nodes

Supervisor & Specialist Patterns

Supervisor routing, specialist agents, delegation patterns. How to build a boss agent that knows which worker to call.

Supervisor Specialist Delegation

Conditional Edges & Branching

add_conditional_edges, router functions, dynamic routing based on state. Build graphs that make intelligent decisions at every node.

Conditional Edges Router Dynamic Routing

Human-in-the-Loop Checkpoints

interrupt_before, interrupt_after, checkpoint persistence, approval flows. Let humans supervise and approve critical agent decisions.

HITL Checkpoints Interrupts

Parallel Agent Execution

Fan-out/fan-in, parallel node execution, result aggregation. Run multiple agents simultaneously and merge their outputs.

Parallel Fan-out Aggregation

Weekly Build: Supervisor Agent Network

Build a supervisor that routes queries to 3 specialist agents (Knowledge, Research, Action) based on intent classification.

Architecture

User Query + Context
    |
    v
SUPERVISOR NODE (intent classification)
    |
    ├── condition: "knowledge" → KNOWLEDGE AGENT
    ├── condition: "research" → RESEARCH AGENT
    ├── condition: "action"   → ACTION AGENT
    └── condition: "unknown"  → FALLBACK RESPONSE
    |
    v
RESPONSE NODE (format & return)

Key Files

File	Purpose
`agents/supervisor.py`	LangGraph state machine
`agents/knowledge_agent.py`	Knowledge specialist agent
`agents/research_agent.py`	Research specialist agent
`agents/action_agent.py`	Action specialist agent
`graph.py`	Graph definition & compilation

Resources

Required Reading

LangGraph Official Documentation
LangGraph Multi-Agent Tutorial
LangGraph Human-in-the-Loop Guide

Code Repository

Clone the bootcamp repo and switch to the week-05 branch:

git clone https://github.com/softbricks-academy/agentic-ai-bootcamp.git
cd agentic-ai-bootcamp
git checkout week-05

Session Recording

Recording will be available within 24 hours after the live session. Check the WhatsApp group for the link.

Homework

Due before Week 6 live session.

Complete the supervisor agent network — push your code to the bootcamp repo
Add a 4th specialist agent — extend the supervisor to route to a new specialist of your choice
Implement checkpointing — add human-in-the-loop approval before the action agent executes
Test with 10 diverse queries — document routing accuracy and share results in the WhatsApp group

Week 05: LangGraph — Stateful Multi-Agent Systems