Local LLM setup

For this part of Poké-Fun we need to access a Large Language Model (LLM). Although you can use a cloud solution, Ollama gives you the chance of running a simpler one locally, so you don't need to spend any money (nor send any potentially private information to a third party).

The easiest approach is to follow the official download instructions — although in macOS you can also use brew, and in Linux you can use your distribution's package manager. It's not needed to keep Ollama running as a background service, you can just fire it up using ollama serve whenever you need it.

To install a new model, you can use ollama run <model-name>. For Poké-Fun it's enough to use Gemma 4 with Effective 4B parameter (gemma4:e4b), which you can get up and running using ollama run gemma4:e4b. The first time it needs to download a big chunk of data.

By default, you get a small chat interface. You can try some queries, and then exit the prompt by typing /bye.

>>> What is a good name for a workshop about Kotlin, Arrow, Compose Multiplatform, and AI?
Thinking...
Here's a thinking process that leads to the suggested names:

[ ... ]

## 🚀 Catchy & Marketing Names (For Event Websites)
*These are designed to sound exciting, future-proof, and appeal broadly.*

1. **CognitoFlow:** (A play on "cognition" + flow). *High concept, suggests intelligent design.*
2. **The Hyper-Intelligent Stack.**
3. **Digital Brains: Building AI-Powered Multiplatform Apps.**
4. **NextGen UI/UX: Functional Apps Meeting Intelligence.**
5. **Code X Future:** Kotlin, Arrow, and the Age of AI.

[...]

>>> /bye

The model is still waiting in the background for more queries, though; you can see the running models with ollama ps in a terminal, and stop one of them using ollama stop <model-name>.