Last time, we discussed how to prepare the data for our question-generating system. To recap, here was the rough plan:

1. Fetch and prepare Wikipedia data (see more here)
2. Use an LLM to generate questions based on a given Wikipedia article (the focus of today’s post)

There are numerous LLM services available online, but for our purposes, we’ll use an open-source solution that can be run locally on our machine: ollama. This tool acts as a wrapper around popular, freely available models, providing a uniform API and CLI while managing model downloads for us.

I’ve always loved quizzes and trivia. As a kid, one of my favorite games was MindMaze. For those unfamiliar, MindMaze was a trivia game included in Microsoft’s Encarta, a digital multimedia encyclopedia. However, with the advent of Wikipedia, encyclopedias like Encarta became less popular. They were published in fixed, yearly iterations, distributed on CDs, and, most importantly, they cost money. The decline of Encarta also meant the end of MindMaze, which was disappointing for trivia enthusiasts like me.