Skip to content

Project 5 — LangGraph Research Agent

Build a multi-node LangGraph agent that takes a research question, plans sub-questions, researches each one, synthesizes findings, and runs a critic loop to ensure quality. This is the most architecturally complex project — it demonstrates stateful, conditional, multi-agent reasoning.

What you'll build

A LangGraph graph with: - Planner node — decomposes the research question into 3–5 sub-questions - Researcher node — answers each sub-question (in parallel, with asyncio.gather) - Writer node — synthesizes findings into a structured report - Critic node — evaluates quality and provides feedback - Conditional router — loops back to writer if quality is below threshold (max 3 loops) - A FastAPI endpoint that exposes the agent as a streaming API

Skills covered

Skill Where
LangGraph StateGraph 02-implementation
Conditional routing 02-implementation
Parallel node execution 02-implementation
MemorySaver checkpointing 03-advanced-features
Agent quality evaluation 04-evaluation
LangSmith tracing 03-advanced-features

Prerequisites

  • Week 02 Day 03 Part 1 — AI Agents
  • Week 02 Day 03 Part 2 — LangGraph

Tech stack

langchain==0.3.7
langchain-openai==0.2.3
langgraph==0.2.28
langsmith==0.1.147
fastapi==0.115.0
uvicorn==0.30.6
pydantic==2.9.0
python-dotenv==1.0.1

01-setup