LangGraph Agent Patterns

Executable repository demonstrating the fundamental patterns for building agents with LangGraph — from the prebuilt helper to manual construction with state, memory, and routing.

Architecture

Level 1 — Foundations
├── N1A  create_react_agent       Prebuilt agent (2 nodes: agent + tools)
├── N1B  StateGraph manual        Same agent built from scratch
├── N1C  MemorySaver              In-session memory via thread_id
└── N1D  conditional_edges        Deterministic router → specialized nodes

Level 2 — Patterns
├── N2A  State & Routing          MessagesState · Command API · LLM routing · Prompt chaining
├── N2B  HITL · Streaming         interrupt_before · stream modes · SQLite checkpointer
└── N2C  Architectural Patterns   Orchestrator-Worker · Evaluator-Optimizer

Level 3 — Multi-Agent
├── N3A  Subgraph & Supervisor    Subgraph as node · LLM-coordinated specialists
├── N3B  Swarm & Send API         Decentralized handoffs · Dynamic map-reduce fan-out
└── N3C  Long-term Memory         Store · cross-thread / cross-user persistence

Prerequisites

Python 3.10+ and an OpenAI API key.

1. Create a virtual environment

python3 -m venv .venv
source .venv/bin/activate

2. Install dependencies

pip install -r requirements.txt

3. Configure your API key

cp .env.example .env
# Edit .env and add your OPENAI_API_KEY

4. Register the Jupyter kernel

python -m ipykernel install --user --name langgraph-patterns --display-name "LangGraph Patterns"

Notebooks

Run them in order — each level builds on the previous.

Level 1 — Foundations

#	Notebook	What it demonstrates
N1A	N1A_create_agent	create_react_agent prebuilt + 1 custom tool
N1B	N1B_state_graph	Same graph built manually with StateGraph + TypedDict + Reducers
N1C	N1C_memory_saver	MemorySaver — multi-turn memory via thread_id
N1D	N1D_conditional_edges	Deterministic intent router → 3 specialized nodes

Level 2 — Patterns

#	Notebook	What it demonstrates
N2A	N2A_state_routing	MessagesState · Command API · LLM routing with structured output · Prompt chaining with validation
N2B	N2B_hitl_streaming_memory	interrupt_before (HITL) · Streaming (updates / messages) · SqliteSaver persistent memory
N2C	N2C_patterns_advanced	Orchestrator-Worker (parallel decomposition) · Evaluator-Optimizer (iterative quality loop)

Level 3 — Multi-Agent

#	Notebook	What it demonstrates
N3A	N3A_subgraph_supervisor	Subgraph as a node (compiled graph reused inside another) · Supervisor pattern with LLM coordinator + specialist agents
N3B	N3B_swarm_send_api	Swarm — decentralized handoffs via Command(goto=) · Send API — dynamic fan-out to parallel workers (map-reduce)
N3C	N3C_long_term_memory	Store — cross-thread, cross-user memory · namespaces · combining checkpointer (short-term) with store (long-term)

Running

jupyter lab --notebook-dir=notebooks

Select the "LangGraph Patterns" kernel in each notebook. The first cell in every notebook installs dependencies automatically via %pip install.

Project structure

.
├── .env.example          # API key template
├── .gitignore
├── requirements.txt      # All dependencies
├── README.md
└── notebooks/
    ├── N1A_create_agent.ipynb
    ├── N1B_state_graph.ipynb
    ├── N1C_memory_saver.ipynb
    ├── N1D_conditional_edges.ipynb
    ├── N2A_state_routing.ipynb
    ├── N2B_hitl_streaming_memory.ipynb
    ├── N2C_patterns_advanced.ipynb
    ├── N3A_subgraph_supervisor.ipynb
    ├── N3B_swarm_send_api.ipynb
    └── N3C_long_term_memory.ipynb