Steam Game Recommender, with a RAG System That Focuses on Explainability First

During a steam sale event, I always find myself trying to look for new video games to play. If I wanted to find a new game I would try to whittle it down by using steam tags, but the steam tag system is very broad "action". could apply to many many games.

That got me thinking, what aspects do I like about my favorite games?

Well I like Persona 4 because of the city vibes and jazz fusion,

Spore because of the unique character creation and whimsical theme.

Balatro for its unique deck building synergies.

What if I could capture unique tags that identify a game that aren't just "action" and put them into vectors to show the (focus) of a game

For example I could break persona 4 into something like

Game play Focus vector: Day cycle 20% Dungeon crawling 20% Social sim 20%

Tags: Music: jazz fusion Vibe: Small rural town

I find that this system makes searching for games more "fun" now I can see why I like balatro. I like it because of the card synergies not so much for its rogue-like nature.

I also find that this helps find new underrated games, and beats the trap that Collaborative Filtering algorithms that get into where it "feels" like you get recommended the same things.