A practical guide to running Google's Gemma 4 locally with Ollama, choosing the right model size, and building an AI agent with native function calling.
