A hands-on workshop for building a framework-free AI coding agent with tools, an agent loop, and test-based verification.
The ReAct edition: the same LangGraph bug-fixing agent, driven by a model that reasons before it acts. One flag turns thinking on; the consequences — a loop guard for thinking, a lying trace, and the token bill — are the workshop. JetBrains Mellum2 Thinking via Ollama.
The same Python bug-fixing agent, rebuilt on LangGraph: which parts of a hand-written agent a framework absorbs, which parts it leaves you holding, and which parts it quietly breaks. You write the stop condition and the loop guard. Runs locally on JetBrains Mellum2 via Ollama.
Build a Python bug-fixing agent three times — as a plain loop, on LangGraph, and with a reasoning model. A JetBrains Academy course where you write the parts no framework can: the stop condition and the loop guards. Runs locally on JetBrains Mellum2 via Ollama.
To associate your repository with the mellum topic, visit your repo's landing page and select "manage topics."