langgraph

LangGraph Examples

These examples demonstrate the same patterns from PicoAgents implemented using LangGraph. LangGraph is LangChain's framework for building stateful, multi-actor applications with LLMs, using a graph-based approach.

Setup

# Install LangGraph
pip install langgraph langchain-openai python-dotenv

# Option 1: Azure OpenAI (used in examples)
export AZURE_OPENAI_ENDPOINT="https://your-endpoint.openai.azure.com/"
export AZURE_OPENAI_API_KEY="your-key"

# Option 2: OpenAI directly
export OPENAI_API_KEY="your-api-key"

Examples

Agents

Example	PicoAgents Equivalent	Description
agents/basic_agent.py	agents/basic-agent.py	ReAct agent with weather and calculator
agents/memory.py	agents/memory.py	Checkpointer-based conversation memory
agents/structured_output.py	agents/structured-output.py	Pydantic model responses

Workflows

Example	PicoAgents Equivalent	Description
workflows/sequential.py	workflows/sequential.py	Sequential node pipeline

Orchestration

Example	PicoAgents Equivalent	Description
orchestration/round_robin.py	orchestration/round-robin.py	Cyclic graph for agent turns
orchestration/supervisor.py	orchestration/supervisor.py	Supervisor-controlled delegation

Key Differences from PicoAgents

1. Graph-centric: Everything is nodes and edges in a StateGraph
2. Channels & Reducers: State is managed through typed channels with reducers
3. Checkpointing: Memory is handled via checkpointers (MemorySaver, SQLite, etc.)
4. Conditional routing: Edges can be conditional based on state
5. Built-in ReAct: create_react_agent provides tool-calling agents

Running Examples

# From the examples/frameworks/langgraph directory
python agents/basic_agent.py
python workflows/sequential.py
python orchestration/round_robin.py

Model Configuration

Examples use Azure OpenAI with gpt-4.1-mini via AzureChatOpenAI. To use OpenAI directly instead, change the import and client:

# Azure OpenAI (current)
from langchain_openai import AzureChatOpenAI
llm = AzureChatOpenAI(azure_deployment="gpt-4.1-mini", ...)

# OpenAI directly
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(model="gpt-4o-mini", ...)

Compatible models:

• gpt-4o-mini / gpt-4.1-mini - Fast, cost-effective
• gpt-4o - More capable for complex tasks
• Any LangChain-compatible model (Anthropic, Google, etc.)