Local AI Agent with LangGraph + Ollama (Full Tutorial, Qwen3)

🚀 Code: https://github.com/langchain-ai/langchain-academy/ Tired of API bills and vendor lock-in? In this tutorial, you’ll learn how to adapt a LangGraph agent to run entirely locally using the free, open-source Qwen3 model with Ollama. We'll take a standard LangChain Academy example built for GPT-4o and show you how to swap it out for a powerful local alternative, giving you full control over your data and costs. // WHAT YOU'LL LEARN Setting up a Python development environment with UV, a blazingly fast package manager. Installing and running powerful open-source models like Qwen3 locally with Ollama. Building a conversational agent with memory using LangGraph's StateGraph. Defining and binding tools (like a calculator) to your local LLM. Adapting existing LangChain code from proprietary to open-source models. Launching and interacting with your agent using LangGraph Studio. Observing and debugging your agent's thought process with LangSmith. Understanding the core components of an agentic graph: nodes, edges, and state. // RESOURCES LangChain Academy Course: https://academy.langchain.com/ LangGraph Docs: https://docs.langchain.com/oss/python/langgraph/overview Ollama & Qwen3 Model: https://ollama.com/ https://ollama.com/library/qwen3 UV Package Manager: https://github.com/astral-sh/uv // CHAPTERS 00:00 - Intro: Moving Beyond Proprietary Models 01:03 - The Stack: LangGraph, Ollama, and UV 01:14 - Environment Setup with UV Package Manager 01:50 - Installing Ollama & Pulling the Qwen3 Model 04:01 - Creating a Virtual Environment & Project Dependencies 05:35 - Code Deep Dive: From OpenAI to Ollama 06:21 - Building the Agent Graph (Nodes & Edges) 11:00 - Configuring LangSmith for Tracing 12:12 - Running the Agent in LangGraph Studio 12:46 - Live Demo: Testing the Agent's Memory 13:56 - Recap & Next Steps // NEXT STEPS If this helped, give the video a like and subscribe! What should we build next with this local stack? Let me know in the comments below. Discord